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> | |
1da177e4 LT |
122 | #include <linux/security.h> |
123 | #include <linux/slab.h> | |
1da177e4 LT |
124 | #include <linux/syscalls.h> |
125 | #include <linux/time.h> | |
4fb3a538 | 126 | #include <linux/rcupdate.h> |
ab1f1611 | 127 | #include <linux/pid_namespace.h> |
48f74186 | 128 | #include <linux/hashtable.h> |
7012b02a | 129 | #include <linux/percpu.h> |
1da177e4 | 130 | |
62af4f1f JL |
131 | #define CREATE_TRACE_POINTS |
132 | #include <trace/events/filelock.h> | |
133 | ||
7c0f6ba6 | 134 | #include <linux/uaccess.h> |
1da177e4 LT |
135 | |
136 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
137 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
11afe9f7 | 138 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) |
cff2fce5 | 139 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
9d5b86ac | 140 | #define IS_REMOTELCK(fl) (fl->fl_pid <= 0) |
1da177e4 | 141 | |
ab83fa4b BF |
142 | static bool lease_breaking(struct file_lock *fl) |
143 | { | |
778fc546 BF |
144 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
145 | } | |
146 | ||
147 | static int target_leasetype(struct file_lock *fl) | |
148 | { | |
149 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
150 | return F_UNLCK; | |
151 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
152 | return F_RDLCK; | |
153 | return fl->fl_type; | |
ab83fa4b BF |
154 | } |
155 | ||
1da177e4 LT |
156 | int leases_enable = 1; |
157 | int lease_break_time = 45; | |
158 | ||
1c8c601a | 159 | /* |
7012b02a | 160 | * The global file_lock_list is only used for displaying /proc/locks, so we |
7c3f654d PZ |
161 | * keep a list on each CPU, with each list protected by its own spinlock. |
162 | * Global serialization is done using file_rwsem. | |
163 | * | |
164 | * Note that alterations to the list also require that the relevant flc_lock is | |
165 | * held. | |
1c8c601a | 166 | */ |
7c3f654d PZ |
167 | struct file_lock_list_struct { |
168 | spinlock_t lock; | |
169 | struct hlist_head hlist; | |
170 | }; | |
171 | static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list); | |
aba37660 | 172 | DEFINE_STATIC_PERCPU_RWSEM(file_rwsem); |
88974691 | 173 | |
1c8c601a | 174 | /* |
48f74186 | 175 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 176 | * It is protected by blocked_lock_lock. |
48f74186 JL |
177 | * |
178 | * We hash locks by lockowner in order to optimize searching for the lock a | |
179 | * particular lockowner is waiting on. | |
180 | * | |
181 | * FIXME: make this value scale via some heuristic? We generally will want more | |
182 | * buckets when we have more lockowners holding locks, but that's a little | |
183 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 184 | */ |
48f74186 JL |
185 | #define BLOCKED_HASH_BITS 7 |
186 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 187 | |
1c8c601a | 188 | /* |
7b2296af JL |
189 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
190 | * want to be holding this lock. | |
1c8c601a | 191 | * |
ada5c1da N |
192 | * In addition, it also protects the fl->fl_blocked_requests list, and the |
193 | * fl->fl_blocker pointer for file_lock structures that are acting as lock | |
194 | * requests (in contrast to those that are acting as records of acquired locks). | |
1c8c601a JL |
195 | * |
196 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 197 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 198 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 199 | * flc_lock. |
1c8c601a | 200 | */ |
7b2296af | 201 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 202 | |
4a075e39 | 203 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 204 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 205 | |
4a075e39 | 206 | static struct file_lock_context * |
5c1c669a | 207 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 208 | { |
128a3785 | 209 | struct file_lock_context *ctx; |
4a075e39 | 210 | |
128a3785 DV |
211 | /* paired with cmpxchg() below */ |
212 | ctx = smp_load_acquire(&inode->i_flctx); | |
213 | if (likely(ctx) || type == F_UNLCK) | |
4a075e39 JL |
214 | goto out; |
215 | ||
128a3785 DV |
216 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
217 | if (!ctx) | |
4a075e39 JL |
218 | goto out; |
219 | ||
128a3785 DV |
220 | spin_lock_init(&ctx->flc_lock); |
221 | INIT_LIST_HEAD(&ctx->flc_flock); | |
222 | INIT_LIST_HEAD(&ctx->flc_posix); | |
223 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
224 | |
225 | /* | |
226 | * Assign the pointer if it's not already assigned. If it is, then | |
227 | * free the context we just allocated. | |
228 | */ | |
128a3785 DV |
229 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
230 | kmem_cache_free(flctx_cache, ctx); | |
231 | ctx = smp_load_acquire(&inode->i_flctx); | |
232 | } | |
4a075e39 | 233 | out: |
1890910f | 234 | trace_locks_get_lock_context(inode, type, ctx); |
128a3785 | 235 | return ctx; |
4a075e39 JL |
236 | } |
237 | ||
e24dadab JL |
238 | static void |
239 | locks_dump_ctx_list(struct list_head *list, char *list_type) | |
240 | { | |
241 | struct file_lock *fl; | |
242 | ||
243 | list_for_each_entry(fl, list, fl_list) { | |
244 | pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type, fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
245 | } | |
246 | } | |
247 | ||
248 | static void | |
249 | locks_check_ctx_lists(struct inode *inode) | |
250 | { | |
251 | struct file_lock_context *ctx = inode->i_flctx; | |
252 | ||
253 | if (unlikely(!list_empty(&ctx->flc_flock) || | |
254 | !list_empty(&ctx->flc_posix) || | |
255 | !list_empty(&ctx->flc_lease))) { | |
256 | pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n", | |
257 | MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), | |
258 | inode->i_ino); | |
259 | locks_dump_ctx_list(&ctx->flc_flock, "FLOCK"); | |
260 | locks_dump_ctx_list(&ctx->flc_posix, "POSIX"); | |
261 | locks_dump_ctx_list(&ctx->flc_lease, "LEASE"); | |
262 | } | |
263 | } | |
264 | ||
3953704f BC |
265 | static void |
266 | locks_check_ctx_file_list(struct file *filp, struct list_head *list, | |
267 | char *list_type) | |
268 | { | |
269 | struct file_lock *fl; | |
270 | struct inode *inode = locks_inode(filp); | |
271 | ||
272 | list_for_each_entry(fl, list, fl_list) | |
273 | if (fl->fl_file == filp) | |
274 | pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx " | |
275 | " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", | |
276 | list_type, MAJOR(inode->i_sb->s_dev), | |
277 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
278 | fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
279 | } | |
280 | ||
4a075e39 | 281 | void |
f27a0fe0 | 282 | locks_free_lock_context(struct inode *inode) |
4a075e39 | 283 | { |
f27a0fe0 JL |
284 | struct file_lock_context *ctx = inode->i_flctx; |
285 | ||
e24dadab JL |
286 | if (unlikely(ctx)) { |
287 | locks_check_ctx_lists(inode); | |
4a075e39 JL |
288 | kmem_cache_free(flctx_cache, ctx); |
289 | } | |
290 | } | |
291 | ||
ee19cc40 | 292 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 293 | { |
139ca04e | 294 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 295 | INIT_LIST_HEAD(&fl->fl_list); |
ada5c1da N |
296 | INIT_LIST_HEAD(&fl->fl_blocked_requests); |
297 | INIT_LIST_HEAD(&fl->fl_blocked_member); | |
ee19cc40 | 298 | init_waitqueue_head(&fl->fl_wait); |
a51cb91d MS |
299 | } |
300 | ||
1da177e4 | 301 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 302 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 303 | { |
ee19cc40 | 304 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
305 | |
306 | if (fl) | |
ee19cc40 | 307 | locks_init_lock_heads(fl); |
a51cb91d MS |
308 | |
309 | return fl; | |
1da177e4 | 310 | } |
c5b1f0d9 | 311 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 312 | |
a9e61e25 | 313 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
314 | { |
315 | if (fl->fl_ops) { | |
316 | if (fl->fl_ops->fl_release_private) | |
317 | fl->fl_ops->fl_release_private(fl); | |
318 | fl->fl_ops = NULL; | |
319 | } | |
47831f35 | 320 | |
5c97d7b1 | 321 | if (fl->fl_lmops) { |
cae80b30 JL |
322 | if (fl->fl_lmops->lm_put_owner) { |
323 | fl->fl_lmops->lm_put_owner(fl->fl_owner); | |
324 | fl->fl_owner = NULL; | |
325 | } | |
5c97d7b1 KM |
326 | fl->fl_lmops = NULL; |
327 | } | |
47831f35 | 328 | } |
a9e61e25 | 329 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 330 | |
1da177e4 | 331 | /* Free a lock which is not in use. */ |
05fa3135 | 332 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 333 | { |
5ce29646 | 334 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 335 | BUG_ON(!list_empty(&fl->fl_list)); |
ada5c1da N |
336 | BUG_ON(!list_empty(&fl->fl_blocked_requests)); |
337 | BUG_ON(!list_empty(&fl->fl_blocked_member)); | |
139ca04e | 338 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 339 | |
47831f35 | 340 | locks_release_private(fl); |
1da177e4 LT |
341 | kmem_cache_free(filelock_cache, fl); |
342 | } | |
05fa3135 | 343 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 344 | |
ed9814d8 JL |
345 | static void |
346 | locks_dispose_list(struct list_head *dispose) | |
347 | { | |
348 | struct file_lock *fl; | |
349 | ||
350 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
351 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
352 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
353 | locks_free_lock(fl); |
354 | } | |
355 | } | |
356 | ||
1da177e4 LT |
357 | void locks_init_lock(struct file_lock *fl) |
358 | { | |
ee19cc40 MS |
359 | memset(fl, 0, sizeof(struct file_lock)); |
360 | locks_init_lock_heads(fl); | |
1da177e4 LT |
361 | } |
362 | ||
363 | EXPORT_SYMBOL(locks_init_lock); | |
364 | ||
1da177e4 LT |
365 | /* |
366 | * Initialize a new lock from an existing file_lock structure. | |
367 | */ | |
3fe0fff1 | 368 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
369 | { |
370 | new->fl_owner = fl->fl_owner; | |
371 | new->fl_pid = fl->fl_pid; | |
0996905f | 372 | new->fl_file = NULL; |
1da177e4 LT |
373 | new->fl_flags = fl->fl_flags; |
374 | new->fl_type = fl->fl_type; | |
375 | new->fl_start = fl->fl_start; | |
376 | new->fl_end = fl->fl_end; | |
f328296e | 377 | new->fl_lmops = fl->fl_lmops; |
0996905f | 378 | new->fl_ops = NULL; |
f328296e KM |
379 | |
380 | if (fl->fl_lmops) { | |
381 | if (fl->fl_lmops->lm_get_owner) | |
cae80b30 | 382 | fl->fl_lmops->lm_get_owner(fl->fl_owner); |
f328296e | 383 | } |
0996905f | 384 | } |
3fe0fff1 | 385 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
386 | |
387 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
388 | { | |
566709bd JL |
389 | /* "new" must be a freshly-initialized lock */ |
390 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 391 | |
3fe0fff1 | 392 | locks_copy_conflock(new, fl); |
f328296e | 393 | |
0996905f | 394 | new->fl_file = fl->fl_file; |
1da177e4 | 395 | new->fl_ops = fl->fl_ops; |
47831f35 | 396 | |
f328296e KM |
397 | if (fl->fl_ops) { |
398 | if (fl->fl_ops->fl_copy_lock) | |
399 | fl->fl_ops->fl_copy_lock(new, fl); | |
400 | } | |
1da177e4 LT |
401 | } |
402 | ||
403 | EXPORT_SYMBOL(locks_copy_lock); | |
404 | ||
5946c431 N |
405 | static void locks_move_blocks(struct file_lock *new, struct file_lock *fl) |
406 | { | |
407 | struct file_lock *f; | |
408 | ||
409 | /* | |
410 | * As ctx->flc_lock is held, new requests cannot be added to | |
411 | * ->fl_blocked_requests, so we don't need a lock to check if it | |
412 | * is empty. | |
413 | */ | |
414 | if (list_empty(&fl->fl_blocked_requests)) | |
415 | return; | |
416 | spin_lock(&blocked_lock_lock); | |
417 | list_splice_init(&fl->fl_blocked_requests, &new->fl_blocked_requests); | |
418 | list_for_each_entry(f, &fl->fl_blocked_requests, fl_blocked_member) | |
419 | f->fl_blocker = new; | |
420 | spin_unlock(&blocked_lock_lock); | |
421 | } | |
422 | ||
1da177e4 LT |
423 | static inline int flock_translate_cmd(int cmd) { |
424 | if (cmd & LOCK_MAND) | |
425 | return cmd & (LOCK_MAND | LOCK_RW); | |
426 | switch (cmd) { | |
427 | case LOCK_SH: | |
428 | return F_RDLCK; | |
429 | case LOCK_EX: | |
430 | return F_WRLCK; | |
431 | case LOCK_UN: | |
432 | return F_UNLCK; | |
433 | } | |
434 | return -EINVAL; | |
435 | } | |
436 | ||
437 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 | 438 | static struct file_lock * |
d6367d62 | 439 | flock_make_lock(struct file *filp, unsigned int cmd, struct file_lock *fl) |
1da177e4 | 440 | { |
1da177e4 | 441 | int type = flock_translate_cmd(cmd); |
6e129d00 | 442 | |
1da177e4 | 443 | if (type < 0) |
6e129d00 | 444 | return ERR_PTR(type); |
1da177e4 | 445 | |
d6367d62 N |
446 | if (fl == NULL) { |
447 | fl = locks_alloc_lock(); | |
448 | if (fl == NULL) | |
449 | return ERR_PTR(-ENOMEM); | |
450 | } else { | |
451 | locks_init_lock(fl); | |
452 | } | |
1da177e4 LT |
453 | |
454 | fl->fl_file = filp; | |
73a8f5f7 | 455 | fl->fl_owner = filp; |
1da177e4 LT |
456 | fl->fl_pid = current->tgid; |
457 | fl->fl_flags = FL_FLOCK; | |
458 | fl->fl_type = type; | |
