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. | |
fd7732e0 N |
115 | * |
116 | * Locking conflicts and dependencies: | |
117 | * If multiple threads attempt to lock the same byte (or flock the same file) | |
118 | * only one can be granted the lock, and other must wait their turn. | |
119 | * The first lock has been "applied" or "granted", the others are "waiting" | |
120 | * and are "blocked" by the "applied" lock.. | |
121 | * | |
122 | * Waiting and applied locks are all kept in trees whose properties are: | |
123 | * | |
124 | * - the root of a tree may be an applied or waiting lock. | |
125 | * - every other node in the tree is a waiting lock that | |
126 | * conflicts with every ancestor of that node. | |
127 | * | |
128 | * Every such tree begins life as a waiting singleton which obviously | |
129 | * satisfies the above properties. | |
130 | * | |
131 | * The only ways we modify trees preserve these properties: | |
132 | * | |
133 | * 1. We may add a new leaf node, but only after first verifying that it | |
134 | * conflicts with all of its ancestors. | |
135 | * 2. We may remove the root of a tree, creating a new singleton | |
136 | * tree from the root and N new trees rooted in the immediate | |
137 | * children. | |
138 | * 3. If the root of a tree is not currently an applied lock, we may | |
139 | * apply it (if possible). | |
140 | * 4. We may upgrade the root of the tree (either extend its range, | |
141 | * or upgrade its entire range from read to write). | |
142 | * | |
143 | * When an applied lock is modified in a way that reduces or downgrades any | |
144 | * part of its range, we remove all its children (2 above). This particularly | |
145 | * happens when a lock is unlocked. | |
146 | * | |
147 | * For each of those child trees we "wake up" the thread which is | |
148 | * waiting for the lock so it can continue handling as follows: if the | |
149 | * root of the tree applies, we do so (3). If it doesn't, it must | |
150 | * conflict with some applied lock. We remove (wake up) all of its children | |
151 | * (2), and add it is a new leaf to the tree rooted in the applied | |
152 | * lock (1). We then repeat the process recursively with those | |
153 | * children. | |
154 | * | |
1da177e4 LT |
155 | */ |
156 | ||
157 | #include <linux/capability.h> | |
158 | #include <linux/file.h> | |
9f3acc31 | 159 | #include <linux/fdtable.h> |
1da177e4 LT |
160 | #include <linux/fs.h> |
161 | #include <linux/init.h> | |
1da177e4 LT |
162 | #include <linux/security.h> |
163 | #include <linux/slab.h> | |
1da177e4 LT |
164 | #include <linux/syscalls.h> |
165 | #include <linux/time.h> | |
4fb3a538 | 166 | #include <linux/rcupdate.h> |
ab1f1611 | 167 | #include <linux/pid_namespace.h> |
48f74186 | 168 | #include <linux/hashtable.h> |
7012b02a | 169 | #include <linux/percpu.h> |
1da177e4 | 170 | |
62af4f1f JL |
171 | #define CREATE_TRACE_POINTS |
172 | #include <trace/events/filelock.h> | |
173 | ||
7c0f6ba6 | 174 | #include <linux/uaccess.h> |
1da177e4 LT |
175 | |
176 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
177 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
11afe9f7 | 178 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) |
cff2fce5 | 179 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
9d5b86ac | 180 | #define IS_REMOTELCK(fl) (fl->fl_pid <= 0) |
1da177e4 | 181 | |
ab83fa4b BF |
182 | static bool lease_breaking(struct file_lock *fl) |
183 | { | |
778fc546 BF |
184 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
185 | } | |
186 | ||
187 | static int target_leasetype(struct file_lock *fl) | |
188 | { | |
189 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
190 | return F_UNLCK; | |
191 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
192 | return F_RDLCK; | |
193 | return fl->fl_type; | |
ab83fa4b BF |
194 | } |
195 | ||
1da177e4 LT |
196 | int leases_enable = 1; |
197 | int lease_break_time = 45; | |
198 | ||
1c8c601a | 199 | /* |
7012b02a | 200 | * The global file_lock_list is only used for displaying /proc/locks, so we |
7c3f654d PZ |
201 | * keep a list on each CPU, with each list protected by its own spinlock. |
202 | * Global serialization is done using file_rwsem. | |
203 | * | |
204 | * Note that alterations to the list also require that the relevant flc_lock is | |
205 | * held. | |
1c8c601a | 206 | */ |
7c3f654d PZ |
207 | struct file_lock_list_struct { |
208 | spinlock_t lock; | |
209 | struct hlist_head hlist; | |
210 | }; | |
211 | static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list); | |
aba37660 | 212 | DEFINE_STATIC_PERCPU_RWSEM(file_rwsem); |
88974691 | 213 | |
1c8c601a | 214 | /* |
48f74186 | 215 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 216 | * It is protected by blocked_lock_lock. |
48f74186 JL |
217 | * |
218 | * We hash locks by lockowner in order to optimize searching for the lock a | |
219 | * particular lockowner is waiting on. | |
220 | * | |
221 | * FIXME: make this value scale via some heuristic? We generally will want more | |
222 | * buckets when we have more lockowners holding locks, but that's a little | |
223 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 224 | */ |
48f74186 JL |
225 | #define BLOCKED_HASH_BITS 7 |
226 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 227 | |
1c8c601a | 228 | /* |
7b2296af JL |
229 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
230 | * want to be holding this lock. | |
1c8c601a | 231 | * |
ada5c1da N |
232 | * In addition, it also protects the fl->fl_blocked_requests list, and the |
233 | * fl->fl_blocker pointer for file_lock structures that are acting as lock | |
234 | * requests (in contrast to those that are acting as records of acquired locks). | |
1c8c601a JL |
235 | * |
236 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 237 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 238 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 239 | * flc_lock. |
1c8c601a | 240 | */ |
7b2296af | 241 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 242 | |
4a075e39 | 243 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 244 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 245 | |
4a075e39 | 246 | static struct file_lock_context * |
5c1c669a | 247 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 248 | { |
128a3785 | 249 | struct file_lock_context *ctx; |
4a075e39 | 250 | |
128a3785 DV |
251 | /* paired with cmpxchg() below */ |
252 | ctx = smp_load_acquire(&inode->i_flctx); | |
253 | if (likely(ctx) || type == F_UNLCK) | |
4a075e39 JL |
254 | goto out; |
255 | ||
128a3785 DV |
256 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
257 | if (!ctx) | |
4a075e39 JL |
258 | goto out; |
259 | ||
128a3785 DV |
260 | spin_lock_init(&ctx->flc_lock); |
261 | INIT_LIST_HEAD(&ctx->flc_flock); | |
262 | INIT_LIST_HEAD(&ctx->flc_posix); | |
263 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
264 | |
265 | /* | |
266 | * Assign the pointer if it's not already assigned. If it is, then | |
267 | * free the context we just allocated. | |
268 | */ | |
128a3785 DV |
269 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
270 | kmem_cache_free(flctx_cache, ctx); | |
271 | ctx = smp_load_acquire(&inode->i_flctx); | |
272 | } | |
4a075e39 | 273 | out: |
1890910f | 274 | trace_locks_get_lock_context(inode, type, ctx); |
128a3785 | 275 | return ctx; |
4a075e39 JL |
276 | } |
277 | ||
e24dadab JL |
278 | static void |
279 | locks_dump_ctx_list(struct list_head *list, char *list_type) | |
280 | { | |
281 | struct file_lock *fl; | |
282 | ||
283 | list_for_each_entry(fl, list, fl_list) { | |
284 | pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type, fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
285 | } | |
286 | } | |
287 | ||
288 | static void | |
289 | locks_check_ctx_lists(struct inode *inode) | |
290 | { | |
291 | struct file_lock_context *ctx = inode->i_flctx; | |
