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); | |
88974691 | 167 | |
1c8c601a | 168 | /* |
48f74186 | 169 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 170 | * It is protected by blocked_lock_lock. |
48f74186 JL |
171 | * |
172 | * We hash locks by lockowner in order to optimize searching for the lock a | |
173 | * particular lockowner is waiting on. | |
174 | * | |
175 | * FIXME: make this value scale via some heuristic? We generally will want more | |
176 | * buckets when we have more lockowners holding locks, but that's a little | |
177 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 178 | */ |
48f74186 JL |
179 | #define BLOCKED_HASH_BITS 7 |
180 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 181 | |
1c8c601a | 182 | /* |
7b2296af JL |
183 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
184 | * want to be holding this lock. | |
1c8c601a JL |
185 | * |
186 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
187 | * pointer for file_lock structures that are acting as lock requests (in | |
188 | * contrast to those that are acting as records of acquired locks). | |
189 | * | |
190 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 191 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 192 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 193 | * flc_lock. |
1c8c601a JL |
194 | * |
195 | * In particular, adding an entry to the fl_block list requires that you hold | |
6109c850 JL |
196 | * both the flc_lock and the blocked_lock_lock (acquired in that order). |
197 | * Deleting an entry from the list however only requires the file_lock_lock. | |
1c8c601a | 198 | */ |
7b2296af | 199 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 200 | |
4a075e39 | 201 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 202 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 203 | |
4a075e39 | 204 | static struct file_lock_context * |
5c1c669a | 205 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 206 | { |
128a3785 | 207 | struct file_lock_context *ctx; |
4a075e39 | 208 | |
128a3785 DV |
209 | /* paired with cmpxchg() below */ |
210 | ctx = smp_load_acquire(&inode->i_flctx); | |
211 | if (likely(ctx) || type == F_UNLCK) | |
4a075e39 JL |
212 | goto out; |
213 | ||
128a3785 DV |
214 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
215 | if (!ctx) | |
4a075e39 JL |
216 | goto out; |
217 | ||
128a3785 DV |
218 | spin_lock_init(&ctx->flc_lock); |
219 | INIT_LIST_HEAD(&ctx->flc_flock); | |
220 | INIT_LIST_HEAD(&ctx->flc_posix); | |
221 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
222 | |
223 | /* | |
224 | * Assign the pointer if it's not already assigned. If it is, then | |
225 | * free the context we just allocated. | |
226 | */ | |
128a3785 DV |
227 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
228 | kmem_cache_free(flctx_cache, ctx); | |
229 | ctx = smp_load_acquire(&inode->i_flctx); | |
230 | } | |
4a075e39 | 231 | out: |
128a3785 | 232 | return ctx; |
4a075e39 JL |
233 | } |
234 | ||
235 | void | |
236 | locks_free_lock_context(struct file_lock_context *ctx) | |
237 | { | |
238 | if (ctx) { | |
239 | WARN_ON_ONCE(!list_empty(&ctx->flc_flock)); | |
bd61e0a9 | 240 | WARN_ON_ONCE(!list_empty(&ctx->flc_posix)); |
8634b51f | 241 | WARN_ON_ONCE(!list_empty(&ctx->flc_lease)); |
4a075e39 JL |
242 | kmem_cache_free(flctx_cache, ctx); |
243 | } | |
244 | } | |
245 | ||
ee19cc40 | 246 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 247 | { |
139ca04e | 248 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 249 | INIT_LIST_HEAD(&fl->fl_list); |
ee19cc40 MS |
250 | INIT_LIST_HEAD(&fl->fl_block); |
251 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
252 | } |
253 | ||
1da177e4 | 254 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 255 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 256 | { |
ee19cc40 | 257 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
258 | |
259 | if (fl) | |
ee19cc40 | 260 | locks_init_lock_heads(fl); |
a51cb91d MS |
261 | |
262 | return fl; | |
1da177e4 | 263 | } |
c5b1f0d9 | 264 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 265 | |
a9e61e25 | 266 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
267 | { |
268 | if (fl->fl_ops) { | |
269 | if (fl->fl_ops->fl_release_private) | |
270 | fl->fl_ops->fl_release_private(fl); | |
271 | fl->fl_ops = NULL; | |
272 | } | |
47831f35 | 273 | |
5c97d7b1 | 274 | if (fl->fl_lmops) { |
cae80b30 JL |
275 | if (fl->fl_lmops->lm_put_owner) { |
276 | fl->fl_lmops->lm_put_owner(fl->fl_owner); | |
277 | fl->fl_owner = NULL; | |
278 | } | |
5c97d7b1 KM |
279 | fl->fl_lmops = NULL; |
280 | } | |
47831f35 | 281 | } |
a9e61e25 | 282 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 283 | |
1da177e4 | 284 | /* Free a lock which is not in use. */ |
05fa3135 | 285 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 286 | { |
5ce29646 | 287 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 288 | BUG_ON(!list_empty(&fl->fl_list)); |
5ce29646 | 289 | BUG_ON(!list_empty(&fl->fl_block)); |
139ca04e | 290 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 291 | |
47831f35 | 292 | locks_release_private(fl); |
1da177e4 LT |
293 | kmem_cache_free(filelock_cache, fl); |
294 | } | |
05fa3135 | 295 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 296 | |
ed9814d8 JL |
297 | static void |
298 | locks_dispose_list(struct list_head *dispose) | |
299 | { | |
300 | struct file_lock *fl; | |
301 | ||
302 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
303 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
304 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
305 | locks_free_lock(fl); |
306 | } | |
307 | } | |
308 | ||
1da177e4 LT |
309 | void locks_init_lock(struct file_lock *fl) |
310 | { | |
ee19cc40 MS |
311 | memset(fl, 0, sizeof(struct file_lock)); |
312 | locks_init_lock_heads(fl); | |
1da177e4 LT |
313 | } |
314 | ||
315 | EXPORT_SYMBOL(locks_init_lock); | |
316 | ||
1da177e4 LT |
317 | /* |
318 | * Initialize a new lock from an existing file_lock structure. | |
319 | */ | |
3fe0fff1 | 320 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
321 | { |
322 | new->fl_owner = fl->fl_owner; | |
323 | new->fl_pid = fl->fl_pid; | |
0996905f | 324 | new->fl_file = NULL; |
1da177e4 LT |
325 | new->fl_flags = fl->fl_flags; |
326 | new->fl_type = fl->fl_type; | |
327 | new->fl_start = fl->fl_start; | |
328 | new->fl_end = fl->fl_end; | |
f328296e | 329 | new->fl_lmops = fl->fl_lmops; |
0996905f | 330 | new->fl_ops = NULL; |
f328296e KM |
331 | |
332 | if (fl->fl_lmops) { | |
333 | if (fl->fl_lmops->lm_get_owner) | |
cae80b30 | 334 | fl->fl_lmops->lm_get_owner(fl->fl_owner); |
f328296e | 335 | } |
0996905f | 336 | } |
3fe0fff1 | 337 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
338 | |
339 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
340 | { | |
566709bd JL |
341 | /* "new" must be a freshly-initialized lock */ |
342 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 343 | |
3fe0fff1 | 344 | locks_copy_conflock(new, fl); |
f328296e | 345 | |
0996905f | 346 | new->fl_file = fl->fl_file; |
1da177e4 | 347 | new->fl_ops = fl->fl_ops; |
47831f35 | 348 | |
f328296e KM |
349 | if (fl->fl_ops) { |
350 | if (fl->fl_ops->fl_copy_lock) | |
351 | fl->fl_ops->fl_copy_lock(new, fl); | |
352 | } | |
1da177e4 LT |
353 | } |
354 | ||
355 | EXPORT_SYMBOL(locks_copy_lock); | |
356 | ||
357 | static inline int flock_translate_cmd(int cmd) { | |
358 | if (cmd & LOCK_MAND) | |
359 | return cmd & (LOCK_MAND | LOCK_RW); | |
360 | switch (cmd) { | |
361 | case LOCK_SH: | |
362 | return F_RDLCK; | |
363 | case LOCK_EX: | |
364 | return F_WRLCK; | |
365 | case LOCK_UN: | |
366 | return F_UNLCK; | |
367 | } | |
368 | return -EINVAL; | |
369 | } | |
370 | ||
371 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 JL |
372 | static struct file_lock * |
373 | flock_make_lock(struct file *filp, unsigned int cmd) | |
1da177e4 LT |
374 | { |
375 | struct file_lock *fl; | |
376 | int type = flock_translate_cmd(cmd); | |
6e129d00 | 377 | |
1da177e4 | 378 | if (type < 0) |
6e129d00 | 379 | return ERR_PTR(type); |
1da177e4 LT |
380 | |
381 | fl = locks_alloc_lock(); | |
382 | if (fl == NULL) | |
6e129d00 | 383 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
384 | |
385 | fl->fl_file = filp; | |
73a8f5f7 | 386 | fl->fl_owner = filp; |
1da177e4 LT |
387 | fl->fl_pid = current->tgid; |
388 | fl->fl_flags = FL_FLOCK; | |
389 | fl->fl_type = type; | |
390 | fl->fl_end = OFFSET_MAX; | |
391 | ||
6e129d00 | 392 | return fl; |
1da177e4 LT |
393 | } |
394 | ||
0ec4f431 | 395 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
396 | { |
397 | switch (type) { | |
398 | case F_RDLCK: | |
399 | case F_WRLCK: | |
400 | case F_UNLCK: | |
401 | fl->fl_type = type; | |
402 | break; | |
403 | default: | |
404 | return -EINVAL; | |
405 | } | |
406 | return 0; | |
407 | } | |
408 | ||
ef12e72a BF |
409 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
410 | struct flock64 *l) | |
1da177e4 | 411 | { |
1da177e4 | 412 | switch (l->l_whence) { |
f5579f8c | 413 | case SEEK_SET: |
ef12e72a | 414 | fl->fl_start = 0; |
1da177e4 | 415 | break; |
f5579f8c | 416 | case SEEK_CUR: |
ef12e72a | 417 | fl->fl_start = filp->f_pos; |
1da177e4 | 418 | break; |
f5579f8c | 419 | case SEEK_END: |
ef12e72a | 420 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
421 | break; |
422 | default: | |
423 | return -EINVAL; | |
424 | } | |
ef12e72a BF |
425 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
426 | return -EOVERFLOW; | |
427 | fl->fl_start += l->l_start; | |
428 | if (fl->fl_start < 0) | |
429 | return -EINVAL; | |
1da177e4 LT |
430 | |
431 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
432 | POSIX-2001 defines it. */ | |
4c780a46 | 433 | if (l->l_len > 0) { |
ef12e72a BF |
434 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
435 | return -EOVERFLOW; | |
436 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
437 | ||
4c780a46 | 438 | } else if (l->l_len < 0) { |
ef12e72a | 439 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 440 | return -EINVAL; |
ef12e72a BF |
441 | fl->fl_end = fl->fl_start - 1; |
442 | fl->fl_start += l->l_len; | |
443 | } else | |
444 | fl->fl_end = OFFSET_MAX; | |
445 | ||
1da177e4 LT |
446 | fl->fl_owner = current->files; |
447 | fl->fl_pid = current->tgid; | |
448 | fl->fl_file = filp; | |
449 | fl->fl_flags = FL_POSIX; | |
450 | fl->fl_ops = NULL; | |
451 | fl->fl_lmops = NULL; | |
452 | ||
453 | return assign_type(fl, l->l_type); | |
454 | } | |
455 | ||
ef12e72a BF |
456 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
457 | * style lock. | |
458 | */ | |
459 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
460 | struct flock *l) | |
1da177e4 | 461 | { |
ef12e72a BF |
462 | struct flock64 ll = { |
463 | .l_type = l->l_type, | |
464 | .l_whence = l->l_whence, | |
465 | .l_start = l->l_start, | |
466 | .l_len = l->l_len, | |
467 | }; | |
468 | ||
469 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 470 | } |
1da177e4 LT |
471 | |
472 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
473 | static bool |
474 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
475 | { |
476 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 477 | return false; |
1da177e4 LT |
478 | } |
479 | ||
1c7dd2ff JL |
480 | static void |
481 | lease_setup(struct file_lock *fl, void **priv) | |
482 | { | |
483 | struct file *filp = fl->fl_file; | |
484 | struct fasync_struct *fa = *priv; | |
485 | ||
486 | /* | |
487 | * fasync_insert_entry() returns the old entry if any. If there was no | |
488 | * old entry, then it used "priv" and inserted it into the fasync list. | |
489 | * Clear the pointer to indicate that it shouldn't be freed. | |
490 | */ | |
491 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
492 | *priv = NULL; | |
493 | ||
