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 JL |
129 | #include <linux/percpu.h> |
130 | #include <linux/lglock.h> | |
1da177e4 | 131 | |
62af4f1f JL |
132 | #define CREATE_TRACE_POINTS |
133 | #include <trace/events/filelock.h> | |
134 | ||
1da177e4 LT |
135 | #include <asm/uaccess.h> |
136 | ||
137 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
138 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
11afe9f7 | 139 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) |
cff2fce5 | 140 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
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 JL |
160 | * The global file_lock_list is only used for displaying /proc/locks, so we |
161 | * keep a list on each CPU, with each list protected by its own spinlock via | |
162 | * the file_lock_lglock. Note that alterations to the list also require that | |
6109c850 | 163 | * the relevant flc_lock is held. |
1c8c601a | 164 | */ |
7012b02a JL |
165 | DEFINE_STATIC_LGLOCK(file_lock_lglock); |
166 | static DEFINE_PER_CPU(struct hlist_head, file_lock_list); | |
aba37660 | 167 | DEFINE_STATIC_PERCPU_RWSEM(file_rwsem); |
88974691 | 168 | |
1c8c601a | 169 | /* |
48f74186 | 170 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 171 | * It is protected by blocked_lock_lock. |
48f74186 JL |
172 | * |
173 | * We hash locks by lockowner in order to optimize searching for the lock a | |
174 | * particular lockowner is waiting on. | |
175 | * | |
176 | * FIXME: make this value scale via some heuristic? We generally will want more | |
177 | * buckets when we have more lockowners holding locks, but that's a little | |
178 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 179 | */ |
48f74186 JL |
180 | #define BLOCKED_HASH_BITS 7 |
181 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 182 | |
1c8c601a | 183 | /* |
7b2296af JL |
184 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
185 | * want to be holding this lock. | |
1c8c601a JL |
186 | * |
187 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
188 | * pointer for file_lock structures that are acting as lock requests (in | |
189 | * contrast to those that are acting as records of acquired locks). | |
190 | * | |
191 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 192 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 193 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 194 | * flc_lock. |
1c8c601a JL |
195 | * |
196 | * In particular, adding an entry to the fl_block list requires that you hold | |
6109c850 JL |
197 | * both the flc_lock and the blocked_lock_lock (acquired in that order). |
198 | * Deleting an entry from the list however only requires the file_lock_lock. | |
1c8c601a | 199 | */ |
7b2296af | 200 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 201 | |
4a075e39 | 202 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 203 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 204 | |
4a075e39 | 205 | static struct file_lock_context * |
5c1c669a | 206 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 207 | { |
128a3785 | 208 | struct file_lock_context *ctx; |
4a075e39 | 209 | |
128a3785 DV |
210 | /* paired with cmpxchg() below */ |
211 | ctx = smp_load_acquire(&inode->i_flctx); | |
212 | if (likely(ctx) || type == F_UNLCK) | |
4a075e39 JL |
213 | goto out; |
214 | ||
128a3785 DV |
215 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
216 | if (!ctx) | |
4a075e39 JL |
217 | goto out; |
218 | ||
128a3785 DV |
219 | spin_lock_init(&ctx->flc_lock); |
220 | INIT_LIST_HEAD(&ctx->flc_flock); | |
221 | INIT_LIST_HEAD(&ctx->flc_posix); | |
222 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
223 | |
224 | /* | |
225 | * Assign the pointer if it's not already assigned. If it is, then | |
226 | * free the context we just allocated. | |
227 | */ | |
128a3785 DV |
228 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
229 | kmem_cache_free(flctx_cache, ctx); | |
230 | ctx = smp_load_acquire(&inode->i_flctx); | |
231 | } | |
4a075e39 | 232 | out: |
1890910f | 233 | trace_locks_get_lock_context(inode, type, ctx); |
128a3785 | 234 | return ctx; |
4a075e39 JL |
235 | } |
236 | ||
e24dadab JL |
237 | static void |
238 | locks_dump_ctx_list(struct list_head *list, char *list_type) | |
239 | { | |
240 | struct file_lock *fl; | |
241 | ||
242 | list_for_each_entry(fl, list, fl_list) { | |
243 | 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); | |
244 | } | |
245 | } | |
246 | ||
247 | static void | |
248 | locks_check_ctx_lists(struct inode *inode) | |
249 | { | |
250 | struct file_lock_context *ctx = inode->i_flctx; | |
251 | ||
252 | if (unlikely(!list_empty(&ctx->flc_flock) || | |
253 | !list_empty(&ctx->flc_posix) || | |
254 | !list_empty(&ctx->flc_lease))) { | |
255 | pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n", | |
256 | MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), | |
257 | inode->i_ino); | |
258 | locks_dump_ctx_list(&ctx->flc_flock, "FLOCK"); | |
259 | locks_dump_ctx_list(&ctx->flc_posix, "POSIX"); | |
260 | locks_dump_ctx_list(&ctx->flc_lease, "LEASE"); | |
261 | } | |
262 | } | |
263 | ||
4a075e39 | 264 | void |
f27a0fe0 | 265 | locks_free_lock_context(struct inode *inode) |
4a075e39 | 266 | { |
f27a0fe0 JL |
267 | struct file_lock_context *ctx = inode->i_flctx; |
268 | ||
e24dadab JL |
269 | if (unlikely(ctx)) { |
270 | locks_check_ctx_lists(inode); | |
4a075e39 JL |
271 | kmem_cache_free(flctx_cache, ctx); |
272 | } | |
273 | } | |
274 | ||
ee19cc40 | 275 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 276 | { |
139ca04e | 277 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 278 | INIT_LIST_HEAD(&fl->fl_list); |
ee19cc40 MS |
279 | INIT_LIST_HEAD(&fl->fl_block); |
280 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
281 | } |
282 | ||
1da177e4 | 283 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 284 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 285 | { |
ee19cc40 | 286 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
287 | |
288 | if (fl) | |
ee19cc40 | 289 | locks_init_lock_heads(fl); |
a51cb91d MS |
290 | |
291 | return fl; | |
1da177e4 | 292 | } |
c5b1f0d9 | 293 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 294 | |
a9e61e25 | 295 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
296 | { |
297 | if (fl->fl_ops) { | |
298 | if (fl->fl_ops->fl_release_private) | |
299 | fl->fl_ops->fl_release_private(fl); | |
300 | fl->fl_ops = NULL; | |
301 | } | |
47831f35 | 302 | |
5c97d7b1 | 303 | if (fl->fl_lmops) { |
cae80b30 JL |
304 | if (fl->fl_lmops->lm_put_owner) { |
305 | fl->fl_lmops->lm_put_owner(fl->fl_owner); | |
306 | fl->fl_owner = NULL; | |
307 | } | |
5c97d7b1 KM |
308 | fl->fl_lmops = NULL; |
309 | } | |
47831f35 | 310 | } |
a9e61e25 | 311 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 312 | |
1da177e4 | 313 | /* Free a lock which is not in use. */ |
05fa3135 | 314 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 315 | { |
5ce29646 | 316 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 317 | BUG_ON(!list_empty(&fl->fl_list)); |
5ce29646 | 318 | BUG_ON(!list_empty(&fl->fl_block)); |
139ca04e | 319 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 320 | |
47831f35 | 321 | locks_release_private(fl); |
1da177e4 LT |
322 | kmem_cache_free(filelock_cache, fl); |
323 | } | |
05fa3135 | 324 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 325 | |
ed9814d8 JL |
326 | static void |
327 | locks_dispose_list(struct list_head *dispose) | |
328 | { | |
329 | struct file_lock *fl; | |
330 | ||
331 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
332 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
333 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
334 | locks_free_lock(fl); |
335 | } | |
336 | } | |
337 | ||
1da177e4 LT |
338 | void locks_init_lock(struct file_lock *fl) |
339 | { | |
ee19cc40 MS |
340 | memset(fl, 0, sizeof(struct file_lock)); |
341 | locks_init_lock_heads(fl); | |
1da177e4 LT |
342 | } |
343 | ||
344 | EXPORT_SYMBOL(locks_init_lock); | |
345 | ||
1da177e4 LT |
346 | /* |
347 | * Initialize a new lock from an existing file_lock structure. | |
348 | */ | |
3fe0fff1 | 349 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
350 | { |
351 | new->fl_owner = fl->fl_owner; | |
352 | new->fl_pid = fl->fl_pid; | |
0996905f | 353 | new->fl_file = NULL; |
1da177e4 LT |
354 | new->fl_flags = fl->fl_flags; |
355 | new->fl_type = fl->fl_type; | |
356 | new->fl_start = fl->fl_start; | |
357 | new->fl_end = fl->fl_end; | |
f328296e | 358 | new->fl_lmops = fl->fl_lmops; |
0996905f | 359 | new->fl_ops = NULL; |
f328296e KM |
360 | |
361 | if (fl->fl_lmops) { | |
362 | if (fl->fl_lmops->lm_get_owner) | |
cae80b30 | 363 | fl->fl_lmops->lm_get_owner(fl->fl_owner); |
f328296e | 364 | } |
0996905f | 365 | } |
3fe0fff1 | 366 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
367 | |
368 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
369 | { | |
566709bd JL |
370 | /* "new" must be a freshly-initialized lock */ |
371 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 372 | |
3fe0fff1 | 373 | locks_copy_conflock(new, fl); |
f328296e | 374 | |
0996905f | 375 | new->fl_file = fl->fl_file; |
1da177e4 | 376 | new->fl_ops = fl->fl_ops; |
47831f35 | 377 | |
f328296e KM |
378 | if (fl->fl_ops) { |
379 | if (fl->fl_ops->fl_copy_lock) | |
380 | fl->fl_ops->fl_copy_lock(new, fl); | |
381 | } | |
1da177e4 LT |
382 | } |
383 | ||
384 | EXPORT_SYMBOL(locks_copy_lock); | |
385 | ||
386 | static inline int flock_translate_cmd(int cmd) { | |
387 | if (cmd & LOCK_MAND) | |
388 | return cmd & (LOCK_MAND | LOCK_RW); | |
389 | switch (cmd) { | |
390 | case LOCK_SH: | |
391 | return F_RDLCK; | |
392 | case LOCK_EX: | |
393 | return F_WRLCK; | |
394 | case LOCK_UN: | |
395 | return F_UNLCK; | |
396 | } | |
397 | return -EINVAL; | |
398 | } | |
399 | ||
400 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 JL |
401 | static struct file_lock * |
402 | flock_make_lock(struct file *filp, unsigned int cmd) | |
1da177e4 LT |
403 | { |
404 | struct file_lock *fl; | |
405 | int type = flock_translate_cmd(cmd); | |
6e129d00 | 406 | |
1da177e4 | 407 | if (type < 0) |
6e129d00 | 408 | return ERR_PTR(type); |
1da177e4 LT |
409 | |
410 | fl = locks_alloc_lock(); | |
411 | if (fl == NULL) | |
6e129d00 | 412 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
413 | |
414 | fl->fl_file = filp; | |
73a8f5f7 | 415 | fl->fl_owner = filp; |
1da177e4 LT |
416 | fl->fl_pid = current->tgid; |
417 | fl->fl_flags = FL_FLOCK; | |
418 | fl->fl_type = type; | |
419 | fl->fl_end = OFFSET_MAX; | |
420 | ||
6e129d00 | 421 | return fl; |
1da177e4 LT |
422 | } |
423 | ||
0ec4f431 | 424 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
425 | { |
426 | switch (type) { | |
427 | case F_RDLCK: | |
428 | case F_WRLCK: | |
429 | case F_UNLCK: | |
430 | fl->fl_type = type; | |
431 | break; | |
432 | default: | |
433 | return -EINVAL; | |
434 | } | |
435 | return 0; | |
436 | } | |
437 | ||
ef12e72a BF |
438 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
439 | struct flock64 *l) | |
1da177e4 | 440 | { |
1da177e4 | 441 | switch (l->l_whence) { |
f5579f8c | 442 | case SEEK_SET: |
ef12e72a | 443 | fl->fl_start = 0; |
1da177e4 | 444 | break; |
f5579f8c | 445 | case SEEK_CUR: |
ef12e72a | 446 | fl->fl_start = filp->f_pos; |
1da177e4 | 447 | break; |
f5579f8c | 448 | case SEEK_END: |
ef12e72a | 449 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
450 | break; |
451 | default: | |
452 | return -EINVAL; | |
453 | } | |
ef12e72a BF |
454 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
455 | return -EOVERFLOW; | |
456 | fl->fl_start += l->l_start; | |
457 | if (fl->fl_start < 0) | |
458 | return -EINVAL; | |
1da177e4 LT |
459 | |
460 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
461 | POSIX-2001 defines it. */ | |
4c780a46 | 462 | if (l->l_len > 0) { |
ef12e72a BF |
463 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
464 | return -EOVERFLOW; | |
465 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
466 | ||
4c780a46 | 467 | } else if (l->l_len < 0) { |
ef12e72a | 468 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 469 | return -EINVAL; |
ef12e72a BF |
470 | fl->fl_end = fl->fl_start - 1; |
471 | fl->fl_start += l->l_len; | |
472 | } else | |
473 | fl->fl_end = OFFSET_MAX; | |
474 | ||
1da177e4 LT |