459 | fl->fl_end = OFFSET_MAX; | |
460 | ||
6e129d00 | 461 | return fl; |
1da177e4 LT |
462 | } |
463 | ||
0ec4f431 | 464 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
465 | { |
466 | switch (type) { | |
467 | case F_RDLCK: | |
468 | case F_WRLCK: | |
469 | case F_UNLCK: | |
470 | fl->fl_type = type; | |
471 | break; | |
472 | default: | |
473 | return -EINVAL; | |
474 | } | |
475 | return 0; | |
476 | } | |
477 | ||
ef12e72a BF |
478 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
479 | struct flock64 *l) | |
1da177e4 | 480 | { |
1da177e4 | 481 | switch (l->l_whence) { |
f5579f8c | 482 | case SEEK_SET: |
ef12e72a | 483 | fl->fl_start = 0; |
1da177e4 | 484 | break; |
f5579f8c | 485 | case SEEK_CUR: |
ef12e72a | 486 | fl->fl_start = filp->f_pos; |
1da177e4 | 487 | break; |
f5579f8c | 488 | case SEEK_END: |
ef12e72a | 489 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
490 | break; |
491 | default: | |
492 | return -EINVAL; | |
493 | } | |
ef12e72a BF |
494 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
495 | return -EOVERFLOW; | |
496 | fl->fl_start += l->l_start; | |
497 | if (fl->fl_start < 0) | |
498 | return -EINVAL; | |
1da177e4 LT |
499 | |
500 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
501 | POSIX-2001 defines it. */ | |
4c780a46 | 502 | if (l->l_len > 0) { |
ef12e72a BF |
503 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
504 | return -EOVERFLOW; | |
505 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
506 | ||
4c780a46 | 507 | } else if (l->l_len < 0) { |
ef12e72a | 508 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 509 | return -EINVAL; |
ef12e72a BF |
510 | fl->fl_end = fl->fl_start - 1; |
511 | fl->fl_start += l->l_len; | |
512 | } else | |
513 | fl->fl_end = OFFSET_MAX; | |
514 | ||
1da177e4 LT |
515 | fl->fl_owner = current->files; |
516 | fl->fl_pid = current->tgid; | |
517 | fl->fl_file = filp; | |
518 | fl->fl_flags = FL_POSIX; | |
519 | fl->fl_ops = NULL; | |
520 | fl->fl_lmops = NULL; | |
521 | ||
522 | return assign_type(fl, l->l_type); | |
523 | } | |
524 | ||
ef12e72a BF |
525 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
526 | * style lock. | |
527 | */ | |
528 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
529 | struct flock *l) | |
1da177e4 | 530 | { |
ef12e72a BF |
531 | struct flock64 ll = { |
532 | .l_type = l->l_type, | |
533 | .l_whence = l->l_whence, | |
534 | .l_start = l->l_start, | |
535 | .l_len = l->l_len, | |
536 | }; | |
537 | ||
538 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 539 | } |
1da177e4 LT |
540 | |
541 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
542 | static bool |
543 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
544 | { |
545 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 546 | return false; |
1da177e4 LT |
547 | } |
548 | ||
1c7dd2ff JL |
549 | static void |
550 | lease_setup(struct file_lock *fl, void **priv) | |
551 | { | |
552 | struct file *filp = fl->fl_file; | |
553 | struct fasync_struct *fa = *priv; | |
554 | ||
555 | /* | |
556 | * fasync_insert_entry() returns the old entry if any. If there was no | |
557 | * old entry, then it used "priv" and inserted it into the fasync list. | |
558 | * Clear the pointer to indicate that it shouldn't be freed. | |
559 | */ | |
560 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
561 | *priv = NULL; | |
562 | ||
01919134 | 563 | __f_setown(filp, task_pid(current), PIDTYPE_TGID, 0); |
1c7dd2ff JL |
564 | } |
565 | ||
7b021967 | 566 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 567 | .lm_break = lease_break_callback, |
8fb47a4f | 568 | .lm_change = lease_modify, |
1c7dd2ff | 569 | .lm_setup = lease_setup, |
1da177e4 LT |
570 | }; |
571 | ||
572 | /* | |
573 | * Initialize a lease, use the default lock manager operations | |
574 | */ | |
0ec4f431 | 575 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
447a5647 | 576 | { |
75dff55a TM |
577 | if (assign_type(fl, type) != 0) |
578 | return -EINVAL; | |
579 | ||
7ca76311 | 580 | fl->fl_owner = filp; |
1da177e4 LT |
581 | fl->fl_pid = current->tgid; |
582 | ||
583 | fl->fl_file = filp; | |
584 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
585 | fl->fl_start = 0; |
586 | fl->fl_end = OFFSET_MAX; | |
587 | fl->fl_ops = NULL; | |
588 | fl->fl_lmops = &lease_manager_ops; | |
589 | return 0; | |
590 | } | |
591 | ||
592 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 593 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
594 | { |
595 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 596 | int error = -ENOMEM; |
1da177e4 LT |
597 | |
598 | if (fl == NULL) | |
e32b8ee2 | 599 | return ERR_PTR(error); |
1da177e4 LT |
600 | |
601 | error = lease_init(filp, type, fl); | |
75dff55a TM |
602 | if (error) { |
603 | locks_free_lock(fl); | |
e32b8ee2 | 604 | return ERR_PTR(error); |
75dff55a | 605 | } |
e32b8ee2 | 606 | return fl; |
1da177e4 LT |
607 | } |
608 | ||
609 | /* Check if two locks overlap each other. | |
610 | */ | |
611 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
612 | { | |
613 | return ((fl1->fl_end >= fl2->fl_start) && | |
614 | (fl2->fl_end >= fl1->fl_start)); | |
615 | } | |
616 | ||
617 | /* | |
618 | * Check whether two locks have the same owner. | |
619 | */ | |
33443c42 | 620 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 621 | { |
8fb47a4f | 622 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 623 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 624 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
625 | return fl1->fl_owner == fl2->fl_owner; |
626 | } | |
627 | ||
6109c850 | 628 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 629 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 630 | { |
7c3f654d PZ |
631 | struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list); |
632 | ||
aba37660 PZ |
633 | percpu_rwsem_assert_held(&file_rwsem); |
634 | ||
7c3f654d | 635 | spin_lock(&fll->lock); |
7012b02a | 636 | fl->fl_link_cpu = smp_processor_id(); |
7c3f654d PZ |
637 | hlist_add_head(&fl->fl_link, &fll->hlist); |
638 | spin_unlock(&fll->lock); | |
88974691 JL |
639 | } |
640 | ||
6109c850 | 641 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 642 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 643 | { |
7c3f654d PZ |
644 | struct file_lock_list_struct *fll; |
645 | ||
aba37660 PZ |
646 | percpu_rwsem_assert_held(&file_rwsem); |
647 | ||
7012b02a JL |
648 | /* |
649 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 650 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
651 | * also require that it be held. |
652 | */ | |
653 | if (hlist_unhashed(&fl->fl_link)) | |
654 | return; | |
7c3f654d PZ |
655 | |
656 | fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu); | |
657 | spin_lock(&fll->lock); | |
139ca04e | 658 | hlist_del_init(&fl->fl_link); |
7c3f654d | 659 | spin_unlock(&fll->lock); |
88974691 JL |
660 | } |
661 | ||
3999e493 JL |
662 | static unsigned long |
663 | posix_owner_key(struct file_lock *fl) | |
664 | { | |
665 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
666 | return fl->fl_lmops->lm_owner_key(fl); | |
667 | return (unsigned long)fl->fl_owner; | |
668 | } | |
669 | ||
6ca10ed8 | 670 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 671 | { |
663d5af7 DW |
672 | lockdep_assert_held(&blocked_lock_lock); |
673 | ||
3999e493 | 674 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
675 | } |
676 | ||
6ca10ed8 | 677 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 678 | { |
663d5af7 DW |
679 | lockdep_assert_held(&blocked_lock_lock); |
680 | ||
48f74186 | 681 | hash_del(&waiter->fl_link); |
88974691 JL |
682 | } |
683 | ||
1da177e4 LT |
684 | /* Remove waiter from blocker's block list. |
685 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 686 | * |
7b2296af | 687 | * Must be called with blocked_lock_lock held. |
1da177e4 | 688 | */ |
33443c42 | 689 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 690 | { |
88974691 | 691 | locks_delete_global_blocked(waiter); |
ada5c1da N |
692 | list_del_init(&waiter->fl_blocked_member); |
693 | waiter->fl_blocker = NULL; | |
1da177e4 LT |
694 | } |
695 | ||
ad6bbd8b N |
696 | static void __locks_wake_up_blocks(struct file_lock *blocker) |
697 | { | |
698 | while (!list_empty(&blocker->fl_blocked_requests)) { | |
699 | struct file_lock *waiter; | |
700 | ||
701 | waiter = list_first_entry(&blocker->fl_blocked_requests, | |
702 | struct file_lock, fl_blocked_member); | |
703 | __locks_delete_block(waiter); | |
704 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) | |
705 | waiter->fl_lmops->lm_notify(waiter); | |
706 | else | |
707 | wake_up(&waiter->fl_wait); | |
708 | } | |
709 | } | |
710 | ||
1a9e64a7 | 711 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 712 | { |
7b2296af | 713 | spin_lock(&blocked_lock_lock); |
5946c431 | 714 | __locks_wake_up_blocks(waiter); |
1da177e4 | 715 | __locks_delete_block(waiter); |
7b2296af | 716 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
717 | } |
718 | ||
719 | /* Insert waiter into blocker's block list. | |
720 | * We use a circular list so that processes can be easily woken up in | |
721 | * the order they blocked. The documentation doesn't require this but | |
722 | * it seems like the reasonable thing to do. | |
1c8c601a | 723 | * |
6109c850 | 724 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
ada5c1da N |
725 | * fl_blocked_requests list itself is protected by the blocked_lock_lock, |
726 | * but by ensuring that the flc_lock is also held on insertions we can avoid | |
727 | * taking the blocked_lock_lock in some cases when we see that the | |
728 | * fl_blocked_requests list is empty. | |
1da177e4 | 729 | */ |
1c8c601a JL |
730 | static void __locks_insert_block(struct file_lock *blocker, |
731 | struct file_lock *waiter) | |
1da177e4 | 732 | { |
ada5c1da N |
733 | BUG_ON(!list_empty(&waiter->fl_blocked_member)); |
734 | waiter->fl_blocker = blocker; | |
735 | list_add_tail(&waiter->fl_blocked_member, &blocker->fl_blocked_requests); | |
cff2fce5 | 736 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a | 737 | locks_insert_global_blocked(waiter); |
5946c431 N |
738 | |
739 | /* The requests in waiter->fl_blocked are known to conflict with | |
740 | * waiter, but might not conflict with blocker, or the requests | |
741 | * and lock which block it. So they all need to be woken. | |
742 | */ | |
743 | __locks_wake_up_blocks(waiter); | |
1c8c601a JL |
744 | } |
745 | ||
6109c850 | 746 | /* Must be called with flc_lock held. */ |
1c8c601a JL |
747 | static void locks_insert_block(struct file_lock *blocker, |
748 | struct file_lock *waiter) | |
749 | { | |
7b2296af | 750 | spin_lock(&blocked_lock_lock); |
1c8c601a | 751 | __locks_insert_block(blocker, waiter); |
7b2296af | 752 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
753 | } |
754 | ||
1cb36012 JL |
755 | /* |
756 | * Wake up processes blocked waiting for blocker. | |
757 | * | |
6109c850 | 758 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
759 | */ |
760 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
761 | { | |
4e8c765d JL |
762 | /* |
763 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 | 764 | * blocked requests are only added to the list under the flc_lock, and |
ada5c1da N |
765 | * the flc_lock is always held here. Note that removal from the |
766 | * fl_blocked_requests list does not require the flc_lock, so we must | |
767 | * recheck list_empty() after acquiring the blocked_lock_lock. | |
4e8c765d | 768 | */ |
ada5c1da | 769 | if (list_empty(&blocker->fl_blocked_requests)) |
4e8c765d JL |
770 | return; |
771 | ||
7b2296af | 772 | spin_lock(&blocked_lock_lock); |
ad6bbd8b | 773 | __locks_wake_up_blocks(blocker); |
7b2296af | 774 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
775 | } |
776 | ||
5263e31e | 777 | static void |
e084c1bd | 778 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e | 779 | { |
5263e31e JL |
780 | list_add_tail(&fl->fl_list, before); |
781 | locks_insert_global_locks(fl); | |
782 | } | |
783 | ||
8634b51f | 784 | static void |
e084c1bd | 785 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 786 | { |
88974691 | 787 | locks_delete_global_locks(fl); |
8634b51f | 788 | list_del_init(&fl->fl_list); |
1da177e4 | 789 | locks_wake_up_blocks(fl); |
24cbe784 JL |
790 | } |
791 | ||
8634b51f | 792 | static void |
e084c1bd | 793 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 794 | { |
e084c1bd | 795 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 796 | if (dispose) |
6dee60f6 | 797 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
798 | else |
799 | locks_free_lock(fl); | |
1da177e4 LT |
800 | } |
801 | ||
802 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
803 | * checks for shared/exclusive status of overlapping locks. | |
804 | */ | |
805 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
806 | { | |
807 | if (sys_fl->fl_type == F_WRLCK) | |
808 | return 1; | |
809 | if (caller_fl->fl_type == F_WRLCK) | |