292 | ||
293 | if (unlikely(!list_empty(&ctx->flc_flock) || | |
294 | !list_empty(&ctx->flc_posix) || | |
295 | !list_empty(&ctx->flc_lease))) { | |
296 | pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n", | |
297 | MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), | |
298 | inode->i_ino); | |
299 | locks_dump_ctx_list(&ctx->flc_flock, "FLOCK"); | |
300 | locks_dump_ctx_list(&ctx->flc_posix, "POSIX"); | |
301 | locks_dump_ctx_list(&ctx->flc_lease, "LEASE"); | |
302 | } | |
303 | } | |
304 | ||
3953704f BC |
305 | static void |
306 | locks_check_ctx_file_list(struct file *filp, struct list_head *list, | |
307 | char *list_type) | |
308 | { | |
309 | struct file_lock *fl; | |
310 | struct inode *inode = locks_inode(filp); | |
311 | ||
312 | list_for_each_entry(fl, list, fl_list) | |
313 | if (fl->fl_file == filp) | |
314 | pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx " | |
315 | " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", | |
316 | list_type, MAJOR(inode->i_sb->s_dev), | |
317 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
318 | fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
319 | } | |
320 | ||
4a075e39 | 321 | void |
f27a0fe0 | 322 | locks_free_lock_context(struct inode *inode) |
4a075e39 | 323 | { |
f27a0fe0 JL |
324 | struct file_lock_context *ctx = inode->i_flctx; |
325 | ||
e24dadab JL |
326 | if (unlikely(ctx)) { |
327 | locks_check_ctx_lists(inode); | |
4a075e39 JL |
328 | kmem_cache_free(flctx_cache, ctx); |
329 | } | |
330 | } | |
331 | ||
ee19cc40 | 332 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 333 | { |
139ca04e | 334 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 335 | INIT_LIST_HEAD(&fl->fl_list); |
ada5c1da N |
336 | INIT_LIST_HEAD(&fl->fl_blocked_requests); |
337 | INIT_LIST_HEAD(&fl->fl_blocked_member); | |
ee19cc40 | 338 | init_waitqueue_head(&fl->fl_wait); |
a51cb91d MS |
339 | } |
340 | ||
1da177e4 | 341 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 342 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 343 | { |
ee19cc40 | 344 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
345 | |
346 | if (fl) | |
ee19cc40 | 347 | locks_init_lock_heads(fl); |
a51cb91d MS |
348 | |
349 | return fl; | |
1da177e4 | 350 | } |
c5b1f0d9 | 351 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 352 | |
a9e61e25 | 353 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
354 | { |
355 | if (fl->fl_ops) { | |
356 | if (fl->fl_ops->fl_release_private) | |
357 | fl->fl_ops->fl_release_private(fl); | |
358 | fl->fl_ops = NULL; | |
359 | } | |
47831f35 | 360 | |
5c97d7b1 | 361 | if (fl->fl_lmops) { |
cae80b30 JL |
362 | if (fl->fl_lmops->lm_put_owner) { |
363 | fl->fl_lmops->lm_put_owner(fl->fl_owner); | |
364 | fl->fl_owner = NULL; | |
365 | } | |
5c97d7b1 KM |
366 | fl->fl_lmops = NULL; |
367 | } | |
47831f35 | 368 | } |
a9e61e25 | 369 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 370 | |
1da177e4 | 371 | /* Free a lock which is not in use. */ |
05fa3135 | 372 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 373 | { |
5ce29646 | 374 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 375 | BUG_ON(!list_empty(&fl->fl_list)); |
ada5c1da N |
376 | BUG_ON(!list_empty(&fl->fl_blocked_requests)); |
377 | BUG_ON(!list_empty(&fl->fl_blocked_member)); | |
139ca04e | 378 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 379 | |
47831f35 | 380 | locks_release_private(fl); |
1da177e4 LT |
381 | kmem_cache_free(filelock_cache, fl); |
382 | } | |
05fa3135 | 383 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 384 | |
ed9814d8 JL |
385 | static void |
386 | locks_dispose_list(struct list_head *dispose) | |
387 | { | |
388 | struct file_lock *fl; | |
389 | ||
390 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
391 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
392 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
393 | locks_free_lock(fl); |
394 | } | |
395 | } | |
396 | ||
1da177e4 LT |
397 | void locks_init_lock(struct file_lock *fl) |
398 | { | |
ee19cc40 MS |
399 | memset(fl, 0, sizeof(struct file_lock)); |
400 | locks_init_lock_heads(fl); | |
1da177e4 LT |
401 | } |
402 | ||
403 | EXPORT_SYMBOL(locks_init_lock); | |
404 | ||
1da177e4 LT |
405 | /* |
406 | * Initialize a new lock from an existing file_lock structure. | |
407 | */ | |
3fe0fff1 | 408 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
409 | { |
410 | new->fl_owner = fl->fl_owner; | |
411 | new->fl_pid = fl->fl_pid; | |
0996905f | 412 | new->fl_file = NULL; |
1da177e4 LT |
413 | new->fl_flags = fl->fl_flags; |
414 | new->fl_type = fl->fl_type; | |
415 | new->fl_start = fl->fl_start; | |
416 | new->fl_end = fl->fl_end; | |
f328296e | 417 | new->fl_lmops = fl->fl_lmops; |
0996905f | 418 | new->fl_ops = NULL; |
f328296e KM |
419 | |
420 | if (fl->fl_lmops) { | |
421 | if (fl->fl_lmops->lm_get_owner) | |
cae80b30 | 422 | fl->fl_lmops->lm_get_owner(fl->fl_owner); |
f328296e | 423 | } |
0996905f | 424 | } |
3fe0fff1 | 425 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
426 | |
427 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
428 | { | |
566709bd JL |
429 | /* "new" must be a freshly-initialized lock */ |
430 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 431 | |
3fe0fff1 | 432 | locks_copy_conflock(new, fl); |
f328296e | 433 | |
0996905f | 434 | new->fl_file = fl->fl_file; |
1da177e4 | 435 | new->fl_ops = fl->fl_ops; |
47831f35 | 436 | |
f328296e KM |
437 | if (fl->fl_ops) { |
438 | if (fl->fl_ops->fl_copy_lock) | |
439 | fl->fl_ops->fl_copy_lock(new, fl); | |
440 | } | |
1da177e4 LT |
441 | } |
442 | ||
443 | EXPORT_SYMBOL(locks_copy_lock); | |
444 | ||
5946c431 N |
445 | static void locks_move_blocks(struct file_lock *new, struct file_lock *fl) |
446 | { | |
447 | struct file_lock *f; | |
448 | ||
449 | /* | |
450 | * As ctx->flc_lock is held, new requests cannot be added to | |
451 | * ->fl_blocked_requests, so we don't need a lock to check if it | |
452 | * is empty. | |
453 | */ | |
454 | if (list_empty(&fl->fl_blocked_requests)) | |
455 | return; | |
456 | spin_lock(&blocked_lock_lock); | |
457 | list_splice_init(&fl->fl_blocked_requests, &new->fl_blocked_requests); | |
458 | list_for_each_entry(f, &fl->fl_blocked_requests, fl_blocked_member) | |
459 | f->fl_blocker = new; | |
460 | spin_unlock(&blocked_lock_lock); | |
461 | } | |
462 | ||
1da177e4 LT |
463 | static inline int flock_translate_cmd(int cmd) { |
464 | if (cmd & LOCK_MAND) | |
465 | return cmd & (LOCK_MAND | LOCK_RW); | |
466 | switch (cmd) { | |
467 | case LOCK_SH: | |
468 | return F_RDLCK; | |
469 | case LOCK_EX: | |
470 | return F_WRLCK; | |
471 | case LOCK_UN: | |
472 | return F_UNLCK; | |
473 | } | |
474 | return -EINVAL; | |
475 | } | |
476 | ||
477 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 | 478 | static struct file_lock * |
d6367d62 | 479 | flock_make_lock(struct file *filp, unsigned int cmd, struct file_lock *fl) |
1da177e4 | 480 | { |
1da177e4 | 481 | int type = flock_translate_cmd(cmd); |
6e129d00 | 482 | |
1da177e4 | 483 | if (type < 0) |
6e129d00 | 484 | return ERR_PTR(type); |
1da177e4 | 485 | |
d6367d62 N |
486 | if (fl == NULL) { |
487 | fl = locks_alloc_lock(); | |
488 | if (fl == NULL) | |
489 | return ERR_PTR(-ENOMEM); | |
490 | } else { | |
491 | locks_init_lock(fl); | |
492 | } | |
1da177e4 LT |
493 | |
494 | fl->fl_file = filp; | |
73a8f5f7 | 495 | fl->fl_owner = filp; |
1da177e4 LT |
496 | fl->fl_pid = current->tgid; |
497 | fl->fl_flags = FL_FLOCK; | |
498 | fl->fl_type = type; | |
499 | fl->fl_end = OFFSET_MAX; | |
500 | ||
6e129d00 | 501 | return fl; |
1da177e4 LT |
502 | } |
503 | ||
0ec4f431 | 504 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
505 | { |
506 | switch (type) { | |
507 | case F_RDLCK: | |
508 | case F_WRLCK: | |
509 | case F_UNLCK: | |
510 | fl->fl_type = type; | |
511 | break; | |
512 | default: | |
513 | return -EINVAL; | |
514 | } | |