494 | __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
495 | } | |
496 | ||
7b021967 | 497 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 498 | .lm_break = lease_break_callback, |
8fb47a4f | 499 | .lm_change = lease_modify, |
1c7dd2ff | 500 | .lm_setup = lease_setup, |
1da177e4 LT |
501 | }; |
502 | ||
503 | /* | |
504 | * Initialize a lease, use the default lock manager operations | |
505 | */ | |
0ec4f431 | 506 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 507 | { |
75dff55a TM |
508 | if (assign_type(fl, type) != 0) |
509 | return -EINVAL; | |
510 | ||
7ca76311 | 511 | fl->fl_owner = filp; |
1da177e4 LT |
512 | fl->fl_pid = current->tgid; |
513 | ||
514 | fl->fl_file = filp; | |
515 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
516 | fl->fl_start = 0; |
517 | fl->fl_end = OFFSET_MAX; | |
518 | fl->fl_ops = NULL; | |
519 | fl->fl_lmops = &lease_manager_ops; | |
520 | return 0; | |
521 | } | |
522 | ||
523 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 524 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
525 | { |
526 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 527 | int error = -ENOMEM; |
1da177e4 LT |
528 | |
529 | if (fl == NULL) | |
e32b8ee2 | 530 | return ERR_PTR(error); |
1da177e4 LT |
531 | |
532 | error = lease_init(filp, type, fl); | |
75dff55a TM |
533 | if (error) { |
534 | locks_free_lock(fl); | |
e32b8ee2 | 535 | return ERR_PTR(error); |
75dff55a | 536 | } |
e32b8ee2 | 537 | return fl; |
1da177e4 LT |
538 | } |
539 | ||
540 | /* Check if two locks overlap each other. | |
541 | */ | |
542 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
543 | { | |
544 | return ((fl1->fl_end >= fl2->fl_start) && | |
545 | (fl2->fl_end >= fl1->fl_start)); | |
546 | } | |
547 | ||
548 | /* | |
549 | * Check whether two locks have the same owner. | |
550 | */ | |
33443c42 | 551 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 552 | { |
8fb47a4f | 553 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 554 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 555 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
556 | return fl1->fl_owner == fl2->fl_owner; |
557 | } | |
558 | ||
6109c850 | 559 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 560 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 561 | { |
7012b02a JL |
562 | lg_local_lock(&file_lock_lglock); |
563 | fl->fl_link_cpu = smp_processor_id(); | |
564 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
565 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
566 | } |
567 | ||
6109c850 | 568 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 569 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 570 | { |
7012b02a JL |
571 | /* |
572 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 573 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
574 | * also require that it be held. |
575 | */ | |
576 | if (hlist_unhashed(&fl->fl_link)) | |
577 | return; | |
578 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 579 | hlist_del_init(&fl->fl_link); |
7012b02a | 580 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
581 | } |
582 | ||
3999e493 JL |
583 | static unsigned long |
584 | posix_owner_key(struct file_lock *fl) | |
585 | { | |
586 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
587 | return fl->fl_lmops->lm_owner_key(fl); | |
588 | return (unsigned long)fl->fl_owner; | |
589 | } | |
590 | ||
6ca10ed8 | 591 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 592 | { |
663d5af7 DW |
593 | lockdep_assert_held(&blocked_lock_lock); |
594 | ||
3999e493 | 595 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
596 | } |
597 | ||
6ca10ed8 | 598 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 599 | { |
663d5af7 DW |
600 | lockdep_assert_held(&blocked_lock_lock); |
601 | ||
48f74186 | 602 | hash_del(&waiter->fl_link); |
88974691 JL |
603 | } |
604 | ||
1da177e4 LT |
605 | /* Remove waiter from blocker's block list. |
606 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 607 | * |
7b2296af | 608 | * Must be called with blocked_lock_lock held. |
1da177e4 | 609 | */ |
33443c42 | 610 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 611 | { |
88974691 | 612 | locks_delete_global_blocked(waiter); |
1da177e4 | 613 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
614 | waiter->fl_next = NULL; |
615 | } | |
616 | ||
1a9e64a7 | 617 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 618 | { |
7b2296af | 619 | spin_lock(&blocked_lock_lock); |
1da177e4 | 620 | __locks_delete_block(waiter); |
7b2296af | 621 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
622 | } |
623 | ||
624 | /* Insert waiter into blocker's block list. | |
625 | * We use a circular list so that processes can be easily woken up in | |
626 | * the order they blocked. The documentation doesn't require this but | |
627 | * it seems like the reasonable thing to do. | |
1c8c601a | 628 | * |
6109c850 JL |
629 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
630 | * fl_block list itself is protected by the blocked_lock_lock, but by ensuring | |
631 | * that the flc_lock is also held on insertions we can avoid taking the | |
632 | * blocked_lock_lock in some cases when we see that the fl_block list is empty. | |
1da177e4 | 633 | */ |
1c8c601a JL |
634 | static void __locks_insert_block(struct file_lock *blocker, |
635 | struct file_lock *waiter) | |
1da177e4 | 636 | { |
6dc0fe8f | 637 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 638 | waiter->fl_next = blocker; |
88974691 | 639 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 640 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
641 | locks_insert_global_blocked(waiter); |
642 | } | |
643 | ||
6109c850 | 644 | /* Must be called with flc_lock held. */ |
1c8c601a JL |
645 | static void locks_insert_block(struct file_lock *blocker, |
646 | struct file_lock *waiter) | |
647 | { | |
7b2296af | 648 | spin_lock(&blocked_lock_lock); |
1c8c601a | 649 | __locks_insert_block(blocker, waiter); |
7b2296af | 650 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
651 | } |
652 | ||
1cb36012 JL |
653 | /* |
654 | * Wake up processes blocked waiting for blocker. | |
655 | * | |
6109c850 | 656 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
657 | */ |
658 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
659 | { | |
4e8c765d JL |
660 | /* |
661 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 JL |
662 | * blocked requests are only added to the list under the flc_lock, and |
663 | * the flc_lock is always held here. Note that removal from the fl_block | |
664 | * list does not require the flc_lock, so we must recheck list_empty() | |
7b2296af | 665 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
666 | */ |
667 | if (list_empty(&blocker->fl_block)) | |
668 | return; | |
669 | ||
7b2296af | 670 | spin_lock(&blocked_lock_lock); |
1da177e4 | 671 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
672 | struct file_lock *waiter; |
673 | ||
674 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
675 | struct file_lock, fl_block); |
676 | __locks_delete_block(waiter); | |
8fb47a4f BF |
677 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
678 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
679 | else |
680 | wake_up(&waiter->fl_wait); | |
681 | } | |
7b2296af | 682 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
683 | } |
684 | ||
5263e31e | 685 | static void |
e084c1bd | 686 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e JL |
687 | { |
688 | fl->fl_nspid = get_pid(task_tgid(current)); | |
689 | list_add_tail(&fl->fl_list, before); | |
690 | locks_insert_global_locks(fl); | |
691 | } | |
692 | ||
8634b51f | 693 | static void |
e084c1bd | 694 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 695 | { |
88974691 | 696 | locks_delete_global_locks(fl); |
8634b51f | 697 | list_del_init(&fl->fl_list); |
ab1f1611 VG |
698 | if (fl->fl_nspid) { |
699 | put_pid(fl->fl_nspid); | |
700 | fl->fl_nspid = NULL; | |
701 | } | |
1da177e4 | 702 | locks_wake_up_blocks(fl); |
24cbe784 JL |
703 | } |
704 | ||
8634b51f | 705 | static void |
e084c1bd | 706 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 707 | { |
e084c1bd | 708 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 709 | if (dispose) |
6dee60f6 | 710 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
711 | else |
712 | locks_free_lock(fl); | |
1da177e4 LT |
713 | } |
714 | ||
715 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
716 | * checks for shared/exclusive status of overlapping locks. | |
717 | */ | |
718 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
719 | { | |
720 | if (sys_fl->fl_type == F_WRLCK) | |
721 | return 1; | |
722 | if (caller_fl->fl_type == F_WRLCK) | |
723 | return 1; | |
724 | return 0; | |
725 | } | |
726 | ||
727 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
728 | * checking before calling the locks_conflict(). | |
729 | */ | |
730 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
731 | { | |
732 | /* POSIX locks owned by the same process do not conflict with | |
733 | * each other. | |
734 | */ | |
9b8c8695 | 735 | if (posix_same_owner(caller_fl, sys_fl)) |
1da177e4 LT |
736 | return (0); |
737 | ||
738 | /* Check whether they overlap */ | |
739 | if (!locks_overlap(caller_fl, sys_fl)) | |
740 | return 0; | |
741 | ||
742 | return (locks_conflict(caller_fl, sys_fl)); | |
743 | } | |
744 | ||
745 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
746 | * checking before calling the locks_conflict(). | |
747 | */ | |
748 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
749 | { | |
750 | /* FLOCK locks referring to the same filp do not conflict with | |
751 | * each other. | |
752 | */ | |
9b8c8695 | 753 | if (caller_fl->fl_file == sys_fl->fl_file) |
1da177e4 LT |
754 | return (0); |
755 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
756 | return 0; | |
757 | ||
758 | return (locks_conflict(caller_fl, sys_fl)); | |
759 | } | |
760 | ||
6d34ac19 | 761 | void |
9d6a8c5c | 762 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
763 | { |
764 | struct file_lock *cfl; | |
bd61e0a9 | 765 | struct file_lock_context *ctx; |
1c8c601a | 766 | struct inode *inode = file_inode(filp); |
1da177e4 | 767 | |
128a3785 | 768 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
769 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
770 | fl->fl_type = F_UNLCK; | |
771 | return; | |
772 | } | |
773 | ||
6109c850 | 774 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
775 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
776 | if (posix_locks_conflict(fl, cfl)) { | |
777 | locks_copy_conflock(fl, cfl); | |
778 | if (cfl->fl_nspid) | |
779 | fl->fl_pid = pid_vnr(cfl->fl_nspid); | |
780 | goto out; | |
781 | } | |
1da177e4 | 782 | } |
bd61e0a9 JL |
783 | fl->fl_type = F_UNLCK; |
784 | out: | |
6109c850 | 785 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 786 | return; |
1da177e4 | 787 | } |
1da177e4 LT |
788 | EXPORT_SYMBOL(posix_test_lock); |
789 | ||
b533184f BF |
790 | /* |
791 | * Deadlock detection: | |
792 | * | |
793 | * We attempt to detect deadlocks that are due purely to posix file | |
794 | * locks. | |
1da177e4 | 795 | * |
b533184f BF |
796 | * We assume that a task can be waiting for at most one lock at a time. |
797 | * So for any acquired lock, the process holding that lock may be | |
798 | * waiting on at most one other lock. That lock in turns may be held by | |
799 | * someone waiting for at most one other lock. Given a requested lock | |
800 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
801 | * follow this chain of waiters to ensure we are not about to create a | |
802 | * cycle. | |
1da177e4 | 803 | * |
b533184f BF |
804 | * Since we do this before we ever put a process to sleep on a lock, we |
805 | * are ensured that there is never a cycle; that is what guarantees that | |