810 | return 1; | |
811 | return 0; | |
812 | } | |
813 | ||
814 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
815 | * checking before calling the locks_conflict(). | |
816 | */ | |
817 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
818 | { | |
819 | /* POSIX locks owned by the same process do not conflict with | |
820 | * each other. | |
821 | */ | |
9b8c8695 | 822 | if (posix_same_owner(caller_fl, sys_fl)) |
1da177e4 LT |
823 | return (0); |
824 | ||
825 | /* Check whether they overlap */ | |
826 | if (!locks_overlap(caller_fl, sys_fl)) | |
827 | return 0; | |
828 | ||
829 | return (locks_conflict(caller_fl, sys_fl)); | |
830 | } | |
831 | ||
832 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
833 | * checking before calling the locks_conflict(). | |
834 | */ | |
835 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
836 | { | |
837 | /* FLOCK locks referring to the same filp do not conflict with | |
838 | * each other. | |
839 | */ | |
9b8c8695 | 840 | if (caller_fl->fl_file == sys_fl->fl_file) |
1da177e4 LT |
841 | return (0); |
842 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
843 | return 0; | |
844 | ||
845 | return (locks_conflict(caller_fl, sys_fl)); | |
846 | } | |
847 | ||
6d34ac19 | 848 | void |
9d6a8c5c | 849 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
850 | { |
851 | struct file_lock *cfl; | |
bd61e0a9 | 852 | struct file_lock_context *ctx; |
c568d683 | 853 | struct inode *inode = locks_inode(filp); |
1da177e4 | 854 | |
128a3785 | 855 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
856 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
857 | fl->fl_type = F_UNLCK; | |
858 | return; | |
859 | } | |
860 | ||
6109c850 | 861 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
862 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
863 | if (posix_locks_conflict(fl, cfl)) { | |
864 | locks_copy_conflock(fl, cfl); | |
bd61e0a9 JL |
865 | goto out; |
866 | } | |
1da177e4 | 867 | } |
bd61e0a9 JL |
868 | fl->fl_type = F_UNLCK; |
869 | out: | |
6109c850 | 870 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 871 | return; |
1da177e4 | 872 | } |
1da177e4 LT |
873 | EXPORT_SYMBOL(posix_test_lock); |
874 | ||
b533184f BF |
875 | /* |
876 | * Deadlock detection: | |
877 | * | |
878 | * We attempt to detect deadlocks that are due purely to posix file | |
879 | * locks. | |
1da177e4 | 880 | * |
b533184f BF |
881 | * We assume that a task can be waiting for at most one lock at a time. |
882 | * So for any acquired lock, the process holding that lock may be | |
883 | * waiting on at most one other lock. That lock in turns may be held by | |
884 | * someone waiting for at most one other lock. Given a requested lock | |
885 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
886 | * follow this chain of waiters to ensure we are not about to create a | |
887 | * cycle. | |
1da177e4 | 888 | * |
b533184f BF |
889 | * Since we do this before we ever put a process to sleep on a lock, we |
890 | * are ensured that there is never a cycle; that is what guarantees that | |
891 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 892 | * |
b533184f BF |
893 | * Note: the above assumption may not be true when handling lock |
894 | * requests from a broken NFS client. It may also fail in the presence | |
895 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 896 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 897 | * |
cff2fce5 | 898 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
899 | * Because the owner is not even nominally tied to a thread of |
900 | * execution, the deadlock detection below can't reasonably work well. Just | |
901 | * skip it for those. | |
902 | * | |
cff2fce5 | 903 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
904 | * locks that just checks for the case where two tasks are attempting to |
905 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 906 | */ |
97855b49 BF |
907 | |
908 | #define MAX_DEADLK_ITERATIONS 10 | |
909 | ||
b533184f BF |
910 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
911 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
912 | { | |
913 | struct file_lock *fl; | |
914 | ||
3999e493 | 915 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
5946c431 N |
916 | if (posix_same_owner(fl, block_fl)) { |
917 | while (fl->fl_blocker) | |
918 | fl = fl->fl_blocker; | |
919 | return fl; | |
920 | } | |
b533184f BF |
921 | } |
922 | return NULL; | |
923 | } | |
924 | ||
7b2296af | 925 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 926 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
927 | struct file_lock *block_fl) |
928 | { | |
97855b49 | 929 | int i = 0; |
1da177e4 | 930 | |
663d5af7 DW |
931 | lockdep_assert_held(&blocked_lock_lock); |
932 | ||
57b65325 JL |
933 | /* |
934 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 935 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 936 | */ |
cff2fce5 | 937 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
938 | return 0; |
939 | ||
b533184f BF |
940 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
941 | if (i++ > MAX_DEADLK_ITERATIONS) | |
942 | return 0; | |
943 | if (posix_same_owner(caller_fl, block_fl)) | |
944 | return 1; | |
1da177e4 LT |
945 | } |
946 | return 0; | |
947 | } | |
948 | ||
1da177e4 | 949 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 950 | * after any leases, but before any posix locks. |
f475ae95 TM |
951 | * |
952 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
953 | * whether or not a lock was successfully freed by testing the return | |
954 | * value for -ENOENT. | |
1da177e4 | 955 | */ |
bcd7f78d | 956 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 957 | { |
993dfa87 | 958 | struct file_lock *new_fl = NULL; |
5263e31e JL |
959 | struct file_lock *fl; |
960 | struct file_lock_context *ctx; | |
1da177e4 | 961 | int error = 0; |
5263e31e | 962 | bool found = false; |
ed9814d8 | 963 | LIST_HEAD(dispose); |
1da177e4 | 964 | |
5c1c669a JL |
965 | ctx = locks_get_lock_context(inode, request->fl_type); |
966 | if (!ctx) { | |
967 | if (request->fl_type != F_UNLCK) | |
968 | return -ENOMEM; | |
969 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
970 | } | |
5263e31e | 971 | |
b89f4321 | 972 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 973 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
974 | if (!new_fl) |
975 | return -ENOMEM; | |
84d535ad PE |
976 | } |
977 | ||
87709e28 | 978 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 979 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
980 | if (request->fl_flags & FL_ACCESS) |
981 | goto find_conflict; | |
982 | ||
5263e31e | 983 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 984 | if (request->fl_file != fl->fl_file) |
1da177e4 | 985 | continue; |
993dfa87 | 986 | if (request->fl_type == fl->fl_type) |
1da177e4 | 987 | goto out; |
5263e31e | 988 | found = true; |
e084c1bd | 989 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
990 | break; |
991 | } | |
1da177e4 | 992 | |
f475ae95 TM |
993 | if (request->fl_type == F_UNLCK) { |
994 | if ((request->fl_flags & FL_EXISTS) && !found) | |
995 | error = -ENOENT; | |
993dfa87 | 996 | goto out; |
f475ae95 | 997 | } |
1da177e4 | 998 | |
f07f18dd | 999 | find_conflict: |
5263e31e | 1000 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 1001 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
1002 | continue; |
1003 | error = -EAGAIN; | |
bde74e4b MS |
1004 | if (!(request->fl_flags & FL_SLEEP)) |
1005 | goto out; | |
1006 | error = FILE_LOCK_DEFERRED; | |
1007 | locks_insert_block(fl, request); | |
1da177e4 LT |
1008 | goto out; |
1009 | } | |
f07f18dd TM |
1010 | if (request->fl_flags & FL_ACCESS) |
1011 | goto out; | |
993dfa87 | 1012 | locks_copy_lock(new_fl, request); |
5946c431 | 1013 | locks_move_blocks(new_fl, request); |
e084c1bd | 1014 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 1015 | new_fl = NULL; |
9cedc194 | 1016 | error = 0; |
1da177e4 LT |
1017 | |
1018 | out: | |
6109c850 | 1019 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1020 | percpu_up_read_preempt_enable(&file_rwsem); |
993dfa87 TM |
1021 | if (new_fl) |
1022 | locks_free_lock(new_fl); | |
ed9814d8 | 1023 | locks_dispose_list(&dispose); |
c883da31 | 1024 | trace_flock_lock_inode(inode, request, error); |
1da177e4 LT |
1025 | return error; |
1026 | } | |
1027 | ||
b4d629a3 JL |
1028 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
1029 | struct file_lock *conflock) | |
1da177e4 | 1030 | { |
bd61e0a9 | 1031 | struct file_lock *fl, *tmp; |
39005d02 MS |
1032 | struct file_lock *new_fl = NULL; |
1033 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
1034 | struct file_lock *left = NULL; |
1035 | struct file_lock *right = NULL; | |
bd61e0a9 | 1036 | struct file_lock_context *ctx; |
b9746ef8 JL |
1037 | int error; |
1038 | bool added = false; | |
ed9814d8 | 1039 | LIST_HEAD(dispose); |
1da177e4 | 1040 | |
5c1c669a | 1041 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 1042 | if (!ctx) |
5c1c669a | 1043 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 1044 | |
1da177e4 LT |
1045 | /* |
1046 | * We may need two file_lock structures for this operation, | |
1047 | * so we get them in advance to avoid races. | |
39005d02 MS |
1048 | * |
1049 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1050 | */ |
39005d02 MS |
1051 | if (!(request->fl_flags & FL_ACCESS) && |
1052 | (request->fl_type != F_UNLCK || | |
1053 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1054 | new_fl = locks_alloc_lock(); | |
1055 | new_fl2 = locks_alloc_lock(); | |
1056 | } | |
1da177e4 | 1057 | |
87709e28 | 1058 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1059 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1060 | /* |
1061 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1062 | * there are any, either return error or put the request on the | |
48f74186 | 1063 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1064 | */ |
1da177e4 | 1065 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1066 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1067 | if (!posix_locks_conflict(request, fl)) |
1068 | continue; | |
5842add2 | 1069 | if (conflock) |
3fe0fff1 | 1070 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1071 | error = -EAGAIN; |
1072 | if (!(request->fl_flags & FL_SLEEP)) | |
1073 | goto out; | |
1c8c601a JL |
1074 | /* |
1075 | * Deadlock detection and insertion into the blocked | |
1076 | * locks list must be done while holding the same lock! | |
1077 | */ | |
1da177e4 | 1078 | error = -EDEADLK; |
7b2296af | 1079 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
1080 | if (likely(!posix_locks_deadlock(request, fl))) { |
1081 | error = FILE_LOCK_DEFERRED; | |
1082 | __locks_insert_block(fl, request); | |
1083 | } | |
7b2296af | 1084 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1085 | goto out; |
1086 | } | |
1087 | } | |
1088 | ||
1089 | /* If we're just looking for a conflict, we're done. */ | |
1090 | error = 0; | |
1091 | if (request->fl_flags & FL_ACCESS) | |
1092 | goto out; | |
1093 | ||
bd61e0a9 JL |
1094 | /* Find the first old lock with the same owner as the new lock */ |
1095 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1096 | if (posix_same_owner(request, fl)) | |
1097 | break; | |
1da177e4 LT |
1098 | } |
1099 | ||
1cb36012 | 1100 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1101 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1102 | if (!posix_same_owner(request, fl)) | |
1103 | break; | |
1104 | ||
1105 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1106 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1107 | /* In all comparisons of start vs end, use |
1108 | * "start - 1" rather than "end + 1". If end | |
1109 | * is OFFSET_MAX, end + 1 will become negative. | |
1110 | */ | |
1da177e4 | 1111 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1112 | continue; |
1da177e4 LT |
1113 | /* If the next lock in the list has entirely bigger |
1114 | * addresses than the new one, insert the lock here. | |
1115 | */ | |
449231d6 | 1116 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1117 | break; |
1118 | ||
1119 | /* If we come here, the new and old lock are of the | |
1120 | * same type and adjacent or overlapping. Make one | |
1121 | * lock yielding from the lower start address of both | |
1122 | * locks to the higher end address. | |
1123 | */ | |
1124 | if (fl->fl_start > request->fl_start) | |
1125 | fl->fl_start = request->fl_start; | |
1126 | else | |
1127 | request->fl_start = fl->fl_start; | |
1128 | if (fl->fl_end < request->fl_end) | |
1129 | fl->fl_end = request->fl_end; | |
1130 | else | |
1131 | request->fl_end = fl->fl_end; | |
1132 | if (added) { | |
e084c1bd | 1133 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1134 | continue; |
1135 | } | |
1136 | request = fl; | |
b9746ef8 | 1137 | added = true; |
bd61e0a9 | 1138 | } else { |
1da177e4 LT |
1139 | /* Processing for different lock types is a bit |
