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 | { |
16306a61 N |
713 | /* |
714 | * If fl_blocker is NULL, it won't be set again as this thread | |
715 | * "owns" the lock and is the only one that might try to claim | |
716 | * the lock. So it is safe to test fl_blocker locklessly. | |
717 | * Also if fl_blocker is NULL, this waiter is not listed on | |
718 | * fl_blocked_requests for some lock, so no other request can | |
719 | * be added to the list of fl_blocked_requests for this | |
720 | * request. So if fl_blocker is NULL, it is safe to | |
721 | * locklessly check if fl_blocked_requests is empty. If both | |
722 | * of these checks succeed, there is no need to take the lock. | |
723 | */ | |
724 | if (waiter->fl_blocker == NULL && | |
725 | list_empty(&waiter->fl_blocked_requests)) | |
726 | return; | |
7b2296af | 727 | spin_lock(&blocked_lock_lock); |
5946c431 | 728 | __locks_wake_up_blocks(waiter); |
1da177e4 | 729 | __locks_delete_block(waiter); |
7b2296af | 730 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
731 | } |
732 | ||
733 | /* Insert waiter into blocker's block list. | |
734 | * We use a circular list so that processes can be easily woken up in | |
735 | * the order they blocked. The documentation doesn't require this but | |
736 | * it seems like the reasonable thing to do. | |
1c8c601a | 737 | * |
6109c850 | 738 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
ada5c1da N |
739 | * fl_blocked_requests list itself is protected by the blocked_lock_lock, |
740 | * but by ensuring that the flc_lock is also held on insertions we can avoid | |
741 | * taking the blocked_lock_lock in some cases when we see that the | |
742 | * fl_blocked_requests list is empty. | |
1da177e4 | 743 | */ |
1c8c601a JL |
744 | static void __locks_insert_block(struct file_lock *blocker, |
745 | struct file_lock *waiter) | |
1da177e4 | 746 | { |
ada5c1da N |
747 | BUG_ON(!list_empty(&waiter->fl_blocked_member)); |
748 | waiter->fl_blocker = blocker; | |
749 | list_add_tail(&waiter->fl_blocked_member, &blocker->fl_blocked_requests); | |
cff2fce5 | 750 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a | 751 | locks_insert_global_blocked(waiter); |
5946c431 N |
752 | |
753 | /* The requests in waiter->fl_blocked are known to conflict with | |
754 | * waiter, but might not conflict with blocker, or the requests | |
755 | * and lock which block it. So they all need to be woken. | |
756 | */ | |
757 | __locks_wake_up_blocks(waiter); | |
1c8c601a JL |
758 | } |
759 | ||
6109c850 | 760 | /* Must be called with flc_lock held. */ |
1c8c601a JL |
761 | static void locks_insert_block(struct file_lock *blocker, |
762 | struct file_lock *waiter) | |
763 | { | |
7b2296af | 764 | spin_lock(&blocked_lock_lock); |
1c8c601a | 765 | __locks_insert_block(blocker, waiter); |
7b2296af | 766 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
767 | } |
768 | ||
1cb36012 JL |
769 | /* |
770 | * Wake up processes blocked waiting for blocker. | |
771 | * | |
6109c850 | 772 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
773 | */ |
774 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
775 | { | |
4e8c765d JL |
776 | /* |
777 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 | 778 | * blocked requests are only added to the list under the flc_lock, and |
ada5c1da N |
779 | * the flc_lock is always held here. Note that removal from the |
780 | * fl_blocked_requests list does not require the flc_lock, so we must | |
781 | * recheck list_empty() after acquiring the blocked_lock_lock. | |
4e8c765d | 782 | */ |
ada5c1da | 783 | if (list_empty(&blocker->fl_blocked_requests)) |
4e8c765d JL |
784 | return; |
785 | ||
7b2296af | 786 | spin_lock(&blocked_lock_lock); |
ad6bbd8b | 787 | __locks_wake_up_blocks(blocker); |
7b2296af | 788 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
789 | } |
790 | ||
5263e31e | 791 | static void |
e084c1bd | 792 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e | 793 | { |
5263e31e JL |
794 | list_add_tail(&fl->fl_list, before); |
795 | locks_insert_global_locks(fl); | |
796 | } | |
797 | ||
8634b51f | 798 | static void |
e084c1bd | 799 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 800 | { |
88974691 | 801 | locks_delete_global_locks(fl); |
8634b51f | 802 | list_del_init(&fl->fl_list); |
1da177e4 | 803 | locks_wake_up_blocks(fl); |
24cbe784 JL |
804 | } |
805 | ||
8634b51f | 806 | static void |
e084c1bd | 807 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 808 | { |
e084c1bd | 809 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 810 | if (dispose) |
6dee60f6 | 811 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
812 | else |
813 | locks_free_lock(fl); | |
1da177e4 LT |
814 | } |
815 | ||
816 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
817 | * checks for shared/exclusive status of overlapping locks. | |
818 | */ | |
c0e15908 N |
819 | static bool locks_conflict(struct file_lock *caller_fl, |
820 | struct file_lock *sys_fl) | |
1da177e4 LT |
821 | { |
822 | if (sys_fl->fl_type == F_WRLCK) | |
c0e15908 | 823 | return true; |
1da177e4 | 824 | if (caller_fl->fl_type == F_WRLCK) |
c0e15908 N |
825 | return true; |
826 | return false; | |
1da177e4 LT |
827 | } |
828 | ||
829 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
830 | * checking before calling the locks_conflict(). | |
831 | */ | |
c0e15908 N |
832 | static bool posix_locks_conflict(struct file_lock *caller_fl, |
833 | struct file_lock *sys_fl) | |
1da177e4 LT |
834 | { |
835 | /* POSIX locks owned by the same process do not conflict with | |
836 | * each other. | |
837 | */ | |
9b8c8695 | 838 | if (posix_same_owner(caller_fl, sys_fl)) |
c0e15908 | 839 | return false; |
1da177e4 LT |
840 | |
841 | /* Check whether they overlap */ | |
842 | if (!locks_overlap(caller_fl, sys_fl)) | |
c0e15908 | 843 | return false; |
1da177e4 | 844 | |
c0e15908 | 845 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
846 | } |
847 | ||
848 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
849 | * checking before calling the locks_conflict(). | |
850 | */ | |
c0e15908 N |
851 | static bool flock_locks_conflict(struct file_lock *caller_fl, |
852 | struct file_lock *sys_fl) | |
1da177e4 LT |
853 | { |
854 | /* FLOCK locks referring to the same filp do not conflict with | |
855 | * each other. | |
856 | */ | |
9b8c8695 | 857 | if (caller_fl->fl_file == sys_fl->fl_file) |
c0e15908 | 858 | return false; |
1da177e4 | 859 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) |
c0e15908 | 860 | return false; |
1da177e4 | 861 | |
c0e15908 | 862 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
863 | } |
864 | ||
6d34ac19 | 865 | void |
9d6a8c5c | 866 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
867 | { |
868 | struct file_lock *cfl; | |
bd61e0a9 | 869 | struct file_lock_context *ctx; |
c568d683 | 870 | struct inode *inode = locks_inode(filp); |
1da177e4 | 871 | |
128a3785 | 872 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
873 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
874 | fl->fl_type = F_UNLCK; | |
875 | return; | |
876 | } | |
877 | ||
6109c850 | 878 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
879 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
880 | if (posix_locks_conflict(fl, cfl)) { | |
881 | locks_copy_conflock(fl, cfl); | |
bd61e0a9 JL |
882 | goto out; |
883 | } | |
1da177e4 | 884 | } |
bd61e0a9 JL |
885 | fl->fl_type = F_UNLCK; |
886 | out: | |
6109c850 | 887 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 888 | return; |
1da177e4 | 889 | } |
1da177e4 LT |
890 | EXPORT_SYMBOL(posix_test_lock); |
891 | ||
b533184f BF |
892 | /* |
893 | * Deadlock detection: | |
894 | * | |
895 | * We attempt to detect deadlocks that are due purely to posix file | |
896 | * locks. | |
1da177e4 | 897 | * |
b533184f BF |
898 | * We assume that a task can be waiting for at most one lock at a time. |
899 | * So for any acquired lock, the process holding that lock may be | |
900 | * waiting on at most one other lock. That lock in turns may be held by | |
901 | * someone waiting for at most one other lock. Given a requested lock | |
902 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
903 | * follow this chain of waiters to ensure we are not about to create a | |
904 | * cycle. | |
1da177e4 | 905 | * |
b533184f BF |
906 | * Since we do this before we ever put a process to sleep on a lock, we |
907 | * are ensured that there is never a cycle; that is what guarantees that | |
908 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 909 | * |
b533184f BF |
910 | * Note: the above assumption may not be true when handling lock |
911 | * requests from a broken NFS client. It may also fail in the presence | |
912 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 913 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 914 | * |
cff2fce5 | 915 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
916 | * Because the owner is not even nominally tied to a thread of |
917 | * execution, the deadlock detection below can't reasonably work well. Just | |
918 | * skip it for those. | |
919 | * | |
cff2fce5 | 920 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
921 | * locks that just checks for the case where two tasks are attempting to |
922 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 923 | */ |
97855b49 BF |
924 | |
925 | #define MAX_DEADLK_ITERATIONS 10 | |
926 | ||
b533184f BF |
927 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
928 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
929 | { | |
930 | struct file_lock *fl; | |
931 | ||
3999e493 | 932 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
5946c431 N |
933 | if (posix_same_owner(fl, block_fl)) { |
934 | while (fl->fl_blocker) | |
935 | fl = fl->fl_blocker; | |
936 | return fl; | |
937 | } | |
b533184f BF |
938 | } |
939 | return NULL; | |
940 | } | |
941 | ||
7b2296af | 942 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 943 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
944 | struct file_lock *block_fl) |
945 | { | |
97855b49 | 946 | int i = 0; |
1da177e4 | 947 | |
663d5af7 DW |
948 | lockdep_assert_held(&blocked_lock_lock); |
949 | ||
57b65325 JL |
950 | /* |
951 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 952 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 953 | */ |
cff2fce5 | 954 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
955 | return 0; |