515 | return 0; | |
516 | } | |
517 | ||
ef12e72a BF |
518 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
519 | struct flock64 *l) | |
1da177e4 | 520 | { |
1da177e4 | 521 | switch (l->l_whence) { |
f5579f8c | 522 | case SEEK_SET: |
ef12e72a | 523 | fl->fl_start = 0; |
1da177e4 | 524 | break; |
f5579f8c | 525 | case SEEK_CUR: |
ef12e72a | 526 | fl->fl_start = filp->f_pos; |
1da177e4 | 527 | break; |
f5579f8c | 528 | case SEEK_END: |
ef12e72a | 529 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
530 | break; |
531 | default: | |
532 | return -EINVAL; | |
533 | } | |
ef12e72a BF |
534 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
535 | return -EOVERFLOW; | |
536 | fl->fl_start += l->l_start; | |
537 | if (fl->fl_start < 0) | |
538 | return -EINVAL; | |
1da177e4 LT |
539 | |
540 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
541 | POSIX-2001 defines it. */ | |
4c780a46 | 542 | if (l->l_len > 0) { |
ef12e72a BF |
543 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
544 | return -EOVERFLOW; | |
545 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
546 | ||
4c780a46 | 547 | } else if (l->l_len < 0) { |
ef12e72a | 548 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 549 | return -EINVAL; |
ef12e72a BF |
550 | fl->fl_end = fl->fl_start - 1; |
551 | fl->fl_start += l->l_len; | |
552 | } else | |
553 | fl->fl_end = OFFSET_MAX; | |
554 | ||
1da177e4 LT |
555 | fl->fl_owner = current->files; |
556 | fl->fl_pid = current->tgid; | |
557 | fl->fl_file = filp; | |
558 | fl->fl_flags = FL_POSIX; | |
559 | fl->fl_ops = NULL; | |
560 | fl->fl_lmops = NULL; | |
561 | ||
562 | return assign_type(fl, l->l_type); | |
563 | } | |
564 | ||
ef12e72a BF |
565 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
566 | * style lock. | |
567 | */ | |
568 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
569 | struct flock *l) | |
1da177e4 | 570 | { |
ef12e72a BF |
571 | struct flock64 ll = { |
572 | .l_type = l->l_type, | |
573 | .l_whence = l->l_whence, | |
574 | .l_start = l->l_start, | |
575 | .l_len = l->l_len, | |
576 | }; | |
577 | ||
578 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 579 | } |
1da177e4 LT |
580 | |
581 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
582 | static bool |
583 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
584 | { |
585 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 586 | return false; |
1da177e4 LT |
587 | } |
588 | ||
1c7dd2ff JL |
589 | static void |
590 | lease_setup(struct file_lock *fl, void **priv) | |
591 | { | |
592 | struct file *filp = fl->fl_file; | |
593 | struct fasync_struct *fa = *priv; | |
594 | ||
595 | /* | |
596 | * fasync_insert_entry() returns the old entry if any. If there was no | |
597 | * old entry, then it used "priv" and inserted it into the fasync list. | |
598 | * Clear the pointer to indicate that it shouldn't be freed. | |
599 | */ | |
600 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
601 | *priv = NULL; | |
602 | ||
01919134 | 603 | __f_setown(filp, task_pid(current), PIDTYPE_TGID, 0); |
1c7dd2ff JL |
604 | } |
605 | ||
7b021967 | 606 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 607 | .lm_break = lease_break_callback, |
8fb47a4f | 608 | .lm_change = lease_modify, |
1c7dd2ff | 609 | .lm_setup = lease_setup, |
1da177e4 LT |
610 | }; |
611 | ||
612 | /* | |
613 | * Initialize a lease, use the default lock manager operations | |
614 | */ | |
0ec4f431 | 615 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
447a5647 | 616 | { |
75dff55a TM |
617 | if (assign_type(fl, type) != 0) |
618 | return -EINVAL; | |
619 | ||
7ca76311 | 620 | fl->fl_owner = filp; |
1da177e4 LT |
621 | fl->fl_pid = current->tgid; |
622 | ||
623 | fl->fl_file = filp; | |
624 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
625 | fl->fl_start = 0; |
626 | fl->fl_end = OFFSET_MAX; | |
627 | fl->fl_ops = NULL; | |
628 | fl->fl_lmops = &lease_manager_ops; | |
629 | return 0; | |
630 | } | |
631 | ||
632 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 633 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
634 | { |
635 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 636 | int error = -ENOMEM; |
1da177e4 LT |
637 | |
638 | if (fl == NULL) | |
e32b8ee2 | 639 | return ERR_PTR(error); |
1da177e4 LT |
640 | |
641 | error = lease_init(filp, type, fl); | |
75dff55a TM |
642 | if (error) { |
643 | locks_free_lock(fl); | |
e32b8ee2 | 644 | return ERR_PTR(error); |
75dff55a | 645 | } |
e32b8ee2 | 646 | return fl; |
1da177e4 LT |
647 | } |
648 | ||
649 | /* Check if two locks overlap each other. | |
650 | */ | |
651 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
652 | { | |
653 | return ((fl1->fl_end >= fl2->fl_start) && | |
654 | (fl2->fl_end >= fl1->fl_start)); | |
655 | } | |
656 | ||
657 | /* | |
658 | * Check whether two locks have the same owner. | |
659 | */ | |
33443c42 | 660 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 661 | { |
8fb47a4f | 662 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 663 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 664 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
665 | return fl1->fl_owner == fl2->fl_owner; |
666 | } | |
667 | ||
6109c850 | 668 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 669 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 670 | { |
7c3f654d PZ |
671 | struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list); |
672 | ||
aba37660 PZ |
673 | percpu_rwsem_assert_held(&file_rwsem); |
674 | ||
7c3f654d | 675 | spin_lock(&fll->lock); |
7012b02a | 676 | fl->fl_link_cpu = smp_processor_id(); |
7c3f654d PZ |
677 | hlist_add_head(&fl->fl_link, &fll->hlist); |
678 | spin_unlock(&fll->lock); | |
88974691 JL |
679 | } |
680 | ||
6109c850 | 681 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 682 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 683 | { |
7c3f654d PZ |
684 | struct file_lock_list_struct *fll; |
685 | ||
aba37660 PZ |
686 | percpu_rwsem_assert_held(&file_rwsem); |
687 | ||
7012b02a JL |
688 | /* |
689 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 690 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
691 | * also require that it be held. |
692 | */ | |
693 | if (hlist_unhashed(&fl->fl_link)) | |
694 | return; | |
7c3f654d PZ |
695 | |
696 | fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu); | |
697 | spin_lock(&fll->lock); | |
139ca04e | 698 | hlist_del_init(&fl->fl_link); |
7c3f654d | 699 | spin_unlock(&fll->lock); |
88974691 JL |
700 | } |
701 | ||
3999e493 JL |
702 | static unsigned long |
703 | posix_owner_key(struct file_lock *fl) | |
704 | { | |
705 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
706 | return fl->fl_lmops->lm_owner_key(fl); | |
707 | return (unsigned long)fl->fl_owner; | |
708 | } | |
709 | ||
6ca10ed8 | 710 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 711 | { |
663d5af7 DW |
712 | lockdep_assert_held(&blocked_lock_lock); |
713 | ||
3999e493 | 714 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
715 | } |
716 | ||
6ca10ed8 | 717 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 718 | { |
663d5af7 DW |
719 | lockdep_assert_held(&blocked_lock_lock); |
720 | ||
48f74186 | 721 | hash_del(&waiter->fl_link); |
88974691 JL |
722 | } |
723 | ||
1da177e4 LT |
724 | /* Remove waiter from blocker's block list. |
725 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 726 | * |
7b2296af | 727 | * Must be called with blocked_lock_lock held. |
1da177e4 | 728 | */ |
33443c42 | 729 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 730 | { |
88974691 | 731 | locks_delete_global_blocked(waiter); |
ada5c1da N |
732 | list_del_init(&waiter->fl_blocked_member); |
733 | waiter->fl_blocker = NULL; | |
1da177e4 LT |
734 | } |
735 | ||
ad6bbd8b N |
736 | static void __locks_wake_up_blocks(struct file_lock *blocker) |
737 | { | |
738 | while (!list_empty(&blocker->fl_blocked_requests)) { | |