1140 | * more complex. | |
1141 | */ | |
1142 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1143 | continue; |
1da177e4 LT |
1144 | if (fl->fl_start > request->fl_end) |
1145 | break; | |
1146 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1147 | added = true; |
1da177e4 LT |
1148 | if (fl->fl_start < request->fl_start) |
1149 | left = fl; | |
1150 | /* If the next lock in the list has a higher end | |
1151 | * address than the new one, insert the new one here. | |
1152 | */ | |
1153 | if (fl->fl_end > request->fl_end) { | |
1154 | right = fl; | |
1155 | break; | |
1156 | } | |
1157 | if (fl->fl_start >= request->fl_start) { | |
1158 | /* The new lock completely replaces an old | |
1159 | * one (This may happen several times). | |
1160 | */ | |
1161 | if (added) { | |
e084c1bd | 1162 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1163 | continue; |
1164 | } | |
b84d49f9 JL |
1165 | /* |
1166 | * Replace the old lock with new_fl, and | |
1167 | * remove the old one. It's safe to do the | |
1168 | * insert here since we know that we won't be | |
1169 | * using new_fl later, and that the lock is | |
1170 | * just replacing an existing lock. | |
1da177e4 | 1171 | */ |
b84d49f9 JL |
1172 | error = -ENOLCK; |
1173 | if (!new_fl) | |
1174 | goto out; | |
1175 | locks_copy_lock(new_fl, request); | |
1176 | request = new_fl; | |
1177 | new_fl = NULL; | |
e084c1bd JL |
1178 | locks_insert_lock_ctx(request, &fl->fl_list); |
1179 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1180 | added = true; |
1da177e4 LT |
1181 | } |
1182 | } | |
1da177e4 LT |
1183 | } |
1184 | ||
0d9a490a | 1185 | /* |
1cb36012 JL |
1186 | * The above code only modifies existing locks in case of merging or |
1187 | * replacing. If new lock(s) need to be inserted all modifications are | |
1188 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1189 | */ |
1190 | error = -ENOLCK; /* "no luck" */ | |
1191 | if (right && left == right && !new_fl2) | |
1192 | goto out; | |
1193 | ||
1da177e4 LT |
1194 | error = 0; |
1195 | if (!added) { | |
f475ae95 TM |
1196 | if (request->fl_type == F_UNLCK) { |
1197 | if (request->fl_flags & FL_EXISTS) | |
1198 | error = -ENOENT; | |
1da177e4 | 1199 | goto out; |
f475ae95 | 1200 | } |
0d9a490a MS |
1201 | |
1202 | if (!new_fl) { | |
1203 | error = -ENOLCK; | |
1204 | goto out; | |
1205 | } | |
1da177e4 | 1206 | locks_copy_lock(new_fl, request); |
5946c431 | 1207 | locks_move_blocks(new_fl, request); |
e084c1bd | 1208 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
2e2f756f | 1209 | fl = new_fl; |
1da177e4 LT |
1210 | new_fl = NULL; |
1211 | } | |
1212 | if (right) { | |
1213 | if (left == right) { | |
1214 | /* The new lock breaks the old one in two pieces, | |
1215 | * so we have to use the second new lock. | |
1216 | */ | |
1217 | left = new_fl2; | |
1218 | new_fl2 = NULL; | |
1219 | locks_copy_lock(left, right); | |
e084c1bd | 1220 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1221 | } |
1222 | right->fl_start = request->fl_end + 1; | |
1223 | locks_wake_up_blocks(right); | |
1224 | } | |
1225 | if (left) { | |
1226 | left->fl_end = request->fl_start - 1; | |
1227 | locks_wake_up_blocks(left); | |
1228 | } | |
1229 | out: | |
6109c850 | 1230 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1231 | percpu_up_read_preempt_enable(&file_rwsem); |
1da177e4 LT |
1232 | /* |
1233 | * Free any unused locks. | |
1234 | */ | |
1235 | if (new_fl) | |
1236 | locks_free_lock(new_fl); | |
1237 | if (new_fl2) | |
1238 | locks_free_lock(new_fl2); | |
ed9814d8 | 1239 | locks_dispose_list(&dispose); |
1890910f JL |
1240 | trace_posix_lock_inode(inode, request, error); |
1241 | ||
1da177e4 LT |
1242 | return error; |
1243 | } | |
1244 | ||
1245 | /** | |
1246 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1247 | * @filp: The file to apply the lock to | |
1248 | * @fl: The lock to be applied | |
150b3934 | 1249 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1250 | * |
1251 | * Add a POSIX style lock to a file. | |
1252 | * We merge adjacent & overlapping locks whenever possible. | |
1253 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1254 | * |
1255 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1256 | * whether or not a lock was successfully freed by testing the return | |
1257 | * value for -ENOENT. | |
1da177e4 | 1258 | */ |
150b3934 | 1259 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1260 | struct file_lock *conflock) |
1261 | { | |
c568d683 | 1262 | return posix_lock_inode(locks_inode(filp), fl, conflock); |
1da177e4 | 1263 | } |
150b3934 | 1264 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1265 | |
1266 | /** | |
29d01b22 JL |
1267 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1268 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1269 | * @fl: The lock to be applied |
1270 | * | |
616fb38f | 1271 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1272 | */ |
616fb38f | 1273 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1274 | { |
1275 | int error; | |
1276 | might_sleep (); | |
1277 | for (;;) { | |
b4d629a3 | 1278 | error = posix_lock_inode(inode, fl, NULL); |
bde74e4b | 1279 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1280 | break; |
ada5c1da | 1281 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
1da177e4 LT |
1282 | if (!error) |
1283 | continue; | |
1284 | ||
1285 | locks_delete_block(fl); | |
1286 | break; | |
1287 | } | |
1288 | return error; | |
1289 | } | |
29d01b22 | 1290 | |
9e8925b6 | 1291 | #ifdef CONFIG_MANDATORY_FILE_LOCKING |
1da177e4 LT |
1292 | /** |
1293 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1294 | * @file: the file to check |
1da177e4 LT |
1295 | * |
1296 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1297 | * This function is called from locks_verify_locked() only. | |
1298 | */ | |
d7a06983 | 1299 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1300 | { |
bd61e0a9 | 1301 | int ret; |
c568d683 | 1302 | struct inode *inode = locks_inode(file); |
bd61e0a9 | 1303 | struct file_lock_context *ctx; |
1da177e4 LT |
1304 | struct file_lock *fl; |
1305 | ||
128a3785 | 1306 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
1307 | if (!ctx || list_empty_careful(&ctx->flc_posix)) |
1308 | return 0; | |
1309 | ||
1da177e4 LT |
1310 | /* |
1311 | * Search the lock list for this inode for any POSIX locks. | |
1312 | */ | |
6109c850 | 1313 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
1314 | ret = 0; |
1315 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1316 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1317 | fl->fl_owner != file) { |
1318 | ret = -EAGAIN; | |
1da177e4 | 1319 | break; |
bd61e0a9 | 1320 | } |
1da177e4 | 1321 | } |
6109c850 | 1322 | spin_unlock(&ctx->flc_lock); |
bd61e0a9 | 1323 | return ret; |
1da177e4 LT |
1324 | } |
1325 | ||
1326 | /** | |
1327 | * locks_mandatory_area - Check for a conflicting lock | |
acc15575 | 1328 | * @inode: the file to check |
1da177e4 | 1329 | * @filp: how the file was opened (if it was) |
acc15575 CH |
1330 | * @start: first byte in the file to check |
1331 | * @end: lastbyte in the file to check | |
1332 | * @type: %F_WRLCK for a write lock, else %F_RDLCK | |
1da177e4 LT |
1333 | * |
1334 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1da177e4 | 1335 | */ |
acc15575 CH |
1336 | int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start, |
1337 | loff_t end, unsigned char type) | |
1da177e4 LT |
1338 | { |
1339 | struct file_lock fl; | |
1340 | int error; | |
29723ade | 1341 | bool sleep = false; |
1da177e4 LT |
1342 | |
1343 | locks_init_lock(&fl); | |
1da177e4 LT |
1344 | fl.fl_pid = current->tgid; |
1345 | fl.fl_file = filp; | |
1346 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1347 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1348 | sleep = true; |
acc15575 CH |
1349 | fl.fl_type = type; |
1350 | fl.fl_start = start; | |
1351 | fl.fl_end = end; | |
1da177e4 LT |
1352 | |
1353 | for (;;) { | |
29723ade | 1354 | if (filp) { |
73a8f5f7 | 1355 | fl.fl_owner = filp; |
29723ade | 1356 | fl.fl_flags &= ~FL_SLEEP; |
b4d629a3 | 1357 | error = posix_lock_inode(inode, &fl, NULL); |
29723ade JL |
1358 | if (!error) |
1359 | break; | |
1360 | } | |
1361 | ||
1362 | if (sleep) | |
1363 | fl.fl_flags |= FL_SLEEP; | |
1364 | fl.fl_owner = current->files; | |
b4d629a3 | 1365 | error = posix_lock_inode(inode, &fl, NULL); |
bde74e4b | 1366 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1367 | break; |
ada5c1da | 1368 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_blocker); |
1da177e4 LT |
1369 | if (!error) { |
1370 | /* | |
1371 | * If we've been sleeping someone might have | |
1372 | * changed the permissions behind our back. | |
1373 | */ | |
a16877ca | 1374 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1375 | continue; |
1376 | } | |
1377 | ||
1378 | locks_delete_block(&fl); | |
1379 | break; | |
1380 | } | |
1381 | ||
1382 | return error; | |
1383 | } | |
1384 | ||
1385 | EXPORT_SYMBOL(locks_mandatory_area); | |
9e8925b6 | 1386 | #endif /* CONFIG_MANDATORY_FILE_LOCKING */ |
1da177e4 | 1387 | |
778fc546 BF |
1388 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1389 | { | |
1390 | switch (arg) { | |
1391 | case F_UNLCK: | |
1392 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1393 | /* fall through: */ | |
1394 | case F_RDLCK: | |
1395 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1396 | } | |
1397 | } | |
1398 | ||
1da177e4 | 1399 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1400 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1401 | { |
1da177e4 LT |
1402 | int error = assign_type(fl, arg); |
1403 | ||
1404 | if (error) | |
1405 | return error; | |
778fc546 | 1406 | lease_clear_pending(fl, arg); |
1da177e4 | 1407 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1408 | if (arg == F_UNLCK) { |
1409 | struct file *filp = fl->fl_file; | |
1410 | ||
1411 | f_delown(filp); | |
1412 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1413 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1414 | if (fl->fl_fasync != NULL) { | |
1415 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1416 | fl->fl_fasync = NULL; | |
1417 | } | |
e084c1bd | 1418 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1419 | } |
1da177e4 LT |
1420 | return 0; |
1421 | } | |
1da177e4 LT |
1422 | EXPORT_SYMBOL(lease_modify); |
1423 | ||
778fc546 BF |
1424 | static bool past_time(unsigned long then) |
1425 | { | |
1426 | if (!then) | |
1427 | /* 0 is a special value meaning "this never expires": */ | |
1428 | return false; | |
1429 | return time_after(jiffies, then); | |
1430 | } | |
1431 | ||
c45198ed | 1432 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1433 | { |
8634b51f JL |
1434 | struct file_lock_context *ctx = inode->i_flctx; |
1435 | struct file_lock *fl, *tmp; | |
1da177e4 | 1436 | |
6109c850 | 1437 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1438 | |
8634b51f | 1439 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1440 | trace_time_out_leases(inode, fl); |
778fc546 | 1441 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1442 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1443 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1444 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1445 | } |
1446 | } | |
1447 | ||
df4e8d2c BF |
1448 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1449 | { | |
11afe9f7 CH |
1450 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) |
1451 | return false; | |
df4e8d2c BF |
1452 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) |
1453 | return false; | |
1454 | return locks_conflict(breaker, lease); | |
1455 | } | |
1456 | ||
03d12ddf JL |
1457 | static bool |
1458 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1459 | { | |
8634b51f | 1460 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1461 | struct file_lock *fl; |
1462 | ||
6109c850 | 1463 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1464 | |
8634b51f | 1465 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1466 | if (leases_conflict(fl, breaker)) |
1467 | return true; | |
1468 | } | |
1469 | return false; | |
1470 | } | |
1471 | ||
1da177e4 LT |
1472 | /** |
1473 | * __break_lease - revoke all outstanding leases on file | |
1474 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1475 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1476 | * break all leases | |
1477 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1478 | * only delegations | |
1da177e4 | 1479 | * |
87250dd2 | 1480 | * break_lease (inlined for speed) has checked there already is at least |
1481 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1482 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1483 | * specified %O_NONBLOCK to your open(). |
1484 | */ | |
df4e8d2c | 1485 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1486 | { |
778fc546 | 1487 | int error = 0; |
128a3785 | 1488 | struct file_lock_context *ctx; |