475 | fl->fl_owner = current->files; |
476 | fl->fl_pid = current->tgid; | |
477 | fl->fl_file = filp; | |
478 | fl->fl_flags = FL_POSIX; | |
479 | fl->fl_ops = NULL; | |
480 | fl->fl_lmops = NULL; | |
481 | ||
482 | return assign_type(fl, l->l_type); | |
483 | } | |
484 | ||
ef12e72a BF |
485 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
486 | * style lock. | |
487 | */ | |
488 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
489 | struct flock *l) | |
1da177e4 | 490 | { |
ef12e72a BF |
491 | struct flock64 ll = { |
492 | .l_type = l->l_type, | |
493 | .l_whence = l->l_whence, | |
494 | .l_start = l->l_start, | |
495 | .l_len = l->l_len, | |
496 | }; | |
497 | ||
498 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 499 | } |
1da177e4 LT |
500 | |
501 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
502 | static bool |
503 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
504 | { |
505 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 506 | return false; |
1da177e4 LT |
507 | } |
508 | ||
1c7dd2ff JL |
509 | static void |
510 | lease_setup(struct file_lock *fl, void **priv) | |
511 | { | |
512 | struct file *filp = fl->fl_file; | |
513 | struct fasync_struct *fa = *priv; | |
514 | ||
515 | /* | |
516 | * fasync_insert_entry() returns the old entry if any. If there was no | |
517 | * old entry, then it used "priv" and inserted it into the fasync list. | |
518 | * Clear the pointer to indicate that it shouldn't be freed. | |
519 | */ | |
520 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
521 | *priv = NULL; | |
522 | ||
523 | __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
524 | } | |
525 | ||
7b021967 | 526 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 527 | .lm_break = lease_break_callback, |
8fb47a4f | 528 | .lm_change = lease_modify, |
1c7dd2ff | 529 | .lm_setup = lease_setup, |
1da177e4 LT |
530 | }; |
531 | ||
532 | /* | |
533 | * Initialize a lease, use the default lock manager operations | |
534 | */ | |
0ec4f431 | 535 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 536 | { |
75dff55a TM |
537 | if (assign_type(fl, type) != 0) |
538 | return -EINVAL; | |
539 | ||
7ca76311 | 540 | fl->fl_owner = filp; |
1da177e4 LT |
541 | fl->fl_pid = current->tgid; |
542 | ||
543 | fl->fl_file = filp; | |
544 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
545 | fl->fl_start = 0; |
546 | fl->fl_end = OFFSET_MAX; | |
547 | fl->fl_ops = NULL; | |
548 | fl->fl_lmops = &lease_manager_ops; | |
549 | return 0; | |
550 | } | |
551 | ||
552 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 553 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
554 | { |
555 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 556 | int error = -ENOMEM; |
1da177e4 LT |
557 | |
558 | if (fl == NULL) | |
e32b8ee2 | 559 | return ERR_PTR(error); |
1da177e4 LT |
560 | |
561 | error = lease_init(filp, type, fl); | |
75dff55a TM |
562 | if (error) { |
563 | locks_free_lock(fl); | |
e32b8ee2 | 564 | return ERR_PTR(error); |
75dff55a | 565 | } |
e32b8ee2 | 566 | return fl; |
1da177e4 LT |
567 | } |
568 | ||
569 | /* Check if two locks overlap each other. | |
570 | */ | |
571 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
572 | { | |
573 | return ((fl1->fl_end >= fl2->fl_start) && | |
574 | (fl2->fl_end >= fl1->fl_start)); | |
575 | } | |
576 | ||
577 | /* | |
578 | * Check whether two locks have the same owner. | |
579 | */ | |
33443c42 | 580 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 581 | { |
8fb47a4f | 582 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 583 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 584 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
585 | return fl1->fl_owner == fl2->fl_owner; |
586 | } | |
587 | ||
6109c850 | 588 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 589 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 590 | { |
aba37660 PZ |
591 | percpu_rwsem_assert_held(&file_rwsem); |
592 | ||
7012b02a JL |
593 | lg_local_lock(&file_lock_lglock); |
594 | fl->fl_link_cpu = smp_processor_id(); | |
595 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
596 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
597 | } |
598 | ||
6109c850 | 599 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 600 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 601 | { |
aba37660 PZ |
602 | percpu_rwsem_assert_held(&file_rwsem); |
603 | ||
7012b02a JL |
604 | /* |
605 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 606 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
607 | * also require that it be held. |
608 | */ | |
609 | if (hlist_unhashed(&fl->fl_link)) | |
610 | return; | |
611 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 612 | hlist_del_init(&fl->fl_link); |
7012b02a | 613 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
614 | } |
615 | ||
3999e493 JL |
616 | static unsigned long |
617 | posix_owner_key(struct file_lock *fl) | |
618 | { | |
619 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
620 | return fl->fl_lmops->lm_owner_key(fl); | |
621 | return (unsigned long)fl->fl_owner; | |
622 | } | |
623 | ||
6ca10ed8 | 624 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 625 | { |
663d5af7 DW |
626 | lockdep_assert_held(&blocked_lock_lock); |
627 | ||
3999e493 | 628 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
629 | } |
630 | ||
6ca10ed8 | 631 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 632 | { |
663d5af7 DW |
633 | lockdep_assert_held(&blocked_lock_lock); |
634 | ||
48f74186 | 635 | hash_del(&waiter->fl_link); |
88974691 JL |
636 | } |
637 | ||
1da177e4 LT |
638 | /* Remove waiter from blocker's block list. |
639 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 640 | * |
7b2296af | 641 | * Must be called with blocked_lock_lock held. |
1da177e4 | 642 | */ |
33443c42 | 643 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 644 | { |
88974691 | 645 | locks_delete_global_blocked(waiter); |
1da177e4 | 646 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
647 | waiter->fl_next = NULL; |
648 | } | |
649 | ||
1a9e64a7 | 650 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 651 | { |
7b2296af | 652 | spin_lock(&blocked_lock_lock); |
1da177e4 | 653 | __locks_delete_block(waiter); |
7b2296af | 654 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
655 | } |
656 | ||
657 | /* Insert waiter into blocker's block list. | |
658 | * We use a circular list so that processes can be easily woken up in | |
659 | * the order they blocked. The documentation doesn't require this but | |
660 | * it seems like the reasonable thing to do. | |
1c8c601a | 661 | * |
6109c850 JL |
662 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
663 | * fl_block list itself is protected by the blocked_lock_lock, but by ensuring | |
664 | * that the flc_lock is also held on insertions we can avoid taking the | |
665 | * blocked_lock_lock in some cases when we see that the fl_block list is empty. | |
1da177e4 | 666 | */ |
1c8c601a JL |
667 | static void __locks_insert_block(struct file_lock *blocker, |
668 | struct file_lock *waiter) | |
1da177e4 | 669 | { |
6dc0fe8f | 670 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 671 | waiter->fl_next = blocker; |
88974691 | 672 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 673 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
674 | locks_insert_global_blocked(waiter); |
675 | } | |
676 | ||
6109c850 | 677 | /* Must be called with flc_lock held. */ |
1c8c601a JL |
678 | static void locks_insert_block(struct file_lock *blocker, |
679 | struct file_lock *waiter) | |
680 | { | |
7b2296af | 681 | spin_lock(&blocked_lock_lock); |
1c8c601a | 682 | __locks_insert_block(blocker, waiter); |
7b2296af | 683 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
684 | } |
685 | ||
1cb36012 JL |
686 | /* |
687 | * Wake up processes blocked waiting for blocker. | |
688 | * | |
6109c850 | 689 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
690 | */ |
691 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
692 | { | |
4e8c765d JL |
693 | /* |
694 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 JL |
695 | * blocked requests are only added to the list under the flc_lock, and |
696 | * the flc_lock is always held here. Note that removal from the fl_block | |
697 | * list does not require the flc_lock, so we must recheck list_empty() | |
7b2296af | 698 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
699 | */ |
700 | if (list_empty(&blocker->fl_block)) | |
701 | return; | |
702 | ||
7b2296af | 703 | spin_lock(&blocked_lock_lock); |
1da177e4 | 704 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
705 | struct file_lock *waiter; |
706 | ||
707 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
708 | struct file_lock, fl_block); |
709 | __locks_delete_block(waiter); | |
8fb47a4f BF |
710 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
711 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
712 | else |
713 | wake_up(&waiter->fl_wait); | |
714 | } | |
7b2296af | 715 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
716 | } |
717 | ||
5263e31e | 718 | static void |
e084c1bd | 719 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e JL |
720 | { |
721 | fl->fl_nspid = get_pid(task_tgid(current)); | |
722 | list_add_tail(&fl->fl_list, before); | |
723 | locks_insert_global_locks(fl); | |
724 | } | |
725 | ||
8634b51f | 726 | static void |
e084c1bd | 727 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 728 | { |
88974691 | 729 | locks_delete_global_locks(fl); |
8634b51f | 730 | list_del_init(&fl->fl_list); |
ab1f1611 VG |
731 | if (fl->fl_nspid) { |
732 | put_pid(fl->fl_nspid); | |
733 | fl->fl_nspid = NULL; | |
734 | } | |
1da177e4 | 735 | locks_wake_up_blocks(fl); |
24cbe784 JL |
736 | } |
737 | ||
8634b51f | 738 | static void |
e084c1bd | 739 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 740 | { |
e084c1bd | 741 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 742 | if (dispose) |
6dee60f6 | 743 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
744 | else |
745 | locks_free_lock(fl); | |
1da177e4 LT |
746 | } |
747 | ||
748 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
749 | * checks for shared/exclusive status of overlapping locks. | |
750 | */ | |
751 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
752 | { | |
753 | if (sys_fl->fl_type == F_WRLCK) | |
754 | return 1; | |
755 | if (caller_fl->fl_type == F_WRLCK) | |
756 | return 1; | |
757 | return 0; | |
758 | } | |
759 | ||
760 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
761 | * checking before calling the locks_conflict(). | |
762 | */ | |
763 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
764 | { | |
765 | /* POSIX locks owned by the same process do not conflict with | |
766 | * each other. | |
767 | */ | |
9b8c8695 | 768 | if (posix_same_owner(caller_fl, sys_fl)) |
1da177e4 LT |
769 | return (0); |
770 | ||
771 | /* Check whether they overlap */ | |
772 | if (!locks_overlap(caller_fl, sys_fl)) | |
773 | return 0; | |
774 | ||
775 | return (locks_conflict(caller_fl, sys_fl)); | |
776 | } | |
777 | ||
778 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
779 | * checking before calling the locks_conflict(). | |
780 | */ | |
781 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
782 | { | |
783 | /* FLOCK locks referring to the same filp do not conflict with | |
784 | * each other. | |
785 | */ | |
9b8c8695 | 786 | if (caller_fl->fl_file == sys_fl->fl_file) |
1da177e4 LT |
787 | return (0); |
788 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
789 | return 0; | |
790 | ||
791 | return (locks_conflict(caller_fl, sys_fl)); | |
792 | } | |
793 | ||
6d34ac19 | 794 | void |
9d6a8c5c | 795 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
796 | { |
797 | struct file_lock *cfl; | |
bd61e0a9 | 798 | struct file_lock_context *ctx; |
1c8c601a | 799 | struct inode *inode = file_inode(filp); |
1da177e4 | 800 | |
128a3785 | 801 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
802 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
803 | fl->fl_type = F_UNLCK; | |
804 | return; | |
805 | } | |
806 | ||
6109c850 | 807 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
808 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
809 | if (posix_locks_conflict(fl, cfl)) { | |
810 | locks_copy_conflock(fl, cfl); | |