956 | ||
b533184f BF |
957 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
958 | if (i++ > MAX_DEADLK_ITERATIONS) | |
959 | return 0; | |
960 | if (posix_same_owner(caller_fl, block_fl)) | |
961 | return 1; | |
1da177e4 LT |
962 | } |
963 | return 0; | |
964 | } | |
965 | ||
1da177e4 | 966 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 967 | * after any leases, but before any posix locks. |
f475ae95 TM |
968 | * |
969 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
970 | * whether or not a lock was successfully freed by testing the return | |
971 | * value for -ENOENT. | |
1da177e4 | 972 | */ |
bcd7f78d | 973 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 974 | { |
993dfa87 | 975 | struct file_lock *new_fl = NULL; |
5263e31e JL |
976 | struct file_lock *fl; |
977 | struct file_lock_context *ctx; | |
1da177e4 | 978 | int error = 0; |
5263e31e | 979 | bool found = false; |
ed9814d8 | 980 | LIST_HEAD(dispose); |
1da177e4 | 981 | |
5c1c669a JL |
982 | ctx = locks_get_lock_context(inode, request->fl_type); |
983 | if (!ctx) { | |
984 | if (request->fl_type != F_UNLCK) | |
985 | return -ENOMEM; | |
986 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
987 | } | |
5263e31e | 988 | |
b89f4321 | 989 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 990 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
991 | if (!new_fl) |
992 | return -ENOMEM; | |
84d535ad PE |
993 | } |
994 | ||
87709e28 | 995 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 996 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
997 | if (request->fl_flags & FL_ACCESS) |
998 | goto find_conflict; | |
999 | ||
5263e31e | 1000 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 1001 | if (request->fl_file != fl->fl_file) |
1da177e4 | 1002 | continue; |
993dfa87 | 1003 | if (request->fl_type == fl->fl_type) |
1da177e4 | 1004 | goto out; |
5263e31e | 1005 | found = true; |
e084c1bd | 1006 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1007 | break; |
1008 | } | |
1da177e4 | 1009 | |
f475ae95 TM |
1010 | if (request->fl_type == F_UNLCK) { |
1011 | if ((request->fl_flags & FL_EXISTS) && !found) | |
1012 | error = -ENOENT; | |
993dfa87 | 1013 | goto out; |
f475ae95 | 1014 | } |
1da177e4 | 1015 | |
f07f18dd | 1016 | find_conflict: |
5263e31e | 1017 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 1018 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
1019 | continue; |
1020 | error = -EAGAIN; | |
bde74e4b MS |
1021 | if (!(request->fl_flags & FL_SLEEP)) |
1022 | goto out; | |
1023 | error = FILE_LOCK_DEFERRED; | |
1024 | locks_insert_block(fl, request); | |
1da177e4 LT |
1025 | goto out; |
1026 | } | |
f07f18dd TM |
1027 | if (request->fl_flags & FL_ACCESS) |
1028 | goto out; | |
993dfa87 | 1029 | locks_copy_lock(new_fl, request); |
5946c431 | 1030 | locks_move_blocks(new_fl, request); |
e084c1bd | 1031 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 1032 | new_fl = NULL; |
9cedc194 | 1033 | error = 0; |
1da177e4 LT |
1034 | |
1035 | out: | |
6109c850 | 1036 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1037 | percpu_up_read_preempt_enable(&file_rwsem); |
993dfa87 TM |
1038 | if (new_fl) |
1039 | locks_free_lock(new_fl); | |
ed9814d8 | 1040 | locks_dispose_list(&dispose); |
c883da31 | 1041 | trace_flock_lock_inode(inode, request, error); |
1da177e4 LT |
1042 | return error; |
1043 | } | |
1044 | ||
b4d629a3 JL |
1045 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
1046 | struct file_lock *conflock) | |
1da177e4 | 1047 | { |
bd61e0a9 | 1048 | struct file_lock *fl, *tmp; |
39005d02 MS |
1049 | struct file_lock *new_fl = NULL; |
1050 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
1051 | struct file_lock *left = NULL; |
1052 | struct file_lock *right = NULL; | |
bd61e0a9 | 1053 | struct file_lock_context *ctx; |
b9746ef8 JL |
1054 | int error; |
1055 | bool added = false; | |
ed9814d8 | 1056 | LIST_HEAD(dispose); |
1da177e4 | 1057 | |
5c1c669a | 1058 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 1059 | if (!ctx) |
5c1c669a | 1060 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 1061 | |
1da177e4 LT |
1062 | /* |
1063 | * We may need two file_lock structures for this operation, | |
1064 | * so we get them in advance to avoid races. | |
39005d02 MS |
1065 | * |
1066 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1067 | */ |
39005d02 MS |
1068 | if (!(request->fl_flags & FL_ACCESS) && |
1069 | (request->fl_type != F_UNLCK || | |
1070 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1071 | new_fl = locks_alloc_lock(); | |
1072 | new_fl2 = locks_alloc_lock(); | |
1073 | } | |
1da177e4 | 1074 | |
87709e28 | 1075 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1076 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1077 | /* |
1078 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1079 | * there are any, either return error or put the request on the | |
48f74186 | 1080 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1081 | */ |
1da177e4 | 1082 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1083 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1084 | if (!posix_locks_conflict(request, fl)) |
1085 | continue; | |
5842add2 | 1086 | if (conflock) |
3fe0fff1 | 1087 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1088 | error = -EAGAIN; |
1089 | if (!(request->fl_flags & FL_SLEEP)) | |
1090 | goto out; | |
1c8c601a JL |
1091 | /* |
1092 | * Deadlock detection and insertion into the blocked | |
1093 | * locks list must be done while holding the same lock! | |
1094 | */ | |
1da177e4 | 1095 | error = -EDEADLK; |
7b2296af | 1096 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
1097 | if (likely(!posix_locks_deadlock(request, fl))) { |
1098 | error = FILE_LOCK_DEFERRED; | |
1099 | __locks_insert_block(fl, request); | |
1100 | } | |
7b2296af | 1101 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1102 | goto out; |
1103 | } | |
1104 | } | |
1105 | ||
1106 | /* If we're just looking for a conflict, we're done. */ | |
1107 | error = 0; | |
1108 | if (request->fl_flags & FL_ACCESS) | |
1109 | goto out; | |
1110 | ||
bd61e0a9 JL |
1111 | /* Find the first old lock with the same owner as the new lock */ |
1112 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1113 | if (posix_same_owner(request, fl)) | |
1114 | break; | |
1da177e4 LT |
1115 | } |
1116 | ||
1cb36012 | 1117 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1118 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1119 | if (!posix_same_owner(request, fl)) | |
1120 | break; | |
1121 | ||
1122 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1123 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1124 | /* In all comparisons of start vs end, use |
1125 | * "start - 1" rather than "end + 1". If end | |
1126 | * is OFFSET_MAX, end + 1 will become negative. | |
1127 | */ | |
1da177e4 | 1128 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1129 | continue; |
1da177e4 LT |
1130 | /* If the next lock in the list has entirely bigger |
1131 | * addresses than the new one, insert the lock here. | |
1132 | */ | |
449231d6 | 1133 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1134 | break; |
1135 | ||
1136 | /* If we come here, the new and old lock are of the | |
1137 | * same type and adjacent or overlapping. Make one | |
1138 | * lock yielding from the lower start address of both | |
1139 | * locks to the higher end address. | |
1140 | */ | |
1141 | if (fl->fl_start > request->fl_start) | |
1142 | fl->fl_start = request->fl_start; | |
1143 | else | |
1144 | request->fl_start = fl->fl_start; | |
1145 | if (fl->fl_end < request->fl_end) | |
1146 | fl->fl_end = request->fl_end; | |
1147 | else | |
1148 | request->fl_end = fl->fl_end; | |
1149 | if (added) { | |
e084c1bd | 1150 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1151 | continue; |
1152 | } | |
1153 | request = fl; | |
b9746ef8 | 1154 | added = true; |
bd61e0a9 | 1155 | } else { |
1da177e4 LT |
1156 | /* Processing for different lock types is a bit |
1157 | * more complex. | |
1158 | */ | |
1159 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1160 | continue; |
1da177e4 LT |
1161 | if (fl->fl_start > request->fl_end) |
1162 | break; | |
1163 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1164 | added = true; |
1da177e4 LT |
1165 | if (fl->fl_start < request->fl_start) |
1166 | left = fl; | |
1167 | /* If the next lock in the list has a higher end | |
1168 | * address than the new one, insert the new one here. | |
1169 | */ | |
1170 | if (fl->fl_end > request->fl_end) { | |
1171 | right = fl; | |
1172 | break; | |
1173 | } | |
1174 | if (fl->fl_start >= request->fl_start) { | |
1175 | /* The new lock completely replaces an old | |
1176 | * one (This may happen several times). | |
1177 | */ | |
1178 | if (added) { | |
e084c1bd | 1179 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1180 | continue; |
1181 | } | |
b84d49f9 JL |
1182 | /* |
1183 | * Replace the old lock with new_fl, and | |
1184 | * remove the old one. It's safe to do the | |
1185 | * insert here since we know that we won't be | |
1186 | * using new_fl later, and that the lock is | |
1187 | * just replacing an existing lock. | |
1da177e4 | 1188 | */ |
b84d49f9 JL |
1189 | error = -ENOLCK; |
1190 | if (!new_fl) | |
1191 | goto out; | |
1192 | locks_copy_lock(new_fl, request); | |
1193 | request = new_fl; | |
1194 | new_fl = NULL; | |
e084c1bd JL |
1195 | locks_insert_lock_ctx(request, &fl->fl_list); |
1196 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1197 | added = true; |
1da177e4 LT |
1198 | } |
1199 | } | |
1da177e4 LT |
1200 | } |
1201 | ||
0d9a490a | 1202 | /* |
1cb36012 JL |
1203 | * The above code only modifies existing locks in case of merging or |
1204 | * replacing. If new lock(s) need to be inserted all modifications are | |
1205 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1206 | */ |
1207 | error = -ENOLCK; /* "no luck" */ | |
1208 | if (right && left == right && !new_fl2) | |
1209 | goto out; | |
1210 | ||
1da177e4 LT |
1211 | error = 0; |
1212 | if (!added) { | |
f475ae95 TM |
1213 | if (request->fl_type == F_UNLCK) { |
1214 | if (request->fl_flags & FL_EXISTS) | |
1215 | error = -ENOENT; | |
1da177e4 | 1216 | goto out; |
f475ae95 | 1217 | } |
0d9a490a MS |
1218 | |
1219 | if (!new_fl) { | |
1220 | error = -ENOLCK; | |
1221 | goto out; | |
1222 | } | |
1da177e4 | 1223 | locks_copy_lock(new_fl, request); |