739 | struct file_lock *waiter; | |
740 | ||
741 | waiter = list_first_entry(&blocker->fl_blocked_requests, | |
742 | struct file_lock, fl_blocked_member); | |
743 | __locks_delete_block(waiter); | |
744 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) | |
745 | waiter->fl_lmops->lm_notify(waiter); | |
746 | else | |
747 | wake_up(&waiter->fl_wait); | |
748 | } | |
749 | } | |
750 | ||
cb03f94f N |
751 | /** |
752 | * locks_delete_lock - stop waiting for a file lock | |
753 | * @waiter: the lock which was waiting | |
754 | * | |
755 | * lockd/nfsd need to disconnect the lock while working on it. | |
756 | */ | |
757 | int locks_delete_block(struct file_lock *waiter) | |
1da177e4 | 758 | { |
cb03f94f N |
759 | int status = -ENOENT; |
760 | ||
16306a61 N |
761 | /* |
762 | * If fl_blocker is NULL, it won't be set again as this thread | |
763 | * "owns" the lock and is the only one that might try to claim | |
764 | * the lock. So it is safe to test fl_blocker locklessly. | |
765 | * Also if fl_blocker is NULL, this waiter is not listed on | |
766 | * fl_blocked_requests for some lock, so no other request can | |
767 | * be added to the list of fl_blocked_requests for this | |
768 | * request. So if fl_blocker is NULL, it is safe to | |
769 | * locklessly check if fl_blocked_requests is empty. If both | |
770 | * of these checks succeed, there is no need to take the lock. | |
771 | */ | |
772 | if (waiter->fl_blocker == NULL && | |
773 | list_empty(&waiter->fl_blocked_requests)) | |
cb03f94f | 774 | return status; |
7b2296af | 775 | spin_lock(&blocked_lock_lock); |
cb03f94f N |
776 | if (waiter->fl_blocker) |
777 | status = 0; | |
5946c431 | 778 | __locks_wake_up_blocks(waiter); |
1da177e4 | 779 | __locks_delete_block(waiter); |
7b2296af | 780 | spin_unlock(&blocked_lock_lock); |
cb03f94f | 781 | return status; |
1da177e4 | 782 | } |
cb03f94f | 783 | EXPORT_SYMBOL(locks_delete_block); |
1da177e4 LT |
784 | |
785 | /* Insert waiter into blocker's block list. | |
786 | * We use a circular list so that processes can be easily woken up in | |
787 | * the order they blocked. The documentation doesn't require this but | |
788 | * it seems like the reasonable thing to do. | |
1c8c601a | 789 | * |
6109c850 | 790 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
ada5c1da N |
791 | * fl_blocked_requests list itself is protected by the blocked_lock_lock, |
792 | * but by ensuring that the flc_lock is also held on insertions we can avoid | |
793 | * taking the blocked_lock_lock in some cases when we see that the | |
794 | * fl_blocked_requests list is empty. | |
fd7732e0 N |
795 | * |
796 | * Rather than just adding to the list, we check for conflicts with any existing | |
797 | * waiters, and add beneath any waiter that blocks the new waiter. | |
798 | * Thus wakeups don't happen until needed. | |
1da177e4 | 799 | */ |
1c8c601a | 800 | static void __locks_insert_block(struct file_lock *blocker, |
fd7732e0 N |
801 | struct file_lock *waiter, |
802 | bool conflict(struct file_lock *, | |
803 | struct file_lock *)) | |
1da177e4 | 804 | { |
fd7732e0 | 805 | struct file_lock *fl; |
ada5c1da | 806 | BUG_ON(!list_empty(&waiter->fl_blocked_member)); |
fd7732e0 N |
807 | |
808 | new_blocker: | |
809 | list_for_each_entry(fl, &blocker->fl_blocked_requests, fl_blocked_member) | |
810 | if (conflict(fl, waiter)) { | |
811 | blocker = fl; | |
812 | goto new_blocker; | |
813 | } | |
ada5c1da N |
814 | waiter->fl_blocker = blocker; |
815 | list_add_tail(&waiter->fl_blocked_member, &blocker->fl_blocked_requests); | |
cff2fce5 | 816 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a | 817 | locks_insert_global_blocked(waiter); |
5946c431 N |
818 | |
819 | /* The requests in waiter->fl_blocked are known to conflict with | |
820 | * waiter, but might not conflict with blocker, or the requests | |
821 | * and lock which block it. So they all need to be woken. | |
822 | */ | |
823 | __locks_wake_up_blocks(waiter); | |
1c8c601a JL |
824 | } |
825 | ||
6109c850 | 826 | /* Must be called with flc_lock held. */ |
1c8c601a | 827 | static void locks_insert_block(struct file_lock *blocker, |
fd7732e0 N |
828 | struct file_lock *waiter, |
829 | bool conflict(struct file_lock *, | |
830 | struct file_lock *)) | |
1c8c601a | 831 | { |
7b2296af | 832 | spin_lock(&blocked_lock_lock); |
fd7732e0 | 833 | __locks_insert_block(blocker, waiter, conflict); |
7b2296af | 834 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
835 | } |
836 | ||
1cb36012 JL |
837 | /* |
838 | * Wake up processes blocked waiting for blocker. | |
839 | * | |
6109c850 | 840 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
841 | */ |
842 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
843 | { | |
4e8c765d JL |
844 | /* |
845 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 | 846 | * blocked requests are only added to the list under the flc_lock, and |
ada5c1da N |
847 | * the flc_lock is always held here. Note that removal from the |
848 | * fl_blocked_requests list does not require the flc_lock, so we must | |
849 | * recheck list_empty() after acquiring the blocked_lock_lock. | |
4e8c765d | 850 | */ |
ada5c1da | 851 | if (list_empty(&blocker->fl_blocked_requests)) |
4e8c765d JL |
852 | return; |
853 | ||
7b2296af | 854 | spin_lock(&blocked_lock_lock); |
ad6bbd8b | 855 | __locks_wake_up_blocks(blocker); |
7b2296af | 856 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
857 | } |
858 | ||
5263e31e | 859 | static void |
e084c1bd | 860 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e | 861 | { |
5263e31e JL |
862 | list_add_tail(&fl->fl_list, before); |
863 | locks_insert_global_locks(fl); | |
864 | } | |
865 | ||
8634b51f | 866 | static void |
e084c1bd | 867 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 868 | { |
88974691 | 869 | locks_delete_global_locks(fl); |
8634b51f | 870 | list_del_init(&fl->fl_list); |
1da177e4 | 871 | locks_wake_up_blocks(fl); |
24cbe784 JL |
872 | } |
873 | ||
8634b51f | 874 | static void |
e084c1bd | 875 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 876 | { |
e084c1bd | 877 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 878 | if (dispose) |
6dee60f6 | 879 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
880 | else |
881 | locks_free_lock(fl); | |
1da177e4 LT |
882 | } |
883 | ||
884 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
885 | * checks for shared/exclusive status of overlapping locks. | |
886 | */ | |
c0e15908 N |
887 | static bool locks_conflict(struct file_lock *caller_fl, |
888 | struct file_lock *sys_fl) | |
1da177e4 LT |
889 | { |
890 | if (sys_fl->fl_type == F_WRLCK) | |
c0e15908 | 891 | return true; |
1da177e4 | 892 | if (caller_fl->fl_type == F_WRLCK) |
c0e15908 N |
893 | return true; |
894 | return false; | |
1da177e4 LT |
895 | } |
896 | ||
897 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
898 | * checking before calling the locks_conflict(). | |
899 | */ | |
c0e15908 N |
900 | static bool posix_locks_conflict(struct file_lock *caller_fl, |
901 | struct file_lock *sys_fl) | |
1da177e4 LT |
902 | { |
903 | /* POSIX locks owned by the same process do not conflict with | |
904 | * each other. | |
905 | */ | |
9b8c8695 | 906 | if (posix_same_owner(caller_fl, sys_fl)) |
c0e15908 | 907 | return false; |
1da177e4 LT |
908 | |
909 | /* Check whether they overlap */ | |
910 | if (!locks_overlap(caller_fl, sys_fl)) | |
c0e15908 | 911 | return false; |
1da177e4 | 912 | |
c0e15908 | 913 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
914 | } |
915 | ||
916 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
917 | * checking before calling the locks_conflict(). | |
918 | */ | |
c0e15908 N |
919 | static bool flock_locks_conflict(struct file_lock *caller_fl, |
920 | struct file_lock *sys_fl) | |
1da177e4 LT |
921 | { |
922 | /* FLOCK locks referring to the same filp do not conflict with | |
923 | * each other. | |