a901125c | 1489 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1490 | unsigned long break_time; |
8737c930 | 1491 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1492 | LIST_HEAD(dispose); |
1da177e4 | 1493 | |
8737c930 | 1494 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1495 | if (IS_ERR(new_fl)) |
1496 | return PTR_ERR(new_fl); | |
df4e8d2c | 1497 | new_fl->fl_flags = type; |
1da177e4 | 1498 | |
8634b51f | 1499 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1500 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1501 | if (!ctx) { |
1502 | WARN_ON_ONCE(1); | |
1503 | return error; | |
1504 | } | |
1505 | ||
87709e28 | 1506 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1507 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1508 | |
c45198ed | 1509 | time_out_leases(inode, &dispose); |
1da177e4 | 1510 | |
03d12ddf | 1511 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1512 | goto out; |
1513 | ||
1da177e4 LT |
1514 | break_time = 0; |
1515 | if (lease_break_time > 0) { | |
1516 | break_time = jiffies + lease_break_time * HZ; | |
1517 | if (break_time == 0) | |
1518 | break_time++; /* so that 0 means no break time */ | |
1519 | } | |
1520 | ||
a901125c | 1521 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1522 | if (!leases_conflict(fl, new_fl)) |
1523 | continue; | |
778fc546 BF |
1524 | if (want_write) { |
1525 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1526 | continue; | |
1527 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1528 | fl->fl_break_time = break_time; |
778fc546 | 1529 | } else { |
8634b51f | 1530 | if (lease_breaking(fl)) |
778fc546 BF |
1531 | continue; |
1532 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1533 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1534 | } |
4d01b7f5 | 1535 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1536 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1537 | } |
1538 | ||
8634b51f | 1539 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1540 | goto out; |
1541 | ||
843c6b2f | 1542 | if (mode & O_NONBLOCK) { |
62af4f1f | 1543 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1544 | error = -EWOULDBLOCK; |
1545 | goto out; | |
1546 | } | |
1547 | ||
1548 | restart: | |
8634b51f JL |
1549 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1550 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1551 | if (break_time != 0) |
1da177e4 | 1552 | break_time -= jiffies; |
f1c6bb2c JL |
1553 | if (break_time == 0) |
1554 | break_time++; | |
8634b51f | 1555 | locks_insert_block(fl, new_fl); |
62af4f1f | 1556 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1557 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1558 | percpu_up_read_preempt_enable(&file_rwsem); |
aba37660 | 1559 | |
c45198ed | 1560 | locks_dispose_list(&dispose); |
4321e01e | 1561 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
ada5c1da | 1562 | !new_fl->fl_blocker, break_time); |
aba37660 | 1563 | |
87709e28 | 1564 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1565 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1566 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1567 | locks_delete_block(new_fl); |
1da177e4 | 1568 | if (error >= 0) { |
778fc546 BF |
1569 | /* |
1570 | * Wait for the next conflicting lease that has not been | |
1571 | * broken yet | |
1572 | */ | |
03d12ddf JL |
1573 | if (error == 0) |
1574 | time_out_leases(inode, &dispose); | |
1575 | if (any_leases_conflict(inode, new_fl)) | |
1576 | goto restart; | |
1da177e4 LT |
1577 | error = 0; |
1578 | } | |
1da177e4 | 1579 | out: |
6109c850 | 1580 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1581 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1582 | locks_dispose_list(&dispose); |
6d4b9e38 | 1583 | locks_free_lock(new_fl); |
1da177e4 LT |
1584 | return error; |
1585 | } | |
1586 | ||
1587 | EXPORT_SYMBOL(__break_lease); | |
1588 | ||
1589 | /** | |
76c47948 | 1590 | * lease_get_mtime - update modified time of an inode with exclusive lease |
1da177e4 | 1591 | * @inode: the inode |
76c47948 | 1592 | * @time: pointer to a timespec which contains the last modified time |
1da177e4 LT |
1593 | * |
1594 | * This is to force NFS clients to flush their caches for files with | |
1595 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1596 | * exclusive lease, then they could be modifying it. |
1da177e4 | 1597 | */ |
95582b00 | 1598 | void lease_get_mtime(struct inode *inode, struct timespec64 *time) |
1da177e4 | 1599 | { |
bfe86024 | 1600 | bool has_lease = false; |
128a3785 | 1601 | struct file_lock_context *ctx; |
8634b51f | 1602 | struct file_lock *fl; |
bfe86024 | 1603 | |
128a3785 | 1604 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1605 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1606 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1607 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1608 | struct file_lock, fl_list); | |
1609 | if (fl && (fl->fl_type == F_WRLCK)) | |
1610 | has_lease = true; | |
6109c850 | 1611 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1612 | } |
1613 | ||
1614 | if (has_lease) | |
c2050a45 | 1615 | *time = current_time(inode); |
1da177e4 LT |
1616 | } |
1617 | ||
1618 | EXPORT_SYMBOL(lease_get_mtime); | |
1619 | ||
1620 | /** | |
1621 | * fcntl_getlease - Enquire what lease is currently active | |
1622 | * @filp: the file | |
1623 | * | |
1624 | * The value returned by this function will be one of | |
1625 | * (if no lease break is pending): | |
1626 | * | |
1627 | * %F_RDLCK to indicate a shared lease is held. | |
1628 | * | |
1629 | * %F_WRLCK to indicate an exclusive lease is held. | |
1630 | * | |
1631 | * %F_UNLCK to indicate no lease is held. | |
1632 | * | |
1633 | * (if a lease break is pending): | |
1634 | * | |
1635 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1636 | * changed to a shared lease (or removed). | |
1637 | * | |
1638 | * %F_UNLCK to indicate the lease needs to be removed. | |
1639 | * | |
1640 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1641 | * should be returned to userspace. | |
1642 | */ | |
1643 | int fcntl_getlease(struct file *filp) | |
1644 | { | |
1645 | struct file_lock *fl; | |
c568d683 | 1646 | struct inode *inode = locks_inode(filp); |
128a3785 | 1647 | struct file_lock_context *ctx; |
1da177e4 | 1648 | int type = F_UNLCK; |
c45198ed | 1649 | LIST_HEAD(dispose); |
1da177e4 | 1650 | |
128a3785 | 1651 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1652 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
5f43086b | 1653 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1654 | spin_lock(&ctx->flc_lock); |
c568d683 | 1655 | time_out_leases(inode, &dispose); |
8634b51f JL |
1656 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
1657 | if (fl->fl_file != filp) | |
1658 | continue; | |
778fc546 | 1659 | type = target_leasetype(fl); |
1da177e4 LT |
1660 | break; |
1661 | } | |
6109c850 | 1662 | spin_unlock(&ctx->flc_lock); |
5f43086b PZ |
1663 | percpu_up_read_preempt_enable(&file_rwsem); |
1664 | ||
8634b51f | 1665 | locks_dispose_list(&dispose); |
1da177e4 | 1666 | } |
1da177e4 LT |
1667 | return type; |
1668 | } | |
1669 | ||
24cbe784 JL |
1670 | /** |
1671 | * check_conflicting_open - see if the given dentry points to a file that has | |
1672 | * an existing open that would conflict with the | |
1673 | * desired lease. | |
1674 | * @dentry: dentry to check | |
1675 | * @arg: type of lease that we're trying to acquire | |
7fadc59c | 1676 | * @flags: current lock flags |
24cbe784 JL |
1677 | * |
1678 | * Check to see if there's an existing open fd on this file that would | |
1679 | * conflict with the lease we're trying to set. | |
1680 | */ | |
1681 | static int | |
11afe9f7 | 1682 | check_conflicting_open(const struct dentry *dentry, const long arg, int flags) |
24cbe784 JL |
1683 | { |
1684 | int ret = 0; | |
1685 | struct inode *inode = dentry->d_inode; | |
1686 | ||
11afe9f7 CH |
1687 | if (flags & FL_LAYOUT) |
1688 | return 0; | |
1689 | ||
8cf9ee50 | 1690 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
24cbe784 JL |
1691 | return -EAGAIN; |
1692 | ||
1693 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1694 | (atomic_read(&inode->i_count) > 1))) | |
1695 | ret = -EAGAIN; | |
1696 | ||
1697 | return ret; | |
1698 | } | |
1699 | ||
e6f5c789 JL |
1700 | static int |
1701 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1702 | { |
8634b51f | 1703 | struct file_lock *fl, *my_fl = NULL, *lease; |
0f7fc9e4 | 1704 | struct dentry *dentry = filp->f_path.dentry; |
c568d683 | 1705 | struct inode *inode = dentry->d_inode; |
8634b51f | 1706 | struct file_lock_context *ctx; |
df4e8d2c | 1707 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1708 | int error; |
c45198ed | 1709 | LIST_HEAD(dispose); |
1da177e4 | 1710 | |
096657b6 | 1711 | lease = *flp; |
62af4f1f JL |
1712 | trace_generic_add_lease(inode, lease); |
1713 | ||
5c1c669a JL |
1714 | /* Note that arg is never F_UNLCK here */ |
1715 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1716 | if (!ctx) |
1717 | return -ENOMEM; | |
1718 | ||
df4e8d2c BF |
1719 | /* |
1720 | * In the delegation case we need mutual exclusion with | |
1721 | * a number of operations that take the i_mutex. We trylock | |
1722 | * because delegations are an optional optimization, and if | |
1723 | * there's some chance of a conflict--we'd rather not | |
1724 | * bother, maybe that's a sign this just isn't a good file to | |
1725 | * hand out a delegation on. | |
1726 | */ | |
5955102c | 1727 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1728 | return -EAGAIN; |
1729 | ||
1730 | if (is_deleg && arg == F_WRLCK) { | |
1731 | /* Write delegations are not currently supported: */ | |
5955102c | 1732 | inode_unlock(inode); |
df4e8d2c BF |
1733 | WARN_ON_ONCE(1); |
1734 | return -EINVAL; | |
1735 | } | |
096657b6 | 1736 | |
87709e28 | 1737 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1738 | spin_lock(&ctx->flc_lock); |
c45198ed | 1739 | time_out_leases(inode, &dispose); |
11afe9f7 | 1740 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
24cbe784 | 1741 | if (error) |
096657b6 | 1742 | goto out; |
6d5e8b05 | 1743 | |
1da177e4 LT |
1744 | /* |
1745 | * At this point, we know that if there is an exclusive | |
1746 | * lease on this file, then we hold it on this filp | |
1747 | * (otherwise our open of this file would have blocked). | |
1748 | * And if we are trying to acquire an exclusive lease, | |
1749 | * then the file is not open by anyone (including us) | |
1750 | * except for this filp. | |
1751 | */ | |
c1f24ef4 | 1752 | error = -EAGAIN; |
8634b51f | 1753 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1754 | if (fl->fl_file == filp && |
1755 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1756 | my_fl = fl; |
c1f24ef4 BF |
1757 | continue; |
1758 | } | |
8634b51f | 1759 | |
c1f24ef4 BF |
1760 | /* |
1761 | * No exclusive leases if someone else has a lease on | |
1762 | * this file: | |
1763 | */ | |
1764 | if (arg == F_WRLCK) | |
1765 | goto out; | |
1766 | /* | |
1767 | * Modifying our existing lease is OK, but no getting a | |
1768 | * new lease if someone else is opening for write: | |
1769 | */ | |
1770 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1771 | goto out; | |
1da177e4 LT |
1772 | } |
1773 | ||
8634b51f | 1774 | if (my_fl != NULL) { |
0164bf02 JL |
1775 | lease = my_fl; |
1776 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1777 | if (error) |
1778 | goto out; | |
1779 | goto out_setup; | |
1da177e4 LT |
1780 | } |
1781 | ||
1da177e4 LT |
1782 | error = -EINVAL; |
1783 | if (!leases_enable) | |
1784 | goto out; | |
1785 | ||
e084c1bd | 1786 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1787 | /* |
1788 | * The check in break_lease() is lockless. It's possible for another | |
1789 | * open to race in after we did the earlier check for a conflicting | |
1790 | * open but before the lease was inserted. Check again for a | |
1791 | * conflicting open and cancel the lease if there is one. | |
1792 | * | |
1793 | * We also add a barrier here to ensure that the insertion of the lock | |
1794 | * precedes these checks. | |
1795 | */ | |
1796 | smp_mb(); | |
11afe9f7 | 1797 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
8634b51f | 1798 | if (error) { |
e084c1bd | 1799 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1800 | goto out; |
1801 | } | |
1c7dd2ff JL |
1802 | |
1803 | out_setup: | |
1804 | if (lease->fl_lmops->lm_setup) | |
1805 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1806 | out: |
6109c850 | 1807 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1808 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1809 | locks_dispose_list(&dispose); |
df4e8d2c | 1810 | if (is_deleg) |
5955102c | 1811 | inode_unlock(inode); |
8634b51f | 1812 | if (!error && !my_fl) |
1c7dd2ff | 1813 | *flp = NULL; |
1da177e4 LT |
1814 | return error; |
1815 | } | |
8335ebd9 | 1816 | |
2ab99ee1 | 1817 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1818 | { |