5946c431 | 1224 | locks_move_blocks(new_fl, request); |
e084c1bd | 1225 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
2e2f756f | 1226 | fl = new_fl; |
1da177e4 LT |
1227 | new_fl = NULL; |
1228 | } | |
1229 | if (right) { | |
1230 | if (left == right) { | |
1231 | /* The new lock breaks the old one in two pieces, | |
1232 | * so we have to use the second new lock. | |
1233 | */ | |
1234 | left = new_fl2; | |
1235 | new_fl2 = NULL; | |
1236 | locks_copy_lock(left, right); | |
e084c1bd | 1237 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1238 | } |
1239 | right->fl_start = request->fl_end + 1; | |
1240 | locks_wake_up_blocks(right); | |
1241 | } | |
1242 | if (left) { | |
1243 | left->fl_end = request->fl_start - 1; | |
1244 | locks_wake_up_blocks(left); | |
1245 | } | |
1246 | out: | |
6109c850 | 1247 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1248 | percpu_up_read_preempt_enable(&file_rwsem); |
1da177e4 LT |
1249 | /* |
1250 | * Free any unused locks. | |
1251 | */ | |
1252 | if (new_fl) | |
1253 | locks_free_lock(new_fl); | |
1254 | if (new_fl2) | |
1255 | locks_free_lock(new_fl2); | |
ed9814d8 | 1256 | locks_dispose_list(&dispose); |
1890910f JL |
1257 | trace_posix_lock_inode(inode, request, error); |
1258 | ||
1da177e4 LT |
1259 | return error; |
1260 | } | |
1261 | ||
1262 | /** | |
1263 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1264 | * @filp: The file to apply the lock to | |
1265 | * @fl: The lock to be applied | |
150b3934 | 1266 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1267 | * |
1268 | * Add a POSIX style lock to a file. | |
1269 | * We merge adjacent & overlapping locks whenever possible. | |
1270 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1271 | * |
1272 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1273 | * whether or not a lock was successfully freed by testing the return | |
1274 | * value for -ENOENT. | |
1da177e4 | 1275 | */ |
150b3934 | 1276 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1277 | struct file_lock *conflock) |
1278 | { | |
c568d683 | 1279 | return posix_lock_inode(locks_inode(filp), fl, conflock); |
1da177e4 | 1280 | } |
150b3934 | 1281 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1282 | |
1283 | /** | |
29d01b22 JL |
1284 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1285 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1286 | * @fl: The lock to be applied |
1287 | * | |
616fb38f | 1288 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1289 | */ |
616fb38f | 1290 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1291 | { |
1292 | int error; | |
1293 | might_sleep (); | |
1294 | for (;;) { | |
b4d629a3 | 1295 | error = posix_lock_inode(inode, fl, NULL); |
bde74e4b | 1296 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1297 | break; |
ada5c1da | 1298 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
1299 | if (error) |
1300 | break; | |
1da177e4 | 1301 | } |
16306a61 | 1302 | locks_delete_block(fl); |
1da177e4 LT |
1303 | return error; |
1304 | } | |
29d01b22 | 1305 | |
9e8925b6 | 1306 | #ifdef CONFIG_MANDATORY_FILE_LOCKING |
1da177e4 LT |
1307 | /** |
1308 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1309 | * @file: the file to check |
1da177e4 LT |
1310 | * |
1311 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1312 | * This function is called from locks_verify_locked() only. | |
1313 | */ | |
d7a06983 | 1314 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1315 | { |
bd61e0a9 | 1316 | int ret; |
c568d683 | 1317 | struct inode *inode = locks_inode(file); |
bd61e0a9 | 1318 | struct file_lock_context *ctx; |
1da177e4 LT |
1319 | struct file_lock *fl; |
1320 | ||
128a3785 | 1321 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
1322 | if (!ctx || list_empty_careful(&ctx->flc_posix)) |
1323 | return 0; | |
1324 | ||
1da177e4 LT |
1325 | /* |
1326 | * Search the lock list for this inode for any POSIX locks. | |
1327 | */ | |
6109c850 | 1328 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
1329 | ret = 0; |
1330 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1331 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1332 | fl->fl_owner != file) { |
1333 | ret = -EAGAIN; | |
1da177e4 | 1334 | break; |
bd61e0a9 | 1335 | } |
1da177e4 | 1336 | } |
6109c850 | 1337 | spin_unlock(&ctx->flc_lock); |
bd61e0a9 | 1338 | return ret; |
1da177e4 LT |
1339 | } |
1340 | ||
1341 | /** | |
1342 | * locks_mandatory_area - Check for a conflicting lock | |
acc15575 | 1343 | * @inode: the file to check |
1da177e4 | 1344 | * @filp: how the file was opened (if it was) |
acc15575 CH |
1345 | * @start: first byte in the file to check |
1346 | * @end: lastbyte in the file to check | |
1347 | * @type: %F_WRLCK for a write lock, else %F_RDLCK | |
1da177e4 LT |
1348 | * |
1349 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1da177e4 | 1350 | */ |
acc15575 CH |
1351 | int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start, |
1352 | loff_t end, unsigned char type) | |
1da177e4 LT |
1353 | { |
1354 | struct file_lock fl; | |
1355 | int error; | |
29723ade | 1356 | bool sleep = false; |
1da177e4 LT |
1357 | |
1358 | locks_init_lock(&fl); | |
1da177e4 LT |
1359 | fl.fl_pid = current->tgid; |
1360 | fl.fl_file = filp; | |
1361 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1362 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1363 | sleep = true; |
acc15575 CH |
1364 | fl.fl_type = type; |
1365 | fl.fl_start = start; | |
1366 | fl.fl_end = end; | |
1da177e4 LT |
1367 | |
1368 | for (;;) { | |
29723ade | 1369 | if (filp) { |
73a8f5f7 | 1370 | fl.fl_owner = filp; |
29723ade | 1371 | fl.fl_flags &= ~FL_SLEEP; |
b4d629a3 | 1372 | error = posix_lock_inode(inode, &fl, NULL); |
29723ade JL |
1373 | if (!error) |
1374 | break; | |
1375 | } | |
1376 | ||
1377 | if (sleep) | |
1378 | fl.fl_flags |= FL_SLEEP; | |
1379 | fl.fl_owner = current->files; | |
b4d629a3 | 1380 | error = posix_lock_inode(inode, &fl, NULL); |
bde74e4b | 1381 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1382 | break; |
ada5c1da | 1383 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_blocker); |
1da177e4 LT |
1384 | if (!error) { |
1385 | /* | |
1386 | * If we've been sleeping someone might have | |
1387 | * changed the permissions behind our back. | |
1388 | */ | |
a16877ca | 1389 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1390 | continue; |
1391 | } | |
1392 | ||
1da177e4 LT |
1393 | break; |
1394 | } | |
16306a61 | 1395 | locks_delete_block(&fl); |
1da177e4 LT |
1396 | |
1397 | return error; | |
1398 | } | |
1399 | ||
1400 | EXPORT_SYMBOL(locks_mandatory_area); | |
9e8925b6 | 1401 | #endif /* CONFIG_MANDATORY_FILE_LOCKING */ |
1da177e4 | 1402 | |
778fc546 BF |
1403 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1404 | { | |
1405 | switch (arg) { | |
1406 | case F_UNLCK: | |
1407 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1408 | /* fall through: */ | |
1409 | case F_RDLCK: | |
1410 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1411 | } | |
1412 | } | |
1413 | ||
1da177e4 | 1414 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1415 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1416 | { |
1da177e4 LT |
1417 | int error = assign_type(fl, arg); |
1418 | ||
1419 | if (error) | |
1420 | return error; | |
778fc546 | 1421 | lease_clear_pending(fl, arg); |
1da177e4 | 1422 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1423 | if (arg == F_UNLCK) { |
1424 | struct file *filp = fl->fl_file; | |
1425 | ||
1426 | f_delown(filp); | |
1427 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1428 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1429 | if (fl->fl_fasync != NULL) { | |
1430 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1431 | fl->fl_fasync = NULL; | |
1432 | } | |
e084c1bd | 1433 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1434 | } |
1da177e4 LT |
1435 | return 0; |
1436 | } | |
1da177e4 LT |
1437 | EXPORT_SYMBOL(lease_modify); |
1438 | ||
778fc546 BF |
1439 | static bool past_time(unsigned long then) |
1440 | { | |
1441 | if (!then) | |
1442 | /* 0 is a special value meaning "this never expires": */ | |
1443 | return false; | |
1444 | return time_after(jiffies, then); | |
1445 | } | |
1446 | ||
c45198ed | 1447 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1448 | { |
8634b51f JL |
1449 | struct file_lock_context *ctx = inode->i_flctx; |
1450 | struct file_lock *fl, *tmp; | |
1da177e4 | 1451 | |
6109c850 | 1452 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1453 | |
8634b51f | 1454 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1455 | trace_time_out_leases(inode, fl); |
778fc546 | 1456 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1457 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1458 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1459 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1460 | } |
1461 | } | |
1462 | ||
df4e8d2c BF |
1463 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1464 | { | |
11afe9f7 CH |
1465 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) |
1466 | return false; | |
df4e8d2c BF |
1467 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) |
1468 | return false; | |
1469 | return locks_conflict(breaker, lease); | |
1470 | } | |
1471 | ||
03d12ddf JL |
1472 | static bool |
1473 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1474 | { | |
8634b51f | 1475 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1476 | struct file_lock *fl; |
1477 | ||
6109c850 | 1478 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1479 | |
8634b51f | 1480 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1481 | if (leases_conflict(fl, breaker)) |
1482 | return true; | |
1483 | } | |
1484 | return false; | |
1485 | } | |
1486 | ||
1da177e4 LT |
1487 | /** |
1488 | * __break_lease - revoke all outstanding leases on file | |
1489 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1490 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1491 | * break all leases | |
1492 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1493 | * only delegations | |
1da177e4 | 1494 | * |
87250dd2 | 1495 | * break_lease (inlined for speed) has checked there already is at least |
1496 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1497 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1498 | * specified %O_NONBLOCK to your open(). |
1499 | */ | |