811 | if (cfl->fl_nspid) | |
812 | fl->fl_pid = pid_vnr(cfl->fl_nspid); | |
813 | goto out; | |
814 | } | |
1da177e4 | 815 | } |
bd61e0a9 JL |
816 | fl->fl_type = F_UNLCK; |
817 | out: | |
6109c850 | 818 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 819 | return; |
1da177e4 | 820 | } |
1da177e4 LT |
821 | EXPORT_SYMBOL(posix_test_lock); |
822 | ||
b533184f BF |
823 | /* |
824 | * Deadlock detection: | |
825 | * | |
826 | * We attempt to detect deadlocks that are due purely to posix file | |
827 | * locks. | |
1da177e4 | 828 | * |
b533184f BF |
829 | * We assume that a task can be waiting for at most one lock at a time. |
830 | * So for any acquired lock, the process holding that lock may be | |
831 | * waiting on at most one other lock. That lock in turns may be held by | |
832 | * someone waiting for at most one other lock. Given a requested lock | |
833 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
834 | * follow this chain of waiters to ensure we are not about to create a | |
835 | * cycle. | |
1da177e4 | 836 | * |
b533184f BF |
837 | * Since we do this before we ever put a process to sleep on a lock, we |
838 | * are ensured that there is never a cycle; that is what guarantees that | |
839 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 840 | * |
b533184f BF |
841 | * Note: the above assumption may not be true when handling lock |
842 | * requests from a broken NFS client. It may also fail in the presence | |
843 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 844 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 845 | * |
cff2fce5 | 846 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
847 | * Because the owner is not even nominally tied to a thread of |
848 | * execution, the deadlock detection below can't reasonably work well. Just | |
849 | * skip it for those. | |
850 | * | |
cff2fce5 | 851 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
852 | * locks that just checks for the case where two tasks are attempting to |
853 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 854 | */ |
97855b49 BF |
855 | |
856 | #define MAX_DEADLK_ITERATIONS 10 | |
857 | ||
b533184f BF |
858 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
859 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
860 | { | |
861 | struct file_lock *fl; | |
862 | ||
3999e493 | 863 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
864 | if (posix_same_owner(fl, block_fl)) |
865 | return fl->fl_next; | |
866 | } | |
867 | return NULL; | |
868 | } | |
869 | ||
7b2296af | 870 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 871 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
872 | struct file_lock *block_fl) |
873 | { | |
97855b49 | 874 | int i = 0; |
1da177e4 | 875 | |
663d5af7 DW |
876 | lockdep_assert_held(&blocked_lock_lock); |
877 | ||
57b65325 JL |
878 | /* |
879 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 880 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 881 | */ |
cff2fce5 | 882 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
883 | return 0; |
884 | ||
b533184f BF |
885 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
886 | if (i++ > MAX_DEADLK_ITERATIONS) | |
887 | return 0; | |
888 | if (posix_same_owner(caller_fl, block_fl)) | |
889 | return 1; | |
1da177e4 LT |
890 | } |
891 | return 0; | |
892 | } | |
893 | ||
1da177e4 | 894 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 895 | * after any leases, but before any posix locks. |
f475ae95 TM |
896 | * |
897 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
898 | * whether or not a lock was successfully freed by testing the return | |
899 | * value for -ENOENT. | |
1da177e4 | 900 | */ |
bcd7f78d | 901 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 902 | { |
993dfa87 | 903 | struct file_lock *new_fl = NULL; |
5263e31e JL |
904 | struct file_lock *fl; |
905 | struct file_lock_context *ctx; | |
1da177e4 | 906 | int error = 0; |
5263e31e | 907 | bool found = false; |
ed9814d8 | 908 | LIST_HEAD(dispose); |
1da177e4 | 909 | |
5c1c669a JL |
910 | ctx = locks_get_lock_context(inode, request->fl_type); |
911 | if (!ctx) { | |
912 | if (request->fl_type != F_UNLCK) | |
913 | return -ENOMEM; | |
914 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
915 | } | |
5263e31e | 916 | |
b89f4321 | 917 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 918 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
919 | if (!new_fl) |
920 | return -ENOMEM; | |
84d535ad PE |
921 | } |
922 | ||
aba37660 | 923 | percpu_down_read(&file_rwsem); |
6109c850 | 924 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
925 | if (request->fl_flags & FL_ACCESS) |
926 | goto find_conflict; | |
927 | ||
5263e31e | 928 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 929 | if (request->fl_file != fl->fl_file) |
1da177e4 | 930 | continue; |
993dfa87 | 931 | if (request->fl_type == fl->fl_type) |
1da177e4 | 932 | goto out; |
5263e31e | 933 | found = true; |
e084c1bd | 934 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
935 | break; |
936 | } | |
1da177e4 | 937 | |
f475ae95 TM |
938 | if (request->fl_type == F_UNLCK) { |
939 | if ((request->fl_flags & FL_EXISTS) && !found) | |
940 | error = -ENOENT; | |
993dfa87 | 941 | goto out; |
f475ae95 | 942 | } |
1da177e4 | 943 | |
f07f18dd | 944 | find_conflict: |
5263e31e | 945 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 946 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
947 | continue; |
948 | error = -EAGAIN; | |
bde74e4b MS |
949 | if (!(request->fl_flags & FL_SLEEP)) |
950 | goto out; | |
951 | error = FILE_LOCK_DEFERRED; | |
952 | locks_insert_block(fl, request); | |
1da177e4 LT |
953 | goto out; |
954 | } | |
f07f18dd TM |
955 | if (request->fl_flags & FL_ACCESS) |
956 | goto out; | |
993dfa87 | 957 | locks_copy_lock(new_fl, request); |
e084c1bd | 958 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 959 | new_fl = NULL; |
9cedc194 | 960 | error = 0; |
1da177e4 LT |
961 | |
962 | out: | |
6109c850 | 963 | spin_unlock(&ctx->flc_lock); |
aba37660 | 964 | percpu_up_read(&file_rwsem); |
993dfa87 TM |
965 | if (new_fl) |
966 | locks_free_lock(new_fl); | |
ed9814d8 | 967 | locks_dispose_list(&dispose); |
1da177e4 LT |
968 | return error; |
969 | } | |
970 | ||
b4d629a3 JL |
971 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
972 | struct file_lock *conflock) | |
1da177e4 | 973 | { |
bd61e0a9 | 974 | struct file_lock *fl, *tmp; |
39005d02 MS |
975 | struct file_lock *new_fl = NULL; |
976 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
977 | struct file_lock *left = NULL; |
978 | struct file_lock *right = NULL; | |
bd61e0a9 | 979 | struct file_lock_context *ctx; |
b9746ef8 JL |
980 | int error; |
981 | bool added = false; | |
ed9814d8 | 982 | LIST_HEAD(dispose); |
1da177e4 | 983 | |
5c1c669a | 984 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 985 | if (!ctx) |
5c1c669a | 986 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 987 | |
1da177e4 LT |
988 | /* |
989 | * We may need two file_lock structures for this operation, | |
990 | * so we get them in advance to avoid races. | |
39005d02 MS |
991 | * |
992 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 993 | */ |
39005d02 MS |
994 | if (!(request->fl_flags & FL_ACCESS) && |
995 | (request->fl_type != F_UNLCK || | |
996 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
997 | new_fl = locks_alloc_lock(); | |
998 | new_fl2 = locks_alloc_lock(); | |
999 | } | |
1da177e4 | 1000 | |
aba37660 | 1001 | percpu_down_read(&file_rwsem); |
6109c850 | 1002 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1003 | /* |
1004 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1005 | * there are any, either return error or put the request on the | |
48f74186 | 1006 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1007 | */ |
1da177e4 | 1008 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1009 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1010 | if (!posix_locks_conflict(request, fl)) |
1011 | continue; | |
5842add2 | 1012 | if (conflock) |
3fe0fff1 | 1013 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1014 | error = -EAGAIN; |
1015 | if (!(request->fl_flags & FL_SLEEP)) | |
1016 | goto out; | |
1c8c601a JL |
1017 | /* |
1018 | * Deadlock detection and insertion into the blocked | |
1019 | * locks list must be done while holding the same lock! | |
1020 | */ | |
1da177e4 | 1021 | error = -EDEADLK; |
7b2296af | 1022 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
1023 | if (likely(!posix_locks_deadlock(request, fl))) { |
1024 | error = FILE_LOCK_DEFERRED; | |
1025 | __locks_insert_block(fl, request); | |
1026 | } | |
7b2296af | 1027 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1028 | goto out; |
1029 | } | |
1030 | } | |
1031 | ||
1032 | /* If we're just looking for a conflict, we're done. */ | |
1033 | error = 0; | |
1034 | if (request->fl_flags & FL_ACCESS) | |
1035 | goto out; | |
1036 | ||
bd61e0a9 JL |
1037 | /* Find the first old lock with the same owner as the new lock */ |
1038 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1039 | if (posix_same_owner(request, fl)) | |
1040 | break; | |
1da177e4 LT |
1041 | } |
1042 | ||
1cb36012 | 1043 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1044 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1045 | if (!posix_same_owner(request, fl)) | |
1046 | break; | |
1047 | ||
1048 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1049 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1050 | /* In all comparisons of start vs end, use |
1051 | * "start - 1" rather than "end + 1". If end | |
1052 | * is OFFSET_MAX, end + 1 will become negative. | |
1053 | */ | |
1da177e4 | 1054 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1055 | continue; |
1da177e4 LT |
1056 | /* If the next lock in the list has entirely bigger |
1057 | * addresses than the new one, insert the lock here. | |
1058 | */ | |
449231d6 | 1059 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1060 | break; |
1061 | ||
1062 | /* If we come here, the new and old lock are of the | |
1063 | * same type and adjacent or overlapping. Make one | |
1064 | * lock yielding from the lower start address of both | |
1065 | * locks to the higher end address. | |
1066 | */ | |
1067 | if (fl->fl_start > request->fl_start) | |
1068 | fl->fl_start = request->fl_start; | |
1069 | else | |
1070 | request->fl_start = fl->fl_start; | |
1071 | if (fl->fl_end < request->fl_end) | |
1072 | fl->fl_end = request->fl_end; | |
1073 | else | |
1074 | request->fl_end = fl->fl_end; | |
1075 | if (added) { | |
e084c1bd | 1076 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1077 | continue; |
1078 | } | |
1079 | request = fl; | |
b9746ef8 | 1080 | added = true; |
bd61e0a9 | 1081 | } else { |
1da177e4 LT |
1082 | /* Processing for different lock types is a bit |
1083 | * more complex. | |
1084 | */ | |
1085 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1086 | continue; |
1da177e4 LT |
1087 | if (fl->fl_start > request->fl_end) |
1088 | break; | |
1089 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1090 | added = true; |
1da177e4 LT |
1091 | if (fl->fl_start < request->fl_start) |
1092 | left = fl; | |
1093 | /* If the next lock in the list has a higher end | |
1094 | * address than the new one, insert the new one here. | |
1095 | */ | |
1096 | if (fl->fl_end > request->fl_end) { | |
1097 | right = fl; | |
1098 | break; | |
1099 | } | |
1100 | if (fl->fl_start >= request->fl_start) { | |
1101 | /* The new lock completely replaces an old | |
1102 | * one (This may happen several times). | |
1103 | */ | |
1104 | if (added) { | |
e084c1bd | 1105 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1106 | continue; |
1107 | } | |
b84d49f9 JL |
1108 | /* |
1109 | * Replace the old lock with new_fl, and | |
1110 | * remove the old one. It's safe to do the | |
1111 | * insert here since we know that we won't be | |
1112 | * using new_fl later, and that the lock is | |
1113 | * just replacing an existing lock. | |
1da177e4 | 1114 | */ |
b84d49f9 JL |
1115 | error = -ENOLCK; |
1116 | if (!new_fl) | |
1117 | goto out; | |
1118 | locks_copy_lock(new_fl, request); | |
1119 | request = new_fl; | |
1120 | new_fl = NULL; | |
e084c1bd JL |
1121 | locks_insert_lock_ctx(request, &fl->fl_list); |
1122 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1123 | added = true; |