806 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 807 | * |
b533184f BF |
808 | * Note: the above assumption may not be true when handling lock |
809 | * requests from a broken NFS client. It may also fail in the presence | |
810 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 811 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 812 | * |
cff2fce5 | 813 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
814 | * Because the owner is not even nominally tied to a thread of |
815 | * execution, the deadlock detection below can't reasonably work well. Just | |
816 | * skip it for those. | |
817 | * | |
cff2fce5 | 818 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
819 | * locks that just checks for the case where two tasks are attempting to |
820 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 821 | */ |
97855b49 BF |
822 | |
823 | #define MAX_DEADLK_ITERATIONS 10 | |
824 | ||
b533184f BF |
825 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
826 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
827 | { | |
828 | struct file_lock *fl; | |
829 | ||
3999e493 | 830 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
831 | if (posix_same_owner(fl, block_fl)) |
832 | return fl->fl_next; | |
833 | } | |
834 | return NULL; | |
835 | } | |
836 | ||
7b2296af | 837 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 838 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
839 | struct file_lock *block_fl) |
840 | { | |
97855b49 | 841 | int i = 0; |
1da177e4 | 842 | |
663d5af7 DW |
843 | lockdep_assert_held(&blocked_lock_lock); |
844 | ||
57b65325 JL |
845 | /* |
846 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 847 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 848 | */ |
cff2fce5 | 849 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
850 | return 0; |
851 | ||
b533184f BF |
852 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
853 | if (i++ > MAX_DEADLK_ITERATIONS) | |
854 | return 0; | |
855 | if (posix_same_owner(caller_fl, block_fl)) | |
856 | return 1; | |
1da177e4 LT |
857 | } |
858 | return 0; | |
859 | } | |
860 | ||
1da177e4 | 861 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 862 | * after any leases, but before any posix locks. |
f475ae95 TM |
863 | * |
864 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
865 | * whether or not a lock was successfully freed by testing the return | |
866 | * value for -ENOENT. | |
1da177e4 | 867 | */ |
bcd7f78d | 868 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 869 | { |
993dfa87 | 870 | struct file_lock *new_fl = NULL; |
5263e31e JL |
871 | struct file_lock *fl; |
872 | struct file_lock_context *ctx; | |
1da177e4 | 873 | int error = 0; |
5263e31e | 874 | bool found = false; |
ed9814d8 | 875 | LIST_HEAD(dispose); |
1da177e4 | 876 | |
5c1c669a JL |
877 | ctx = locks_get_lock_context(inode, request->fl_type); |
878 | if (!ctx) { | |
879 | if (request->fl_type != F_UNLCK) | |
880 | return -ENOMEM; | |
881 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
882 | } | |
5263e31e | 883 | |
b89f4321 | 884 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 885 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
886 | if (!new_fl) |
887 | return -ENOMEM; | |
84d535ad PE |
888 | } |
889 | ||
6109c850 | 890 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
891 | if (request->fl_flags & FL_ACCESS) |
892 | goto find_conflict; | |
893 | ||
5263e31e | 894 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 895 | if (request->fl_file != fl->fl_file) |
1da177e4 | 896 | continue; |
993dfa87 | 897 | if (request->fl_type == fl->fl_type) |
1da177e4 | 898 | goto out; |
5263e31e | 899 | found = true; |
e084c1bd | 900 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
901 | break; |
902 | } | |
1da177e4 | 903 | |
f475ae95 TM |
904 | if (request->fl_type == F_UNLCK) { |
905 | if ((request->fl_flags & FL_EXISTS) && !found) | |
906 | error = -ENOENT; | |
993dfa87 | 907 | goto out; |
f475ae95 | 908 | } |
1da177e4 | 909 | |
f07f18dd | 910 | find_conflict: |
5263e31e | 911 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 912 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
913 | continue; |
914 | error = -EAGAIN; | |
bde74e4b MS |
915 | if (!(request->fl_flags & FL_SLEEP)) |
916 | goto out; | |
917 | error = FILE_LOCK_DEFERRED; | |
918 | locks_insert_block(fl, request); | |
1da177e4 LT |
919 | goto out; |
920 | } | |
f07f18dd TM |
921 | if (request->fl_flags & FL_ACCESS) |
922 | goto out; | |
993dfa87 | 923 | locks_copy_lock(new_fl, request); |
e084c1bd | 924 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 925 | new_fl = NULL; |
9cedc194 | 926 | error = 0; |
1da177e4 LT |
927 | |
928 | out: | |
6109c850 | 929 | spin_unlock(&ctx->flc_lock); |
993dfa87 TM |
930 | if (new_fl) |
931 | locks_free_lock(new_fl); | |
ed9814d8 | 932 | locks_dispose_list(&dispose); |
1da177e4 LT |
933 | return error; |
934 | } | |
935 | ||
150b3934 | 936 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 | 937 | { |
bd61e0a9 | 938 | struct file_lock *fl, *tmp; |
39005d02 MS |
939 | struct file_lock *new_fl = NULL; |
940 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
941 | struct file_lock *left = NULL; |
942 | struct file_lock *right = NULL; | |
bd61e0a9 | 943 | struct file_lock_context *ctx; |
b9746ef8 JL |
944 | int error; |
945 | bool added = false; | |
ed9814d8 | 946 | LIST_HEAD(dispose); |
1da177e4 | 947 | |
5c1c669a | 948 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 949 | if (!ctx) |
5c1c669a | 950 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 951 | |
1da177e4 LT |
952 | /* |
953 | * We may need two file_lock structures for this operation, | |
954 | * so we get them in advance to avoid races. | |
39005d02 MS |
955 | * |
956 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 957 | */ |
39005d02 MS |
958 | if (!(request->fl_flags & FL_ACCESS) && |
959 | (request->fl_type != F_UNLCK || | |
960 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
961 | new_fl = locks_alloc_lock(); | |
962 | new_fl2 = locks_alloc_lock(); | |
963 | } | |
1da177e4 | 964 | |
6109c850 | 965 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
966 | /* |
967 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
968 | * there are any, either return error or put the request on the | |
48f74186 | 969 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 970 | */ |
1da177e4 | 971 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 972 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
973 | if (!posix_locks_conflict(request, fl)) |
974 | continue; | |
5842add2 | 975 | if (conflock) |
3fe0fff1 | 976 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
977 | error = -EAGAIN; |
978 | if (!(request->fl_flags & FL_SLEEP)) | |
979 | goto out; | |
1c8c601a JL |
980 | /* |
981 | * Deadlock detection and insertion into the blocked | |
982 | * locks list must be done while holding the same lock! | |
983 | */ | |
1da177e4 | 984 | error = -EDEADLK; |
7b2296af | 985 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
986 | if (likely(!posix_locks_deadlock(request, fl))) { |
987 | error = FILE_LOCK_DEFERRED; | |
988 | __locks_insert_block(fl, request); | |
989 | } | |
7b2296af | 990 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
991 | goto out; |
992 | } | |
993 | } | |
994 | ||
995 | /* If we're just looking for a conflict, we're done. */ | |
996 | error = 0; | |
997 | if (request->fl_flags & FL_ACCESS) | |
998 | goto out; | |
999 | ||
bd61e0a9 JL |
1000 | /* Find the first old lock with the same owner as the new lock */ |
1001 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1002 | if (posix_same_owner(request, fl)) | |
1003 | break; | |
1da177e4 LT |
1004 | } |
1005 | ||
1cb36012 | 1006 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1007 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1008 | if (!posix_same_owner(request, fl)) | |
1009 | break; | |
1010 | ||
1011 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1012 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1013 | /* In all comparisons of start vs end, use |
1014 | * "start - 1" rather than "end + 1". If end | |
1015 | * is OFFSET_MAX, end + 1 will become negative. | |
1016 | */ | |
1da177e4 | 1017 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1018 | continue; |
1da177e4 LT |
1019 | /* If the next lock in the list has entirely bigger |
1020 | * addresses than the new one, insert the lock here. | |
1021 | */ | |
449231d6 | 1022 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1023 | break; |
1024 | ||
1025 | /* If we come here, the new and old lock are of the | |
1026 | * same type and adjacent or overlapping. Make one | |
1027 | * lock yielding from the lower start address of both | |
1028 | * locks to the higher end address. | |
1029 | */ | |
1030 | if (fl->fl_start > request->fl_start) | |
1031 | fl->fl_start = request->fl_start; | |
1032 | else | |
1033 | request->fl_start = fl->fl_start; | |
1034 | if (fl->fl_end < request->fl_end) | |
1035 | fl->fl_end = request->fl_end; | |
1036 | else | |
1037 | request->fl_end = fl->fl_end; | |
1038 | if (added) { | |
e084c1bd | 1039 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1040 | continue; |
1041 | } | |
1042 | request = fl; | |
b9746ef8 | 1043 | added = true; |
bd61e0a9 | 1044 | } else { |
1da177e4 LT |
1045 | /* Processing for different lock types is a bit |
1046 | * more complex. | |
1047 | */ | |
1048 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1049 | continue; |
1da177e4 LT |
1050 | if (fl->fl_start > request->fl_end) |
1051 | break; | |
1052 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1053 | added = true; |
1da177e4 LT |
1054 | if (fl->fl_start < request->fl_start) |
1055 | left = fl; | |
1056 | /* If the next lock in the list has a higher end | |
1057 | * address than the new one, insert the new one here. | |
1058 | */ | |
1059 | if (fl->fl_end > request->fl_end) { | |
1060 | right = fl; | |
1061 | break; | |
1062 | } | |
1063 | if (fl->fl_start >= request->fl_start) { | |
1064 | /* The new lock completely replaces an old | |
1065 | * one (This may happen several times). | |
1066 | */ | |
1067 | if (added) { | |
e084c1bd | 1068 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1069 | continue; |
1070 | } | |
b84d49f9 JL |
1071 | /* |
1072 | * Replace the old lock with new_fl, and | |
1073 | * remove the old one. It's safe to do the | |
1074 | * insert here since we know that we won't be | |
1075 | * using new_fl later, and that the lock is | |
1076 | * just replacing an existing lock. | |
1da177e4 | 1077 | */ |
b84d49f9 JL |
1078 | error = -ENOLCK; |
1079 | if (!new_fl) | |
1080 | goto out; | |
1081 | locks_copy_lock(new_fl, request); | |
1082 | request = new_fl; | |
1083 | new_fl = NULL; | |
e084c1bd JL |
1084 | locks_insert_lock_ctx(request, &fl->fl_list); |
1085 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1086 | added = true; |
1da177e4 LT |
1087 | } |
1088 | } | |
1da177e4 LT |
1089 | } |
1090 | ||
0d9a490a | 1091 | /* |
1cb36012 JL |
1092 | * The above code only modifies existing locks in case of merging or |
1093 | * replacing. If new lock(s) need to be inserted all modifications are | |
1094 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1095 | */ |
1096 | error = -ENOLCK; /* "no luck" */ | |
1097 | if (right && left == right && !new_fl2) | |
1098 | goto out; | |
1099 | ||
1da177e4 LT |
1100 | error = 0; |
1101 | if (!added) { | |
f475ae95 TM |
1102 | if (request->fl_type == F_UNLCK) { |
1103 | if (request->fl_flags & FL_EXISTS) | |
1104 | error = -ENOENT; | |
1da177e4 | 1105 | goto out; |
f475ae95 | 1106 | } |
0d9a490a MS |
1107 | |
1108 | if (!new_fl) { | |
1109 | error = -ENOLCK; | |
1110 | goto out; | |
1111 | } | |
1da177e4 | 1112 | locks_copy_lock(new_fl, request); |
e084c1bd | 1113 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