df4e8d2c | 1500 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1501 | { |
778fc546 | 1502 | int error = 0; |
128a3785 | 1503 | struct file_lock_context *ctx; |
a901125c | 1504 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1505 | unsigned long break_time; |
8737c930 | 1506 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1507 | LIST_HEAD(dispose); |
1da177e4 | 1508 | |
8737c930 | 1509 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1510 | if (IS_ERR(new_fl)) |
1511 | return PTR_ERR(new_fl); | |
df4e8d2c | 1512 | new_fl->fl_flags = type; |
1da177e4 | 1513 | |
8634b51f | 1514 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1515 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1516 | if (!ctx) { |
1517 | WARN_ON_ONCE(1); | |
1518 | return error; | |
1519 | } | |
1520 | ||
87709e28 | 1521 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1522 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1523 | |
c45198ed | 1524 | time_out_leases(inode, &dispose); |
1da177e4 | 1525 | |
03d12ddf | 1526 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1527 | goto out; |
1528 | ||
1da177e4 LT |
1529 | break_time = 0; |
1530 | if (lease_break_time > 0) { | |
1531 | break_time = jiffies + lease_break_time * HZ; | |
1532 | if (break_time == 0) | |
1533 | break_time++; /* so that 0 means no break time */ | |
1534 | } | |
1535 | ||
a901125c | 1536 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1537 | if (!leases_conflict(fl, new_fl)) |
1538 | continue; | |
778fc546 BF |
1539 | if (want_write) { |
1540 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1541 | continue; | |
1542 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1543 | fl->fl_break_time = break_time; |
778fc546 | 1544 | } else { |
8634b51f | 1545 | if (lease_breaking(fl)) |
778fc546 BF |
1546 | continue; |
1547 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1548 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1549 | } |
4d01b7f5 | 1550 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1551 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1552 | } |
1553 | ||
8634b51f | 1554 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1555 | goto out; |
1556 | ||
843c6b2f | 1557 | if (mode & O_NONBLOCK) { |
62af4f1f | 1558 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1559 | error = -EWOULDBLOCK; |
1560 | goto out; | |
1561 | } | |
1562 | ||
1563 | restart: | |
8634b51f JL |
1564 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1565 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1566 | if (break_time != 0) |
1da177e4 | 1567 | break_time -= jiffies; |
f1c6bb2c JL |
1568 | if (break_time == 0) |
1569 | break_time++; | |
8634b51f | 1570 | locks_insert_block(fl, new_fl); |
62af4f1f | 1571 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1572 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1573 | percpu_up_read_preempt_enable(&file_rwsem); |
aba37660 | 1574 | |
c45198ed | 1575 | locks_dispose_list(&dispose); |
4321e01e | 1576 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
ada5c1da | 1577 | !new_fl->fl_blocker, break_time); |
aba37660 | 1578 | |
87709e28 | 1579 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1580 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1581 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1582 | locks_delete_block(new_fl); |
1da177e4 | 1583 | if (error >= 0) { |
778fc546 BF |
1584 | /* |
1585 | * Wait for the next conflicting lease that has not been | |
1586 | * broken yet | |
1587 | */ | |
03d12ddf JL |
1588 | if (error == 0) |
1589 | time_out_leases(inode, &dispose); | |
1590 | if (any_leases_conflict(inode, new_fl)) | |
1591 | goto restart; | |
1da177e4 LT |
1592 | error = 0; |
1593 | } | |
1da177e4 | 1594 | out: |
6109c850 | 1595 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1596 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1597 | locks_dispose_list(&dispose); |
6d4b9e38 | 1598 | locks_free_lock(new_fl); |
1da177e4 LT |
1599 | return error; |
1600 | } | |
1601 | ||
1602 | EXPORT_SYMBOL(__break_lease); | |
1603 | ||
1604 | /** | |
76c47948 | 1605 | * lease_get_mtime - update modified time of an inode with exclusive lease |
1da177e4 | 1606 | * @inode: the inode |
76c47948 | 1607 | * @time: pointer to a timespec which contains the last modified time |
1da177e4 LT |
1608 | * |
1609 | * This is to force NFS clients to flush their caches for files with | |
1610 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1611 | * exclusive lease, then they could be modifying it. |
1da177e4 | 1612 | */ |
95582b00 | 1613 | void lease_get_mtime(struct inode *inode, struct timespec64 *time) |
1da177e4 | 1614 | { |
bfe86024 | 1615 | bool has_lease = false; |
128a3785 | 1616 | struct file_lock_context *ctx; |
8634b51f | 1617 | struct file_lock *fl; |
bfe86024 | 1618 | |
128a3785 | 1619 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1620 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1621 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1622 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1623 | struct file_lock, fl_list); | |
1624 | if (fl && (fl->fl_type == F_WRLCK)) | |
1625 | has_lease = true; | |
6109c850 | 1626 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1627 | } |
1628 | ||
1629 | if (has_lease) | |
c2050a45 | 1630 | *time = current_time(inode); |
1da177e4 LT |
1631 | } |
1632 | ||
1633 | EXPORT_SYMBOL(lease_get_mtime); | |
1634 | ||
1635 | /** | |
1636 | * fcntl_getlease - Enquire what lease is currently active | |
1637 | * @filp: the file | |
1638 | * | |
1639 | * The value returned by this function will be one of | |
1640 | * (if no lease break is pending): | |
1641 | * | |
1642 | * %F_RDLCK to indicate a shared lease is held. | |
1643 | * | |
1644 | * %F_WRLCK to indicate an exclusive lease is held. | |
1645 | * | |
1646 | * %F_UNLCK to indicate no lease is held. | |
1647 | * | |
1648 | * (if a lease break is pending): | |
1649 | * | |
1650 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1651 | * changed to a shared lease (or removed). | |
1652 | * | |
1653 | * %F_UNLCK to indicate the lease needs to be removed. | |
1654 | * | |
1655 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1656 | * should be returned to userspace. | |
1657 | */ | |
1658 | int fcntl_getlease(struct file *filp) | |
1659 | { | |
1660 | struct file_lock *fl; | |
c568d683 | 1661 | struct inode *inode = locks_inode(filp); |
128a3785 | 1662 | struct file_lock_context *ctx; |
1da177e4 | 1663 | int type = F_UNLCK; |
c45198ed | 1664 | LIST_HEAD(dispose); |
1da177e4 | 1665 | |
128a3785 | 1666 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1667 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
5f43086b | 1668 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1669 | spin_lock(&ctx->flc_lock); |
c568d683 | 1670 | time_out_leases(inode, &dispose); |
8634b51f JL |
1671 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
1672 | if (fl->fl_file != filp) | |
1673 | continue; | |
778fc546 | 1674 | type = target_leasetype(fl); |
1da177e4 LT |
1675 | break; |
1676 | } | |
6109c850 | 1677 | spin_unlock(&ctx->flc_lock); |
5f43086b PZ |
1678 | percpu_up_read_preempt_enable(&file_rwsem); |
1679 | ||
8634b51f | 1680 | locks_dispose_list(&dispose); |
1da177e4 | 1681 | } |
1da177e4 LT |
1682 | return type; |
1683 | } | |
1684 | ||
24cbe784 JL |
1685 | /** |
1686 | * check_conflicting_open - see if the given dentry points to a file that has | |
1687 | * an existing open that would conflict with the | |
1688 | * desired lease. | |
1689 | * @dentry: dentry to check | |
1690 | * @arg: type of lease that we're trying to acquire | |
7fadc59c | 1691 | * @flags: current lock flags |
24cbe784 JL |
1692 | * |
1693 | * Check to see if there's an existing open fd on this file that would | |
1694 | * conflict with the lease we're trying to set. | |
1695 | */ | |
1696 | static int | |
11afe9f7 | 1697 | check_conflicting_open(const struct dentry *dentry, const long arg, int flags) |
24cbe784 JL |
1698 | { |
1699 | int ret = 0; | |
1700 | struct inode *inode = dentry->d_inode; | |
1701 | ||
11afe9f7 CH |
1702 | if (flags & FL_LAYOUT) |
1703 | return 0; | |
1704 | ||
8cf9ee50 | 1705 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
24cbe784 JL |
1706 | return -EAGAIN; |
1707 | ||
1708 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1709 | (atomic_read(&inode->i_count) > 1))) | |
1710 | ret = -EAGAIN; | |
1711 | ||
1712 | return ret; | |
1713 | } | |
1714 | ||
e6f5c789 JL |
1715 | static int |
1716 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1717 | { |
8634b51f | 1718 | struct file_lock *fl, *my_fl = NULL, *lease; |
0f7fc9e4 | 1719 | struct dentry *dentry = filp->f_path.dentry; |
c568d683 | 1720 | struct inode *inode = dentry->d_inode; |
8634b51f | 1721 | struct file_lock_context *ctx; |
df4e8d2c | 1722 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1723 | int error; |
c45198ed | 1724 | LIST_HEAD(dispose); |
1da177e4 | 1725 | |
096657b6 | 1726 | lease = *flp; |
62af4f1f JL |
1727 | trace_generic_add_lease(inode, lease); |
1728 | ||
5c1c669a JL |
1729 | /* Note that arg is never F_UNLCK here */ |
1730 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1731 | if (!ctx) |
1732 | return -ENOMEM; | |
1733 | ||
df4e8d2c BF |
1734 | /* |
1735 | * In the delegation case we need mutual exclusion with | |
1736 | * a number of operations that take the i_mutex. We trylock | |
1737 | * because delegations are an optional optimization, and if | |
1738 | * there's some chance of a conflict--we'd rather not | |
1739 | * bother, maybe that's a sign this just isn't a good file to | |
1740 | * hand out a delegation on. | |
1741 | */ | |
5955102c | 1742 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1743 | return -EAGAIN; |
1744 | ||
1745 | if (is_deleg && arg == F_WRLCK) { | |
1746 | /* Write delegations are not currently supported: */ | |
5955102c | 1747 | inode_unlock(inode); |
df4e8d2c BF |
1748 | WARN_ON_ONCE(1); |
1749 | return -EINVAL; | |
1750 | } | |
096657b6 | 1751 | |
87709e28 | 1752 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1753 | spin_lock(&ctx->flc_lock); |
c45198ed | 1754 | time_out_leases(inode, &dispose); |
11afe9f7 | 1755 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
24cbe784 | 1756 | if (error) |
096657b6 | 1757 | goto out; |
6d5e8b05 | 1758 | |
1da177e4 LT |
1759 | /* |
1760 | * At this point, we know that if there is an exclusive | |