1da177e4 LT |
1124 | } |
1125 | } | |
1da177e4 LT |
1126 | } |
1127 | ||
0d9a490a | 1128 | /* |
1cb36012 JL |
1129 | * The above code only modifies existing locks in case of merging or |
1130 | * replacing. If new lock(s) need to be inserted all modifications are | |
1131 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1132 | */ |
1133 | error = -ENOLCK; /* "no luck" */ | |
1134 | if (right && left == right && !new_fl2) | |
1135 | goto out; | |
1136 | ||
1da177e4 LT |
1137 | error = 0; |
1138 | if (!added) { | |
f475ae95 TM |
1139 | if (request->fl_type == F_UNLCK) { |
1140 | if (request->fl_flags & FL_EXISTS) | |
1141 | error = -ENOENT; | |
1da177e4 | 1142 | goto out; |
f475ae95 | 1143 | } |
0d9a490a MS |
1144 | |
1145 | if (!new_fl) { | |
1146 | error = -ENOLCK; | |
1147 | goto out; | |
1148 | } | |
1da177e4 | 1149 | locks_copy_lock(new_fl, request); |
e084c1bd | 1150 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
2e2f756f | 1151 | fl = new_fl; |
1da177e4 LT |
1152 | new_fl = NULL; |
1153 | } | |
1154 | if (right) { | |
1155 | if (left == right) { | |
1156 | /* The new lock breaks the old one in two pieces, | |
1157 | * so we have to use the second new lock. | |
1158 | */ | |
1159 | left = new_fl2; | |
1160 | new_fl2 = NULL; | |
1161 | locks_copy_lock(left, right); | |
e084c1bd | 1162 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1163 | } |
1164 | right->fl_start = request->fl_end + 1; | |
1165 | locks_wake_up_blocks(right); | |
1166 | } | |
1167 | if (left) { | |
1168 | left->fl_end = request->fl_start - 1; | |
1169 | locks_wake_up_blocks(left); | |
1170 | } | |
1171 | out: | |
6109c850 | 1172 | spin_unlock(&ctx->flc_lock); |
aba37660 | 1173 | percpu_up_read(&file_rwsem); |
1da177e4 LT |
1174 | /* |
1175 | * Free any unused locks. | |
1176 | */ | |
1177 | if (new_fl) | |
1178 | locks_free_lock(new_fl); | |
1179 | if (new_fl2) | |
1180 | locks_free_lock(new_fl2); | |
ed9814d8 | 1181 | locks_dispose_list(&dispose); |
1890910f JL |
1182 | trace_posix_lock_inode(inode, request, error); |
1183 | ||
1da177e4 LT |
1184 | return error; |
1185 | } | |
1186 | ||
1187 | /** | |
1188 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1189 | * @filp: The file to apply the lock to | |
1190 | * @fl: The lock to be applied | |
150b3934 | 1191 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1192 | * |
1193 | * Add a POSIX style lock to a file. | |
1194 | * We merge adjacent & overlapping locks whenever possible. | |
1195 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1196 | * |
1197 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1198 | * whether or not a lock was successfully freed by testing the return | |
1199 | * value for -ENOENT. | |
1da177e4 | 1200 | */ |
150b3934 | 1201 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1202 | struct file_lock *conflock) |
1203 | { | |
b4d629a3 | 1204 | return posix_lock_inode(file_inode(filp), fl, conflock); |
1da177e4 | 1205 | } |
150b3934 | 1206 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1207 | |
1208 | /** | |
29d01b22 JL |
1209 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1210 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1211 | * @fl: The lock to be applied |
1212 | * | |
616fb38f | 1213 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1214 | */ |
616fb38f | 1215 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1216 | { |
1217 | int error; | |
1218 | might_sleep (); | |
1219 | for (;;) { | |
b4d629a3 | 1220 | error = posix_lock_inode(inode, fl, NULL); |
bde74e4b | 1221 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1222 | break; |
1223 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1224 | if (!error) | |
1225 | continue; | |
1226 | ||
1227 | locks_delete_block(fl); | |
1228 | break; | |
1229 | } | |
1230 | return error; | |
1231 | } | |
29d01b22 | 1232 | |
9e8925b6 | 1233 | #ifdef CONFIG_MANDATORY_FILE_LOCKING |
1da177e4 LT |
1234 | /** |
1235 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1236 | * @file: the file to check |
1da177e4 LT |
1237 | * |
1238 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1239 | * This function is called from locks_verify_locked() only. | |
1240 | */ | |
d7a06983 | 1241 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1242 | { |
bd61e0a9 | 1243 | int ret; |
d7a06983 | 1244 | struct inode *inode = file_inode(file); |
bd61e0a9 | 1245 | struct file_lock_context *ctx; |
1da177e4 LT |
1246 | struct file_lock *fl; |
1247 | ||
128a3785 | 1248 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
1249 | if (!ctx || list_empty_careful(&ctx->flc_posix)) |
1250 | return 0; | |
1251 | ||
1da177e4 LT |
1252 | /* |
1253 | * Search the lock list for this inode for any POSIX locks. | |
1254 | */ | |
6109c850 | 1255 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
1256 | ret = 0; |
1257 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1258 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1259 | fl->fl_owner != file) { |
1260 | ret = -EAGAIN; | |
1da177e4 | 1261 | break; |
bd61e0a9 | 1262 | } |
1da177e4 | 1263 | } |
6109c850 | 1264 | spin_unlock(&ctx->flc_lock); |
bd61e0a9 | 1265 | return ret; |
1da177e4 LT |
1266 | } |
1267 | ||
1268 | /** | |
1269 | * locks_mandatory_area - Check for a conflicting lock | |
acc15575 | 1270 | * @inode: the file to check |
1da177e4 | 1271 | * @filp: how the file was opened (if it was) |
acc15575 CH |
1272 | * @start: first byte in the file to check |
1273 | * @end: lastbyte in the file to check | |
1274 | * @type: %F_WRLCK for a write lock, else %F_RDLCK | |
1da177e4 LT |
1275 | * |
1276 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1da177e4 | 1277 | */ |
acc15575 CH |
1278 | int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start, |
1279 | loff_t end, unsigned char type) | |
1da177e4 LT |
1280 | { |
1281 | struct file_lock fl; | |
1282 | int error; | |
29723ade | 1283 | bool sleep = false; |
1da177e4 LT |
1284 | |
1285 | locks_init_lock(&fl); | |
1da177e4 LT |
1286 | fl.fl_pid = current->tgid; |
1287 | fl.fl_file = filp; | |
1288 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1289 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1290 | sleep = true; |
acc15575 CH |
1291 | fl.fl_type = type; |
1292 | fl.fl_start = start; | |
1293 | fl.fl_end = end; | |
1da177e4 LT |
1294 | |
1295 | for (;;) { | |
29723ade | 1296 | if (filp) { |
73a8f5f7 | 1297 | fl.fl_owner = filp; |
29723ade | 1298 | fl.fl_flags &= ~FL_SLEEP; |
b4d629a3 | 1299 | error = posix_lock_inode(inode, &fl, NULL); |
29723ade JL |
1300 | if (!error) |
1301 | break; | |
1302 | } | |
1303 | ||
1304 | if (sleep) | |
1305 | fl.fl_flags |= FL_SLEEP; | |
1306 | fl.fl_owner = current->files; | |
b4d629a3 | 1307 | error = posix_lock_inode(inode, &fl, NULL); |
bde74e4b | 1308 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1309 | break; |
1310 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1311 | if (!error) { | |
1312 | /* | |
1313 | * If we've been sleeping someone might have | |
1314 | * changed the permissions behind our back. | |
1315 | */ | |
a16877ca | 1316 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1317 | continue; |
1318 | } | |
1319 | ||
1320 | locks_delete_block(&fl); | |
1321 | break; | |
1322 | } | |
1323 | ||
1324 | return error; | |
1325 | } | |
1326 | ||
1327 | EXPORT_SYMBOL(locks_mandatory_area); | |
9e8925b6 | 1328 | #endif /* CONFIG_MANDATORY_FILE_LOCKING */ |
1da177e4 | 1329 | |
778fc546 BF |
1330 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1331 | { | |
1332 | switch (arg) { | |
1333 | case F_UNLCK: | |
1334 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1335 | /* fall through: */ | |
1336 | case F_RDLCK: | |
1337 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1338 | } | |
1339 | } | |
1340 | ||
1da177e4 | 1341 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1342 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1343 | { |
1da177e4 LT |
1344 | int error = assign_type(fl, arg); |
1345 | ||
1346 | if (error) | |
1347 | return error; | |
778fc546 | 1348 | lease_clear_pending(fl, arg); |
1da177e4 | 1349 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1350 | if (arg == F_UNLCK) { |
1351 | struct file *filp = fl->fl_file; | |
1352 | ||
1353 | f_delown(filp); | |
1354 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1355 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1356 | if (fl->fl_fasync != NULL) { | |
1357 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1358 | fl->fl_fasync = NULL; | |
1359 | } | |
e084c1bd | 1360 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1361 | } |
1da177e4 LT |
1362 | return 0; |
1363 | } | |
1da177e4 LT |
1364 | EXPORT_SYMBOL(lease_modify); |
1365 | ||
778fc546 BF |
1366 | static bool past_time(unsigned long then) |
1367 | { | |
1368 | if (!then) | |
1369 | /* 0 is a special value meaning "this never expires": */ | |
1370 | return false; | |
1371 | return time_after(jiffies, then); | |
1372 | } | |
1373 | ||
c45198ed | 1374 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1375 | { |
8634b51f JL |
1376 | struct file_lock_context *ctx = inode->i_flctx; |
1377 | struct file_lock *fl, *tmp; | |
1da177e4 | 1378 | |
6109c850 | 1379 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1380 | |
8634b51f | 1381 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1382 | trace_time_out_leases(inode, fl); |
778fc546 | 1383 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1384 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1385 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1386 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1387 | } |
1388 | } | |
1389 | ||
df4e8d2c BF |
1390 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1391 | { | |
11afe9f7 CH |
1392 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) |
1393 | return false; | |
df4e8d2c BF |
1394 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) |
1395 | return false; | |
1396 | return locks_conflict(breaker, lease); | |
1397 | } | |
1398 | ||
03d12ddf JL |
1399 | static bool |
1400 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1401 | { | |
8634b51f | 1402 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1403 | struct file_lock *fl; |
1404 | ||
6109c850 | 1405 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1406 | |
8634b51f | 1407 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1408 | if (leases_conflict(fl, breaker)) |
1409 | return true; | |
1410 | } | |
1411 | return false; | |
1412 | } | |
1413 | ||
1da177e4 LT |
1414 | /** |
1415 | * __break_lease - revoke all outstanding leases on file | |
1416 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1417 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1418 | * break all leases | |
1419 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1420 | * only delegations | |
1da177e4 | 1421 | * |
87250dd2 | 1422 | * break_lease (inlined for speed) has checked there already is at least |
1423 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1424 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1425 | * specified %O_NONBLOCK to your open(). |
1426 | */ | |
df4e8d2c | 1427 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1428 | { |
778fc546 | 1429 | int error = 0; |
128a3785 | 1430 | struct file_lock_context *ctx; |
a901125c | 1431 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1432 | unsigned long break_time; |
8737c930 | 1433 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1434 | LIST_HEAD(dispose); |
1da177e4 | 1435 | |
8737c930 | 1436 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1437 | if (IS_ERR(new_fl)) |
1438 | return PTR_ERR(new_fl); | |
df4e8d2c | 1439 | new_fl->fl_flags = type; |
1da177e4 | 1440 | |
8634b51f | 1441 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1442 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1443 | if (!ctx) { |
1444 | WARN_ON_ONCE(1); | |
1445 | return error; | |
1446 | } | |
1447 | ||
aba37660 | 1448 | percpu_down_read(&file_rwsem); |
6109c850 | 1449 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1450 | |
c45198ed | 1451 | time_out_leases(inode, &dispose); |
1da177e4 | 1452 | |
03d12ddf | 1453 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1454 | goto out; |
1455 | ||
1da177e4 LT |
1456 | break_time = 0; |
1457 | if (lease_break_time > 0) { | |
1458 | break_time = jiffies + lease_break_time * HZ; | |