0efaa7e8 | 1819 | int error = -EAGAIN; |
8634b51f | 1820 | struct file_lock *fl, *victim = NULL; |
c568d683 | 1821 | struct inode *inode = locks_inode(filp); |
128a3785 | 1822 | struct file_lock_context *ctx; |
c45198ed | 1823 | LIST_HEAD(dispose); |
8335ebd9 | 1824 | |
128a3785 | 1825 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1826 | if (!ctx) { |
1827 | trace_generic_delete_lease(inode, NULL); | |
1828 | return error; | |
1829 | } | |
1830 | ||
87709e28 | 1831 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1832 | spin_lock(&ctx->flc_lock); |
8634b51f | 1833 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1834 | if (fl->fl_file == filp && |
1835 | fl->fl_owner == owner) { | |
8634b51f | 1836 | victim = fl; |
0efaa7e8 | 1837 | break; |
8634b51f | 1838 | } |
8335ebd9 | 1839 | } |
a9b1b455 | 1840 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1841 | if (victim) |
7448cc37 | 1842 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1843 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1844 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1845 | locks_dispose_list(&dispose); |
0efaa7e8 | 1846 | return error; |
8335ebd9 BF |
1847 | } |
1848 | ||
1849 | /** | |
1850 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1851 | * @filp: file pointer |
1852 | * @arg: type of lease to obtain | |
1853 | * @flp: input - file_lock to use, output - file_lock inserted | |
1854 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1855 | * doesn't require it) | |
8335ebd9 BF |
1856 | * |
1857 | * The (input) flp->fl_lmops->lm_break function is required | |
1858 | * by break_lease(). | |
8335ebd9 | 1859 | */ |
e6f5c789 JL |
1860 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1861 | void **priv) | |
8335ebd9 | 1862 | { |
c568d683 | 1863 | struct inode *inode = locks_inode(filp); |
8335ebd9 BF |
1864 | int error; |
1865 | ||
8e96e3b7 | 1866 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1867 | return -EACCES; |
1868 | if (!S_ISREG(inode->i_mode)) | |
1869 | return -EINVAL; | |
1870 | error = security_file_lock(filp, arg); | |
1871 | if (error) | |
1872 | return error; | |
1873 | ||
8335ebd9 BF |
1874 | switch (arg) { |
1875 | case F_UNLCK: | |
2ab99ee1 | 1876 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1877 | case F_RDLCK: |
1878 | case F_WRLCK: | |
0efaa7e8 JL |
1879 | if (!(*flp)->fl_lmops->lm_break) { |
1880 | WARN_ON_ONCE(1); | |
1881 | return -ENOLCK; | |
1882 | } | |
11afe9f7 | 1883 | |
e6f5c789 | 1884 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1885 | default: |
8d657eb3 | 1886 | return -EINVAL; |
8335ebd9 BF |
1887 | } |
1888 | } | |
0af1a450 | 1889 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1890 | |
b89f4321 | 1891 | /** |
e51673aa | 1892 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1893 | * @filp: file pointer |
1894 | * @arg: type of lease to obtain | |
1895 | * @lease: file_lock to use when adding a lease | |
1896 | * @priv: private info for lm_setup when adding a lease (may be | |
1897 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1898 | * |
1899 | * Call this to establish a lease on the file. The "lease" argument is not | |
1900 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
80b79dd0 MCC |
1901 | * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be |
1902 | * set; if not, this function will return -ENOLCK (and generate a scary-looking | |
e51673aa | 1903 | * stack trace). |
1c7dd2ff JL |
1904 | * |
1905 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1906 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1907 | */ |
e6f5c789 JL |
1908 | int |
1909 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1910 | { |
de2a4a50 | 1911 | if (filp->f_op->setlease) |
f82b4b67 | 1912 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1913 | else |
f82b4b67 | 1914 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1915 | } |
a9933cea | 1916 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1917 | |
0ceaf6c7 | 1918 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1919 | { |
1c7dd2ff | 1920 | struct file_lock *fl; |
f7347ce4 | 1921 | struct fasync_struct *new; |
1da177e4 LT |
1922 | int error; |
1923 | ||
c5b1f0d9 AB |
1924 | fl = lease_alloc(filp, arg); |
1925 | if (IS_ERR(fl)) | |
1926 | return PTR_ERR(fl); | |
1da177e4 | 1927 | |
f7347ce4 LT |
1928 | new = fasync_alloc(); |
1929 | if (!new) { | |
1930 | locks_free_lock(fl); | |
1931 | return -ENOMEM; | |
1932 | } | |
1c7dd2ff | 1933 | new->fa_fd = fd; |
f7347ce4 | 1934 | |
1c7dd2ff | 1935 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1936 | if (fl) |
1937 | locks_free_lock(fl); | |
f7347ce4 LT |
1938 | if (new) |
1939 | fasync_free(new); | |
1da177e4 LT |
1940 | return error; |
1941 | } | |
1942 | ||
0ceaf6c7 BF |
1943 | /** |
1944 | * fcntl_setlease - sets a lease on an open file | |
1945 | * @fd: open file descriptor | |
1946 | * @filp: file pointer | |
1947 | * @arg: type of lease to obtain | |
1948 | * | |
1949 | * Call this fcntl to establish a lease on the file. | |
1950 | * Note that you also need to call %F_SETSIG to | |
1951 | * receive a signal when the lease is broken. | |
1952 | */ | |
1953 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1954 | { | |
1955 | if (arg == F_UNLCK) | |
2ab99ee1 | 1956 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
1957 | return do_fcntl_add_lease(fd, filp, arg); |
1958 | } | |
1959 | ||
1da177e4 | 1960 | /** |
29d01b22 JL |
1961 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
1962 | * @inode: inode of the file to apply to | |
1da177e4 LT |
1963 | * @fl: The lock to be applied |
1964 | * | |
29d01b22 | 1965 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 1966 | */ |
616fb38f | 1967 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1968 | { |
1969 | int error; | |
1970 | might_sleep(); | |
1971 | for (;;) { | |
29d01b22 | 1972 | error = flock_lock_inode(inode, fl); |
bde74e4b | 1973 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1974 | break; |
ada5c1da | 1975 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
1da177e4 LT |
1976 | if (!error) |
1977 | continue; | |
1978 | ||
1979 | locks_delete_block(fl); | |
1980 | break; | |
1981 | } | |
1982 | return error; | |
1983 | } | |
1984 | ||
e55c34a6 BC |
1985 | /** |
1986 | * locks_lock_inode_wait - Apply a lock to an inode | |
1987 | * @inode: inode of the file to apply to | |
1988 | * @fl: The lock to be applied | |
1989 | * | |
1990 | * Apply a POSIX or FLOCK style lock request to an inode. | |
1991 | */ | |
1992 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
1993 | { | |
1994 | int res = 0; | |
1995 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
1996 | case FL_POSIX: | |
1997 | res = posix_lock_inode_wait(inode, fl); | |
1998 | break; | |
1999 | case FL_FLOCK: | |
2000 | res = flock_lock_inode_wait(inode, fl); | |
2001 | break; | |
2002 | default: | |
2003 | BUG(); | |
2004 | } | |
2005 | return res; | |
2006 | } | |
2007 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
2008 | ||
1da177e4 LT |
2009 | /** |
2010 | * sys_flock: - flock() system call. | |
2011 | * @fd: the file descriptor to lock. | |
2012 | * @cmd: the type of lock to apply. | |
2013 | * | |
2014 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
80b79dd0 | 2015 | * The @cmd can be one of: |
1da177e4 | 2016 | * |
80b79dd0 MCC |
2017 | * - %LOCK_SH -- a shared lock. |
2018 | * - %LOCK_EX -- an exclusive lock. | |
2019 | * - %LOCK_UN -- remove an existing lock. | |
2020 | * - %LOCK_MAND -- a 'mandatory' flock. | |
2021 | * This exists to emulate Windows Share Modes. | |
1da177e4 LT |
2022 | * |
2023 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
2024 | * processes read and write access respectively. | |
2025 | */ | |
002c8976 | 2026 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 2027 | { |
2903ff01 | 2028 | struct fd f = fdget(fd); |
1da177e4 LT |
2029 | struct file_lock *lock; |
2030 | int can_sleep, unlock; | |
2031 | int error; | |
2032 | ||
2033 | error = -EBADF; | |
2903ff01 | 2034 | if (!f.file) |
1da177e4 LT |
2035 | goto out; |
2036 | ||
2037 | can_sleep = !(cmd & LOCK_NB); | |
2038 | cmd &= ~LOCK_NB; | |
2039 | unlock = (cmd == LOCK_UN); | |
2040 | ||
aeb5d727 | 2041 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 2042 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
2043 | goto out_putf; |
2044 | ||
d6367d62 | 2045 | lock = flock_make_lock(f.file, cmd, NULL); |
6e129d00 JL |
2046 | if (IS_ERR(lock)) { |
2047 | error = PTR_ERR(lock); | |
1da177e4 | 2048 | goto out_putf; |
6e129d00 JL |
2049 | } |
2050 | ||
1da177e4 LT |
2051 | if (can_sleep) |
2052 | lock->fl_flags |= FL_SLEEP; | |
2053 | ||
2903ff01 | 2054 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
2055 | if (error) |
2056 | goto out_free; | |
2057 | ||
de2a4a50 | 2058 | if (f.file->f_op->flock) |
2903ff01 | 2059 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
2060 | (can_sleep) ? F_SETLKW : F_SETLK, |
2061 | lock); | |
2062 | else | |
4f656367 | 2063 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
2064 | |
2065 | out_free: | |
993dfa87 | 2066 | locks_free_lock(lock); |
1da177e4 LT |
2067 | |
2068 | out_putf: | |
2903ff01 | 2069 | fdput(f); |
1da177e4 LT |
2070 | out: |
2071 | return error; | |
2072 | } | |
2073 | ||
3ee17abd BF |
2074 | /** |
2075 | * vfs_test_lock - test file byte range lock | |
2076 | * @filp: The file to test lock for | |
6924c554 | 2077 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2078 | * |
2079 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2080 | * setting conf->fl_type to something other than F_UNLCK. | |
2081 | */ | |
2082 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2083 | { | |
de2a4a50 | 2084 | if (filp->f_op->lock) |
3ee17abd BF |
2085 | return filp->f_op->lock(filp, F_GETLK, fl); |
2086 | posix_test_lock(filp, fl); | |
2087 | return 0; | |
2088 | } | |
2089 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2090 | ||
9d5b86ac BC |
2091 | /** |
2092 | * locks_translate_pid - translate a file_lock's fl_pid number into a namespace | |
2093 | * @fl: The file_lock who's fl_pid should be translated | |
2094 | * @ns: The namespace into which the pid should be translated | |
2095 | * | |
2096 | * Used to tranlate a fl_pid into a namespace virtual pid number | |
2097 | */ | |
2098 | static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns) | |
2099 | { | |
2100 | pid_t vnr; | |
2101 | struct pid *pid; | |
2102 | ||
2103 | if (IS_OFDLCK(fl)) | |
2104 | return -1; | |
2105 | if (IS_REMOTELCK(fl)) | |
2106 | return fl->fl_pid; | |
826d7bc9 KK |
2107 | /* |
2108 | * If the flock owner process is dead and its pid has been already | |
2109 | * freed, the translation below won't work, but we still want to show | |
2110 | * flock owner pid number in init pidns. | |
2111 | */ | |
2112 | if (ns == &init_pid_ns) | |
2113 | return (pid_t)fl->fl_pid; | |
9d5b86ac BC |
2114 | |
2115 | rcu_read_lock(); | |
2116 | pid = find_pid_ns(fl->fl_pid, &init_pid_ns); | |
2117 | vnr = pid_nr_ns(pid, ns); | |
2118 | rcu_read_unlock(); | |
2119 | return vnr; | |
2120 | } | |
2121 | ||
c2fa1b8a BF |
2122 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2123 | { | |
9d5b86ac | 2124 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2125 | #if BITS_PER_LONG == 32 |
2126 | /* | |
2127 | * Make sure we can represent the posix lock via | |
2128 | * legacy 32bit flock. | |
2129 | */ | |
2130 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2131 | return -EOVERFLOW; | |
2132 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2133 | return -EOVERFLOW; | |
2134 | #endif | |
2135 | flock->l_start = fl->fl_start; | |
2136 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2137 | fl->fl_end - fl->fl_start + 1; | |
2138 | flock->l_whence = 0; | |
129a84de | 2139 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2140 | return 0; |
2141 | } | |
2142 | ||
2143 | #if BITS_PER_LONG == 32 | |
2144 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2145 | { | |
9d5b86ac | 2146 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2147 | flock->l_start = fl->fl_start; |
2148 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2149 | fl->fl_end - fl->fl_start + 1; | |
2150 | flock->l_whence = 0; | |
2151 | flock->l_type = fl->fl_type; | |
2152 | } | |
2153 | #endif | |
2154 | ||
1da177e4 LT |
2155 | /* Report the first existing lock that would conflict with l. |
2156 | * This implements the F_GETLK command of fcntl(). | |
2157 | */ | |
a75d30c7 | 2158 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock) |
1da177e4 | 2159 | { |
52306e88 | 2160 | struct file_lock *fl; |
1da177e4 LT |
2161 | int error; |
2162 | ||
52306e88 BC |
2163 | fl = locks_alloc_lock(); |
2164 | if (fl == NULL) | |
2165 | return -ENOMEM; | |
1da177e4 | 2166 | error = -EINVAL; |
a75d30c7 | 2167 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2168 | goto out; |
2169 | ||
52306e88 | 2170 | error = flock_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2171 | if (error) |