1761 | * lease on this file, then we hold it on this filp | |
1762 | * (otherwise our open of this file would have blocked). | |
1763 | * And if we are trying to acquire an exclusive lease, | |
1764 | * then the file is not open by anyone (including us) | |
1765 | * except for this filp. | |
1766 | */ | |
c1f24ef4 | 1767 | error = -EAGAIN; |
8634b51f | 1768 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1769 | if (fl->fl_file == filp && |
1770 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1771 | my_fl = fl; |
c1f24ef4 BF |
1772 | continue; |
1773 | } | |
8634b51f | 1774 | |
c1f24ef4 BF |
1775 | /* |
1776 | * No exclusive leases if someone else has a lease on | |
1777 | * this file: | |
1778 | */ | |
1779 | if (arg == F_WRLCK) | |
1780 | goto out; | |
1781 | /* | |
1782 | * Modifying our existing lease is OK, but no getting a | |
1783 | * new lease if someone else is opening for write: | |
1784 | */ | |
1785 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1786 | goto out; | |
1da177e4 LT |
1787 | } |
1788 | ||
8634b51f | 1789 | if (my_fl != NULL) { |
0164bf02 JL |
1790 | lease = my_fl; |
1791 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1792 | if (error) |
1793 | goto out; | |
1794 | goto out_setup; | |
1da177e4 LT |
1795 | } |
1796 | ||
1da177e4 LT |
1797 | error = -EINVAL; |
1798 | if (!leases_enable) | |
1799 | goto out; | |
1800 | ||
e084c1bd | 1801 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1802 | /* |
1803 | * The check in break_lease() is lockless. It's possible for another | |
1804 | * open to race in after we did the earlier check for a conflicting | |
1805 | * open but before the lease was inserted. Check again for a | |
1806 | * conflicting open and cancel the lease if there is one. | |
1807 | * | |
1808 | * We also add a barrier here to ensure that the insertion of the lock | |
1809 | * precedes these checks. | |
1810 | */ | |
1811 | smp_mb(); | |
11afe9f7 | 1812 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
8634b51f | 1813 | if (error) { |
e084c1bd | 1814 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1815 | goto out; |
1816 | } | |
1c7dd2ff JL |
1817 | |
1818 | out_setup: | |
1819 | if (lease->fl_lmops->lm_setup) | |
1820 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1821 | out: |
6109c850 | 1822 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1823 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1824 | locks_dispose_list(&dispose); |
df4e8d2c | 1825 | if (is_deleg) |
5955102c | 1826 | inode_unlock(inode); |
8634b51f | 1827 | if (!error && !my_fl) |
1c7dd2ff | 1828 | *flp = NULL; |
1da177e4 LT |
1829 | return error; |
1830 | } | |
8335ebd9 | 1831 | |
2ab99ee1 | 1832 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1833 | { |
0efaa7e8 | 1834 | int error = -EAGAIN; |
8634b51f | 1835 | struct file_lock *fl, *victim = NULL; |
c568d683 | 1836 | struct inode *inode = locks_inode(filp); |
128a3785 | 1837 | struct file_lock_context *ctx; |
c45198ed | 1838 | LIST_HEAD(dispose); |
8335ebd9 | 1839 | |
128a3785 | 1840 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1841 | if (!ctx) { |
1842 | trace_generic_delete_lease(inode, NULL); | |
1843 | return error; | |
1844 | } | |
1845 | ||
87709e28 | 1846 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1847 | spin_lock(&ctx->flc_lock); |
8634b51f | 1848 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1849 | if (fl->fl_file == filp && |
1850 | fl->fl_owner == owner) { | |
8634b51f | 1851 | victim = fl; |
0efaa7e8 | 1852 | break; |
8634b51f | 1853 | } |
8335ebd9 | 1854 | } |
a9b1b455 | 1855 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1856 | if (victim) |
7448cc37 | 1857 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1858 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1859 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1860 | locks_dispose_list(&dispose); |
0efaa7e8 | 1861 | return error; |
8335ebd9 BF |
1862 | } |
1863 | ||
1864 | /** | |
1865 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1866 | * @filp: file pointer |
1867 | * @arg: type of lease to obtain | |
1868 | * @flp: input - file_lock to use, output - file_lock inserted | |
1869 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1870 | * doesn't require it) | |
8335ebd9 BF |
1871 | * |
1872 | * The (input) flp->fl_lmops->lm_break function is required | |
1873 | * by break_lease(). | |
8335ebd9 | 1874 | */ |
e6f5c789 JL |
1875 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1876 | void **priv) | |
8335ebd9 | 1877 | { |
c568d683 | 1878 | struct inode *inode = locks_inode(filp); |
8335ebd9 BF |
1879 | int error; |
1880 | ||
8e96e3b7 | 1881 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1882 | return -EACCES; |
1883 | if (!S_ISREG(inode->i_mode)) | |
1884 | return -EINVAL; | |
1885 | error = security_file_lock(filp, arg); | |
1886 | if (error) | |
1887 | return error; | |
1888 | ||
8335ebd9 BF |
1889 | switch (arg) { |
1890 | case F_UNLCK: | |
2ab99ee1 | 1891 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1892 | case F_RDLCK: |
1893 | case F_WRLCK: | |
0efaa7e8 JL |
1894 | if (!(*flp)->fl_lmops->lm_break) { |
1895 | WARN_ON_ONCE(1); | |
1896 | return -ENOLCK; | |
1897 | } | |
11afe9f7 | 1898 | |
e6f5c789 | 1899 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1900 | default: |
8d657eb3 | 1901 | return -EINVAL; |
8335ebd9 BF |
1902 | } |
1903 | } | |
0af1a450 | 1904 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1905 | |
b89f4321 | 1906 | /** |
e51673aa | 1907 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1908 | * @filp: file pointer |
1909 | * @arg: type of lease to obtain | |
1910 | * @lease: file_lock to use when adding a lease | |
1911 | * @priv: private info for lm_setup when adding a lease (may be | |
1912 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1913 | * |
1914 | * Call this to establish a lease on the file. The "lease" argument is not | |
1915 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
80b79dd0 MCC |
1916 | * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be |
1917 | * set; if not, this function will return -ENOLCK (and generate a scary-looking | |
e51673aa | 1918 | * stack trace). |
1c7dd2ff JL |
1919 | * |
1920 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1921 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1922 | */ |
e6f5c789 JL |
1923 | int |
1924 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1925 | { |
de2a4a50 | 1926 | if (filp->f_op->setlease) |
f82b4b67 | 1927 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1928 | else |
f82b4b67 | 1929 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1930 | } |
a9933cea | 1931 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1932 | |
0ceaf6c7 | 1933 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1934 | { |
1c7dd2ff | 1935 | struct file_lock *fl; |
f7347ce4 | 1936 | struct fasync_struct *new; |
1da177e4 LT |
1937 | int error; |
1938 | ||
c5b1f0d9 AB |
1939 | fl = lease_alloc(filp, arg); |
1940 | if (IS_ERR(fl)) | |
1941 | return PTR_ERR(fl); | |
1da177e4 | 1942 | |
f7347ce4 LT |
1943 | new = fasync_alloc(); |
1944 | if (!new) { | |
1945 | locks_free_lock(fl); | |
1946 | return -ENOMEM; | |
1947 | } | |
1c7dd2ff | 1948 | new->fa_fd = fd; |
f7347ce4 | 1949 | |
1c7dd2ff | 1950 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1951 | if (fl) |
1952 | locks_free_lock(fl); | |
f7347ce4 LT |
1953 | if (new) |
1954 | fasync_free(new); | |
1da177e4 LT |
1955 | return error; |
1956 | } | |
1957 | ||
0ceaf6c7 BF |
1958 | /** |
1959 | * fcntl_setlease - sets a lease on an open file | |
1960 | * @fd: open file descriptor | |
1961 | * @filp: file pointer | |
1962 | * @arg: type of lease to obtain | |
1963 | * | |
1964 | * Call this fcntl to establish a lease on the file. | |
1965 | * Note that you also need to call %F_SETSIG to | |
1966 | * receive a signal when the lease is broken. | |
1967 | */ | |
1968 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1969 | { | |
1970 | if (arg == F_UNLCK) | |
2ab99ee1 | 1971 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
1972 | return do_fcntl_add_lease(fd, filp, arg); |
1973 | } | |
1974 | ||
1da177e4 | 1975 | /** |
29d01b22 JL |
1976 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
1977 | * @inode: inode of the file to apply to | |
1da177e4 LT |
1978 | * @fl: The lock to be applied |
1979 | * | |
29d01b22 | 1980 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 1981 | */ |
616fb38f | 1982 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1983 | { |
1984 | int error; | |
1985 | might_sleep(); | |
1986 | for (;;) { | |
29d01b22 | 1987 | error = flock_lock_inode(inode, fl); |
bde74e4b | 1988 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1989 | break; |
ada5c1da | 1990 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
1991 | if (error) |
1992 | break; | |
1da177e4 | 1993 | } |
16306a61 | 1994 | locks_delete_block(fl); |
1da177e4 LT |
1995 | return error; |
1996 | } | |
1997 | ||
e55c34a6 BC |
1998 | /** |
1999 | * locks_lock_inode_wait - Apply a lock to an inode | |
2000 | * @inode: inode of the file to apply to | |
2001 | * @fl: The lock to be applied | |
2002 | * | |
2003 | * Apply a POSIX or FLOCK style lock request to an inode. | |
2004 | */ | |
2005 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
2006 | { | |
2007 | int res = 0; | |
2008 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
2009 | case FL_POSIX: | |
2010 | res = posix_lock_inode_wait(inode, fl); | |
2011 | break; | |
2012 | case FL_FLOCK: | |
2013 | res = flock_lock_inode_wait(inode, fl); | |
2014 | break; | |
2015 | default: | |
2016 | BUG(); | |
2017 | } | |
2018 | return res; | |
2019 | } | |
2020 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
2021 | ||
1da177e4 LT |
2022 | /** |
2023 | * sys_flock: - flock() system call. | |
2024 | * @fd: the file descriptor to lock. | |
2025 | * @cmd: the type of lock to apply. | |
2026 | * | |
2027 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
80b79dd0 | 2028 | * The @cmd can be one of: |
1da177e4 | 2029 | * |
80b79dd0 MCC |
2030 | * - %LOCK_SH -- a shared lock. |
2031 | * - %LOCK_EX -- an exclusive lock. | |
2032 | * - %LOCK_UN -- remove an existing lock. | |
2033 | * - %LOCK_MAND -- a 'mandatory' flock. | |
2034 | * This exists to emulate Windows Share Modes. | |
1da177e4 LT |
2035 | * |
2036 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