1459 | if (break_time == 0) | |
1460 | break_time++; /* so that 0 means no break time */ | |
1461 | } | |
1462 | ||
a901125c | 1463 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1464 | if (!leases_conflict(fl, new_fl)) |
1465 | continue; | |
778fc546 BF |
1466 | if (want_write) { |
1467 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1468 | continue; | |
1469 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1470 | fl->fl_break_time = break_time; |
778fc546 | 1471 | } else { |
8634b51f | 1472 | if (lease_breaking(fl)) |
778fc546 BF |
1473 | continue; |
1474 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1475 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1476 | } |
4d01b7f5 | 1477 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1478 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1479 | } |
1480 | ||
8634b51f | 1481 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1482 | goto out; |
1483 | ||
843c6b2f | 1484 | if (mode & O_NONBLOCK) { |
62af4f1f | 1485 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1486 | error = -EWOULDBLOCK; |
1487 | goto out; | |
1488 | } | |
1489 | ||
1490 | restart: | |
8634b51f JL |
1491 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1492 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1493 | if (break_time != 0) |
1da177e4 | 1494 | break_time -= jiffies; |
f1c6bb2c JL |
1495 | if (break_time == 0) |
1496 | break_time++; | |
8634b51f | 1497 | locks_insert_block(fl, new_fl); |
62af4f1f | 1498 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1499 | spin_unlock(&ctx->flc_lock); |
aba37660 PZ |
1500 | percpu_up_read(&file_rwsem); |
1501 | ||
c45198ed | 1502 | locks_dispose_list(&dispose); |
4321e01e MW |
1503 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1504 | !new_fl->fl_next, break_time); | |
aba37660 PZ |
1505 | |
1506 | percpu_down_read(&file_rwsem); | |
6109c850 | 1507 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1508 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1509 | locks_delete_block(new_fl); |
1da177e4 | 1510 | if (error >= 0) { |
778fc546 BF |
1511 | /* |
1512 | * Wait for the next conflicting lease that has not been | |
1513 | * broken yet | |
1514 | */ | |
03d12ddf JL |
1515 | if (error == 0) |
1516 | time_out_leases(inode, &dispose); | |
1517 | if (any_leases_conflict(inode, new_fl)) | |
1518 | goto restart; | |
1da177e4 LT |
1519 | error = 0; |
1520 | } | |
1da177e4 | 1521 | out: |
6109c850 | 1522 | spin_unlock(&ctx->flc_lock); |
aba37660 | 1523 | percpu_up_read(&file_rwsem); |
c45198ed | 1524 | locks_dispose_list(&dispose); |
6d4b9e38 | 1525 | locks_free_lock(new_fl); |
1da177e4 LT |
1526 | return error; |
1527 | } | |
1528 | ||
1529 | EXPORT_SYMBOL(__break_lease); | |
1530 | ||
1531 | /** | |
a6b91919 | 1532 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1533 | * @inode: the inode |
1534 | * @time: pointer to a timespec which will contain the last modified time | |
1535 | * | |
1536 | * This is to force NFS clients to flush their caches for files with | |
1537 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1538 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1539 | */ |
1540 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1541 | { | |
bfe86024 | 1542 | bool has_lease = false; |
128a3785 | 1543 | struct file_lock_context *ctx; |
8634b51f | 1544 | struct file_lock *fl; |
bfe86024 | 1545 | |
128a3785 | 1546 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1547 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1548 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1549 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1550 | struct file_lock, fl_list); | |
1551 | if (fl && (fl->fl_type == F_WRLCK)) | |
1552 | has_lease = true; | |
6109c850 | 1553 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1554 | } |
1555 | ||
1556 | if (has_lease) | |
1da177e4 LT |
1557 | *time = current_fs_time(inode->i_sb); |
1558 | else | |
1559 | *time = inode->i_mtime; | |
1560 | } | |
1561 | ||
1562 | EXPORT_SYMBOL(lease_get_mtime); | |
1563 | ||
1564 | /** | |
1565 | * fcntl_getlease - Enquire what lease is currently active | |
1566 | * @filp: the file | |
1567 | * | |
1568 | * The value returned by this function will be one of | |
1569 | * (if no lease break is pending): | |
1570 | * | |
1571 | * %F_RDLCK to indicate a shared lease is held. | |
1572 | * | |
1573 | * %F_WRLCK to indicate an exclusive lease is held. | |
1574 | * | |
1575 | * %F_UNLCK to indicate no lease is held. | |
1576 | * | |
1577 | * (if a lease break is pending): | |
1578 | * | |
1579 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1580 | * changed to a shared lease (or removed). | |
1581 | * | |
1582 | * %F_UNLCK to indicate the lease needs to be removed. | |
1583 | * | |
1584 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1585 | * should be returned to userspace. | |
1586 | */ | |
1587 | int fcntl_getlease(struct file *filp) | |
1588 | { | |
1589 | struct file_lock *fl; | |
1c8c601a | 1590 | struct inode *inode = file_inode(filp); |
128a3785 | 1591 | struct file_lock_context *ctx; |
1da177e4 | 1592 | int type = F_UNLCK; |
c45198ed | 1593 | LIST_HEAD(dispose); |
1da177e4 | 1594 | |
128a3785 | 1595 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1596 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1597 | spin_lock(&ctx->flc_lock); |
8634b51f JL |
1598 | time_out_leases(file_inode(filp), &dispose); |
1599 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { | |
1600 | if (fl->fl_file != filp) | |
1601 | continue; | |
778fc546 | 1602 | type = target_leasetype(fl); |
1da177e4 LT |
1603 | break; |
1604 | } | |
6109c850 | 1605 | spin_unlock(&ctx->flc_lock); |
8634b51f | 1606 | locks_dispose_list(&dispose); |
1da177e4 | 1607 | } |
1da177e4 LT |
1608 | return type; |
1609 | } | |
1610 | ||
24cbe784 JL |
1611 | /** |
1612 | * check_conflicting_open - see if the given dentry points to a file that has | |
1613 | * an existing open that would conflict with the | |
1614 | * desired lease. | |
1615 | * @dentry: dentry to check | |
1616 | * @arg: type of lease that we're trying to acquire | |
7fadc59c | 1617 | * @flags: current lock flags |
24cbe784 JL |
1618 | * |
1619 | * Check to see if there's an existing open fd on this file that would | |
1620 | * conflict with the lease we're trying to set. | |
1621 | */ | |
1622 | static int | |
11afe9f7 | 1623 | check_conflicting_open(const struct dentry *dentry, const long arg, int flags) |
24cbe784 JL |
1624 | { |
1625 | int ret = 0; | |
1626 | struct inode *inode = dentry->d_inode; | |
1627 | ||
11afe9f7 CH |
1628 | if (flags & FL_LAYOUT) |
1629 | return 0; | |
1630 | ||
24cbe784 JL |
1631 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
1632 | return -EAGAIN; | |
1633 | ||
1634 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1635 | (atomic_read(&inode->i_count) > 1))) | |
1636 | ret = -EAGAIN; | |
1637 | ||
1638 | return ret; | |
1639 | } | |
1640 | ||
e6f5c789 JL |
1641 | static int |
1642 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1643 | { |
8634b51f | 1644 | struct file_lock *fl, *my_fl = NULL, *lease; |
0f7fc9e4 | 1645 | struct dentry *dentry = filp->f_path.dentry; |
6343a212 | 1646 | struct inode *inode = file_inode(filp); |
8634b51f | 1647 | struct file_lock_context *ctx; |
df4e8d2c | 1648 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1649 | int error; |
c45198ed | 1650 | LIST_HEAD(dispose); |
1da177e4 | 1651 | |
096657b6 | 1652 | lease = *flp; |
62af4f1f JL |
1653 | trace_generic_add_lease(inode, lease); |
1654 | ||
5c1c669a JL |
1655 | /* Note that arg is never F_UNLCK here */ |
1656 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1657 | if (!ctx) |
1658 | return -ENOMEM; | |
1659 | ||
df4e8d2c BF |
1660 | /* |
1661 | * In the delegation case we need mutual exclusion with | |
1662 | * a number of operations that take the i_mutex. We trylock | |
1663 | * because delegations are an optional optimization, and if | |
1664 | * there's some chance of a conflict--we'd rather not | |
1665 | * bother, maybe that's a sign this just isn't a good file to | |
1666 | * hand out a delegation on. | |
1667 | */ | |
5955102c | 1668 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1669 | return -EAGAIN; |
1670 | ||
1671 | if (is_deleg && arg == F_WRLCK) { | |
1672 | /* Write delegations are not currently supported: */ | |
5955102c | 1673 | inode_unlock(inode); |
df4e8d2c BF |
1674 | WARN_ON_ONCE(1); |
1675 | return -EINVAL; | |
1676 | } | |
096657b6 | 1677 | |
aba37660 | 1678 | percpu_down_read(&file_rwsem); |
6109c850 | 1679 | spin_lock(&ctx->flc_lock); |
c45198ed | 1680 | time_out_leases(inode, &dispose); |
11afe9f7 | 1681 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
24cbe784 | 1682 | if (error) |
096657b6 | 1683 | goto out; |
6d5e8b05 | 1684 | |
1da177e4 LT |
1685 | /* |
1686 | * At this point, we know that if there is an exclusive | |
1687 | * lease on this file, then we hold it on this filp | |
1688 | * (otherwise our open of this file would have blocked). | |
1689 | * And if we are trying to acquire an exclusive lease, | |
1690 | * then the file is not open by anyone (including us) | |
1691 | * except for this filp. | |
1692 | */ | |
c1f24ef4 | 1693 | error = -EAGAIN; |
8634b51f | 1694 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1695 | if (fl->fl_file == filp && |
1696 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1697 | my_fl = fl; |
c1f24ef4 BF |
1698 | continue; |
1699 | } | |
8634b51f | 1700 | |
c1f24ef4 BF |
1701 | /* |
1702 | * No exclusive leases if someone else has a lease on | |
1703 | * this file: | |
1704 | */ | |
1705 | if (arg == F_WRLCK) | |
1706 | goto out; | |
1707 | /* | |
1708 | * Modifying our existing lease is OK, but no getting a | |
1709 | * new lease if someone else is opening for write: | |
1710 | */ | |
1711 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1712 | goto out; | |
1da177e4 LT |
1713 | } |
1714 | ||
8634b51f | 1715 | if (my_fl != NULL) { |
0164bf02 JL |
1716 | lease = my_fl; |
1717 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1718 | if (error) |
1719 | goto out; | |
1720 | goto out_setup; | |
1da177e4 LT |
1721 | } |
1722 | ||
1da177e4 LT |
1723 | error = -EINVAL; |
1724 | if (!leases_enable) | |
1725 | goto out; | |
1726 | ||
e084c1bd | 1727 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1728 | /* |
1729 | * The check in break_lease() is lockless. It's possible for another | |
1730 | * open to race in after we did the earlier check for a conflicting | |
1731 | * open but before the lease was inserted. Check again for a | |
1732 | * conflicting open and cancel the lease if there is one. | |
1733 | * | |
1734 | * We also add a barrier here to ensure that the insertion of the lock | |
1735 | * precedes these checks. | |
1736 | */ | |
1737 | smp_mb(); | |
11afe9f7 | 1738 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
8634b51f | 1739 | if (error) { |
e084c1bd | 1740 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1741 | goto out; |
1742 | } | |
1c7dd2ff JL |
1743 | |
1744 | out_setup: | |
1745 | if (lease->fl_lmops->lm_setup) | |
1746 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1747 | out: |
6109c850 | 1748 | spin_unlock(&ctx->flc_lock); |
aba37660 | 1749 | percpu_up_read(&file_rwsem); |
c45198ed | 1750 | locks_dispose_list(&dispose); |
df4e8d2c | 1751 | if (is_deleg) |
5955102c | 1752 | inode_unlock(inode); |
8634b51f | 1753 | if (!error && !my_fl) |
1c7dd2ff | 1754 | *flp = NULL; |
1da177e4 LT |
1755 | return error; |
1756 | } | |
8335ebd9 | 1757 | |
2ab99ee1 | 1758 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1759 | { |
0efaa7e8 | 1760 | int error = -EAGAIN; |
8634b51f | 1761 | struct file_lock *fl, *victim = NULL; |
6ca7d910 | 1762 | struct inode *inode = file_inode(filp); |
128a3785 | 1763 | struct file_lock_context *ctx; |
c45198ed | 1764 | LIST_HEAD(dispose); |
8335ebd9 | 1765 | |
128a3785 | 1766 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1767 | if (!ctx) { |
1768 | trace_generic_delete_lease(inode, NULL); | |
1769 | return error; | |
1770 | } | |
1771 | ||
aba37660 | 1772 | percpu_down_read(&file_rwsem); |
6109c850 | 1773 | spin_lock(&ctx->flc_lock); |
8634b51f | 1774 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1775 | if (fl->fl_file == filp && |
1776 | fl->fl_owner == owner) { | |