2e2f756f | 1114 | fl = new_fl; |
1da177e4 LT |
1115 | new_fl = NULL; |
1116 | } | |
1117 | if (right) { | |
1118 | if (left == right) { | |
1119 | /* The new lock breaks the old one in two pieces, | |
1120 | * so we have to use the second new lock. | |
1121 | */ | |
1122 | left = new_fl2; | |
1123 | new_fl2 = NULL; | |
1124 | locks_copy_lock(left, right); | |
e084c1bd | 1125 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1126 | } |
1127 | right->fl_start = request->fl_end + 1; | |
1128 | locks_wake_up_blocks(right); | |
1129 | } | |
1130 | if (left) { | |
1131 | left->fl_end = request->fl_start - 1; | |
1132 | locks_wake_up_blocks(left); | |
1133 | } | |
1134 | out: | |
6109c850 | 1135 | spin_unlock(&ctx->flc_lock); |
1da177e4 LT |
1136 | /* |
1137 | * Free any unused locks. | |
1138 | */ | |
1139 | if (new_fl) | |
1140 | locks_free_lock(new_fl); | |
1141 | if (new_fl2) | |
1142 | locks_free_lock(new_fl2); | |
ed9814d8 | 1143 | locks_dispose_list(&dispose); |
1da177e4 LT |
1144 | return error; |
1145 | } | |
1146 | ||
1147 | /** | |
1148 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1149 | * @filp: The file to apply the lock to | |
1150 | * @fl: The lock to be applied | |
150b3934 | 1151 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1152 | * |
1153 | * Add a POSIX style lock to a file. | |
1154 | * We merge adjacent & overlapping locks whenever possible. | |
1155 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1156 | * |
1157 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1158 | * whether or not a lock was successfully freed by testing the return | |
1159 | * value for -ENOENT. | |
1da177e4 | 1160 | */ |
150b3934 | 1161 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1162 | struct file_lock *conflock) |
1163 | { | |
496ad9aa | 1164 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1165 | } |
150b3934 | 1166 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1167 | |
1168 | /** | |
29d01b22 JL |
1169 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1170 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1171 | * @fl: The lock to be applied |
1172 | * | |
616fb38f | 1173 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1174 | */ |
616fb38f | 1175 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1176 | { |
1177 | int error; | |
1178 | might_sleep (); | |
1179 | for (;;) { | |
29d01b22 | 1180 | error = __posix_lock_file(inode, fl, NULL); |
bde74e4b | 1181 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1182 | break; |
1183 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1184 | if (!error) | |
1185 | continue; | |
1186 | ||
1187 | locks_delete_block(fl); | |
1188 | break; | |
1189 | } | |
1190 | return error; | |
1191 | } | |
29d01b22 | 1192 | |
9e8925b6 | 1193 | #ifdef CONFIG_MANDATORY_FILE_LOCKING |
1da177e4 LT |
1194 | /** |
1195 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1196 | * @file: the file to check |
1da177e4 LT |
1197 | * |
1198 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1199 | * This function is called from locks_verify_locked() only. | |
1200 | */ | |
d7a06983 | 1201 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1202 | { |
bd61e0a9 | 1203 | int ret; |
d7a06983 | 1204 | struct inode *inode = file_inode(file); |
bd61e0a9 | 1205 | struct file_lock_context *ctx; |
1da177e4 LT |
1206 | struct file_lock *fl; |
1207 | ||
128a3785 | 1208 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
1209 | if (!ctx || list_empty_careful(&ctx->flc_posix)) |
1210 | return 0; | |
1211 | ||
1da177e4 LT |
1212 | /* |
1213 | * Search the lock list for this inode for any POSIX locks. | |
1214 | */ | |
6109c850 | 1215 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
1216 | ret = 0; |
1217 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1218 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1219 | fl->fl_owner != file) { |
1220 | ret = -EAGAIN; | |
1da177e4 | 1221 | break; |
bd61e0a9 | 1222 | } |
1da177e4 | 1223 | } |
6109c850 | 1224 | spin_unlock(&ctx->flc_lock); |
bd61e0a9 | 1225 | return ret; |
1da177e4 LT |
1226 | } |
1227 | ||
1228 | /** | |
1229 | * locks_mandatory_area - Check for a conflicting lock | |
1230 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1231 | * for shared | |
1232 | * @inode: the file to check | |
1233 | * @filp: how the file was opened (if it was) | |
1234 | * @offset: start of area to check | |
1235 | * @count: length of area to check | |
1236 | * | |
1237 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1238 | * This function is called from rw_verify_area() and | |
1239 | * locks_verify_truncate(). | |
1240 | */ | |
1241 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1242 | struct file *filp, loff_t offset, | |
1243 | size_t count) | |
1244 | { | |
1245 | struct file_lock fl; | |
1246 | int error; | |
29723ade | 1247 | bool sleep = false; |
1da177e4 LT |
1248 | |
1249 | locks_init_lock(&fl); | |
1da177e4 LT |
1250 | fl.fl_pid = current->tgid; |
1251 | fl.fl_file = filp; | |
1252 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1253 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1254 | sleep = true; |
1da177e4 LT |
1255 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1256 | fl.fl_start = offset; | |
1257 | fl.fl_end = offset + count - 1; | |
1258 | ||
1259 | for (;;) { | |
29723ade | 1260 | if (filp) { |
73a8f5f7 | 1261 | fl.fl_owner = filp; |
29723ade JL |
1262 | fl.fl_flags &= ~FL_SLEEP; |
1263 | error = __posix_lock_file(inode, &fl, NULL); | |
1264 | if (!error) | |
1265 | break; | |
1266 | } | |
1267 | ||
1268 | if (sleep) | |
1269 | fl.fl_flags |= FL_SLEEP; | |
1270 | fl.fl_owner = current->files; | |
150b3934 | 1271 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1272 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1273 | break; |
1274 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1275 | if (!error) { | |
1276 | /* | |
1277 | * If we've been sleeping someone might have | |
1278 | * changed the permissions behind our back. | |
1279 | */ | |
a16877ca | 1280 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1281 | continue; |
1282 | } | |
1283 | ||
1284 | locks_delete_block(&fl); | |
1285 | break; | |
1286 | } | |
1287 | ||
1288 | return error; | |
1289 | } | |
1290 | ||
1291 | EXPORT_SYMBOL(locks_mandatory_area); | |
9e8925b6 | 1292 | #endif /* CONFIG_MANDATORY_FILE_LOCKING */ |
1da177e4 | 1293 | |
778fc546 BF |
1294 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1295 | { | |
1296 | switch (arg) { | |
1297 | case F_UNLCK: | |
1298 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1299 | /* fall through: */ | |
1300 | case F_RDLCK: | |
1301 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1302 | } | |
1303 | } | |
1304 | ||
1da177e4 | 1305 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1306 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1307 | { |
1da177e4 LT |
1308 | int error = assign_type(fl, arg); |
1309 | ||
1310 | if (error) | |
1311 | return error; | |
778fc546 | 1312 | lease_clear_pending(fl, arg); |
1da177e4 | 1313 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1314 | if (arg == F_UNLCK) { |
1315 | struct file *filp = fl->fl_file; | |
1316 | ||
1317 | f_delown(filp); | |
1318 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1319 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1320 | if (fl->fl_fasync != NULL) { | |
1321 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1322 | fl->fl_fasync = NULL; | |
1323 | } | |
e084c1bd | 1324 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1325 | } |
1da177e4 LT |
1326 | return 0; |
1327 | } | |
1da177e4 LT |
1328 | EXPORT_SYMBOL(lease_modify); |
1329 | ||
778fc546 BF |
1330 | static bool past_time(unsigned long then) |
1331 | { | |
1332 | if (!then) | |
1333 | /* 0 is a special value meaning "this never expires": */ | |
1334 | return false; | |
1335 | return time_after(jiffies, then); | |
1336 | } | |
1337 | ||
c45198ed | 1338 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1339 | { |
8634b51f JL |
1340 | struct file_lock_context *ctx = inode->i_flctx; |
1341 | struct file_lock *fl, *tmp; | |
1da177e4 | 1342 | |
6109c850 | 1343 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1344 | |
8634b51f | 1345 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1346 | trace_time_out_leases(inode, fl); |
778fc546 | 1347 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1348 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1349 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1350 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1351 | } |
1352 | } | |
1353 | ||
df4e8d2c BF |
1354 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1355 | { | |
11afe9f7 CH |
1356 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) |
1357 | return false; | |
df4e8d2c BF |
1358 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) |
1359 | return false; | |
1360 | return locks_conflict(breaker, lease); | |
1361 | } | |
1362 | ||
03d12ddf JL |
1363 | static bool |
1364 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1365 | { | |
8634b51f | 1366 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1367 | struct file_lock *fl; |
1368 | ||
6109c850 | 1369 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1370 | |
8634b51f | 1371 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1372 | if (leases_conflict(fl, breaker)) |
1373 | return true; | |
1374 | } | |
1375 | return false; | |
1376 | } | |
1377 | ||
1da177e4 LT |
1378 | /** |
1379 | * __break_lease - revoke all outstanding leases on file | |
1380 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1381 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1382 | * break all leases | |
1383 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1384 | * only delegations | |
1da177e4 | 1385 | * |
87250dd2 | 1386 | * break_lease (inlined for speed) has checked there already is at least |
1387 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1388 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1389 | * specified %O_NONBLOCK to your open(). |
1390 | */ | |
df4e8d2c | 1391 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1392 | { |
778fc546 | 1393 | int error = 0; |
128a3785 | 1394 | struct file_lock_context *ctx; |
a901125c | 1395 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1396 | unsigned long break_time; |
8737c930 | 1397 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1398 | LIST_HEAD(dispose); |
1da177e4 | 1399 | |
8737c930 | 1400 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1401 | if (IS_ERR(new_fl)) |
1402 | return PTR_ERR(new_fl); | |
df4e8d2c | 1403 | new_fl->fl_flags = type; |
1da177e4 | 1404 | |
8634b51f | 1405 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1406 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1407 | if (!ctx) { |
1408 | WARN_ON_ONCE(1); | |
1409 | return error; | |
1410 | } | |
1411 | ||
6109c850 | 1412 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1413 | |
c45198ed | 1414 | time_out_leases(inode, &dispose); |
1da177e4 | 1415 | |
03d12ddf | 1416 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1417 | goto out; |
1418 | ||
1da177e4 LT |
1419 | break_time = 0; |
1420 | if (lease_break_time > 0) { | |
1421 | break_time = jiffies + lease_break_time * HZ; | |
1422 | if (break_time == 0) | |
1423 | break_time++; /* so that 0 means no break time */ | |
1424 | } | |
1425 | ||
a901125c | 1426 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1427 | if (!leases_conflict(fl, new_fl)) |
1428 | continue; | |
778fc546 BF |
1429 | if (want_write) { |
1430 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1431 | continue; | |
1432 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1433 | fl->fl_break_time = break_time; |
778fc546 | 1434 | } else { |
8634b51f | 1435 | if (lease_breaking(fl)) |
778fc546 BF |
1436 | continue; |