2037 | * processes read and write access respectively. | |
2038 | */ | |
002c8976 | 2039 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 2040 | { |
2903ff01 | 2041 | struct fd f = fdget(fd); |
1da177e4 LT |
2042 | struct file_lock *lock; |
2043 | int can_sleep, unlock; | |
2044 | int error; | |
2045 | ||
2046 | error = -EBADF; | |
2903ff01 | 2047 | if (!f.file) |
1da177e4 LT |
2048 | goto out; |
2049 | ||
2050 | can_sleep = !(cmd & LOCK_NB); | |
2051 | cmd &= ~LOCK_NB; | |
2052 | unlock = (cmd == LOCK_UN); | |
2053 | ||
aeb5d727 | 2054 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 2055 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
2056 | goto out_putf; |
2057 | ||
d6367d62 | 2058 | lock = flock_make_lock(f.file, cmd, NULL); |
6e129d00 JL |
2059 | if (IS_ERR(lock)) { |
2060 | error = PTR_ERR(lock); | |
1da177e4 | 2061 | goto out_putf; |
6e129d00 JL |
2062 | } |
2063 | ||
1da177e4 LT |
2064 | if (can_sleep) |
2065 | lock->fl_flags |= FL_SLEEP; | |
2066 | ||
2903ff01 | 2067 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
2068 | if (error) |
2069 | goto out_free; | |
2070 | ||
de2a4a50 | 2071 | if (f.file->f_op->flock) |
2903ff01 | 2072 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
2073 | (can_sleep) ? F_SETLKW : F_SETLK, |
2074 | lock); | |
2075 | else | |
4f656367 | 2076 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
2077 | |
2078 | out_free: | |
993dfa87 | 2079 | locks_free_lock(lock); |
1da177e4 LT |
2080 | |
2081 | out_putf: | |
2903ff01 | 2082 | fdput(f); |
1da177e4 LT |
2083 | out: |
2084 | return error; | |
2085 | } | |
2086 | ||
3ee17abd BF |
2087 | /** |
2088 | * vfs_test_lock - test file byte range lock | |
2089 | * @filp: The file to test lock for | |
6924c554 | 2090 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2091 | * |
2092 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2093 | * setting conf->fl_type to something other than F_UNLCK. | |
2094 | */ | |
2095 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2096 | { | |
de2a4a50 | 2097 | if (filp->f_op->lock) |
3ee17abd BF |
2098 | return filp->f_op->lock(filp, F_GETLK, fl); |
2099 | posix_test_lock(filp, fl); | |
2100 | return 0; | |
2101 | } | |
2102 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2103 | ||
9d5b86ac BC |
2104 | /** |
2105 | * locks_translate_pid - translate a file_lock's fl_pid number into a namespace | |
2106 | * @fl: The file_lock who's fl_pid should be translated | |
2107 | * @ns: The namespace into which the pid should be translated | |
2108 | * | |
2109 | * Used to tranlate a fl_pid into a namespace virtual pid number | |
2110 | */ | |
2111 | static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns) | |
2112 | { | |
2113 | pid_t vnr; | |
2114 | struct pid *pid; | |
2115 | ||
2116 | if (IS_OFDLCK(fl)) | |
2117 | return -1; | |
2118 | if (IS_REMOTELCK(fl)) | |
2119 | return fl->fl_pid; | |
826d7bc9 KK |
2120 | /* |
2121 | * If the flock owner process is dead and its pid has been already | |
2122 | * freed, the translation below won't work, but we still want to show | |
2123 | * flock owner pid number in init pidns. | |
2124 | */ | |
2125 | if (ns == &init_pid_ns) | |
2126 | return (pid_t)fl->fl_pid; | |
9d5b86ac BC |
2127 | |
2128 | rcu_read_lock(); | |
2129 | pid = find_pid_ns(fl->fl_pid, &init_pid_ns); | |
2130 | vnr = pid_nr_ns(pid, ns); | |
2131 | rcu_read_unlock(); | |
2132 | return vnr; | |
2133 | } | |
2134 | ||
c2fa1b8a BF |
2135 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2136 | { | |
9d5b86ac | 2137 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2138 | #if BITS_PER_LONG == 32 |
2139 | /* | |
2140 | * Make sure we can represent the posix lock via | |
2141 | * legacy 32bit flock. | |
2142 | */ | |
2143 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2144 | return -EOVERFLOW; | |
2145 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2146 | return -EOVERFLOW; | |
2147 | #endif | |
2148 | flock->l_start = fl->fl_start; | |
2149 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2150 | fl->fl_end - fl->fl_start + 1; | |
2151 | flock->l_whence = 0; | |
129a84de | 2152 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2153 | return 0; |
2154 | } | |
2155 | ||
2156 | #if BITS_PER_LONG == 32 | |
2157 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2158 | { | |
9d5b86ac | 2159 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2160 | flock->l_start = fl->fl_start; |
2161 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2162 | fl->fl_end - fl->fl_start + 1; | |
2163 | flock->l_whence = 0; | |
2164 | flock->l_type = fl->fl_type; | |
2165 | } | |
2166 | #endif | |
2167 | ||
1da177e4 LT |
2168 | /* Report the first existing lock that would conflict with l. |
2169 | * This implements the F_GETLK command of fcntl(). | |
2170 | */ | |
a75d30c7 | 2171 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock) |
1da177e4 | 2172 | { |
52306e88 | 2173 | struct file_lock *fl; |
1da177e4 LT |
2174 | int error; |
2175 | ||
52306e88 BC |
2176 | fl = locks_alloc_lock(); |
2177 | if (fl == NULL) | |
2178 | return -ENOMEM; | |
1da177e4 | 2179 | error = -EINVAL; |
a75d30c7 | 2180 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2181 | goto out; |
2182 | ||
52306e88 | 2183 | error = flock_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2184 | if (error) |
2185 | goto out; | |
2186 | ||
0d3f7a2d | 2187 | if (cmd == F_OFD_GETLK) { |
90478939 | 2188 | error = -EINVAL; |
a75d30c7 | 2189 | if (flock->l_pid != 0) |
90478939 JL |
2190 | goto out; |
2191 | ||
5d50ffd7 | 2192 | cmd = F_GETLK; |
52306e88 BC |
2193 | fl->fl_flags |= FL_OFDLCK; |
2194 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2195 | } |
2196 | ||
52306e88 | 2197 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2198 | if (error) |
2199 | goto out; | |
1da177e4 | 2200 | |
52306e88 BC |
2201 | flock->l_type = fl->fl_type; |
2202 | if (fl->fl_type != F_UNLCK) { | |
2203 | error = posix_lock_to_flock(flock, fl); | |
c2fa1b8a | 2204 | if (error) |
52306e88 | 2205 | goto out; |
1da177e4 | 2206 | } |
1da177e4 | 2207 | out: |
52306e88 | 2208 | locks_free_lock(fl); |
1da177e4 LT |
2209 | return error; |
2210 | } | |
2211 | ||
7723ec97 ME |
2212 | /** |
2213 | * vfs_lock_file - file byte range lock | |
2214 | * @filp: The file to apply the lock to | |
2215 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2216 | * @fl: The lock to be applied | |
150b3934 ME |
2217 | * @conf: Place to return a copy of the conflicting lock, if found. |
2218 | * | |
2219 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2220 | * as the final argument. | |
2221 | * | |
2222 | * If the filesystem defines a private ->lock() method, then @conf will | |
2223 | * be left unchanged; so a caller that cares should initialize it to | |
2224 | * some acceptable default. | |
2beb6614 ME |
2225 | * |
2226 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2227 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2228 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2229 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2230 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2231 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2232 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2233 | * request completes. |
2234 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2235 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2236 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2237 | * nonzero return code and the file system should release the lock. The file |
2238 | * system is also responsible to keep a corresponding posix lock when it | |
2239 | * grants a lock so the VFS can find out which locks are locally held and do | |
2240 | * the correct lock cleanup when required. | |
2241 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2242 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2243 | * return code. |
7723ec97 | 2244 | */ |
150b3934 | 2245 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2246 | { |
de2a4a50 | 2247 | if (filp->f_op->lock) |
7723ec97 ME |
2248 | return filp->f_op->lock(filp, cmd, fl); |
2249 | else | |
150b3934 | 2250 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2251 | } |
2252 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2253 | ||
b648a6de MS |
2254 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2255 | struct file_lock *fl) | |
2256 | { | |
2257 | int error; | |
2258 | ||
2259 | error = security_file_lock(filp, fl->fl_type); | |
2260 | if (error) | |
2261 | return error; | |
2262 | ||
764c76b3 MS |
2263 | for (;;) { |
2264 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2265 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2266 | break; |
ada5c1da | 2267 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
2268 | if (error) |
2269 | break; | |
b648a6de | 2270 | } |
16306a61 | 2271 | locks_delete_block(fl); |
b648a6de MS |
2272 | |
2273 | return error; | |
2274 | } | |
2275 | ||
6ca7d910 | 2276 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2277 | static int |
2278 | check_fmode_for_setlk(struct file_lock *fl) | |
2279 | { | |
2280 | switch (fl->fl_type) { | |
2281 | case F_RDLCK: | |
2282 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2283 | return -EBADF; | |
2284 | break; | |
2285 | case F_WRLCK: | |
2286 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2287 | return -EBADF; | |
2288 | } | |
2289 | return 0; | |
2290 | } | |
2291 | ||
1da177e4 LT |
2292 | /* Apply the lock described by l to an open file descriptor. |
2293 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2294 | */ | |
c293621b | 2295 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2296 | struct flock *flock) |
1da177e4 LT |
2297 | { |
2298 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2299 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2300 | struct file *f; |
1da177e4 LT |
2301 | int error; |
2302 | ||
2303 | if (file_lock == NULL) | |
2304 | return -ENOLCK; | |
2305 | ||
1da177e4 LT |
2306 | /* Don't allow mandatory locks on files that may be memory mapped |
2307 | * and shared. | |
2308 | */ | |
a16877ca | 2309 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2310 | error = -EAGAIN; |
2311 | goto out; | |
2312 | } | |
2313 | ||
a75d30c7 | 2314 | error = flock_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2315 | if (error) |
2316 | goto out; | |
5d50ffd7 | 2317 | |
cf01f4ee JL |
2318 | error = check_fmode_for_setlk(file_lock); |