8634b51f | 1777 | victim = fl; |
0efaa7e8 | 1778 | break; |
8634b51f | 1779 | } |
8335ebd9 | 1780 | } |
a9b1b455 | 1781 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1782 | if (victim) |
7448cc37 | 1783 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1784 | spin_unlock(&ctx->flc_lock); |
aba37660 | 1785 | percpu_up_read(&file_rwsem); |
c45198ed | 1786 | locks_dispose_list(&dispose); |
0efaa7e8 | 1787 | return error; |
8335ebd9 BF |
1788 | } |
1789 | ||
1790 | /** | |
1791 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1792 | * @filp: file pointer |
1793 | * @arg: type of lease to obtain | |
1794 | * @flp: input - file_lock to use, output - file_lock inserted | |
1795 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1796 | * doesn't require it) | |
8335ebd9 BF |
1797 | * |
1798 | * The (input) flp->fl_lmops->lm_break function is required | |
1799 | * by break_lease(). | |
8335ebd9 | 1800 | */ |
e6f5c789 JL |
1801 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1802 | void **priv) | |
8335ebd9 | 1803 | { |
6ca7d910 | 1804 | struct inode *inode = file_inode(filp); |
8335ebd9 BF |
1805 | int error; |
1806 | ||
8e96e3b7 | 1807 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1808 | return -EACCES; |
1809 | if (!S_ISREG(inode->i_mode)) | |
1810 | return -EINVAL; | |
1811 | error = security_file_lock(filp, arg); | |
1812 | if (error) | |
1813 | return error; | |
1814 | ||
8335ebd9 BF |
1815 | switch (arg) { |
1816 | case F_UNLCK: | |
2ab99ee1 | 1817 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1818 | case F_RDLCK: |
1819 | case F_WRLCK: | |
0efaa7e8 JL |
1820 | if (!(*flp)->fl_lmops->lm_break) { |
1821 | WARN_ON_ONCE(1); | |
1822 | return -ENOLCK; | |
1823 | } | |
11afe9f7 | 1824 | |
e6f5c789 | 1825 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1826 | default: |
8d657eb3 | 1827 | return -EINVAL; |
8335ebd9 BF |
1828 | } |
1829 | } | |
0af1a450 | 1830 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1831 | |
b89f4321 | 1832 | /** |
e51673aa | 1833 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1834 | * @filp: file pointer |
1835 | * @arg: type of lease to obtain | |
1836 | * @lease: file_lock to use when adding a lease | |
1837 | * @priv: private info for lm_setup when adding a lease (may be | |
1838 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1839 | * |
1840 | * Call this to establish a lease on the file. The "lease" argument is not | |
1841 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
1842 | * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set; | |
1843 | * if not, this function will return -ENOLCK (and generate a scary-looking | |
1844 | * stack trace). | |
1c7dd2ff JL |
1845 | * |
1846 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1847 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1848 | */ |
e6f5c789 JL |
1849 | int |
1850 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1851 | { |
1c7dd2ff | 1852 | if (filp->f_op->setlease) |
f82b4b67 | 1853 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1854 | else |
f82b4b67 | 1855 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1856 | } |
a9933cea | 1857 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1858 | |
0ceaf6c7 | 1859 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1860 | { |
1c7dd2ff | 1861 | struct file_lock *fl; |
f7347ce4 | 1862 | struct fasync_struct *new; |
1da177e4 LT |
1863 | int error; |
1864 | ||
c5b1f0d9 AB |
1865 | fl = lease_alloc(filp, arg); |
1866 | if (IS_ERR(fl)) | |
1867 | return PTR_ERR(fl); | |
1da177e4 | 1868 | |
f7347ce4 LT |
1869 | new = fasync_alloc(); |
1870 | if (!new) { | |
1871 | locks_free_lock(fl); | |
1872 | return -ENOMEM; | |
1873 | } | |
1c7dd2ff | 1874 | new->fa_fd = fd; |
f7347ce4 | 1875 | |
1c7dd2ff | 1876 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1877 | if (fl) |
1878 | locks_free_lock(fl); | |
f7347ce4 LT |
1879 | if (new) |
1880 | fasync_free(new); | |
1da177e4 LT |
1881 | return error; |
1882 | } | |
1883 | ||
0ceaf6c7 BF |
1884 | /** |
1885 | * fcntl_setlease - sets a lease on an open file | |
1886 | * @fd: open file descriptor | |
1887 | * @filp: file pointer | |
1888 | * @arg: type of lease to obtain | |
1889 | * | |
1890 | * Call this fcntl to establish a lease on the file. | |
1891 | * Note that you also need to call %F_SETSIG to | |
1892 | * receive a signal when the lease is broken. | |
1893 | */ | |
1894 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1895 | { | |
1896 | if (arg == F_UNLCK) | |
2ab99ee1 | 1897 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
1898 | return do_fcntl_add_lease(fd, filp, arg); |
1899 | } | |
1900 | ||
1da177e4 | 1901 | /** |
29d01b22 JL |
1902 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
1903 | * @inode: inode of the file to apply to | |
1da177e4 LT |
1904 | * @fl: The lock to be applied |
1905 | * | |
29d01b22 | 1906 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 1907 | */ |
616fb38f | 1908 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1909 | { |
1910 | int error; | |
1911 | might_sleep(); | |
1912 | for (;;) { | |
29d01b22 | 1913 | error = flock_lock_inode(inode, fl); |
bde74e4b | 1914 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1915 | break; |
1916 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1917 | if (!error) | |
1918 | continue; | |
1919 | ||
1920 | locks_delete_block(fl); | |
1921 | break; | |
1922 | } | |
1923 | return error; | |
1924 | } | |
1925 | ||
e55c34a6 BC |
1926 | /** |
1927 | * locks_lock_inode_wait - Apply a lock to an inode | |
1928 | * @inode: inode of the file to apply to | |
1929 | * @fl: The lock to be applied | |
1930 | * | |
1931 | * Apply a POSIX or FLOCK style lock request to an inode. | |
1932 | */ | |
1933 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
1934 | { | |
1935 | int res = 0; | |
1936 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
1937 | case FL_POSIX: | |
1938 | res = posix_lock_inode_wait(inode, fl); | |
1939 | break; | |
1940 | case FL_FLOCK: | |
1941 | res = flock_lock_inode_wait(inode, fl); | |
1942 | break; | |
1943 | default: | |
1944 | BUG(); | |
1945 | } | |
1946 | return res; | |
1947 | } | |
1948 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
1949 | ||
1da177e4 LT |
1950 | /** |
1951 | * sys_flock: - flock() system call. | |
1952 | * @fd: the file descriptor to lock. | |
1953 | * @cmd: the type of lock to apply. | |
1954 | * | |
1955 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1956 | * The @cmd can be one of | |
1957 | * | |
1958 | * %LOCK_SH -- a shared lock. | |
1959 | * | |
1960 | * %LOCK_EX -- an exclusive lock. | |
1961 | * | |
1962 | * %LOCK_UN -- remove an existing lock. | |
1963 | * | |
1964 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1965 | * | |
1966 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1967 | * processes read and write access respectively. | |
1968 | */ | |
002c8976 | 1969 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1970 | { |
2903ff01 | 1971 | struct fd f = fdget(fd); |
1da177e4 LT |
1972 | struct file_lock *lock; |
1973 | int can_sleep, unlock; | |
1974 | int error; | |
1975 | ||
1976 | error = -EBADF; | |
2903ff01 | 1977 | if (!f.file) |
1da177e4 LT |
1978 | goto out; |
1979 | ||
1980 | can_sleep = !(cmd & LOCK_NB); | |
1981 | cmd &= ~LOCK_NB; | |
1982 | unlock = (cmd == LOCK_UN); | |
1983 | ||
aeb5d727 | 1984 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1985 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1986 | goto out_putf; |
1987 | ||
6e129d00 JL |
1988 | lock = flock_make_lock(f.file, cmd); |
1989 | if (IS_ERR(lock)) { | |
1990 | error = PTR_ERR(lock); | |
1da177e4 | 1991 | goto out_putf; |
6e129d00 JL |
1992 | } |
1993 | ||
1da177e4 LT |
1994 | if (can_sleep) |
1995 | lock->fl_flags |= FL_SLEEP; | |
1996 | ||
2903ff01 | 1997 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1998 | if (error) |
1999 | goto out_free; | |
2000 | ||
72c2d531 | 2001 | if (f.file->f_op->flock) |
2903ff01 | 2002 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
2003 | (can_sleep) ? F_SETLKW : F_SETLK, |
2004 | lock); | |
2005 | else | |
4f656367 | 2006 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
2007 | |
2008 | out_free: | |
993dfa87 | 2009 | locks_free_lock(lock); |
1da177e4 LT |
2010 | |
2011 | out_putf: | |
2903ff01 | 2012 | fdput(f); |
1da177e4 LT |
2013 | out: |
2014 | return error; | |
2015 | } | |
2016 | ||
3ee17abd BF |
2017 | /** |
2018 | * vfs_test_lock - test file byte range lock | |
2019 | * @filp: The file to test lock for | |
6924c554 | 2020 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2021 | * |
2022 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2023 | * setting conf->fl_type to something other than F_UNLCK. | |
2024 | */ | |
2025 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2026 | { | |
72c2d531 | 2027 | if (filp->f_op->lock) |
3ee17abd BF |
2028 | return filp->f_op->lock(filp, F_GETLK, fl); |
2029 | posix_test_lock(filp, fl); | |
2030 | return 0; | |
2031 | } | |
2032 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2033 | ||
c2fa1b8a BF |
2034 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2035 | { | |
cff2fce5 | 2036 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2037 | #if BITS_PER_LONG == 32 |
2038 | /* | |
2039 | * Make sure we can represent the posix lock via | |
2040 | * legacy 32bit flock. | |
2041 | */ | |
2042 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2043 | return -EOVERFLOW; | |
2044 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2045 | return -EOVERFLOW; | |
2046 | #endif | |
2047 | flock->l_start = fl->fl_start; | |
2048 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2049 | fl->fl_end - fl->fl_start + 1; | |
2050 | flock->l_whence = 0; | |
129a84de | 2051 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2052 | return 0; |
2053 | } | |
2054 | ||
2055 | #if BITS_PER_LONG == 32 | |
2056 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2057 | { | |
cff2fce5 | 2058 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2059 | flock->l_start = fl->fl_start; |
2060 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2061 | fl->fl_end - fl->fl_start + 1; | |
2062 | flock->l_whence = 0; | |
2063 | flock->l_type = fl->fl_type; | |
2064 | } | |
2065 | #endif | |
2066 | ||
1da177e4 LT |
2067 | /* Report the first existing lock that would conflict with l. |
2068 | * This implements the F_GETLK command of fcntl(). | |
2069 | */ | |
c1e62b8f | 2070 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 2071 | { |
9d6a8c5c | 2072 | struct file_lock file_lock; |
1da177e4 LT |
2073 | struct flock flock; |
2074 | int error; | |
2075 | ||
2076 | error = -EFAULT; | |
2077 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2078 | goto out; | |
2079 | error = -EINVAL; | |
2080 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2081 | goto out; | |
2082 | ||
2083 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
2084 | if (error) | |
2085 | goto out; | |
2086 | ||
0d3f7a2d | 2087 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2088 | error = -EINVAL; |
2089 | if (flock.l_pid != 0) | |
2090 | goto out; | |
2091 | ||
5d50ffd7 | 2092 | cmd = F_GETLK; |
cff2fce5 | 2093 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2094 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2095 | } |
2096 | ||
3ee17abd BF |
2097 | error = vfs_test_lock(filp, &file_lock); |
2098 | if (error) | |
2099 | goto out; | |
1da177e4 | 2100 | |
9d6a8c5c ME |
2101 | flock.l_type = file_lock.fl_type; |
2102 | if (file_lock.fl_type != F_UNLCK) { | |
2103 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 2104 | if (error) |
f328296e | 2105 | goto rel_priv; |
1da177e4 LT |
2106 | } |
2107 | error = -EFAULT; | |
2108 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2109 | error = 0; | |
f328296e KM |
2110 | rel_priv: |
2111 | locks_release_private(&file_lock); | |
1da177e4 LT |
2112 | out: |
2113 | return error; | |
2114 | } | |
2115 | ||
7723ec97 ME |
2116 | /** |
2117 | * vfs_lock_file - file byte range lock | |
2118 | * @filp: The file to apply the lock to | |
2119 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2120 | * @fl: The lock to be applied | |
150b3934 ME |
2121 | * @conf: Place to return a copy of the conflicting lock, if found. |
2122 | * | |
2123 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2124 | * as the final argument. | |