1437 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1438 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1439 | } |
4d01b7f5 | 1440 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1441 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1442 | } |
1443 | ||
8634b51f | 1444 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1445 | goto out; |
1446 | ||
843c6b2f | 1447 | if (mode & O_NONBLOCK) { |
62af4f1f | 1448 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1449 | error = -EWOULDBLOCK; |
1450 | goto out; | |
1451 | } | |
1452 | ||
1453 | restart: | |
8634b51f JL |
1454 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1455 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1456 | if (break_time != 0) |
1da177e4 | 1457 | break_time -= jiffies; |
f1c6bb2c JL |
1458 | if (break_time == 0) |
1459 | break_time++; | |
8634b51f | 1460 | locks_insert_block(fl, new_fl); |
62af4f1f | 1461 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1462 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1463 | locks_dispose_list(&dispose); |
4321e01e MW |
1464 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1465 | !new_fl->fl_next, break_time); | |
6109c850 | 1466 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1467 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1468 | locks_delete_block(new_fl); |
1da177e4 | 1469 | if (error >= 0) { |
778fc546 BF |
1470 | /* |
1471 | * Wait for the next conflicting lease that has not been | |
1472 | * broken yet | |
1473 | */ | |
03d12ddf JL |
1474 | if (error == 0) |
1475 | time_out_leases(inode, &dispose); | |
1476 | if (any_leases_conflict(inode, new_fl)) | |
1477 | goto restart; | |
1da177e4 LT |
1478 | error = 0; |
1479 | } | |
1da177e4 | 1480 | out: |
6109c850 | 1481 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1482 | locks_dispose_list(&dispose); |
6d4b9e38 | 1483 | locks_free_lock(new_fl); |
1da177e4 LT |
1484 | return error; |
1485 | } | |
1486 | ||
1487 | EXPORT_SYMBOL(__break_lease); | |
1488 | ||
1489 | /** | |
a6b91919 | 1490 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1491 | * @inode: the inode |
1492 | * @time: pointer to a timespec which will contain the last modified time | |
1493 | * | |
1494 | * This is to force NFS clients to flush their caches for files with | |
1495 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1496 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1497 | */ |
1498 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1499 | { | |
bfe86024 | 1500 | bool has_lease = false; |
128a3785 | 1501 | struct file_lock_context *ctx; |
8634b51f | 1502 | struct file_lock *fl; |
bfe86024 | 1503 | |
128a3785 | 1504 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1505 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1506 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1507 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1508 | struct file_lock, fl_list); | |
1509 | if (fl && (fl->fl_type == F_WRLCK)) | |
1510 | has_lease = true; | |
6109c850 | 1511 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1512 | } |
1513 | ||
1514 | if (has_lease) | |
1da177e4 LT |
1515 | *time = current_fs_time(inode->i_sb); |
1516 | else | |
1517 | *time = inode->i_mtime; | |
1518 | } | |
1519 | ||
1520 | EXPORT_SYMBOL(lease_get_mtime); | |
1521 | ||
1522 | /** | |
1523 | * fcntl_getlease - Enquire what lease is currently active | |
1524 | * @filp: the file | |
1525 | * | |
1526 | * The value returned by this function will be one of | |
1527 | * (if no lease break is pending): | |
1528 | * | |
1529 | * %F_RDLCK to indicate a shared lease is held. | |
1530 | * | |
1531 | * %F_WRLCK to indicate an exclusive lease is held. | |
1532 | * | |
1533 | * %F_UNLCK to indicate no lease is held. | |
1534 | * | |
1535 | * (if a lease break is pending): | |
1536 | * | |
1537 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1538 | * changed to a shared lease (or removed). | |
1539 | * | |
1540 | * %F_UNLCK to indicate the lease needs to be removed. | |
1541 | * | |
1542 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1543 | * should be returned to userspace. | |
1544 | */ | |
1545 | int fcntl_getlease(struct file *filp) | |
1546 | { | |
1547 | struct file_lock *fl; | |
1c8c601a | 1548 | struct inode *inode = file_inode(filp); |
128a3785 | 1549 | struct file_lock_context *ctx; |
1da177e4 | 1550 | int type = F_UNLCK; |
c45198ed | 1551 | LIST_HEAD(dispose); |
1da177e4 | 1552 | |
128a3785 | 1553 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1554 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1555 | spin_lock(&ctx->flc_lock); |
8634b51f JL |
1556 | time_out_leases(file_inode(filp), &dispose); |
1557 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { | |
1558 | if (fl->fl_file != filp) | |
1559 | continue; | |
778fc546 | 1560 | type = target_leasetype(fl); |
1da177e4 LT |
1561 | break; |
1562 | } | |
6109c850 | 1563 | spin_unlock(&ctx->flc_lock); |
8634b51f | 1564 | locks_dispose_list(&dispose); |
1da177e4 | 1565 | } |
1da177e4 LT |
1566 | return type; |
1567 | } | |
1568 | ||
24cbe784 JL |
1569 | /** |
1570 | * check_conflicting_open - see if the given dentry points to a file that has | |
1571 | * an existing open that would conflict with the | |
1572 | * desired lease. | |
1573 | * @dentry: dentry to check | |
1574 | * @arg: type of lease that we're trying to acquire | |
7fadc59c | 1575 | * @flags: current lock flags |
24cbe784 JL |
1576 | * |
1577 | * Check to see if there's an existing open fd on this file that would | |
1578 | * conflict with the lease we're trying to set. | |
1579 | */ | |
1580 | static int | |
11afe9f7 | 1581 | check_conflicting_open(const struct dentry *dentry, const long arg, int flags) |
24cbe784 JL |
1582 | { |
1583 | int ret = 0; | |
1584 | struct inode *inode = dentry->d_inode; | |
1585 | ||
11afe9f7 CH |
1586 | if (flags & FL_LAYOUT) |
1587 | return 0; | |
1588 | ||
24cbe784 JL |
1589 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
1590 | return -EAGAIN; | |
1591 | ||
1592 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1593 | (atomic_read(&inode->i_count) > 1))) | |
1594 | ret = -EAGAIN; | |
1595 | ||
1596 | return ret; | |
1597 | } | |
1598 | ||
e6f5c789 JL |
1599 | static int |
1600 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1601 | { |
8634b51f | 1602 | struct file_lock *fl, *my_fl = NULL, *lease; |
0f7fc9e4 | 1603 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1604 | struct inode *inode = dentry->d_inode; |
8634b51f | 1605 | struct file_lock_context *ctx; |
df4e8d2c | 1606 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1607 | int error; |
c45198ed | 1608 | LIST_HEAD(dispose); |
1da177e4 | 1609 | |
096657b6 | 1610 | lease = *flp; |
62af4f1f JL |
1611 | trace_generic_add_lease(inode, lease); |
1612 | ||
5c1c669a JL |
1613 | /* Note that arg is never F_UNLCK here */ |
1614 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1615 | if (!ctx) |
1616 | return -ENOMEM; | |
1617 | ||
df4e8d2c BF |
1618 | /* |
1619 | * In the delegation case we need mutual exclusion with | |
1620 | * a number of operations that take the i_mutex. We trylock | |
1621 | * because delegations are an optional optimization, and if | |
1622 | * there's some chance of a conflict--we'd rather not | |
1623 | * bother, maybe that's a sign this just isn't a good file to | |
1624 | * hand out a delegation on. | |
1625 | */ | |
1626 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1627 | return -EAGAIN; | |
1628 | ||
1629 | if (is_deleg && arg == F_WRLCK) { | |
1630 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1631 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1632 | WARN_ON_ONCE(1); |
1633 | return -EINVAL; | |
1634 | } | |
096657b6 | 1635 | |
6109c850 | 1636 | spin_lock(&ctx->flc_lock); |
c45198ed | 1637 | time_out_leases(inode, &dispose); |
11afe9f7 | 1638 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
24cbe784 | 1639 | if (error) |
096657b6 | 1640 | goto out; |
6d5e8b05 | 1641 | |
1da177e4 LT |
1642 | /* |
1643 | * At this point, we know that if there is an exclusive | |
1644 | * lease on this file, then we hold it on this filp | |
1645 | * (otherwise our open of this file would have blocked). | |
1646 | * And if we are trying to acquire an exclusive lease, | |
1647 | * then the file is not open by anyone (including us) | |
1648 | * except for this filp. | |
1649 | */ | |
c1f24ef4 | 1650 | error = -EAGAIN; |
8634b51f | 1651 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1652 | if (fl->fl_file == filp && |
1653 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1654 | my_fl = fl; |
c1f24ef4 BF |
1655 | continue; |
1656 | } | |
8634b51f | 1657 | |
c1f24ef4 BF |
1658 | /* |
1659 | * No exclusive leases if someone else has a lease on | |
1660 | * this file: | |
1661 | */ | |
1662 | if (arg == F_WRLCK) | |
1663 | goto out; | |
1664 | /* | |
1665 | * Modifying our existing lease is OK, but no getting a | |
1666 | * new lease if someone else is opening for write: | |
1667 | */ | |
1668 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1669 | goto out; | |
1da177e4 LT |
1670 | } |
1671 | ||
8634b51f | 1672 | if (my_fl != NULL) { |
0164bf02 JL |
1673 | lease = my_fl; |
1674 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1675 | if (error) |
1676 | goto out; | |
1677 | goto out_setup; | |
1da177e4 LT |
1678 | } |
1679 | ||
1da177e4 LT |
1680 | error = -EINVAL; |
1681 | if (!leases_enable) | |
1682 | goto out; | |
1683 | ||
e084c1bd | 1684 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1685 | /* |
1686 | * The check in break_lease() is lockless. It's possible for another | |
1687 | * open to race in after we did the earlier check for a conflicting | |
1688 | * open but before the lease was inserted. Check again for a | |
1689 | * conflicting open and cancel the lease if there is one. | |
1690 | * | |
1691 | * We also add a barrier here to ensure that the insertion of the lock | |
1692 | * precedes these checks. | |
1693 | */ | |
1694 | smp_mb(); | |
11afe9f7 | 1695 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
8634b51f | 1696 | if (error) { |
e084c1bd | 1697 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1698 | goto out; |
1699 | } | |
1c7dd2ff JL |
1700 | |
1701 | out_setup: | |
1702 | if (lease->fl_lmops->lm_setup) | |
1703 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1704 | out: |
6109c850 | 1705 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1706 | locks_dispose_list(&dispose); |
df4e8d2c BF |
1707 | if (is_deleg) |
1708 | mutex_unlock(&inode->i_mutex); | |
8634b51f | 1709 | if (!error && !my_fl) |
1c7dd2ff | 1710 | *flp = NULL; |
1da177e4 LT |
1711 | return error; |
1712 | } | |
8335ebd9 | 1713 | |
2ab99ee1 | 1714 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1715 | { |
0efaa7e8 | 1716 | int error = -EAGAIN; |
8634b51f | 1717 | struct file_lock *fl, *victim = NULL; |
6ca7d910 | 1718 | struct inode *inode = file_inode(filp); |
128a3785 | 1719 | struct file_lock_context *ctx; |
c45198ed | 1720 | LIST_HEAD(dispose); |
8335ebd9 | 1721 | |
128a3785 | 1722 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1723 | if (!ctx) { |
1724 | trace_generic_delete_lease(inode, NULL); | |
1725 | return error; | |
1726 | } | |
1727 | ||
6109c850 | 1728 | spin_lock(&ctx->flc_lock); |
8634b51f | 1729 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1730 | if (fl->fl_file == filp && |
1731 | fl->fl_owner == owner) { | |
8634b51f | 1732 | victim = fl; |
0efaa7e8 | 1733 | break; |
8634b51f | 1734 | } |
8335ebd9 | 1735 | } |
a9b1b455 | 1736 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1737 | if (victim) |
7448cc37 | 1738 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1739 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1740 | locks_dispose_list(&dispose); |
0efaa7e8 | 1741 | return error; |
8335ebd9 BF |
1742 | } |