2172 | goto out; | |
2173 | ||
0d3f7a2d | 2174 | if (cmd == F_OFD_GETLK) { |
90478939 | 2175 | error = -EINVAL; |
a75d30c7 | 2176 | if (flock->l_pid != 0) |
90478939 JL |
2177 | goto out; |
2178 | ||
5d50ffd7 | 2179 | cmd = F_GETLK; |
52306e88 BC |
2180 | fl->fl_flags |= FL_OFDLCK; |
2181 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2182 | } |
2183 | ||
52306e88 | 2184 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2185 | if (error) |
2186 | goto out; | |
1da177e4 | 2187 | |
52306e88 BC |
2188 | flock->l_type = fl->fl_type; |
2189 | if (fl->fl_type != F_UNLCK) { | |
2190 | error = posix_lock_to_flock(flock, fl); | |
c2fa1b8a | 2191 | if (error) |
52306e88 | 2192 | goto out; |
1da177e4 | 2193 | } |
1da177e4 | 2194 | out: |
52306e88 | 2195 | locks_free_lock(fl); |
1da177e4 LT |
2196 | return error; |
2197 | } | |
2198 | ||
7723ec97 ME |
2199 | /** |
2200 | * vfs_lock_file - file byte range lock | |
2201 | * @filp: The file to apply the lock to | |
2202 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2203 | * @fl: The lock to be applied | |
150b3934 ME |
2204 | * @conf: Place to return a copy of the conflicting lock, if found. |
2205 | * | |
2206 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2207 | * as the final argument. | |
2208 | * | |
2209 | * If the filesystem defines a private ->lock() method, then @conf will | |
2210 | * be left unchanged; so a caller that cares should initialize it to | |
2211 | * some acceptable default. | |
2beb6614 ME |
2212 | * |
2213 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2214 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2215 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2216 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2217 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2218 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2219 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2220 | * request completes. |
2221 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2222 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2223 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2224 | * nonzero return code and the file system should release the lock. The file |
2225 | * system is also responsible to keep a corresponding posix lock when it | |
2226 | * grants a lock so the VFS can find out which locks are locally held and do | |
2227 | * the correct lock cleanup when required. | |
2228 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2229 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2230 | * return code. |
7723ec97 | 2231 | */ |
150b3934 | 2232 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2233 | { |
de2a4a50 | 2234 | if (filp->f_op->lock) |
7723ec97 ME |
2235 | return filp->f_op->lock(filp, cmd, fl); |
2236 | else | |
150b3934 | 2237 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2238 | } |
2239 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2240 | ||
b648a6de MS |
2241 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2242 | struct file_lock *fl) | |
2243 | { | |
2244 | int error; | |
2245 | ||
2246 | error = security_file_lock(filp, fl->fl_type); | |
2247 | if (error) | |
2248 | return error; | |
2249 | ||
764c76b3 MS |
2250 | for (;;) { |
2251 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2252 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2253 | break; |
ada5c1da | 2254 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
764c76b3 MS |
2255 | if (!error) |
2256 | continue; | |
2257 | ||
2258 | locks_delete_block(fl); | |
2259 | break; | |
b648a6de MS |
2260 | } |
2261 | ||
2262 | return error; | |
2263 | } | |
2264 | ||
6ca7d910 | 2265 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2266 | static int |
2267 | check_fmode_for_setlk(struct file_lock *fl) | |
2268 | { | |
2269 | switch (fl->fl_type) { | |
2270 | case F_RDLCK: | |
2271 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2272 | return -EBADF; | |
2273 | break; | |
2274 | case F_WRLCK: | |
2275 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2276 | return -EBADF; | |
2277 | } | |
2278 | return 0; | |
2279 | } | |
2280 | ||
1da177e4 LT |
2281 | /* Apply the lock described by l to an open file descriptor. |
2282 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2283 | */ | |
c293621b | 2284 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2285 | struct flock *flock) |
1da177e4 LT |
2286 | { |
2287 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2288 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2289 | struct file *f; |
1da177e4 LT |
2290 | int error; |
2291 | ||
2292 | if (file_lock == NULL) | |
2293 | return -ENOLCK; | |
2294 | ||
1da177e4 LT |
2295 | /* Don't allow mandatory locks on files that may be memory mapped |
2296 | * and shared. | |
2297 | */ | |
a16877ca | 2298 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2299 | error = -EAGAIN; |
2300 | goto out; | |
2301 | } | |
2302 | ||
a75d30c7 | 2303 | error = flock_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2304 | if (error) |
2305 | goto out; | |
5d50ffd7 | 2306 | |
cf01f4ee JL |
2307 | error = check_fmode_for_setlk(file_lock); |
2308 | if (error) | |
2309 | goto out; | |
2310 | ||
5d50ffd7 JL |
2311 | /* |
2312 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2313 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2314 | */ |
2315 | switch (cmd) { | |
0d3f7a2d | 2316 | case F_OFD_SETLK: |
90478939 | 2317 | error = -EINVAL; |
a75d30c7 | 2318 | if (flock->l_pid != 0) |
90478939 JL |
2319 | goto out; |
2320 | ||
5d50ffd7 | 2321 | cmd = F_SETLK; |
cff2fce5 | 2322 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2323 | file_lock->fl_owner = filp; |
5d50ffd7 | 2324 | break; |
0d3f7a2d | 2325 | case F_OFD_SETLKW: |
90478939 | 2326 | error = -EINVAL; |
a75d30c7 | 2327 | if (flock->l_pid != 0) |
90478939 JL |
2328 | goto out; |
2329 | ||
5d50ffd7 | 2330 | cmd = F_SETLKW; |
cff2fce5 | 2331 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2332 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2333 | /* Fallthrough */ |
2334 | case F_SETLKW: | |
1da177e4 LT |
2335 | file_lock->fl_flags |= FL_SLEEP; |
2336 | } | |
5d50ffd7 | 2337 | |
b648a6de | 2338 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2339 | |
c293621b | 2340 | /* |
0752ba80 JL |
2341 | * Attempt to detect a close/fcntl race and recover by releasing the |
2342 | * lock that was just acquired. There is no need to do that when we're | |
2343 | * unlocking though, or for OFD locks. | |
c293621b | 2344 | */ |
0752ba80 JL |
2345 | if (!error && file_lock->fl_type != F_UNLCK && |
2346 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2347 | /* |
2348 | * We need that spin_lock here - it prevents reordering between | |
2349 | * update of i_flctx->flc_posix and check for it done in | |
2350 | * close(). rcu_read_lock() wouldn't do. | |
2351 | */ | |
2352 | spin_lock(¤t->files->file_lock); | |
2353 | f = fcheck(fd); | |
2354 | spin_unlock(¤t->files->file_lock); | |
2355 | if (f != filp) { | |
2356 | file_lock->fl_type = F_UNLCK; | |
2357 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2358 | WARN_ON_ONCE(error); | |
2359 | error = -EBADF; | |
2360 | } | |
1da177e4 | 2361 | } |
c293621b | 2362 | out: |
1890910f | 2363 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2364 | locks_free_lock(file_lock); |
2365 | return error; | |
2366 | } | |
2367 | ||
2368 | #if BITS_PER_LONG == 32 | |
2369 | /* Report the first existing lock that would conflict with l. | |
2370 | * This implements the F_GETLK command of fcntl(). | |
2371 | */ | |
a75d30c7 | 2372 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock) |
1da177e4 | 2373 | { |
52306e88 | 2374 | struct file_lock *fl; |
1da177e4 LT |
2375 | int error; |
2376 | ||
52306e88 BC |
2377 | fl = locks_alloc_lock(); |
2378 | if (fl == NULL) | |
2379 | return -ENOMEM; | |
2380 | ||
1da177e4 | 2381 | error = -EINVAL; |
a75d30c7 | 2382 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2383 | goto out; |
2384 | ||
52306e88 | 2385 | error = flock64_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2386 | if (error) |
2387 | goto out; | |
2388 | ||
0d3f7a2d | 2389 | if (cmd == F_OFD_GETLK) { |
90478939 | 2390 | error = -EINVAL; |
a75d30c7 | 2391 | if (flock->l_pid != 0) |
90478939 JL |
2392 | goto out; |
2393 | ||
5d50ffd7 | 2394 | cmd = F_GETLK64; |
52306e88 BC |
2395 | fl->fl_flags |= FL_OFDLCK; |
2396 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2397 | } |
2398 | ||
52306e88 | 2399 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2400 | if (error) |
2401 | goto out; | |
2402 | ||
52306e88 BC |
2403 | flock->l_type = fl->fl_type; |
2404 | if (fl->fl_type != F_UNLCK) | |
2405 | posix_lock_to_flock64(flock, fl); | |
f328296e | 2406 | |
1da177e4 | 2407 | out: |
52306e88 | 2408 | locks_free_lock(fl); |
1da177e4 LT |
2409 | return error; |
2410 | } | |
2411 | ||
2412 | /* Apply the lock described by l to an open file descriptor. | |
2413 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2414 | */ | |
c293621b | 2415 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2416 | struct flock64 *flock) |
1da177e4 LT |
2417 | { |
2418 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2419 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2420 | struct file *f; |
1da177e4 LT |
2421 | int error; |
2422 | ||
2423 | if (file_lock == NULL) | |
2424 | return -ENOLCK; | |
2425 | ||
1da177e4 LT |
2426 | /* Don't allow mandatory locks on files that may be memory mapped |
2427 | * and shared. | |
2428 | */ | |
a16877ca | 2429 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2430 | error = -EAGAIN; |
2431 | goto out; | |
2432 | } | |
2433 | ||
a75d30c7 | 2434 | error = flock64_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2435 | if (error) |
2436 | goto out; | |
5d50ffd7 | 2437 | |
cf01f4ee JL |
2438 | error = check_fmode_for_setlk(file_lock); |
2439 | if (error) | |
2440 | goto out; | |
2441 | ||
5d50ffd7 JL |
2442 | /* |
2443 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2444 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2445 | */ |
2446 | switch (cmd) { | |
0d3f7a2d | 2447 | case F_OFD_SETLK: |
90478939 | 2448 | error = -EINVAL; |
a75d30c7 | 2449 | if (flock->l_pid != 0) |
90478939 JL |
2450 | goto out; |
2451 | ||
5d50ffd7 | 2452 | cmd = F_SETLK64; |
cff2fce5 | 2453 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2454 | file_lock->fl_owner = filp; |
5d50ffd7 | 2455 | break; |
0d3f7a2d | 2456 | case F_OFD_SETLKW: |
90478939 | 2457 | error = -EINVAL; |
a75d30c7 | 2458 | if (flock->l_pid != 0) |
90478939 JL |
2459 | goto out; |
2460 | ||
5d50ffd7 | 2461 | cmd = F_SETLKW64; |
cff2fce5 | 2462 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2463 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2464 | /* Fallthrough */ |
2465 | case F_SETLKW64: | |
1da177e4 LT |
2466 | file_lock->fl_flags |= FL_SLEEP; |
2467 | } | |
5d50ffd7 | 2468 | |
b648a6de | 2469 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2470 | |
c293621b | 2471 | /* |
0752ba80 JL |
2472 | * Attempt to detect a close/fcntl race and recover by releasing the |
2473 | * lock that was just acquired. There is no need to do that when we're | |
2474 | * unlocking though, or for OFD locks. | |
c293621b | 2475 | */ |
0752ba80 JL |
2476 | if (!error && file_lock->fl_type != F_UNLCK && |
2477 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2478 | /* |
2479 | * We need that spin_lock here - it prevents reordering between | |
2480 | * update of i_flctx->flc_posix and check for it done in | |
2481 | * close(). rcu_read_lock() wouldn't do. | |
2482 | */ | |
2483 | spin_lock(¤t->files->file_lock); | |
2484 | f = fcheck(fd); | |
2485 | spin_unlock(¤t->files->file_lock); | |
2486 | if (f != filp) { | |
2487 | file_lock->fl_type = F_UNLCK; | |
2488 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2489 | WARN_ON_ONCE(error); | |
2490 | error = -EBADF; | |
2491 | } | |
1da177e4 | 2492 | } |
1da177e4 LT |
2493 | out: |
2494 | locks_free_lock(file_lock); | |
2495 | return error; | |
2496 | } | |
2497 | #endif /* BITS_PER_LONG == 32 */ | |
2498 | ||
2499 | /* | |
2500 | * This function is called when the file is being removed | |
2501 | * from the task's fd array. POSIX locks belonging to this task | |
2502 | * are deleted at this time. | |
2503 | */ | |
2504 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2505 | { | |
1890910f | 2506 | int error; |
c568d683 | 2507 | struct inode *inode = locks_inode(filp); |
ff7b86b8 | 2508 | struct file_lock lock; |
128a3785 | 2509 | struct file_lock_context *ctx; |
1da177e4 LT |
2510 | |
2511 | /* | |
2512 | * If there are no locks held on this file, we don't need to call | |
2513 | * posix_lock_file(). Another process could be setting a lock on this | |
2514 | * file at the same time, but we wouldn't remove that lock anyway. | |
2515 | */ | |
c568d683 | 2516 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 | 2517 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2518 | return; |