924 | */ | |
9b8c8695 | 925 | if (caller_fl->fl_file == sys_fl->fl_file) |
c0e15908 | 926 | return false; |
1da177e4 | 927 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) |
c0e15908 | 928 | return false; |
1da177e4 | 929 | |
c0e15908 | 930 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
931 | } |
932 | ||
6d34ac19 | 933 | void |
9d6a8c5c | 934 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
935 | { |
936 | struct file_lock *cfl; | |
bd61e0a9 | 937 | struct file_lock_context *ctx; |
c568d683 | 938 | struct inode *inode = locks_inode(filp); |
1da177e4 | 939 | |
128a3785 | 940 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
941 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
942 | fl->fl_type = F_UNLCK; | |
943 | return; | |
944 | } | |
945 | ||
6109c850 | 946 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
947 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
948 | if (posix_locks_conflict(fl, cfl)) { | |
949 | locks_copy_conflock(fl, cfl); | |
bd61e0a9 JL |
950 | goto out; |
951 | } | |
1da177e4 | 952 | } |
bd61e0a9 JL |
953 | fl->fl_type = F_UNLCK; |
954 | out: | |
6109c850 | 955 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 956 | return; |
1da177e4 | 957 | } |
1da177e4 LT |
958 | EXPORT_SYMBOL(posix_test_lock); |
959 | ||
b533184f BF |
960 | /* |
961 | * Deadlock detection: | |
962 | * | |
963 | * We attempt to detect deadlocks that are due purely to posix file | |
964 | * locks. | |
1da177e4 | 965 | * |
b533184f BF |
966 | * We assume that a task can be waiting for at most one lock at a time. |
967 | * So for any acquired lock, the process holding that lock may be | |
968 | * waiting on at most one other lock. That lock in turns may be held by | |
969 | * someone waiting for at most one other lock. Given a requested lock | |
970 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
971 | * follow this chain of waiters to ensure we are not about to create a | |
972 | * cycle. | |
1da177e4 | 973 | * |
b533184f BF |
974 | * Since we do this before we ever put a process to sleep on a lock, we |
975 | * are ensured that there is never a cycle; that is what guarantees that | |
976 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 977 | * |
b533184f BF |
978 | * Note: the above assumption may not be true when handling lock |
979 | * requests from a broken NFS client. It may also fail in the presence | |
980 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 981 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 982 | * |
cff2fce5 | 983 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
984 | * Because the owner is not even nominally tied to a thread of |
985 | * execution, the deadlock detection below can't reasonably work well. Just | |
986 | * skip it for those. | |
987 | * | |
cff2fce5 | 988 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
989 | * locks that just checks for the case where two tasks are attempting to |
990 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 991 | */ |
97855b49 BF |
992 | |
993 | #define MAX_DEADLK_ITERATIONS 10 | |
994 | ||
b533184f BF |
995 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
996 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
997 | { | |
998 | struct file_lock *fl; | |
999 | ||
3999e493 | 1000 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
5946c431 N |
1001 | if (posix_same_owner(fl, block_fl)) { |
1002 | while (fl->fl_blocker) | |
1003 | fl = fl->fl_blocker; | |
1004 | return fl; | |
1005 | } | |
b533184f BF |
1006 | } |
1007 | return NULL; | |
1008 | } | |
1009 | ||
7b2296af | 1010 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 1011 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
1012 | struct file_lock *block_fl) |
1013 | { | |
97855b49 | 1014 | int i = 0; |
1da177e4 | 1015 | |
663d5af7 DW |
1016 | lockdep_assert_held(&blocked_lock_lock); |
1017 | ||
57b65325 JL |
1018 | /* |
1019 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 1020 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 1021 | */ |
cff2fce5 | 1022 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
1023 | return 0; |
1024 | ||
b533184f BF |
1025 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
1026 | if (i++ > MAX_DEADLK_ITERATIONS) | |
1027 | return 0; | |
1028 | if (posix_same_owner(caller_fl, block_fl)) | |
1029 | return 1; | |
1da177e4 LT |
1030 | } |
1031 | return 0; | |
1032 | } | |
1033 | ||
1da177e4 | 1034 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 1035 | * after any leases, but before any posix locks. |
f475ae95 TM |
1036 | * |
1037 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1038 | * whether or not a lock was successfully freed by testing the return | |
1039 | * value for -ENOENT. | |
1da177e4 | 1040 | */ |
bcd7f78d | 1041 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 1042 | { |
993dfa87 | 1043 | struct file_lock *new_fl = NULL; |
5263e31e JL |
1044 | struct file_lock *fl; |
1045 | struct file_lock_context *ctx; | |
1da177e4 | 1046 | int error = 0; |
5263e31e | 1047 | bool found = false; |
ed9814d8 | 1048 | LIST_HEAD(dispose); |
1da177e4 | 1049 | |
5c1c669a JL |
1050 | ctx = locks_get_lock_context(inode, request->fl_type); |
1051 | if (!ctx) { | |
1052 | if (request->fl_type != F_UNLCK) | |
1053 | return -ENOMEM; | |
1054 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
1055 | } | |
5263e31e | 1056 | |
b89f4321 | 1057 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 1058 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
1059 | if (!new_fl) |
1060 | return -ENOMEM; | |
84d535ad PE |
1061 | } |
1062 | ||
87709e28 | 1063 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1064 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
1065 | if (request->fl_flags & FL_ACCESS) |
1066 | goto find_conflict; | |
1067 | ||
5263e31e | 1068 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 1069 | if (request->fl_file != fl->fl_file) |
1da177e4 | 1070 | continue; |
993dfa87 | 1071 | if (request->fl_type == fl->fl_type) |
1da177e4 | 1072 | goto out; |
5263e31e | 1073 | found = true; |
e084c1bd | 1074 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1075 | break; |
1076 | } | |
1da177e4 | 1077 | |
f475ae95 TM |
1078 | if (request->fl_type == F_UNLCK) { |
1079 | if ((request->fl_flags & FL_EXISTS) && !found) | |
1080 | error = -ENOENT; | |
993dfa87 | 1081 | goto out; |
f475ae95 | 1082 | } |
1da177e4 | 1083 | |
f07f18dd | 1084 | find_conflict: |
5263e31e | 1085 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 1086 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
1087 | continue; |
1088 | error = -EAGAIN; | |
bde74e4b MS |
1089 | if (!(request->fl_flags & FL_SLEEP)) |
1090 | goto out; | |
1091 | error = FILE_LOCK_DEFERRED; | |
fd7732e0 | 1092 | locks_insert_block(fl, request, flock_locks_conflict); |
1da177e4 LT |
1093 | goto out; |
1094 | } | |
f07f18dd TM |
1095 | if (request->fl_flags & FL_ACCESS) |
1096 | goto out; | |
993dfa87 | 1097 | locks_copy_lock(new_fl, request); |
5946c431 | 1098 | locks_move_blocks(new_fl, request); |
e084c1bd | 1099 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 1100 | new_fl = NULL; |
9cedc194 | 1101 | error = 0; |
1da177e4 LT |
1102 | |
1103 | out: | |
6109c850 | 1104 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1105 | percpu_up_read_preempt_enable(&file_rwsem); |
993dfa87 TM |
1106 | if (new_fl) |
1107 | locks_free_lock(new_fl); | |
ed9814d8 | 1108 | locks_dispose_list(&dispose); |
c883da31 | 1109 | trace_flock_lock_inode(inode, request, error); |
1da177e4 LT |
1110 | return error; |
1111 | } | |
1112 | ||
b4d629a3 JL |
1113 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
1114 | struct file_lock *conflock) | |
1da177e4 | 1115 | { |
bd61e0a9 | 1116 | struct file_lock *fl, *tmp; |
39005d02 MS |
1117 | struct file_lock *new_fl = NULL; |
1118 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
1119 | struct file_lock *left = NULL; |