1743 | ||
1744 | /** | |
1745 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1746 | * @filp: file pointer |
1747 | * @arg: type of lease to obtain | |
1748 | * @flp: input - file_lock to use, output - file_lock inserted | |
1749 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1750 | * doesn't require it) | |
8335ebd9 BF |
1751 | * |
1752 | * The (input) flp->fl_lmops->lm_break function is required | |
1753 | * by break_lease(). | |
8335ebd9 | 1754 | */ |
e6f5c789 JL |
1755 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1756 | void **priv) | |
8335ebd9 | 1757 | { |
6ca7d910 | 1758 | struct inode *inode = file_inode(filp); |
8335ebd9 BF |
1759 | int error; |
1760 | ||
8e96e3b7 | 1761 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1762 | return -EACCES; |
1763 | if (!S_ISREG(inode->i_mode)) | |
1764 | return -EINVAL; | |
1765 | error = security_file_lock(filp, arg); | |
1766 | if (error) | |
1767 | return error; | |
1768 | ||
8335ebd9 BF |
1769 | switch (arg) { |
1770 | case F_UNLCK: | |
2ab99ee1 | 1771 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1772 | case F_RDLCK: |
1773 | case F_WRLCK: | |
0efaa7e8 JL |
1774 | if (!(*flp)->fl_lmops->lm_break) { |
1775 | WARN_ON_ONCE(1); | |
1776 | return -ENOLCK; | |
1777 | } | |
11afe9f7 | 1778 | |
e6f5c789 | 1779 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1780 | default: |
8d657eb3 | 1781 | return -EINVAL; |
8335ebd9 BF |
1782 | } |
1783 | } | |
0af1a450 | 1784 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1785 | |
b89f4321 | 1786 | /** |
e51673aa | 1787 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1788 | * @filp: file pointer |
1789 | * @arg: type of lease to obtain | |
1790 | * @lease: file_lock to use when adding a lease | |
1791 | * @priv: private info for lm_setup when adding a lease (may be | |
1792 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1793 | * |
1794 | * Call this to establish a lease on the file. The "lease" argument is not | |
1795 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
1796 | * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set; | |
1797 | * if not, this function will return -ENOLCK (and generate a scary-looking | |
1798 | * stack trace). | |
1c7dd2ff JL |
1799 | * |
1800 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1801 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1802 | */ |
e6f5c789 JL |
1803 | int |
1804 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1805 | { |
1c7dd2ff | 1806 | if (filp->f_op->setlease) |
f82b4b67 | 1807 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1808 | else |
f82b4b67 | 1809 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1810 | } |
a9933cea | 1811 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1812 | |
0ceaf6c7 | 1813 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1814 | { |
1c7dd2ff | 1815 | struct file_lock *fl; |
f7347ce4 | 1816 | struct fasync_struct *new; |
1da177e4 LT |
1817 | int error; |
1818 | ||
c5b1f0d9 AB |
1819 | fl = lease_alloc(filp, arg); |
1820 | if (IS_ERR(fl)) | |
1821 | return PTR_ERR(fl); | |
1da177e4 | 1822 | |
f7347ce4 LT |
1823 | new = fasync_alloc(); |
1824 | if (!new) { | |
1825 | locks_free_lock(fl); | |
1826 | return -ENOMEM; | |
1827 | } | |
1c7dd2ff | 1828 | new->fa_fd = fd; |
f7347ce4 | 1829 | |
1c7dd2ff | 1830 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1831 | if (fl) |
1832 | locks_free_lock(fl); | |
f7347ce4 LT |
1833 | if (new) |
1834 | fasync_free(new); | |
1da177e4 LT |
1835 | return error; |
1836 | } | |
1837 | ||
0ceaf6c7 BF |
1838 | /** |
1839 | * fcntl_setlease - sets a lease on an open file | |
1840 | * @fd: open file descriptor | |
1841 | * @filp: file pointer | |
1842 | * @arg: type of lease to obtain | |
1843 | * | |
1844 | * Call this fcntl to establish a lease on the file. | |
1845 | * Note that you also need to call %F_SETSIG to | |
1846 | * receive a signal when the lease is broken. | |
1847 | */ | |
1848 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1849 | { | |
1850 | if (arg == F_UNLCK) | |
2ab99ee1 | 1851 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
1852 | return do_fcntl_add_lease(fd, filp, arg); |
1853 | } | |
1854 | ||
1da177e4 | 1855 | /** |
29d01b22 JL |
1856 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
1857 | * @inode: inode of the file to apply to | |
1da177e4 LT |
1858 | * @fl: The lock to be applied |
1859 | * | |
29d01b22 | 1860 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 1861 | */ |
616fb38f | 1862 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1863 | { |
1864 | int error; | |
1865 | might_sleep(); | |
1866 | for (;;) { | |
29d01b22 | 1867 | error = flock_lock_inode(inode, fl); |
bde74e4b | 1868 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1869 | break; |
1870 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1871 | if (!error) | |
1872 | continue; | |
1873 | ||
1874 | locks_delete_block(fl); | |
1875 | break; | |
1876 | } | |
1877 | return error; | |
1878 | } | |
1879 | ||
e55c34a6 BC |
1880 | /** |
1881 | * locks_lock_inode_wait - Apply a lock to an inode | |
1882 | * @inode: inode of the file to apply to | |
1883 | * @fl: The lock to be applied | |
1884 | * | |
1885 | * Apply a POSIX or FLOCK style lock request to an inode. | |
1886 | */ | |
1887 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
1888 | { | |
1889 | int res = 0; | |
1890 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
1891 | case FL_POSIX: | |
1892 | res = posix_lock_inode_wait(inode, fl); | |
1893 | break; | |
1894 | case FL_FLOCK: | |
1895 | res = flock_lock_inode_wait(inode, fl); | |
1896 | break; | |
1897 | default: | |
1898 | BUG(); | |
1899 | } | |
1900 | return res; | |
1901 | } | |
1902 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
1903 | ||
1da177e4 LT |
1904 | /** |
1905 | * sys_flock: - flock() system call. | |
1906 | * @fd: the file descriptor to lock. | |
1907 | * @cmd: the type of lock to apply. | |
1908 | * | |
1909 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1910 | * The @cmd can be one of | |
1911 | * | |
1912 | * %LOCK_SH -- a shared lock. | |
1913 | * | |
1914 | * %LOCK_EX -- an exclusive lock. | |
1915 | * | |
1916 | * %LOCK_UN -- remove an existing lock. | |
1917 | * | |
1918 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1919 | * | |
1920 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1921 | * processes read and write access respectively. | |
1922 | */ | |
002c8976 | 1923 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1924 | { |
2903ff01 | 1925 | struct fd f = fdget(fd); |
1da177e4 LT |
1926 | struct file_lock *lock; |
1927 | int can_sleep, unlock; | |
1928 | int error; | |
1929 | ||
1930 | error = -EBADF; | |
2903ff01 | 1931 | if (!f.file) |
1da177e4 LT |
1932 | goto out; |
1933 | ||
1934 | can_sleep = !(cmd & LOCK_NB); | |
1935 | cmd &= ~LOCK_NB; | |
1936 | unlock = (cmd == LOCK_UN); | |
1937 | ||
aeb5d727 | 1938 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1939 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1940 | goto out_putf; |
1941 | ||
6e129d00 JL |
1942 | lock = flock_make_lock(f.file, cmd); |
1943 | if (IS_ERR(lock)) { | |
1944 | error = PTR_ERR(lock); | |
1da177e4 | 1945 | goto out_putf; |
6e129d00 JL |
1946 | } |
1947 | ||
1da177e4 LT |
1948 | if (can_sleep) |
1949 | lock->fl_flags |= FL_SLEEP; | |
1950 | ||
2903ff01 | 1951 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1952 | if (error) |
1953 | goto out_free; | |
1954 | ||
72c2d531 | 1955 | if (f.file->f_op->flock) |
2903ff01 | 1956 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1957 | (can_sleep) ? F_SETLKW : F_SETLK, |
1958 | lock); | |
1959 | else | |
4f656367 | 1960 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
1961 | |
1962 | out_free: | |
993dfa87 | 1963 | locks_free_lock(lock); |
1da177e4 LT |
1964 | |
1965 | out_putf: | |
2903ff01 | 1966 | fdput(f); |
1da177e4 LT |
1967 | out: |
1968 | return error; | |
1969 | } | |
1970 | ||
3ee17abd BF |
1971 | /** |
1972 | * vfs_test_lock - test file byte range lock | |
1973 | * @filp: The file to test lock for | |
6924c554 | 1974 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1975 | * |
1976 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1977 | * setting conf->fl_type to something other than F_UNLCK. | |
1978 | */ | |
1979 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1980 | { | |
72c2d531 | 1981 | if (filp->f_op->lock) |
3ee17abd BF |
1982 | return filp->f_op->lock(filp, F_GETLK, fl); |
1983 | posix_test_lock(filp, fl); | |
1984 | return 0; | |
1985 | } | |
1986 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1987 | ||
c2fa1b8a BF |
1988 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1989 | { | |
cff2fce5 | 1990 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1991 | #if BITS_PER_LONG == 32 |
1992 | /* | |
1993 | * Make sure we can represent the posix lock via | |
1994 | * legacy 32bit flock. | |
1995 | */ | |
1996 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1997 | return -EOVERFLOW; | |
1998 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1999 | return -EOVERFLOW; | |
2000 | #endif | |
2001 | flock->l_start = fl->fl_start; | |
2002 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2003 | fl->fl_end - fl->fl_start + 1; | |
2004 | flock->l_whence = 0; | |
129a84de | 2005 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2006 | return 0; |
2007 | } | |
2008 | ||
2009 | #if BITS_PER_LONG == 32 | |
2010 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2011 | { | |
cff2fce5 | 2012 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2013 | flock->l_start = fl->fl_start; |
2014 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2015 | fl->fl_end - fl->fl_start + 1; | |
2016 | flock->l_whence = 0; | |
2017 | flock->l_type = fl->fl_type; | |
2018 | } | |
2019 | #endif | |
2020 | ||
1da177e4 LT |
2021 | /* Report the first existing lock that would conflict with l. |
2022 | * This implements the F_GETLK command of fcntl(). | |
2023 | */ | |
c1e62b8f | 2024 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 2025 | { |
9d6a8c5c | 2026 | struct file_lock file_lock; |
1da177e4 LT |
2027 | struct flock flock; |
2028 | int error; | |
2029 | ||
2030 | error = -EFAULT; | |
2031 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2032 | goto out; | |
2033 | error = -EINVAL; | |
2034 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2035 | goto out; | |
2036 | ||
2037 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
2038 | if (error) | |
2039 | goto out; | |
2040 | ||
0d3f7a2d | 2041 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2042 | error = -EINVAL; |
2043 | if (flock.l_pid != 0) | |
2044 | goto out; | |
2045 | ||
5d50ffd7 | 2046 | cmd = F_GETLK; |
cff2fce5 | 2047 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2048 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2049 | } |
2050 | ||
3ee17abd BF |
2051 | error = vfs_test_lock(filp, &file_lock); |
2052 | if (error) | |
2053 | goto out; | |
1da177e4 | 2054 | |
9d6a8c5c ME |
2055 | flock.l_type = file_lock.fl_type; |
2056 | if (file_lock.fl_type != F_UNLCK) { | |
2057 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 2058 | if (error) |
f328296e | 2059 | goto rel_priv; |
1da177e4 LT |
2060 | } |
2061 | error = -EFAULT; | |
2062 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2063 | error = 0; | |
f328296e KM |
2064 | rel_priv: |
2065 | locks_release_private(&file_lock); | |
1da177e4 LT |
2066 | out: |
2067 | return error; | |
2068 | } | |
2069 | ||
7723ec97 ME |
2070 | /** |
2071 | * vfs_lock_file - file byte range lock | |
2072 | * @filp: The file to apply the lock to | |
2073 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2074 | * @fl: The lock to be applied | |
150b3934 ME |
2075 | * @conf: Place to return a copy of the conflicting lock, if found. |
2076 | * | |