2125 | * | |
2126 | * If the filesystem defines a private ->lock() method, then @conf will | |
2127 | * be left unchanged; so a caller that cares should initialize it to | |
2128 | * some acceptable default. | |
2beb6614 ME |
2129 | * |
2130 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2131 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2132 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2133 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2134 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2135 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2136 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2137 | * request completes. |
2138 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2139 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2140 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2141 | * nonzero return code and the file system should release the lock. The file |
2142 | * system is also responsible to keep a corresponding posix lock when it | |
2143 | * grants a lock so the VFS can find out which locks are locally held and do | |
2144 | * the correct lock cleanup when required. | |
2145 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2146 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2147 | * return code. |
7723ec97 | 2148 | */ |
150b3934 | 2149 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2150 | { |
72c2d531 | 2151 | if (filp->f_op->lock) |
7723ec97 ME |
2152 | return filp->f_op->lock(filp, cmd, fl); |
2153 | else | |
150b3934 | 2154 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2155 | } |
2156 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2157 | ||
b648a6de MS |
2158 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2159 | struct file_lock *fl) | |
2160 | { | |
2161 | int error; | |
2162 | ||
2163 | error = security_file_lock(filp, fl->fl_type); | |
2164 | if (error) | |
2165 | return error; | |
2166 | ||
764c76b3 MS |
2167 | for (;;) { |
2168 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2169 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2170 | break; |
764c76b3 MS |
2171 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2172 | if (!error) | |
2173 | continue; | |
2174 | ||
2175 | locks_delete_block(fl); | |
2176 | break; | |
b648a6de MS |
2177 | } |
2178 | ||
2179 | return error; | |
2180 | } | |
2181 | ||
6ca7d910 | 2182 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2183 | static int |
2184 | check_fmode_for_setlk(struct file_lock *fl) | |
2185 | { | |
2186 | switch (fl->fl_type) { | |
2187 | case F_RDLCK: | |
2188 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2189 | return -EBADF; | |
2190 | break; | |
2191 | case F_WRLCK: | |
2192 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2193 | return -EBADF; | |
2194 | } | |
2195 | return 0; | |
2196 | } | |
2197 | ||
1da177e4 LT |
2198 | /* Apply the lock described by l to an open file descriptor. |
2199 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2200 | */ | |
c293621b PS |
2201 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2202 | struct flock __user *l) | |
1da177e4 LT |
2203 | { |
2204 | struct file_lock *file_lock = locks_alloc_lock(); | |
2205 | struct flock flock; | |
2206 | struct inode *inode; | |
0b2bac2f | 2207 | struct file *f; |
1da177e4 LT |
2208 | int error; |
2209 | ||
2210 | if (file_lock == NULL) | |
2211 | return -ENOLCK; | |
2212 | ||
1890910f JL |
2213 | inode = file_inode(filp); |
2214 | ||
1da177e4 LT |
2215 | /* |
2216 | * This might block, so we do it before checking the inode. | |
2217 | */ | |
2218 | error = -EFAULT; | |
2219 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2220 | goto out; | |
2221 | ||
1da177e4 LT |
2222 | /* Don't allow mandatory locks on files that may be memory mapped |
2223 | * and shared. | |
2224 | */ | |
a16877ca | 2225 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2226 | error = -EAGAIN; |
2227 | goto out; | |
2228 | } | |
2229 | ||
2230 | error = flock_to_posix_lock(filp, file_lock, &flock); | |
2231 | if (error) | |
2232 | goto out; | |
5d50ffd7 | 2233 | |
cf01f4ee JL |
2234 | error = check_fmode_for_setlk(file_lock); |
2235 | if (error) | |
2236 | goto out; | |
2237 | ||
5d50ffd7 JL |
2238 | /* |
2239 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2240 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2241 | */ |
2242 | switch (cmd) { | |
0d3f7a2d | 2243 | case F_OFD_SETLK: |
90478939 JL |
2244 | error = -EINVAL; |
2245 | if (flock.l_pid != 0) | |
2246 | goto out; | |
2247 | ||
5d50ffd7 | 2248 | cmd = F_SETLK; |
cff2fce5 | 2249 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2250 | file_lock->fl_owner = filp; |
5d50ffd7 | 2251 | break; |
0d3f7a2d | 2252 | case F_OFD_SETLKW: |
90478939 JL |
2253 | error = -EINVAL; |
2254 | if (flock.l_pid != 0) | |
2255 | goto out; | |
2256 | ||
5d50ffd7 | 2257 | cmd = F_SETLKW; |
cff2fce5 | 2258 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2259 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2260 | /* Fallthrough */ |
2261 | case F_SETLKW: | |
1da177e4 LT |
2262 | file_lock->fl_flags |= FL_SLEEP; |
2263 | } | |
5d50ffd7 | 2264 | |
b648a6de | 2265 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2266 | |
c293621b | 2267 | /* |
0752ba80 JL |
2268 | * Attempt to detect a close/fcntl race and recover by releasing the |
2269 | * lock that was just acquired. There is no need to do that when we're | |
2270 | * unlocking though, or for OFD locks. | |
c293621b | 2271 | */ |
0752ba80 JL |
2272 | if (!error && file_lock->fl_type != F_UNLCK && |
2273 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2274 | /* |
2275 | * We need that spin_lock here - it prevents reordering between | |
2276 | * update of i_flctx->flc_posix and check for it done in | |
2277 | * close(). rcu_read_lock() wouldn't do. | |
2278 | */ | |
2279 | spin_lock(¤t->files->file_lock); | |
2280 | f = fcheck(fd); | |
2281 | spin_unlock(¤t->files->file_lock); | |
2282 | if (f != filp) { | |
2283 | file_lock->fl_type = F_UNLCK; | |
2284 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2285 | WARN_ON_ONCE(error); | |
2286 | error = -EBADF; | |
2287 | } | |
1da177e4 | 2288 | } |
c293621b | 2289 | out: |
1890910f | 2290 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2291 | locks_free_lock(file_lock); |
2292 | return error; | |
2293 | } | |
2294 | ||
2295 | #if BITS_PER_LONG == 32 | |
2296 | /* Report the first existing lock that would conflict with l. | |
2297 | * This implements the F_GETLK command of fcntl(). | |
2298 | */ | |
c1e62b8f | 2299 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2300 | { |
9d6a8c5c | 2301 | struct file_lock file_lock; |
1da177e4 LT |
2302 | struct flock64 flock; |
2303 | int error; | |
2304 | ||
2305 | error = -EFAULT; | |
2306 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2307 | goto out; | |
2308 | error = -EINVAL; | |
2309 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2310 | goto out; | |
2311 | ||
2312 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2313 | if (error) | |
2314 | goto out; | |
2315 | ||
0d3f7a2d | 2316 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2317 | error = -EINVAL; |
2318 | if (flock.l_pid != 0) | |
2319 | goto out; | |
2320 | ||
5d50ffd7 | 2321 | cmd = F_GETLK64; |
cff2fce5 | 2322 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2323 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2324 | } |
2325 | ||
3ee17abd BF |
2326 | error = vfs_test_lock(filp, &file_lock); |
2327 | if (error) | |
2328 | goto out; | |
2329 | ||
9d6a8c5c ME |
2330 | flock.l_type = file_lock.fl_type; |
2331 | if (file_lock.fl_type != F_UNLCK) | |
2332 | posix_lock_to_flock64(&flock, &file_lock); | |
2333 | ||
1da177e4 LT |
2334 | error = -EFAULT; |
2335 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2336 | error = 0; | |
f328296e KM |
2337 | |
2338 | locks_release_private(&file_lock); | |
1da177e4 LT |
2339 | out: |
2340 | return error; | |
2341 | } | |
2342 | ||
2343 | /* Apply the lock described by l to an open file descriptor. | |
2344 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2345 | */ | |
c293621b PS |
2346 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2347 | struct flock64 __user *l) | |
1da177e4 LT |
2348 | { |
2349 | struct file_lock *file_lock = locks_alloc_lock(); | |
2350 | struct flock64 flock; | |
2351 | struct inode *inode; | |
0b2bac2f | 2352 | struct file *f; |
1da177e4 LT |
2353 | int error; |
2354 | ||
2355 | if (file_lock == NULL) | |
2356 | return -ENOLCK; | |
2357 | ||
2358 | /* | |
2359 | * This might block, so we do it before checking the inode. | |
2360 | */ | |
2361 | error = -EFAULT; | |
2362 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2363 | goto out; | |
2364 | ||
496ad9aa | 2365 | inode = file_inode(filp); |
1da177e4 LT |
2366 | |
2367 | /* Don't allow mandatory locks on files that may be memory mapped | |
2368 | * and shared. | |
2369 | */ | |
a16877ca | 2370 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2371 | error = -EAGAIN; |
2372 | goto out; | |
2373 | } | |
2374 | ||
2375 | error = flock64_to_posix_lock(filp, file_lock, &flock); | |
2376 | if (error) | |
2377 | goto out; | |
5d50ffd7 | 2378 | |
cf01f4ee JL |
2379 | error = check_fmode_for_setlk(file_lock); |
2380 | if (error) | |
2381 | goto out; | |
2382 | ||
5d50ffd7 JL |
2383 | /* |
2384 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2385 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2386 | */ |
2387 | switch (cmd) { | |
0d3f7a2d | 2388 | case F_OFD_SETLK: |
90478939 JL |
2389 | error = -EINVAL; |
2390 | if (flock.l_pid != 0) | |
2391 | goto out; | |
2392 | ||
5d50ffd7 | 2393 | cmd = F_SETLK64; |
cff2fce5 | 2394 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2395 | file_lock->fl_owner = filp; |
5d50ffd7 | 2396 | break; |
0d3f7a2d | 2397 | case F_OFD_SETLKW: |
90478939 JL |
2398 | error = -EINVAL; |
2399 | if (flock.l_pid != 0) | |
2400 | goto out; | |
2401 | ||
5d50ffd7 | 2402 | cmd = F_SETLKW64; |
cff2fce5 | 2403 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2404 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2405 | /* Fallthrough */ |
2406 | case F_SETLKW64: | |
1da177e4 LT |
2407 | file_lock->fl_flags |= FL_SLEEP; |
2408 | } | |
5d50ffd7 | 2409 | |
b648a6de | 2410 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2411 | |
c293621b | 2412 | /* |
0752ba80 JL |
2413 | * Attempt to detect a close/fcntl race and recover by releasing the |
2414 | * lock that was just acquired. There is no need to do that when we're | |
2415 | * unlocking though, or for OFD locks. | |
c293621b | 2416 | */ |
0752ba80 JL |
2417 | if (!error && file_lock->fl_type != F_UNLCK && |
2418 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2419 | /* |
2420 | * We need that spin_lock here - it prevents reordering between | |
2421 | * update of i_flctx->flc_posix and check for it done in | |
2422 | * close(). rcu_read_lock() wouldn't do. | |
2423 | */ | |
2424 | spin_lock(¤t->files->file_lock); | |
2425 | f = fcheck(fd); | |
2426 | spin_unlock(¤t->files->file_lock); | |
2427 | if (f != filp) { | |
2428 | file_lock->fl_type = F_UNLCK; | |
2429 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2430 | WARN_ON_ONCE(error); | |
2431 | error = -EBADF; | |
2432 | } | |
1da177e4 | 2433 | } |
1da177e4 LT |
2434 | out: |
2435 | locks_free_lock(file_lock); | |
2436 | return error; | |
2437 | } | |
2438 | #endif /* BITS_PER_LONG == 32 */ | |
2439 | ||
2440 | /* | |
2441 | * This function is called when the file is being removed | |
2442 | * from the task's fd array. POSIX locks belonging to this task | |
2443 | * are deleted at this time. | |
2444 | */ | |
2445 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2446 | { | |
1890910f | 2447 | int error; |
ff7b86b8 | 2448 | struct file_lock lock; |
128a3785 | 2449 | struct file_lock_context *ctx; |
1da177e4 LT |
2450 | |
2451 | /* | |
2452 | * If there are no locks held on this file, we don't need to call | |
2453 | * posix_lock_file(). Another process could be setting a lock on this | |
2454 | * file at the same time, but we wouldn't remove that lock anyway. | |
2455 | */ | |
128a3785 | 2456 | ctx = smp_load_acquire(&file_inode(filp)->i_flctx); |
bd61e0a9 | 2457 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2458 | return; |
2459 | ||