2319 | if (error) | |
2320 | goto out; | |
2321 | ||
5d50ffd7 JL |
2322 | /* |
2323 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2324 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2325 | */ |
2326 | switch (cmd) { | |
0d3f7a2d | 2327 | case F_OFD_SETLK: |
90478939 | 2328 | error = -EINVAL; |
a75d30c7 | 2329 | if (flock->l_pid != 0) |
90478939 JL |
2330 | goto out; |
2331 | ||
5d50ffd7 | 2332 | cmd = F_SETLK; |
cff2fce5 | 2333 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2334 | file_lock->fl_owner = filp; |
5d50ffd7 | 2335 | break; |
0d3f7a2d | 2336 | case F_OFD_SETLKW: |
90478939 | 2337 | error = -EINVAL; |
a75d30c7 | 2338 | if (flock->l_pid != 0) |
90478939 JL |
2339 | goto out; |
2340 | ||
5d50ffd7 | 2341 | cmd = F_SETLKW; |
cff2fce5 | 2342 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2343 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2344 | /* Fallthrough */ |
2345 | case F_SETLKW: | |
1da177e4 LT |
2346 | file_lock->fl_flags |= FL_SLEEP; |
2347 | } | |
5d50ffd7 | 2348 | |
b648a6de | 2349 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2350 | |
c293621b | 2351 | /* |
0752ba80 JL |
2352 | * Attempt to detect a close/fcntl race and recover by releasing the |
2353 | * lock that was just acquired. There is no need to do that when we're | |
2354 | * unlocking though, or for OFD locks. | |
c293621b | 2355 | */ |
0752ba80 JL |
2356 | if (!error && file_lock->fl_type != F_UNLCK && |
2357 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2358 | /* |
2359 | * We need that spin_lock here - it prevents reordering between | |
2360 | * update of i_flctx->flc_posix and check for it done in | |
2361 | * close(). rcu_read_lock() wouldn't do. | |
2362 | */ | |
2363 | spin_lock(¤t->files->file_lock); | |
2364 | f = fcheck(fd); | |
2365 | spin_unlock(¤t->files->file_lock); | |
2366 | if (f != filp) { | |
2367 | file_lock->fl_type = F_UNLCK; | |
2368 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2369 | WARN_ON_ONCE(error); | |
2370 | error = -EBADF; | |
2371 | } | |
1da177e4 | 2372 | } |
c293621b | 2373 | out: |
1890910f | 2374 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2375 | locks_free_lock(file_lock); |
2376 | return error; | |
2377 | } | |
2378 | ||
2379 | #if BITS_PER_LONG == 32 | |
2380 | /* Report the first existing lock that would conflict with l. | |
2381 | * This implements the F_GETLK command of fcntl(). | |
2382 | */ | |
a75d30c7 | 2383 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock) |
1da177e4 | 2384 | { |
52306e88 | 2385 | struct file_lock *fl; |
1da177e4 LT |
2386 | int error; |
2387 | ||
52306e88 BC |
2388 | fl = locks_alloc_lock(); |
2389 | if (fl == NULL) | |
2390 | return -ENOMEM; | |
2391 | ||
1da177e4 | 2392 | error = -EINVAL; |
a75d30c7 | 2393 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2394 | goto out; |
2395 | ||
52306e88 | 2396 | error = flock64_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2397 | if (error) |
2398 | goto out; | |
2399 | ||
0d3f7a2d | 2400 | if (cmd == F_OFD_GETLK) { |
90478939 | 2401 | error = -EINVAL; |
a75d30c7 | 2402 | if (flock->l_pid != 0) |
90478939 JL |
2403 | goto out; |
2404 | ||
5d50ffd7 | 2405 | cmd = F_GETLK64; |
52306e88 BC |
2406 | fl->fl_flags |= FL_OFDLCK; |
2407 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2408 | } |
2409 | ||
52306e88 | 2410 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2411 | if (error) |
2412 | goto out; | |
2413 | ||
52306e88 BC |
2414 | flock->l_type = fl->fl_type; |
2415 | if (fl->fl_type != F_UNLCK) | |
2416 | posix_lock_to_flock64(flock, fl); | |
f328296e | 2417 | |
1da177e4 | 2418 | out: |
52306e88 | 2419 | locks_free_lock(fl); |
1da177e4 LT |
2420 | return error; |
2421 | } | |
2422 | ||
2423 | /* Apply the lock described by l to an open file descriptor. | |
2424 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2425 | */ | |
c293621b | 2426 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2427 | struct flock64 *flock) |
1da177e4 LT |
2428 | { |
2429 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2430 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2431 | struct file *f; |
1da177e4 LT |
2432 | int error; |
2433 | ||
2434 | if (file_lock == NULL) | |
2435 | return -ENOLCK; | |
2436 | ||
1da177e4 LT |
2437 | /* Don't allow mandatory locks on files that may be memory mapped |
2438 | * and shared. | |
2439 | */ | |
a16877ca | 2440 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2441 | error = -EAGAIN; |
2442 | goto out; | |
2443 | } | |
2444 | ||
a75d30c7 | 2445 | error = flock64_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2446 | if (error) |
2447 | goto out; | |
5d50ffd7 | 2448 | |
cf01f4ee JL |
2449 | error = check_fmode_for_setlk(file_lock); |
2450 | if (error) | |
2451 | goto out; | |
2452 | ||
5d50ffd7 JL |
2453 | /* |
2454 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2455 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2456 | */ |
2457 | switch (cmd) { | |
0d3f7a2d | 2458 | case F_OFD_SETLK: |
90478939 | 2459 | error = -EINVAL; |
a75d30c7 | 2460 | if (flock->l_pid != 0) |
90478939 JL |
2461 | goto out; |
2462 | ||
5d50ffd7 | 2463 | cmd = F_SETLK64; |
cff2fce5 | 2464 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2465 | file_lock->fl_owner = filp; |
5d50ffd7 | 2466 | break; |
0d3f7a2d | 2467 | case F_OFD_SETLKW: |
90478939 | 2468 | error = -EINVAL; |
a75d30c7 | 2469 | if (flock->l_pid != 0) |
90478939 JL |
2470 | goto out; |
2471 | ||
5d50ffd7 | 2472 | cmd = F_SETLKW64; |
cff2fce5 | 2473 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2474 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2475 | /* Fallthrough */ |
2476 | case F_SETLKW64: | |
1da177e4 LT |
2477 | file_lock->fl_flags |= FL_SLEEP; |
2478 | } | |
5d50ffd7 | 2479 | |
b648a6de | 2480 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2481 | |
c293621b | 2482 | /* |
0752ba80 JL |
2483 | * Attempt to detect a close/fcntl race and recover by releasing the |
2484 | * lock that was just acquired. There is no need to do that when we're | |
2485 | * unlocking though, or for OFD locks. | |
c293621b | 2486 | */ |
0752ba80 JL |
2487 | if (!error && file_lock->fl_type != F_UNLCK && |
2488 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2489 | /* |
2490 | * We need that spin_lock here - it prevents reordering between | |
2491 | * update of i_flctx->flc_posix and check for it done in | |
2492 | * close(). rcu_read_lock() wouldn't do. | |
2493 | */ | |
2494 | spin_lock(¤t->files->file_lock); | |
2495 | f = fcheck(fd); | |
2496 | spin_unlock(¤t->files->file_lock); | |
2497 | if (f != filp) { | |
2498 | file_lock->fl_type = F_UNLCK; | |
2499 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2500 | WARN_ON_ONCE(error); | |
2501 | error = -EBADF; | |
2502 | } | |
1da177e4 | 2503 | } |
1da177e4 LT |
2504 | out: |
2505 | locks_free_lock(file_lock); | |
2506 | return error; | |
2507 | } | |
2508 | #endif /* BITS_PER_LONG == 32 */ | |
2509 | ||
2510 | /* | |
2511 | * This function is called when the file is being removed | |
2512 | * from the task's fd array. POSIX locks belonging to this task | |
2513 | * are deleted at this time. | |
2514 | */ | |
2515 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2516 | { | |
1890910f | 2517 | int error; |
c568d683 | 2518 | struct inode *inode = locks_inode(filp); |
ff7b86b8 | 2519 | struct file_lock lock; |
128a3785 | 2520 | struct file_lock_context *ctx; |
1da177e4 LT |
2521 | |
2522 | /* | |
2523 | * If there are no locks held on this file, we don't need to call | |
2524 | * posix_lock_file(). Another process could be setting a lock on this | |
2525 | * file at the same time, but we wouldn't remove that lock anyway. | |
2526 | */ | |
c568d683 | 2527 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 | 2528 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2529 | return; |
2530 | ||
d6367d62 | 2531 | locks_init_lock(&lock); |
1da177e4 | 2532 | lock.fl_type = F_UNLCK; |
75e1fcc0 | 2533 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2534 | lock.fl_start = 0; |
2535 | lock.fl_end = OFFSET_MAX; | |
2536 | lock.fl_owner = owner; | |
2537 | lock.fl_pid = current->tgid; | |
2538 | lock.fl_file = filp; | |
2539 | lock.fl_ops = NULL; | |
2540 | lock.fl_lmops = NULL; | |
2541 | ||
1890910f | 2542 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2543 | |
1da177e4 LT |
2544 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2545 | lock.fl_ops->fl_release_private(&lock); | |
c568d683 | 2546 | trace_locks_remove_posix(inode, &lock, error); |
1da177e4 LT |
2547 | } |
2548 | ||
2549 | EXPORT_SYMBOL(locks_remove_posix); | |
2550 | ||
3d8e560d | 2551 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2552 | static void |
128a3785 | 2553 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 | 2554 | { |
d6367d62 | 2555 | struct file_lock fl; |
c568d683 | 2556 | struct inode *inode = locks_inode(filp); |
dd459bb1 | 2557 | |
3d8e560d | 2558 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2559 | return; |
2560 | ||
d6367d62 N |
2561 | flock_make_lock(filp, LOCK_UN, &fl); |
2562 | fl.fl_flags |= FL_CLOSE; | |
2563 | ||
de2a4a50 | 2564 | if (filp->f_op->flock) |
dd459bb1 JL |
2565 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2566 | else | |
bcd7f78d | 2567 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2568 | |
2569 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2570 | fl.fl_ops->fl_release_private(&fl); | |
2571 | } | |
2572 | ||
3d8e560d | 2573 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2574 | static void |
128a3785 | 2575 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2576 | { |
8634b51f JL |
2577 | struct file_lock *fl, *tmp; |
2578 | LIST_HEAD(dispose); | |
2579 | ||
3d8e560d | 2580 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2581 | return; |
2582 | ||
5f43086b | 2583 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 2584 | spin_lock(&ctx->flc_lock); |
8634b51f | 2585 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2586 | if (filp == fl->fl_file) |
2587 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2588 | spin_unlock(&ctx->flc_lock); |
5f43086b PZ |
2589 | percpu_up_read_preempt_enable(&file_rwsem); |
2590 | ||
8634b51f JL |
2591 | locks_dispose_list(&dispose); |
2592 | } | |
2593 | ||
1da177e4 LT |
2594 | /* |
2595 | * This function is called on the last close of an open file. | |
2596 | */ | |