2077 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2078 | * as the final argument. | |
2079 | * | |
2080 | * If the filesystem defines a private ->lock() method, then @conf will | |
2081 | * be left unchanged; so a caller that cares should initialize it to | |
2082 | * some acceptable default. | |
2beb6614 ME |
2083 | * |
2084 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2085 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2086 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2087 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2088 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2089 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2090 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2091 | * request completes. |
2092 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2093 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2094 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2095 | * nonzero return code and the file system should release the lock. The file |
2096 | * system is also responsible to keep a corresponding posix lock when it | |
2097 | * grants a lock so the VFS can find out which locks are locally held and do | |
2098 | * the correct lock cleanup when required. | |
2099 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2100 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2101 | * return code. |
7723ec97 | 2102 | */ |
150b3934 | 2103 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2104 | { |
72c2d531 | 2105 | if (filp->f_op->lock) |
7723ec97 ME |
2106 | return filp->f_op->lock(filp, cmd, fl); |
2107 | else | |
150b3934 | 2108 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2109 | } |
2110 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2111 | ||
b648a6de MS |
2112 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2113 | struct file_lock *fl) | |
2114 | { | |
2115 | int error; | |
2116 | ||
2117 | error = security_file_lock(filp, fl->fl_type); | |
2118 | if (error) | |
2119 | return error; | |
2120 | ||
764c76b3 MS |
2121 | for (;;) { |
2122 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2123 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2124 | break; |
764c76b3 MS |
2125 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2126 | if (!error) | |
2127 | continue; | |
2128 | ||
2129 | locks_delete_block(fl); | |
2130 | break; | |
b648a6de MS |
2131 | } |
2132 | ||
2133 | return error; | |
2134 | } | |
2135 | ||
6ca7d910 | 2136 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2137 | static int |
2138 | check_fmode_for_setlk(struct file_lock *fl) | |
2139 | { | |
2140 | switch (fl->fl_type) { | |
2141 | case F_RDLCK: | |
2142 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2143 | return -EBADF; | |
2144 | break; | |
2145 | case F_WRLCK: | |
2146 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2147 | return -EBADF; | |
2148 | } | |
2149 | return 0; | |
2150 | } | |
2151 | ||
1da177e4 LT |
2152 | /* Apply the lock described by l to an open file descriptor. |
2153 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2154 | */ | |
c293621b PS |
2155 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2156 | struct flock __user *l) | |
1da177e4 LT |
2157 | { |
2158 | struct file_lock *file_lock = locks_alloc_lock(); | |
2159 | struct flock flock; | |
2160 | struct inode *inode; | |
0b2bac2f | 2161 | struct file *f; |
1da177e4 LT |
2162 | int error; |
2163 | ||
2164 | if (file_lock == NULL) | |
2165 | return -ENOLCK; | |
2166 | ||
2167 | /* | |
2168 | * This might block, so we do it before checking the inode. | |
2169 | */ | |
2170 | error = -EFAULT; | |
2171 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2172 | goto out; | |
2173 | ||
496ad9aa | 2174 | inode = file_inode(filp); |
1da177e4 LT |
2175 | |
2176 | /* Don't allow mandatory locks on files that may be memory mapped | |
2177 | * and shared. | |
2178 | */ | |
a16877ca | 2179 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2180 | error = -EAGAIN; |
2181 | goto out; | |
2182 | } | |
2183 | ||
c293621b | 2184 | again: |
1da177e4 LT |
2185 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2186 | if (error) | |
2187 | goto out; | |
5d50ffd7 | 2188 | |
cf01f4ee JL |
2189 | error = check_fmode_for_setlk(file_lock); |
2190 | if (error) | |
2191 | goto out; | |
2192 | ||
5d50ffd7 JL |
2193 | /* |
2194 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2195 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2196 | */ |
2197 | switch (cmd) { | |
0d3f7a2d | 2198 | case F_OFD_SETLK: |
90478939 JL |
2199 | error = -EINVAL; |
2200 | if (flock.l_pid != 0) | |
2201 | goto out; | |
2202 | ||
5d50ffd7 | 2203 | cmd = F_SETLK; |
cff2fce5 | 2204 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2205 | file_lock->fl_owner = filp; |
5d50ffd7 | 2206 | break; |
0d3f7a2d | 2207 | case F_OFD_SETLKW: |
90478939 JL |
2208 | error = -EINVAL; |
2209 | if (flock.l_pid != 0) | |
2210 | goto out; | |
2211 | ||
5d50ffd7 | 2212 | cmd = F_SETLKW; |
cff2fce5 | 2213 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2214 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2215 | /* Fallthrough */ |
2216 | case F_SETLKW: | |
1da177e4 LT |
2217 | file_lock->fl_flags |= FL_SLEEP; |
2218 | } | |
5d50ffd7 | 2219 | |
b648a6de | 2220 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2221 | |
c293621b PS |
2222 | /* |
2223 | * Attempt to detect a close/fcntl race and recover by | |
2224 | * releasing the lock that was just acquired. | |
2225 | */ | |
0b2bac2f AV |
2226 | /* |
2227 | * we need that spin_lock here - it prevents reordering between | |
8116bf4c | 2228 | * update of i_flctx->flc_posix and check for it done in close(). |
0b2bac2f AV |
2229 | * rcu_read_lock() wouldn't do. |
2230 | */ | |
2231 | spin_lock(¤t->files->file_lock); | |
2232 | f = fcheck(fd); | |
2233 | spin_unlock(¤t->files->file_lock); | |
2234 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2235 | flock.l_type = F_UNLCK; |
2236 | goto again; | |
1da177e4 LT |
2237 | } |
2238 | ||
c293621b | 2239 | out: |
1da177e4 LT |
2240 | locks_free_lock(file_lock); |
2241 | return error; | |
2242 | } | |
2243 | ||
2244 | #if BITS_PER_LONG == 32 | |
2245 | /* Report the first existing lock that would conflict with l. | |
2246 | * This implements the F_GETLK command of fcntl(). | |
2247 | */ | |
c1e62b8f | 2248 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2249 | { |
9d6a8c5c | 2250 | struct file_lock file_lock; |
1da177e4 LT |
2251 | struct flock64 flock; |
2252 | int error; | |
2253 | ||
2254 | error = -EFAULT; | |
2255 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2256 | goto out; | |
2257 | error = -EINVAL; | |
2258 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2259 | goto out; | |
2260 | ||
2261 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2262 | if (error) | |
2263 | goto out; | |
2264 | ||
0d3f7a2d | 2265 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2266 | error = -EINVAL; |
2267 | if (flock.l_pid != 0) | |
2268 | goto out; | |
2269 | ||
5d50ffd7 | 2270 | cmd = F_GETLK64; |
cff2fce5 | 2271 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2272 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2273 | } |
2274 | ||
3ee17abd BF |
2275 | error = vfs_test_lock(filp, &file_lock); |
2276 | if (error) | |
2277 | goto out; | |
2278 | ||
9d6a8c5c ME |
2279 | flock.l_type = file_lock.fl_type; |
2280 | if (file_lock.fl_type != F_UNLCK) | |
2281 | posix_lock_to_flock64(&flock, &file_lock); | |
2282 | ||
1da177e4 LT |
2283 | error = -EFAULT; |
2284 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2285 | error = 0; | |
f328296e KM |
2286 | |
2287 | locks_release_private(&file_lock); | |
1da177e4 LT |
2288 | out: |
2289 | return error; | |
2290 | } | |
2291 | ||
2292 | /* Apply the lock described by l to an open file descriptor. | |
2293 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2294 | */ | |
c293621b PS |
2295 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2296 | struct flock64 __user *l) | |
1da177e4 LT |
2297 | { |
2298 | struct file_lock *file_lock = locks_alloc_lock(); | |
2299 | struct flock64 flock; | |
2300 | struct inode *inode; | |
0b2bac2f | 2301 | struct file *f; |
1da177e4 LT |
2302 | int error; |
2303 | ||
2304 | if (file_lock == NULL) | |
2305 | return -ENOLCK; | |
2306 | ||
2307 | /* | |
2308 | * This might block, so we do it before checking the inode. | |
2309 | */ | |
2310 | error = -EFAULT; | |
2311 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2312 | goto out; | |
2313 | ||
496ad9aa | 2314 | inode = file_inode(filp); |
1da177e4 LT |
2315 | |
2316 | /* Don't allow mandatory locks on files that may be memory mapped | |
2317 | * and shared. | |
2318 | */ | |
a16877ca | 2319 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2320 | error = -EAGAIN; |
2321 | goto out; | |
2322 | } | |
2323 | ||
c293621b | 2324 | again: |
1da177e4 LT |
2325 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2326 | if (error) | |
2327 | goto out; | |
5d50ffd7 | 2328 | |
cf01f4ee JL |
2329 | error = check_fmode_for_setlk(file_lock); |
2330 | if (error) | |
2331 | goto out; | |
2332 | ||
5d50ffd7 JL |
2333 | /* |
2334 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2335 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2336 | */ |
2337 | switch (cmd) { | |
0d3f7a2d | 2338 | case F_OFD_SETLK: |
90478939 JL |
2339 | error = -EINVAL; |
2340 | if (flock.l_pid != 0) | |
2341 | goto out; | |
2342 | ||
5d50ffd7 | 2343 | cmd = F_SETLK64; |
cff2fce5 | 2344 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2345 | file_lock->fl_owner = filp; |
5d50ffd7 | 2346 | break; |
0d3f7a2d | 2347 | case F_OFD_SETLKW: |
90478939 JL |
2348 | error = -EINVAL; |
2349 | if (flock.l_pid != 0) | |
2350 | goto out; | |
2351 | ||
5d50ffd7 | 2352 | cmd = F_SETLKW64; |
cff2fce5 | 2353 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2354 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2355 | /* Fallthrough */ |
2356 | case F_SETLKW64: | |
1da177e4 LT |
2357 | file_lock->fl_flags |= FL_SLEEP; |
2358 | } | |
5d50ffd7 | 2359 | |
b648a6de | 2360 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2361 | |
c293621b PS |
2362 | /* |
2363 | * Attempt to detect a close/fcntl race and recover by | |
2364 | * releasing the lock that was just acquired. | |
2365 | */ | |
0b2bac2f AV |
2366 | spin_lock(¤t->files->file_lock); |
2367 | f = fcheck(fd); | |
2368 | spin_unlock(¤t->files->file_lock); | |
2369 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2370 | flock.l_type = F_UNLCK; |
2371 | goto again; | |
1da177e4 LT |
2372 | } |
2373 | ||
2374 | out: | |
2375 | locks_free_lock(file_lock); | |
2376 | return error; | |
2377 | } | |
2378 | #endif /* BITS_PER_LONG == 32 */ | |
2379 | ||
2380 | /* | |
2381 | * This function is called when the file is being removed | |
2382 | * from the task's fd array. POSIX locks belonging to this task | |
2383 | * are deleted at this time. | |
2384 | */ | |
2385 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2386 | { | |
ff7b86b8 | 2387 | struct file_lock lock; |
128a3785 | 2388 | struct file_lock_context *ctx; |
1da177e4 LT |
2389 | |
2390 | /* | |
2391 | * If there are no locks held on this file, we don't need to call | |
2392 | * posix_lock_file(). Another process could be setting a lock on this | |
2393 | * file at the same time, but we wouldn't remove that lock anyway. | |
2394 | */ | |
128a3785 | 2395 | ctx = smp_load_acquire(&file_inode(filp)->i_flctx); |
bd61e0a9 | 2396 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2397 | return; |
2398 | ||
2399 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2400 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2401 | lock.fl_start = 0; |
2402 | lock.fl_end = OFFSET_MAX; | |
2403 | lock.fl_owner = owner; | |
2404 | lock.fl_pid = current->tgid; | |
2405 | lock.fl_file = filp; | |
2406 | lock.fl_ops = NULL; | |