2519 | ||
d6367d62 | 2520 | locks_init_lock(&lock); |
1da177e4 | 2521 | lock.fl_type = F_UNLCK; |
75e1fcc0 | 2522 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2523 | lock.fl_start = 0; |
2524 | lock.fl_end = OFFSET_MAX; | |
2525 | lock.fl_owner = owner; | |
2526 | lock.fl_pid = current->tgid; | |
2527 | lock.fl_file = filp; | |
2528 | lock.fl_ops = NULL; | |
2529 | lock.fl_lmops = NULL; | |
2530 | ||
1890910f | 2531 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2532 | |
1da177e4 LT |
2533 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2534 | lock.fl_ops->fl_release_private(&lock); | |
c568d683 | 2535 | trace_locks_remove_posix(inode, &lock, error); |
1da177e4 LT |
2536 | } |
2537 | ||
2538 | EXPORT_SYMBOL(locks_remove_posix); | |
2539 | ||
3d8e560d | 2540 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2541 | static void |
128a3785 | 2542 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 | 2543 | { |
d6367d62 | 2544 | struct file_lock fl; |
c568d683 | 2545 | struct inode *inode = locks_inode(filp); |
dd459bb1 | 2546 | |
3d8e560d | 2547 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2548 | return; |
2549 | ||
d6367d62 N |
2550 | flock_make_lock(filp, LOCK_UN, &fl); |
2551 | fl.fl_flags |= FL_CLOSE; | |
2552 | ||
de2a4a50 | 2553 | if (filp->f_op->flock) |
dd459bb1 JL |
2554 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2555 | else | |
bcd7f78d | 2556 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2557 | |
2558 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2559 | fl.fl_ops->fl_release_private(&fl); | |
2560 | } | |
2561 | ||
3d8e560d | 2562 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2563 | static void |
128a3785 | 2564 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2565 | { |
8634b51f JL |
2566 | struct file_lock *fl, *tmp; |
2567 | LIST_HEAD(dispose); | |
2568 | ||
3d8e560d | 2569 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2570 | return; |
2571 | ||
5f43086b | 2572 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 2573 | spin_lock(&ctx->flc_lock); |
8634b51f | 2574 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2575 | if (filp == fl->fl_file) |
2576 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2577 | spin_unlock(&ctx->flc_lock); |
5f43086b PZ |
2578 | percpu_up_read_preempt_enable(&file_rwsem); |
2579 | ||
8634b51f JL |
2580 | locks_dispose_list(&dispose); |
2581 | } | |
2582 | ||
1da177e4 LT |
2583 | /* |
2584 | * This function is called on the last close of an open file. | |
2585 | */ | |
78ed8a13 | 2586 | void locks_remove_file(struct file *filp) |
1da177e4 | 2587 | { |
128a3785 DV |
2588 | struct file_lock_context *ctx; |
2589 | ||
c568d683 | 2590 | ctx = smp_load_acquire(&locks_inode(filp)->i_flctx); |
128a3785 | 2591 | if (!ctx) |
3d8e560d JL |
2592 | return; |
2593 | ||
dd459bb1 | 2594 | /* remove any OFD locks */ |
73a8f5f7 | 2595 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2596 | |
dd459bb1 | 2597 | /* remove flock locks */ |
128a3785 | 2598 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2599 | |
8634b51f | 2600 | /* remove any leases */ |
128a3785 | 2601 | locks_remove_lease(filp, ctx); |
3953704f BC |
2602 | |
2603 | spin_lock(&ctx->flc_lock); | |
2604 | locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX"); | |
2605 | locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK"); | |
2606 | locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE"); | |
2607 | spin_unlock(&ctx->flc_lock); | |
1da177e4 LT |
2608 | } |
2609 | ||
1da177e4 LT |
2610 | /** |
2611 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2612 | * @waiter: the lock which was waiting |
2613 | * | |
2614 | * lockd needs to block waiting for locks. | |
2615 | */ | |
64a318ee | 2616 | int |
f891a29f | 2617 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2618 | { |
5946c431 | 2619 | int status = -ENOENT; |
64a318ee | 2620 | |
7b2296af | 2621 | spin_lock(&blocked_lock_lock); |
5946c431 N |
2622 | if (waiter->fl_blocker) { |
2623 | __locks_wake_up_blocks(waiter); | |
1da177e4 | 2624 | __locks_delete_block(waiter); |
5946c431 N |
2625 | status = 0; |
2626 | } | |
7b2296af | 2627 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2628 | return status; |
1da177e4 | 2629 | } |
1da177e4 LT |
2630 | EXPORT_SYMBOL(posix_unblock_lock); |
2631 | ||
9b9d2ab4 ME |
2632 | /** |
2633 | * vfs_cancel_lock - file byte range unblock lock | |
2634 | * @filp: The file to apply the unblock to | |
2635 | * @fl: The lock to be unblocked | |
2636 | * | |
2637 | * Used by lock managers to cancel blocked requests | |
2638 | */ | |
2639 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2640 | { | |
de2a4a50 | 2641 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2642 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2643 | return 0; | |
2644 | } | |
2645 | ||
2646 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2647 | ||
7f8ada98 | 2648 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2649 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2650 | #include <linux/seq_file.h> |
2651 | ||
7012b02a JL |
2652 | struct locks_iterator { |
2653 | int li_cpu; | |
2654 | loff_t li_pos; | |
2655 | }; | |
2656 | ||
7f8ada98 | 2657 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2658 | loff_t id, char *pfx) |
1da177e4 LT |
2659 | { |
2660 | struct inode *inode = NULL; | |
ab1f1611 | 2661 | unsigned int fl_pid; |
9d5b86ac | 2662 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
ab1f1611 | 2663 | |
9d5b86ac BC |
2664 | fl_pid = locks_translate_pid(fl, proc_pidns); |
2665 | /* | |
1cf8e5de KK |
2666 | * If lock owner is dead (and pid is freed) or not visible in current |
2667 | * pidns, zero is shown as a pid value. Check lock info from | |
2668 | * init_pid_ns to get saved lock pid value. | |
9d5b86ac | 2669 | */ |
1da177e4 LT |
2670 | |
2671 | if (fl->fl_file != NULL) | |
c568d683 | 2672 | inode = locks_inode(fl->fl_file); |
1da177e4 | 2673 | |
99dc8292 | 2674 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2675 | if (IS_POSIX(fl)) { |
c918d42a | 2676 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2677 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2678 | else if (IS_OFDLCK(fl)) |
5315c26a | 2679 | seq_puts(f, "OFDLCK"); |
c918d42a | 2680 | else |
5315c26a | 2681 | seq_puts(f, "POSIX "); |
c918d42a JL |
2682 | |
2683 | seq_printf(f, " %s ", | |
1da177e4 | 2684 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2685 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2686 | } else if (IS_FLOCK(fl)) { |
2687 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2688 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2689 | } else { |
5315c26a | 2690 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2691 | } |
2692 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2693 | if (fl->fl_flags & FL_DELEG) |
2694 | seq_puts(f, "DELEG "); | |
2695 | else | |
2696 | seq_puts(f, "LEASE "); | |
2697 | ||
ab83fa4b | 2698 | if (lease_breaking(fl)) |
5315c26a | 2699 | seq_puts(f, "BREAKING "); |
1da177e4 | 2700 | else if (fl->fl_file) |
5315c26a | 2701 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2702 | else |
5315c26a | 2703 | seq_puts(f, "BREAKER "); |
1da177e4 | 2704 | } else { |
5315c26a | 2705 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2706 | } |
2707 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2708 | seq_printf(f, "%s ", |
1da177e4 LT |
2709 | (fl->fl_type & LOCK_READ) |
2710 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2711 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2712 | } else { | |
7f8ada98 | 2713 | seq_printf(f, "%s ", |
ab83fa4b | 2714 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2715 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2716 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2717 | } |
2718 | if (inode) { | |
3648888e | 2719 | /* userspace relies on this representation of dev_t */ |
ab1f1611 | 2720 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2721 | MAJOR(inode->i_sb->s_dev), |
2722 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2723 | } else { |
ab1f1611 | 2724 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2725 | } |
2726 | if (IS_POSIX(fl)) { | |
2727 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2728 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2729 | else |
7f8ada98 | 2730 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2731 | } else { |
5315c26a | 2732 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2733 | } |
2734 | } | |
2735 | ||
7f8ada98 | 2736 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2737 | { |
7012b02a | 2738 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2739 | struct file_lock *fl, *bfl; |
d67fd44f | 2740 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
1da177e4 | 2741 | |
139ca04e | 2742 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2743 | |
9d5b86ac | 2744 | if (locks_translate_pid(fl, proc_pidns) == 0) |
d67fd44f NB |
2745 | return 0; |
2746 | ||
7012b02a | 2747 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2748 | |
ada5c1da | 2749 | list_for_each_entry(bfl, &fl->fl_blocked_requests, fl_blocked_member) |
7012b02a | 2750 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2751 | |
7f8ada98 PE |
2752 | return 0; |
2753 | } | |
1da177e4 | 2754 | |
6c8c9031 AV |
2755 | static void __show_fd_locks(struct seq_file *f, |
2756 | struct list_head *head, int *id, | |
2757 | struct file *filp, struct files_struct *files) | |
2758 | { | |
2759 | struct file_lock *fl; | |
2760 | ||
2761 | list_for_each_entry(fl, head, fl_list) { | |
2762 | ||
2763 | if (filp != fl->fl_file) | |
2764 | continue; | |
2765 | if (fl->fl_owner != files && | |
2766 | fl->fl_owner != filp) | |
2767 | continue; | |
2768 | ||
2769 | (*id)++; | |
2770 | seq_puts(f, "lock:\t"); | |
2771 | lock_get_status(f, fl, *id, ""); | |
2772 | } | |
2773 | } | |
2774 | ||
2775 | void show_fd_locks(struct seq_file *f, | |
2776 | struct file *filp, struct files_struct *files) | |
2777 | { | |
c568d683 | 2778 | struct inode *inode = locks_inode(filp); |
6c8c9031 AV |
2779 | struct file_lock_context *ctx; |
2780 | int id = 0; | |
2781 | ||
128a3785 | 2782 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2783 | if (!ctx) |
2784 | return; | |
2785 | ||
2786 | spin_lock(&ctx->flc_lock); | |
2787 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2788 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2789 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2790 | spin_unlock(&ctx->flc_lock); | |
2791 | } | |
2792 | ||
7f8ada98 | 2793 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2794 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2795 | { |
7012b02a | 2796 | struct locks_iterator *iter = f->private; |
99dc8292 | 2797 | |
7012b02a | 2798 | iter->li_pos = *pos + 1; |
aba37660 | 2799 | percpu_down_write(&file_rwsem); |
7b2296af | 2800 | spin_lock(&blocked_lock_lock); |
7c3f654d | 2801 | return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos); |
7f8ada98 | 2802 | } |
1da177e4 | 2803 | |
7f8ada98 PE |
2804 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2805 | { | |
7012b02a JL |
2806 | struct locks_iterator *iter = f->private; |
2807 | ||
2808 | ++iter->li_pos; | |
7c3f654d | 2809 | return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos); |
7f8ada98 | 2810 | } |
1da177e4 | 2811 | |
7f8ada98 | 2812 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2813 | __releases(&blocked_lock_lock) |
7f8ada98 | 2814 | { |
7b2296af | 2815 | spin_unlock(&blocked_lock_lock); |
aba37660 | 2816 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2817 | } |
2818 | ||
d8ba7a36 | 2819 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2820 | .start = locks_start, |
2821 | .next = locks_next, | |
2822 | .stop = locks_stop, | |
2823 | .show = locks_show, | |
2824 | }; | |
d8ba7a36 | 2825 | |
d8ba7a36 AD |
2826 | static int __init proc_locks_init(void) |
2827 | { | |
44414d82 CH |
2828 | proc_create_seq_private("locks", 0, NULL, &locks_seq_operations, |
2829 | sizeof(struct locks_iterator), NULL); | |
d8ba7a36 AD |
2830 | return 0; |
2831 | } | |
91899226 | 2832 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2833 | #endif |
2834 | ||
1da177e4 LT |
2835 | static int __init filelock_init(void) |
2836 | { | |
7012b02a JL |
2837 | int i; |
2838 | ||
4a075e39 JL |
2839 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2840 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2841 | ||
1da177e4 | 2842 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2843 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2844 | ||
7012b02a | 2845 | |
7c3f654d PZ |
2846 | for_each_possible_cpu(i) { |
2847 | struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i); | |
2848 | ||
2849 | spin_lock_init(&fll->lock); | |
2850 | INIT_HLIST_HEAD(&fll->hlist); | |
2851 | } | |
7012b02a | 2852 | |
1da177e4 LT |
2853 | return 0; |
2854 | } | |
2855 | ||
2856 | core_initcall(filelock_init); |