78ed8a13 | 2597 | void locks_remove_file(struct file *filp) |
1da177e4 | 2598 | { |
128a3785 DV |
2599 | struct file_lock_context *ctx; |
2600 | ||
c568d683 | 2601 | ctx = smp_load_acquire(&locks_inode(filp)->i_flctx); |
128a3785 | 2602 | if (!ctx) |
3d8e560d JL |
2603 | return; |
2604 | ||
dd459bb1 | 2605 | /* remove any OFD locks */ |
73a8f5f7 | 2606 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2607 | |
dd459bb1 | 2608 | /* remove flock locks */ |
128a3785 | 2609 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2610 | |
8634b51f | 2611 | /* remove any leases */ |
128a3785 | 2612 | locks_remove_lease(filp, ctx); |
3953704f BC |
2613 | |
2614 | spin_lock(&ctx->flc_lock); | |
2615 | locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX"); | |
2616 | locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK"); | |
2617 | locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE"); | |
2618 | spin_unlock(&ctx->flc_lock); | |
1da177e4 LT |
2619 | } |
2620 | ||
1da177e4 LT |
2621 | /** |
2622 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2623 | * @waiter: the lock which was waiting |
2624 | * | |
2625 | * lockd needs to block waiting for locks. | |
2626 | */ | |
64a318ee | 2627 | int |
f891a29f | 2628 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2629 | { |
5946c431 | 2630 | int status = -ENOENT; |
64a318ee | 2631 | |
7b2296af | 2632 | spin_lock(&blocked_lock_lock); |
5946c431 N |
2633 | if (waiter->fl_blocker) { |
2634 | __locks_wake_up_blocks(waiter); | |
1da177e4 | 2635 | __locks_delete_block(waiter); |
5946c431 N |
2636 | status = 0; |
2637 | } | |
7b2296af | 2638 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2639 | return status; |
1da177e4 | 2640 | } |
1da177e4 LT |
2641 | EXPORT_SYMBOL(posix_unblock_lock); |
2642 | ||
9b9d2ab4 ME |
2643 | /** |
2644 | * vfs_cancel_lock - file byte range unblock lock | |
2645 | * @filp: The file to apply the unblock to | |
2646 | * @fl: The lock to be unblocked | |
2647 | * | |
2648 | * Used by lock managers to cancel blocked requests | |
2649 | */ | |
2650 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2651 | { | |
de2a4a50 | 2652 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2653 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2654 | return 0; | |
2655 | } | |
2656 | ||
2657 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2658 | ||
7f8ada98 | 2659 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2660 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2661 | #include <linux/seq_file.h> |
2662 | ||
7012b02a JL |
2663 | struct locks_iterator { |
2664 | int li_cpu; | |
2665 | loff_t li_pos; | |
2666 | }; | |
2667 | ||
7f8ada98 | 2668 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2669 | loff_t id, char *pfx) |
1da177e4 LT |
2670 | { |
2671 | struct inode *inode = NULL; | |
ab1f1611 | 2672 | unsigned int fl_pid; |
9d5b86ac | 2673 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
ab1f1611 | 2674 | |
9d5b86ac BC |
2675 | fl_pid = locks_translate_pid(fl, proc_pidns); |
2676 | /* | |
1cf8e5de KK |
2677 | * If lock owner is dead (and pid is freed) or not visible in current |
2678 | * pidns, zero is shown as a pid value. Check lock info from | |
2679 | * init_pid_ns to get saved lock pid value. | |
9d5b86ac | 2680 | */ |
1da177e4 LT |
2681 | |
2682 | if (fl->fl_file != NULL) | |
c568d683 | 2683 | inode = locks_inode(fl->fl_file); |
1da177e4 | 2684 | |
99dc8292 | 2685 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2686 | if (IS_POSIX(fl)) { |
c918d42a | 2687 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2688 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2689 | else if (IS_OFDLCK(fl)) |
5315c26a | 2690 | seq_puts(f, "OFDLCK"); |
c918d42a | 2691 | else |
5315c26a | 2692 | seq_puts(f, "POSIX "); |
c918d42a JL |
2693 | |
2694 | seq_printf(f, " %s ", | |
1da177e4 | 2695 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2696 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2697 | } else if (IS_FLOCK(fl)) { |
2698 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2699 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2700 | } else { |
5315c26a | 2701 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2702 | } |
2703 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2704 | if (fl->fl_flags & FL_DELEG) |
2705 | seq_puts(f, "DELEG "); | |
2706 | else | |
2707 | seq_puts(f, "LEASE "); | |
2708 | ||
ab83fa4b | 2709 | if (lease_breaking(fl)) |
5315c26a | 2710 | seq_puts(f, "BREAKING "); |
1da177e4 | 2711 | else if (fl->fl_file) |
5315c26a | 2712 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2713 | else |
5315c26a | 2714 | seq_puts(f, "BREAKER "); |
1da177e4 | 2715 | } else { |
5315c26a | 2716 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2717 | } |
2718 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2719 | seq_printf(f, "%s ", |
1da177e4 LT |
2720 | (fl->fl_type & LOCK_READ) |
2721 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2722 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2723 | } else { | |
7f8ada98 | 2724 | seq_printf(f, "%s ", |
ab83fa4b | 2725 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2726 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2727 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2728 | } |
2729 | if (inode) { | |
3648888e | 2730 | /* userspace relies on this representation of dev_t */ |
ab1f1611 | 2731 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2732 | MAJOR(inode->i_sb->s_dev), |
2733 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2734 | } else { |
ab1f1611 | 2735 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2736 | } |
2737 | if (IS_POSIX(fl)) { | |
2738 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2739 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2740 | else |
7f8ada98 | 2741 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2742 | } else { |
5315c26a | 2743 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2744 | } |
2745 | } | |
2746 | ||
7f8ada98 | 2747 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2748 | { |
7012b02a | 2749 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2750 | struct file_lock *fl, *bfl; |
d67fd44f | 2751 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
1da177e4 | 2752 | |
139ca04e | 2753 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2754 | |
9d5b86ac | 2755 | if (locks_translate_pid(fl, proc_pidns) == 0) |
d67fd44f NB |
2756 | return 0; |
2757 | ||
7012b02a | 2758 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2759 | |
ada5c1da | 2760 | list_for_each_entry(bfl, &fl->fl_blocked_requests, fl_blocked_member) |
7012b02a | 2761 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2762 | |
7f8ada98 PE |
2763 | return 0; |
2764 | } | |
1da177e4 | 2765 | |
6c8c9031 AV |
2766 | static void __show_fd_locks(struct seq_file *f, |
2767 | struct list_head *head, int *id, | |
2768 | struct file *filp, struct files_struct *files) | |
2769 | { | |
2770 | struct file_lock *fl; | |
2771 | ||
2772 | list_for_each_entry(fl, head, fl_list) { | |
2773 | ||
2774 | if (filp != fl->fl_file) | |
2775 | continue; | |
2776 | if (fl->fl_owner != files && | |
2777 | fl->fl_owner != filp) | |
2778 | continue; | |
2779 | ||
2780 | (*id)++; | |
2781 | seq_puts(f, "lock:\t"); | |
2782 | lock_get_status(f, fl, *id, ""); | |
2783 | } | |
2784 | } | |
2785 | ||
2786 | void show_fd_locks(struct seq_file *f, | |
2787 | struct file *filp, struct files_struct *files) | |
2788 | { | |
c568d683 | 2789 | struct inode *inode = locks_inode(filp); |
6c8c9031 AV |
2790 | struct file_lock_context *ctx; |
2791 | int id = 0; | |
2792 | ||
128a3785 | 2793 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2794 | if (!ctx) |
2795 | return; | |
2796 | ||
2797 | spin_lock(&ctx->flc_lock); | |
2798 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2799 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2800 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2801 | spin_unlock(&ctx->flc_lock); | |
2802 | } | |
2803 | ||
7f8ada98 | 2804 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2805 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2806 | { |
7012b02a | 2807 | struct locks_iterator *iter = f->private; |
99dc8292 | 2808 | |
7012b02a | 2809 | iter->li_pos = *pos + 1; |
aba37660 | 2810 | percpu_down_write(&file_rwsem); |
7b2296af | 2811 | spin_lock(&blocked_lock_lock); |
7c3f654d | 2812 | return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos); |
7f8ada98 | 2813 | } |
1da177e4 | 2814 | |
7f8ada98 PE |
2815 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2816 | { | |
7012b02a JL |
2817 | struct locks_iterator *iter = f->private; |
2818 | ||
2819 | ++iter->li_pos; | |
7c3f654d | 2820 | return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos); |
7f8ada98 | 2821 | } |
1da177e4 | 2822 | |
7f8ada98 | 2823 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2824 | __releases(&blocked_lock_lock) |
7f8ada98 | 2825 | { |
7b2296af | 2826 | spin_unlock(&blocked_lock_lock); |
aba37660 | 2827 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2828 | } |
2829 | ||
d8ba7a36 | 2830 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2831 | .start = locks_start, |
2832 | .next = locks_next, | |
2833 | .stop = locks_stop, | |
2834 | .show = locks_show, | |
2835 | }; | |
d8ba7a36 | 2836 | |
d8ba7a36 AD |
2837 | static int __init proc_locks_init(void) |
2838 | { | |
44414d82 CH |
2839 | proc_create_seq_private("locks", 0, NULL, &locks_seq_operations, |
2840 | sizeof(struct locks_iterator), NULL); | |
d8ba7a36 AD |
2841 | return 0; |
2842 | } | |
91899226 | 2843 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2844 | #endif |
2845 | ||
1da177e4 LT |
2846 | static int __init filelock_init(void) |
2847 | { | |
7012b02a JL |
2848 | int i; |
2849 | ||
4a075e39 JL |
2850 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2851 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2852 | ||
1da177e4 | 2853 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2854 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2855 | ||
7012b02a | 2856 | |
7c3f654d PZ |
2857 | for_each_possible_cpu(i) { |
2858 | struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i); | |
2859 | ||
2860 | spin_lock_init(&fll->lock); | |
2861 | INIT_HLIST_HEAD(&fll->hlist); | |
2862 | } | |
7012b02a | 2863 | |
1da177e4 LT |
2864 | return 0; |
2865 | } | |
2866 | ||
2867 | core_initcall(filelock_init); |