2407 | lock.fl_lmops = NULL; | |
2408 | ||
150b3934 | 2409 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2410 | |
1da177e4 LT |
2411 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2412 | lock.fl_ops->fl_release_private(&lock); | |
2413 | } | |
2414 | ||
2415 | EXPORT_SYMBOL(locks_remove_posix); | |
2416 | ||
3d8e560d | 2417 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2418 | static void |
128a3785 | 2419 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 JL |
2420 | { |
2421 | struct file_lock fl = { | |
2422 | .fl_owner = filp, | |
2423 | .fl_pid = current->tgid, | |
2424 | .fl_file = filp, | |
2425 | .fl_flags = FL_FLOCK, | |
2426 | .fl_type = F_UNLCK, | |
2427 | .fl_end = OFFSET_MAX, | |
2428 | }; | |
bcd7f78d | 2429 | struct inode *inode = file_inode(filp); |
dd459bb1 | 2430 | |
3d8e560d | 2431 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2432 | return; |
2433 | ||
2434 | if (filp->f_op->flock) | |
2435 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
2436 | else | |
bcd7f78d | 2437 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2438 | |
2439 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2440 | fl.fl_ops->fl_release_private(&fl); | |
2441 | } | |
2442 | ||
3d8e560d | 2443 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2444 | static void |
128a3785 | 2445 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2446 | { |
8634b51f JL |
2447 | struct file_lock *fl, *tmp; |
2448 | LIST_HEAD(dispose); | |
2449 | ||
3d8e560d | 2450 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2451 | return; |
2452 | ||
6109c850 | 2453 | spin_lock(&ctx->flc_lock); |
8634b51f | 2454 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2455 | if (filp == fl->fl_file) |
2456 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2457 | spin_unlock(&ctx->flc_lock); |
8634b51f JL |
2458 | locks_dispose_list(&dispose); |
2459 | } | |
2460 | ||
1da177e4 LT |
2461 | /* |
2462 | * This function is called on the last close of an open file. | |
2463 | */ | |
78ed8a13 | 2464 | void locks_remove_file(struct file *filp) |
1da177e4 | 2465 | { |
128a3785 DV |
2466 | struct file_lock_context *ctx; |
2467 | ||
2468 | ctx = smp_load_acquire(&file_inode(filp)->i_flctx); | |
2469 | if (!ctx) | |
3d8e560d JL |
2470 | return; |
2471 | ||
dd459bb1 | 2472 | /* remove any OFD locks */ |
73a8f5f7 | 2473 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2474 | |
dd459bb1 | 2475 | /* remove flock locks */ |
128a3785 | 2476 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2477 | |
8634b51f | 2478 | /* remove any leases */ |
128a3785 | 2479 | locks_remove_lease(filp, ctx); |
1da177e4 LT |
2480 | } |
2481 | ||
1da177e4 LT |
2482 | /** |
2483 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2484 | * @waiter: the lock which was waiting |
2485 | * | |
2486 | * lockd needs to block waiting for locks. | |
2487 | */ | |
64a318ee | 2488 | int |
f891a29f | 2489 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2490 | { |
64a318ee BF |
2491 | int status = 0; |
2492 | ||
7b2296af | 2493 | spin_lock(&blocked_lock_lock); |
5996a298 | 2494 | if (waiter->fl_next) |
1da177e4 | 2495 | __locks_delete_block(waiter); |
64a318ee BF |
2496 | else |
2497 | status = -ENOENT; | |
7b2296af | 2498 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2499 | return status; |
1da177e4 | 2500 | } |
1da177e4 LT |
2501 | EXPORT_SYMBOL(posix_unblock_lock); |
2502 | ||
9b9d2ab4 ME |
2503 | /** |
2504 | * vfs_cancel_lock - file byte range unblock lock | |
2505 | * @filp: The file to apply the unblock to | |
2506 | * @fl: The lock to be unblocked | |
2507 | * | |
2508 | * Used by lock managers to cancel blocked requests | |
2509 | */ | |
2510 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2511 | { | |
72c2d531 | 2512 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2513 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2514 | return 0; | |
2515 | } | |
2516 | ||
2517 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2518 | ||
7f8ada98 | 2519 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2520 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2521 | #include <linux/seq_file.h> |
2522 | ||
7012b02a JL |
2523 | struct locks_iterator { |
2524 | int li_cpu; | |
2525 | loff_t li_pos; | |
2526 | }; | |
2527 | ||
7f8ada98 | 2528 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2529 | loff_t id, char *pfx) |
1da177e4 LT |
2530 | { |
2531 | struct inode *inode = NULL; | |
ab1f1611 VG |
2532 | unsigned int fl_pid; |
2533 | ||
2534 | if (fl->fl_nspid) | |
6c5f3e7b | 2535 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2536 | else |
2537 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2538 | |
2539 | if (fl->fl_file != NULL) | |
496ad9aa | 2540 | inode = file_inode(fl->fl_file); |
1da177e4 | 2541 | |
99dc8292 | 2542 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2543 | if (IS_POSIX(fl)) { |
c918d42a | 2544 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2545 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2546 | else if (IS_OFDLCK(fl)) |
5315c26a | 2547 | seq_puts(f, "OFDLCK"); |
c918d42a | 2548 | else |
5315c26a | 2549 | seq_puts(f, "POSIX "); |
c918d42a JL |
2550 | |
2551 | seq_printf(f, " %s ", | |
1da177e4 | 2552 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2553 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2554 | } else if (IS_FLOCK(fl)) { |
2555 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2556 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2557 | } else { |
5315c26a | 2558 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2559 | } |
2560 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2561 | if (fl->fl_flags & FL_DELEG) |
2562 | seq_puts(f, "DELEG "); | |
2563 | else | |
2564 | seq_puts(f, "LEASE "); | |
2565 | ||
ab83fa4b | 2566 | if (lease_breaking(fl)) |
5315c26a | 2567 | seq_puts(f, "BREAKING "); |
1da177e4 | 2568 | else if (fl->fl_file) |
5315c26a | 2569 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2570 | else |
5315c26a | 2571 | seq_puts(f, "BREAKER "); |
1da177e4 | 2572 | } else { |
5315c26a | 2573 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2574 | } |
2575 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2576 | seq_printf(f, "%s ", |
1da177e4 LT |
2577 | (fl->fl_type & LOCK_READ) |
2578 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2579 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2580 | } else { | |
7f8ada98 | 2581 | seq_printf(f, "%s ", |
ab83fa4b | 2582 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2583 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2584 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2585 | } |
2586 | if (inode) { | |
3648888e | 2587 | /* userspace relies on this representation of dev_t */ |
ab1f1611 | 2588 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2589 | MAJOR(inode->i_sb->s_dev), |
2590 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2591 | } else { |
ab1f1611 | 2592 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2593 | } |
2594 | if (IS_POSIX(fl)) { | |
2595 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2596 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2597 | else |
7f8ada98 | 2598 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2599 | } else { |
5315c26a | 2600 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2601 | } |
2602 | } | |
2603 | ||
7f8ada98 | 2604 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2605 | { |
7012b02a | 2606 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2607 | struct file_lock *fl, *bfl; |
1da177e4 | 2608 | |
139ca04e | 2609 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2610 | |
7012b02a | 2611 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2612 | |
7f8ada98 | 2613 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2614 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2615 | |
7f8ada98 PE |
2616 | return 0; |
2617 | } | |
1da177e4 | 2618 | |
6c8c9031 AV |
2619 | static void __show_fd_locks(struct seq_file *f, |
2620 | struct list_head *head, int *id, | |
2621 | struct file *filp, struct files_struct *files) | |
2622 | { | |
2623 | struct file_lock *fl; | |
2624 | ||
2625 | list_for_each_entry(fl, head, fl_list) { | |
2626 | ||
2627 | if (filp != fl->fl_file) | |
2628 | continue; | |
2629 | if (fl->fl_owner != files && | |
2630 | fl->fl_owner != filp) | |
2631 | continue; | |
2632 | ||
2633 | (*id)++; | |
2634 | seq_puts(f, "lock:\t"); | |
2635 | lock_get_status(f, fl, *id, ""); | |
2636 | } | |
2637 | } | |
2638 | ||
2639 | void show_fd_locks(struct seq_file *f, | |
2640 | struct file *filp, struct files_struct *files) | |
2641 | { | |
2642 | struct inode *inode = file_inode(filp); | |
2643 | struct file_lock_context *ctx; | |
2644 | int id = 0; | |
2645 | ||
128a3785 | 2646 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2647 | if (!ctx) |
2648 | return; | |
2649 | ||
2650 | spin_lock(&ctx->flc_lock); | |
2651 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2652 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2653 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2654 | spin_unlock(&ctx->flc_lock); | |
2655 | } | |
2656 | ||
7f8ada98 | 2657 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2658 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2659 | { |
7012b02a | 2660 | struct locks_iterator *iter = f->private; |
99dc8292 | 2661 | |
7012b02a JL |
2662 | iter->li_pos = *pos + 1; |
2663 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2664 | spin_lock(&blocked_lock_lock); |
7012b02a | 2665 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2666 | } |
1da177e4 | 2667 | |
7f8ada98 PE |
2668 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2669 | { | |
7012b02a JL |
2670 | struct locks_iterator *iter = f->private; |
2671 | ||
2672 | ++iter->li_pos; | |
2673 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2674 | } |
1da177e4 | 2675 | |
7f8ada98 | 2676 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2677 | __releases(&blocked_lock_lock) |
7f8ada98 | 2678 | { |
7b2296af | 2679 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2680 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2681 | } |
2682 | ||
d8ba7a36 | 2683 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2684 | .start = locks_start, |
2685 | .next = locks_next, | |
2686 | .stop = locks_stop, | |
2687 | .show = locks_show, | |
2688 | }; | |
d8ba7a36 AD |
2689 | |
2690 | static int locks_open(struct inode *inode, struct file *filp) | |
2691 | { | |
7012b02a JL |
2692 | return seq_open_private(filp, &locks_seq_operations, |
2693 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2694 | } |
2695 | ||
2696 | static const struct file_operations proc_locks_operations = { | |
2697 | .open = locks_open, | |
2698 | .read = seq_read, | |
2699 | .llseek = seq_lseek, | |
99dc8292 | 2700 | .release = seq_release_private, |
d8ba7a36 AD |
2701 | }; |
2702 | ||
2703 | static int __init proc_locks_init(void) | |
2704 | { | |
2705 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2706 | return 0; | |
2707 | } | |
91899226 | 2708 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2709 | #endif |
2710 | ||
1da177e4 LT |
2711 | static int __init filelock_init(void) |
2712 | { | |
7012b02a JL |
2713 | int i; |
2714 | ||
4a075e39 JL |
2715 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2716 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2717 | ||
1da177e4 | 2718 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2719 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2720 | ||
7012b02a JL |
2721 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2722 | ||
2723 | for_each_possible_cpu(i) | |
2724 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2725 | ||
1da177e4 LT |
2726 | return 0; |
2727 | } | |
2728 | ||
2729 | core_initcall(filelock_init); |