1120 | struct file_lock *right = NULL; | |
bd61e0a9 | 1121 | struct file_lock_context *ctx; |
b9746ef8 JL |
1122 | int error; |
1123 | bool added = false; | |
ed9814d8 | 1124 | LIST_HEAD(dispose); |
1da177e4 | 1125 | |
5c1c669a | 1126 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 1127 | if (!ctx) |
5c1c669a | 1128 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 1129 | |
1da177e4 LT |
1130 | /* |
1131 | * We may need two file_lock structures for this operation, | |
1132 | * so we get them in advance to avoid races. | |
39005d02 MS |
1133 | * |
1134 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1135 | */ |
39005d02 MS |
1136 | if (!(request->fl_flags & FL_ACCESS) && |
1137 | (request->fl_type != F_UNLCK || | |
1138 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1139 | new_fl = locks_alloc_lock(); | |
1140 | new_fl2 = locks_alloc_lock(); | |
1141 | } | |
1da177e4 | 1142 | |
87709e28 | 1143 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1144 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1145 | /* |
1146 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1147 | * there are any, either return error or put the request on the | |
48f74186 | 1148 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1149 | */ |
1da177e4 | 1150 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1151 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1152 | if (!posix_locks_conflict(request, fl)) |
1153 | continue; | |
5842add2 | 1154 | if (conflock) |
3fe0fff1 | 1155 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1156 | error = -EAGAIN; |
1157 | if (!(request->fl_flags & FL_SLEEP)) | |
1158 | goto out; | |
1c8c601a JL |
1159 | /* |
1160 | * Deadlock detection and insertion into the blocked | |
1161 | * locks list must be done while holding the same lock! | |
1162 | */ | |
1da177e4 | 1163 | error = -EDEADLK; |
7b2296af | 1164 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
1165 | if (likely(!posix_locks_deadlock(request, fl))) { |
1166 | error = FILE_LOCK_DEFERRED; | |
fd7732e0 N |
1167 | __locks_insert_block(fl, request, |
1168 | posix_locks_conflict); | |
1c8c601a | 1169 | } |
7b2296af | 1170 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1171 | goto out; |
1172 | } | |
1173 | } | |
1174 | ||
1175 | /* If we're just looking for a conflict, we're done. */ | |
1176 | error = 0; | |
1177 | if (request->fl_flags & FL_ACCESS) | |
1178 | goto out; | |
1179 | ||
bd61e0a9 JL |
1180 | /* Find the first old lock with the same owner as the new lock */ |
1181 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1182 | if (posix_same_owner(request, fl)) | |
1183 | break; | |
1da177e4 LT |
1184 | } |
1185 | ||
1cb36012 | 1186 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1187 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1188 | if (!posix_same_owner(request, fl)) | |
1189 | break; | |
1190 | ||
1191 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1192 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1193 | /* In all comparisons of start vs end, use |
1194 | * "start - 1" rather than "end + 1". If end | |
1195 | * is OFFSET_MAX, end + 1 will become negative. | |
1196 | */ | |
1da177e4 | 1197 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1198 | continue; |
1da177e4 LT |
1199 | /* If the next lock in the list has entirely bigger |
1200 | * addresses than the new one, insert the lock here. | |
1201 | */ | |
449231d6 | 1202 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1203 | break; |
1204 | ||
1205 | /* If we come here, the new and old lock are of the | |
1206 | * same type and adjacent or overlapping. Make one | |
1207 | * lock yielding from the lower start address of both | |
1208 | * locks to the higher end address. | |
1209 | */ | |
1210 | if (fl->fl_start > request->fl_start) | |
1211 | fl->fl_start = request->fl_start; | |
1212 | else | |
1213 | request->fl_start = fl->fl_start; | |
1214 | if (fl->fl_end < request->fl_end) | |
1215 | fl->fl_end = request->fl_end; | |
1216 | else | |
1217 | request->fl_end = fl->fl_end; | |
1218 | if (added) { | |
e084c1bd | 1219 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1220 | continue; |
1221 | } | |
1222 | request = fl; | |
b9746ef8 | 1223 | added = true; |
bd61e0a9 | 1224 | } else { |
1da177e4 LT |
1225 | /* Processing for different lock types is a bit |
1226 | * more complex. | |
1227 | */ | |
1228 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1229 | continue; |
1da177e4 LT |
1230 | if (fl->fl_start > request->fl_end) |
1231 | break; | |
1232 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1233 | added = true; |
1da177e4 LT |
1234 | if (fl->fl_start < request->fl_start) |
1235 | left = fl; | |
1236 | /* If the next lock in the list has a higher end | |
1237 | * address than the new one, insert the new one here. | |
1238 | */ | |
1239 | if (fl->fl_end > request->fl_end) { | |
1240 | right = fl; | |
1241 | break; | |
1242 | } | |
1243 | if (fl->fl_start >= request->fl_start) { | |
1244 | /* The new lock completely replaces an old | |
1245 | * one (This may happen several times). | |
1246 | */ | |
1247 | if (added) { | |
e084c1bd | 1248 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1249 | continue; |
1250 | } | |
b84d49f9 JL |
1251 | /* |
1252 | * Replace the old lock with new_fl, and | |
1253 | * remove the old one. It's safe to do the | |
1254 | * insert here since we know that we won't be | |
1255 | * using new_fl later, and that the lock is | |
1256 | * just replacing an existing lock. | |
1da177e4 | 1257 | */ |
b84d49f9 JL |
1258 | error = -ENOLCK; |
1259 | if (!new_fl) | |
1260 | goto out; | |
1261 | locks_copy_lock(new_fl, request); | |
1262 | request = new_fl; | |
1263 | new_fl = NULL; | |
e084c1bd JL |
1264 | locks_insert_lock_ctx(request, &fl->fl_list); |
1265 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1266 | added = true; |
1da177e4 LT |
1267 | } |
1268 | } | |
1da177e4 LT |
1269 | } |
1270 | ||
0d9a490a | 1271 | /* |
1cb36012 JL |
1272 | * The above code only modifies existing locks in case of merging or |
1273 | * replacing. If new lock(s) need to be inserted all modifications are | |
1274 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1275 | */ |
1276 | error = -ENOLCK; /* "no luck" */ | |
1277 | if (right && left == right && !new_fl2) | |
1278 | goto out; | |
1279 | ||
1da177e4 LT |
1280 | error = 0; |
1281 | if (!added) { | |
f475ae95 TM |
1282 | if (request->fl_type == F_UNLCK) { |
1283 | if (request->fl_flags & FL_EXISTS) | |
1284 | error = -ENOENT; | |
1da177e4 | 1285 | goto out; |
f475ae95 | 1286 | } |
0d9a490a MS |
1287 | |
1288 | if (!new_fl) { | |
1289 | error = -ENOLCK; | |
1290 | goto out; | |
1291 | } | |
1da177e4 | 1292 | locks_copy_lock(new_fl, request); |
5946c431 | 1293 | locks_move_blocks(new_fl, request); |
e084c1bd | 1294 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
2e2f756f | 1295 | fl = new_fl; |
1da177e4 LT |
1296 | new_fl = NULL; |
1297 | } | |
1298 | if (right) { | |
1299 | if (left == right) { | |
1300 | /* The new lock breaks the old one in two pieces, | |
1301 | * so we have to use the second new lock. | |
1302 | */ | |
1303 | left = new_fl2; | |
1304 | new_fl2 = NULL; | |
1305 | locks_copy_lock(left, right); | |
e084c1bd | 1306 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1307 | } |
1308 | right->fl_start = request->fl_end + 1; | |
1309 | locks_wake_up_blocks(right); | |
1310 | } | |
1311 | if (left) { | |
1312 | left->fl_end = request->fl_start - 1; | |
1313 | locks_wake_up_blocks(left); | |
1314 | } | |
1315 | out: | |
6109c850 | 1316 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1317 | percpu_up_read_preempt_enable(&file_rwsem); |
1da177e4 LT |
1318 | /* |
1319 | * Free any unused locks. | |
1320 | */ | |
1321 | if (new_fl) | |
1322 | locks_free_lock(new_fl); | |
1323 | if (new_fl2) | |
1324 | locks_free_lock(new_fl2); | |