2460 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2461 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2462 | lock.fl_start = 0; |
2463 | lock.fl_end = OFFSET_MAX; | |
2464 | lock.fl_owner = owner; | |
2465 | lock.fl_pid = current->tgid; | |
2466 | lock.fl_file = filp; | |
2467 | lock.fl_ops = NULL; | |
2468 | lock.fl_lmops = NULL; | |
2469 | ||
1890910f | 2470 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2471 | |
1da177e4 LT |
2472 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2473 | lock.fl_ops->fl_release_private(&lock); | |
1890910f | 2474 | trace_locks_remove_posix(file_inode(filp), &lock, error); |
1da177e4 LT |
2475 | } |
2476 | ||
2477 | EXPORT_SYMBOL(locks_remove_posix); | |
2478 | ||
3d8e560d | 2479 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2480 | static void |
128a3785 | 2481 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 JL |
2482 | { |
2483 | struct file_lock fl = { | |
2484 | .fl_owner = filp, | |
2485 | .fl_pid = current->tgid, | |
2486 | .fl_file = filp, | |
2487 | .fl_flags = FL_FLOCK, | |
2488 | .fl_type = F_UNLCK, | |
2489 | .fl_end = OFFSET_MAX, | |
2490 | }; | |
bcd7f78d | 2491 | struct inode *inode = file_inode(filp); |
dd459bb1 | 2492 | |
3d8e560d | 2493 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2494 | return; |
2495 | ||
2496 | if (filp->f_op->flock) | |
2497 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
2498 | else | |
bcd7f78d | 2499 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2500 | |
2501 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2502 | fl.fl_ops->fl_release_private(&fl); | |
2503 | } | |
2504 | ||
3d8e560d | 2505 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2506 | static void |
128a3785 | 2507 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2508 | { |
8634b51f JL |
2509 | struct file_lock *fl, *tmp; |
2510 | LIST_HEAD(dispose); | |
2511 | ||
3d8e560d | 2512 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2513 | return; |
2514 | ||
6109c850 | 2515 | spin_lock(&ctx->flc_lock); |
8634b51f | 2516 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2517 | if (filp == fl->fl_file) |
2518 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2519 | spin_unlock(&ctx->flc_lock); |
8634b51f JL |
2520 | locks_dispose_list(&dispose); |
2521 | } | |
2522 | ||
1da177e4 LT |
2523 | /* |
2524 | * This function is called on the last close of an open file. | |
2525 | */ | |
78ed8a13 | 2526 | void locks_remove_file(struct file *filp) |
1da177e4 | 2527 | { |
128a3785 DV |
2528 | struct file_lock_context *ctx; |
2529 | ||
2530 | ctx = smp_load_acquire(&file_inode(filp)->i_flctx); | |
2531 | if (!ctx) | |
3d8e560d JL |
2532 | return; |
2533 | ||
dd459bb1 | 2534 | /* remove any OFD locks */ |
73a8f5f7 | 2535 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2536 | |
dd459bb1 | 2537 | /* remove flock locks */ |
128a3785 | 2538 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2539 | |
8634b51f | 2540 | /* remove any leases */ |
128a3785 | 2541 | locks_remove_lease(filp, ctx); |
1da177e4 LT |
2542 | } |
2543 | ||
1da177e4 LT |
2544 | /** |
2545 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2546 | * @waiter: the lock which was waiting |
2547 | * | |
2548 | * lockd needs to block waiting for locks. | |
2549 | */ | |
64a318ee | 2550 | int |
f891a29f | 2551 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2552 | { |
64a318ee BF |
2553 | int status = 0; |
2554 | ||
7b2296af | 2555 | spin_lock(&blocked_lock_lock); |
5996a298 | 2556 | if (waiter->fl_next) |
1da177e4 | 2557 | __locks_delete_block(waiter); |
64a318ee BF |
2558 | else |
2559 | status = -ENOENT; | |
7b2296af | 2560 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2561 | return status; |
1da177e4 | 2562 | } |
1da177e4 LT |
2563 | EXPORT_SYMBOL(posix_unblock_lock); |
2564 | ||
9b9d2ab4 ME |
2565 | /** |
2566 | * vfs_cancel_lock - file byte range unblock lock | |
2567 | * @filp: The file to apply the unblock to | |
2568 | * @fl: The lock to be unblocked | |
2569 | * | |
2570 | * Used by lock managers to cancel blocked requests | |
2571 | */ | |
2572 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2573 | { | |
72c2d531 | 2574 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2575 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2576 | return 0; | |
2577 | } | |
2578 | ||
2579 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2580 | ||
7f8ada98 | 2581 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2582 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2583 | #include <linux/seq_file.h> |
2584 | ||
7012b02a JL |
2585 | struct locks_iterator { |
2586 | int li_cpu; | |
2587 | loff_t li_pos; | |
2588 | }; | |
2589 | ||
7f8ada98 | 2590 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2591 | loff_t id, char *pfx) |
1da177e4 LT |
2592 | { |
2593 | struct inode *inode = NULL; | |
ab1f1611 VG |
2594 | unsigned int fl_pid; |
2595 | ||
2596 | if (fl->fl_nspid) | |
6c5f3e7b | 2597 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2598 | else |
2599 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2600 | |
2601 | if (fl->fl_file != NULL) | |
496ad9aa | 2602 | inode = file_inode(fl->fl_file); |
1da177e4 | 2603 | |
99dc8292 | 2604 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2605 | if (IS_POSIX(fl)) { |
c918d42a | 2606 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2607 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2608 | else if (IS_OFDLCK(fl)) |
5315c26a | 2609 | seq_puts(f, "OFDLCK"); |
c918d42a | 2610 | else |
5315c26a | 2611 | seq_puts(f, "POSIX "); |
c918d42a JL |
2612 | |
2613 | seq_printf(f, " %s ", | |
1da177e4 | 2614 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2615 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2616 | } else if (IS_FLOCK(fl)) { |
2617 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2618 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2619 | } else { |
5315c26a | 2620 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2621 | } |
2622 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2623 | if (fl->fl_flags & FL_DELEG) |
2624 | seq_puts(f, "DELEG "); | |
2625 | else | |
2626 | seq_puts(f, "LEASE "); | |
2627 | ||
ab83fa4b | 2628 | if (lease_breaking(fl)) |
5315c26a | 2629 | seq_puts(f, "BREAKING "); |
1da177e4 | 2630 | else if (fl->fl_file) |
5315c26a | 2631 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2632 | else |
5315c26a | 2633 | seq_puts(f, "BREAKER "); |
1da177e4 | 2634 | } else { |
5315c26a | 2635 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2636 | } |
2637 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2638 | seq_printf(f, "%s ", |
1da177e4 LT |
2639 | (fl->fl_type & LOCK_READ) |
2640 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2641 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2642 | } else { | |
7f8ada98 | 2643 | seq_printf(f, "%s ", |
ab83fa4b | 2644 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2645 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2646 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2647 | } |
2648 | if (inode) { | |
3648888e | 2649 | /* userspace relies on this representation of dev_t */ |
ab1f1611 | 2650 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2651 | MAJOR(inode->i_sb->s_dev), |
2652 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2653 | } else { |
ab1f1611 | 2654 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2655 | } |
2656 | if (IS_POSIX(fl)) { | |
2657 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2658 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2659 | else |
7f8ada98 | 2660 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2661 | } else { |
5315c26a | 2662 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2663 | } |
2664 | } | |
2665 | ||
7f8ada98 | 2666 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2667 | { |
7012b02a | 2668 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2669 | struct file_lock *fl, *bfl; |
1da177e4 | 2670 | |
139ca04e | 2671 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2672 | |
7012b02a | 2673 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2674 | |
7f8ada98 | 2675 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2676 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2677 | |
7f8ada98 PE |
2678 | return 0; |
2679 | } | |
1da177e4 | 2680 | |
6c8c9031 AV |
2681 | static void __show_fd_locks(struct seq_file *f, |
2682 | struct list_head *head, int *id, | |
2683 | struct file *filp, struct files_struct *files) | |
2684 | { | |
2685 | struct file_lock *fl; | |
2686 | ||
2687 | list_for_each_entry(fl, head, fl_list) { | |
2688 | ||
2689 | if (filp != fl->fl_file) | |
2690 | continue; | |
2691 | if (fl->fl_owner != files && | |
2692 | fl->fl_owner != filp) | |
2693 | continue; | |
2694 | ||
2695 | (*id)++; | |
2696 | seq_puts(f, "lock:\t"); | |
2697 | lock_get_status(f, fl, *id, ""); | |
2698 | } | |
2699 | } | |
2700 | ||
2701 | void show_fd_locks(struct seq_file *f, | |
2702 | struct file *filp, struct files_struct *files) | |
2703 | { | |
2704 | struct inode *inode = file_inode(filp); | |
2705 | struct file_lock_context *ctx; | |
2706 | int id = 0; | |
2707 | ||
128a3785 | 2708 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2709 | if (!ctx) |
2710 | return; | |
2711 | ||
2712 | spin_lock(&ctx->flc_lock); | |
2713 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2714 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2715 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2716 | spin_unlock(&ctx->flc_lock); | |
2717 | } | |
2718 | ||
7f8ada98 | 2719 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2720 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2721 | { |
7012b02a | 2722 | struct locks_iterator *iter = f->private; |
99dc8292 | 2723 | |
7012b02a | 2724 | iter->li_pos = *pos + 1; |
aba37660 | 2725 | percpu_down_write(&file_rwsem); |
7012b02a | 2726 | lg_global_lock(&file_lock_lglock); |
7b2296af | 2727 | spin_lock(&blocked_lock_lock); |
7012b02a | 2728 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2729 | } |
1da177e4 | 2730 | |
7f8ada98 PE |
2731 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2732 | { | |
7012b02a JL |
2733 | struct locks_iterator *iter = f->private; |
2734 | ||
2735 | ++iter->li_pos; | |
2736 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2737 | } |
1da177e4 | 2738 | |
7f8ada98 | 2739 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2740 | __releases(&blocked_lock_lock) |
7f8ada98 | 2741 | { |
7b2296af | 2742 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2743 | lg_global_unlock(&file_lock_lglock); |
aba37660 | 2744 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2745 | } |
2746 | ||
d8ba7a36 | 2747 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2748 | .start = locks_start, |
2749 | .next = locks_next, | |
2750 | .stop = locks_stop, | |
2751 | .show = locks_show, | |
2752 | }; | |
d8ba7a36 AD |
2753 | |
2754 | static int locks_open(struct inode *inode, struct file *filp) | |
2755 | { | |
7012b02a JL |
2756 | return seq_open_private(filp, &locks_seq_operations, |
2757 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2758 | } |
2759 | ||
2760 | static const struct file_operations proc_locks_operations = { | |
2761 | .open = locks_open, | |
2762 | .read = seq_read, | |
2763 | .llseek = seq_lseek, | |
99dc8292 | 2764 | .release = seq_release_private, |
d8ba7a36 AD |
2765 | }; |
2766 | ||
2767 | static int __init proc_locks_init(void) | |
2768 | { | |
2769 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2770 | return 0; | |
2771 | } | |
91899226 | 2772 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2773 | #endif |
2774 | ||
1da177e4 LT |
2775 | static int __init filelock_init(void) |
2776 | { | |
7012b02a JL |
2777 | int i; |
2778 | ||
4a075e39 JL |
2779 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2780 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2781 | ||
1da177e4 | 2782 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2783 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2784 | ||
7012b02a JL |
2785 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2786 | ||
2787 | for_each_possible_cpu(i) | |
2788 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2789 | ||
1da177e4 LT |
2790 | return 0; |
2791 | } | |
2792 | ||
2793 | core_initcall(filelock_init); |