ed9814d8 | 1325 | locks_dispose_list(&dispose); |
1890910f JL |
1326 | trace_posix_lock_inode(inode, request, error); |
1327 | ||
1da177e4 LT |
1328 | return error; |
1329 | } | |
1330 | ||
1331 | /** | |
1332 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1333 | * @filp: The file to apply the lock to | |
1334 | * @fl: The lock to be applied | |
150b3934 | 1335 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1336 | * |
1337 | * Add a POSIX style lock to a file. | |
1338 | * We merge adjacent & overlapping locks whenever possible. | |
1339 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1340 | * |
1341 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1342 | * whether or not a lock was successfully freed by testing the return | |
1343 | * value for -ENOENT. | |
1da177e4 | 1344 | */ |
150b3934 | 1345 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1346 | struct file_lock *conflock) |
1347 | { | |
c568d683 | 1348 | return posix_lock_inode(locks_inode(filp), fl, conflock); |
1da177e4 | 1349 | } |
150b3934 | 1350 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1351 | |
1352 | /** | |
29d01b22 JL |
1353 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1354 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1355 | * @fl: The lock to be applied |
1356 | * | |
616fb38f | 1357 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1358 | */ |
616fb38f | 1359 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1360 | { |
1361 | int error; | |
1362 | might_sleep (); | |
1363 | for (;;) { | |
b4d629a3 | 1364 | error = posix_lock_inode(inode, fl, NULL); |
bde74e4b | 1365 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1366 | break; |
ada5c1da | 1367 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
1368 | if (error) |
1369 | break; | |
1da177e4 | 1370 | } |
16306a61 | 1371 | locks_delete_block(fl); |
1da177e4 LT |
1372 | return error; |
1373 | } | |
29d01b22 | 1374 | |
9e8925b6 | 1375 | #ifdef CONFIG_MANDATORY_FILE_LOCKING |
1da177e4 LT |
1376 | /** |
1377 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1378 | * @file: the file to check |
1da177e4 LT |
1379 | * |
1380 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1381 | * This function is called from locks_verify_locked() only. | |
1382 | */ | |
d7a06983 | 1383 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1384 | { |
bd61e0a9 | 1385 | int ret; |
c568d683 | 1386 | struct inode *inode = locks_inode(file); |
bd61e0a9 | 1387 | struct file_lock_context *ctx; |
1da177e4 LT |
1388 | struct file_lock *fl; |
1389 | ||
128a3785 | 1390 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
1391 | if (!ctx || list_empty_careful(&ctx->flc_posix)) |
1392 | return 0; | |
1393 | ||
1da177e4 LT |
1394 | /* |
1395 | * Search the lock list for this inode for any POSIX locks. | |
1396 | */ | |
6109c850 | 1397 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
1398 | ret = 0; |
1399 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1400 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1401 | fl->fl_owner != file) { |
1402 | ret = -EAGAIN; | |
1da177e4 | 1403 | break; |
bd61e0a9 | 1404 | } |
1da177e4 | 1405 | } |
6109c850 | 1406 | spin_unlock(&ctx->flc_lock); |
bd61e0a9 | 1407 | return ret; |
1da177e4 LT |
1408 | } |
1409 | ||
1410 | /** | |
1411 | * locks_mandatory_area - Check for a conflicting lock | |
acc15575 | 1412 | * @inode: the file to check |
1da177e4 | 1413 | * @filp: how the file was opened (if it was) |
acc15575 CH |
1414 | * @start: first byte in the file to check |
1415 | * @end: lastbyte in the file to check | |
1416 | * @type: %F_WRLCK for a write lock, else %F_RDLCK | |
1da177e4 LT |
1417 | * |
1418 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1da177e4 | 1419 | */ |
acc15575 CH |
1420 | int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start, |
1421 | loff_t end, unsigned char type) | |
1da177e4 LT |
1422 | { |
1423 | struct file_lock fl; | |
1424 | int error; | |
29723ade | 1425 | bool sleep = false; |
1da177e4 LT |
1426 | |
1427 | locks_init_lock(&fl); | |
1da177e4 LT |
1428 | fl.fl_pid = current->tgid; |
1429 | fl.fl_file = filp; | |
1430 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1431 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1432 | sleep = true; |
acc15575 CH |
1433 | fl.fl_type = type; |
1434 | fl.fl_start = start; | |
1435 | fl.fl_end = end; | |
1da177e4 LT |
1436 | |
1437 | for (;;) { | |
29723ade | 1438 | if (filp) { |
73a8f5f7 | 1439 | fl.fl_owner = filp; |
29723ade | 1440 | fl.fl_flags &= ~FL_SLEEP; |
b4d629a3 | 1441 | error = posix_lock_inode(inode, &fl, NULL); |
29723ade JL |
1442 | if (!error) |
1443 | break; | |
1444 | } | |
1445 | ||
1446 | if (sleep) | |
1447 | fl.fl_flags |= FL_SLEEP; | |
1448 | fl.fl_owner = current->files; | |
b4d629a3 | 1449 | error = posix_lock_inode(inode, &fl, NULL); |
bde74e4b | 1450 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1451 | break; |
ada5c1da | 1452 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_blocker); |
1da177e4 LT |
1453 | if (!error) { |
1454 | /* | |
1455 | * If we've been sleeping someone might have | |
1456 | * changed the permissions behind our back. | |
1457 | */ | |
a16877ca | 1458 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1459 | continue; |
1460 | } | |
1461 | ||
1da177e4 LT |
1462 | break; |
1463 | } | |
16306a61 | 1464 | locks_delete_block(&fl); |
1da177e4 LT |
1465 | |
1466 | return error; | |
1467 | } | |
1468 | ||
1469 | EXPORT_SYMBOL(locks_mandatory_area); | |
9e8925b6 | 1470 | #endif /* CONFIG_MANDATORY_FILE_LOCKING */ |
1da177e4 | 1471 | |
778fc546 BF |
1472 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1473 | { | |
1474 | switch (arg) { | |
1475 | case F_UNLCK: | |
1476 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1477 | /* fall through: */ | |
1478 | case F_RDLCK: | |
1479 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1480 | } | |
1481 | } | |
1482 | ||
1da177e4 | 1483 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1484 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1485 | { |
1da177e4 LT |
1486 | int error = assign_type(fl, arg); |
1487 | ||
1488 | if (error) | |
1489 | return error; | |
778fc546 | 1490 | lease_clear_pending(fl, arg); |
1da177e4 | 1491 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1492 | if (arg == F_UNLCK) { |
1493 | struct file *filp = fl->fl_file; | |
1494 | ||
1495 | f_delown(filp); | |
1496 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1497 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1498 | if (fl->fl_fasync != NULL) { | |
1499 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1500 | fl->fl_fasync = NULL; | |
1501 | } | |
e084c1bd | 1502 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1503 | } |
1da177e4 LT |
1504 | return 0; |
1505 | } | |
1da177e4 LT |
1506 | EXPORT_SYMBOL(lease_modify); |
1507 | ||
778fc546 BF |
1508 | static bool past_time(unsigned long then) |
1509 | { | |
1510 | if (!then) | |
1511 | /* 0 is a special value meaning "this never expires": */ | |
1512 | return false; | |
1513 | return time_after(jiffies, then); | |
1514 | } | |
1515 | ||
c45198ed | 1516 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1517 | { |
8634b51f JL |
1518 | struct file_lock_context *ctx = inode->i_flctx; |
1519 | struct file_lock *fl, *tmp; | |
1da177e4 | 1520 | |
6109c850 | 1521 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1522 | |
8634b51f | 1523 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1524 | trace_time_out_leases(inode, fl); |
778fc546 | 1525 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1526 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1527 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1528 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1529 | } |
1530 | } | |
1531 | ||
df4e8d2c BF |
1532 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1533 | { | |
11afe9f7 CH |
1534 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) |
1535 | return false; | |
df4e8d2c BF |
1536 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) |
1537 | return false; | |
1538 | return locks_conflict(breaker, lease); | |
1539 | } | |
1540 | ||
03d12ddf JL |
1541 | static bool |
1542 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1543 | { | |
8634b51f | 1544 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1545 | struct file_lock *fl; |
1546 | ||
6109c850 | 1547 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1548 | |
8634b51f | 1549 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1550 | if (leases_conflict(fl, breaker)) |
1551 | return true; | |
1552 | } | |
1553 | return false; | |
1554 | } | |
1555 | ||
1da177e4 LT |
1556 | /** |
1557 | * __break_lease - revoke all outstanding leases on file | |
1558 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1559 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1560 | * break all leases | |
1561 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1562 | * only delegations | |
1da177e4 | 1563 | * |
87250dd2 | 1564 | * break_lease (inlined for speed) has checked there already is at least |
1565 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1566 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1567 | * specified %O_NONBLOCK to your open(). |
1568 | */ | |
df4e8d2c | 1569 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1570 | { |
778fc546 | 1571 | int error = 0; |
128a3785 | 1572 | struct file_lock_context *ctx; |
a901125c | 1573 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1574 | unsigned long break_time; |
8737c930 | 1575 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1576 | LIST_HEAD(dispose); |
1da177e4 | 1577 | |
8737c930 | 1578 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1579 | if (IS_ERR(new_fl)) |
1580 | return PTR_ERR(new_fl); | |
df4e8d2c | 1581 | new_fl->fl_flags = type; |
1da177e4 | 1582 | |
8634b51f | 1583 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1584 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1585 | if (!ctx) { |
1586 | WARN_ON_ONCE(1); | |
1587 | return error; | |
1588 | } | |
1589 | ||
87709e28 | 1590 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1591 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1592 | |
c45198ed | 1593 | time_out_leases(inode, &dispose); |
1da177e4 | 1594 | |
03d12ddf | 1595 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1596 | goto out; |
1597 | ||
1da177e4 LT |
1598 | break_time = 0; |
1599 | if (lease_break_time > 0) { | |
1600 | break_time = jiffies + lease_break_time * HZ; | |
1601 | if (break_time == 0) | |
1602 | break_time++; /* so that 0 means no break time */ | |
1603 | } | |
1604 | ||
a901125c | 1605 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1606 | if (!leases_conflict(fl, new_fl)) |
1607 | continue; | |
778fc546 BF |
1608 | if (want_write) { |
1609 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1610 | continue; | |
1611 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1612 | fl->fl_break_time = break_time; |
778fc546 | 1613 | } else { |
8634b51f | 1614 | if (lease_breaking(fl)) |
778fc546 BF |
1615 | continue; |
1616 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1617 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1618 | } |
4d01b7f5 | 1619 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1620 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1621 | } |
1622 | ||
8634b51f | 1623 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1624 | goto out; |
1625 | ||
843c6b2f | 1626 | if (mode & O_NONBLOCK) { |
62af4f1f | 1627 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1628 | error = -EWOULDBLOCK; |
1629 | goto out; | |
1630 | } | |
1631 | ||
1632 | restart: | |
8634b51f JL |
1633 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1634 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1635 | if (break_time != 0) |
1da177e4 | 1636 | break_time -= jiffies; |
f1c6bb2c JL |
1637 | if (break_time == 0) |
1638 | break_time++; | |
fd7732e0 | 1639 | locks_insert_block(fl, new_fl, leases_conflict); |
62af4f1f | 1640 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1641 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1642 | percpu_up_read_preempt_enable(&file_rwsem); |
aba37660 | 1643 | |
c45198ed | 1644 | locks_dispose_list(&dispose); |
4321e01e | 1645 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
ada5c1da | 1646 | !new_fl->fl_blocker, break_time); |
aba37660 | 1647 | |
87709e28 | 1648 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1649 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1650 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1651 | locks_delete_block(new_fl); |
1da177e4 | 1652 | if (error >= 0) { |
778fc546 BF |
1653 | /* |
1654 | * Wait for the next conflicting lease that has not been | |
1655 | * broken yet | |
1656 | */ | |
03d12ddf JL |
1657 | if (error == 0) |
1658 | time_out_leases(inode, &dispose); | |
1659 | if (any_leases_conflict(inode, new_fl)) | |
1660 | goto restart; | |
1da177e4 LT |
1661 | error = 0; |
1662 | } | |
1da177e4 | 1663 | out: |
6109c850 | 1664 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1665 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1666 | locks_dispose_list(&dispose); |
6d4b9e38 | 1667 | locks_free_lock(new_fl); |
1da177e4 LT |
1668 | return error; |
1669 | } | |
1670 | ||
1671 | EXPORT_SYMBOL(__break_lease); | |
1672 | ||
1673 | /** | |
76c47948 | 1674 | * lease_get_mtime - update modified time of an inode with exclusive lease |
1da177e4 | 1675 | * @inode: the inode |
76c47948 | 1676 | * @time: pointer to a timespec which contains the last modified time |
1da177e4 LT |
1677 | * |
1678 | * This is to force NFS clients to flush their caches for files with | |
1679 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1680 | * exclusive lease, then they could be modifying it. |
1da177e4 | 1681 | */ |
95582b00 | 1682 | void lease_get_mtime(struct inode *inode, struct timespec64 *time) |
1da177e4 | 1683 | { |
bfe86024 | 1684 | bool has_lease = false; |
128a3785 | 1685 | struct file_lock_context *ctx; |
8634b51f | 1686 | struct file_lock *fl; |
bfe86024 | 1687 | |
128a3785 | 1688 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1689 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1690 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1691 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1692 | struct file_lock, fl_list); | |
1693 | if (fl && (fl->fl_type == F_WRLCK)) | |
1694 | has_lease = true; | |
6109c850 | 1695 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1696 | } |
1697 | ||
1698 | if (has_lease) | |
c2050a45 | 1699 | *time = current_time(inode); |
1da177e4 LT |
1700 | } |
1701 | ||
1702 | EXPORT_SYMBOL(lease_get_mtime); | |
1703 | ||
1704 | /** | |
1705 | * fcntl_getlease - Enquire what lease is currently active | |
1706 | * @filp: the file | |
1707 | * | |
1708 | * The value returned by this function will be one of | |
1709 | * (if no lease break is pending): | |
1710 | * | |
1711 | * %F_RDLCK to indicate a shared lease is held. | |
1712 | * | |
1713 | * %F_WRLCK to indicate an exclusive lease is held. | |
1714 | * | |
1715 | * %F_UNLCK to indicate no lease is held. | |
1716 | * | |
1717 | * (if a lease break is pending): | |
1718 | * | |
1719 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1720 | * changed to a shared lease (or removed). | |
1721 | * | |
1722 | * %F_UNLCK to indicate the lease needs to be removed. | |
1723 | * | |
1724 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1725 | * should be returned to userspace. | |
1726 | */ | |
1727 | int fcntl_getlease(struct file *filp) | |
1728 | { | |
1729 | struct file_lock *fl; | |
c568d683 | 1730 | struct inode *inode = locks_inode(filp); |
128a3785 | 1731 | struct file_lock_context *ctx; |
1da177e4 | 1732 | int type = F_UNLCK; |
c45198ed | 1733 | LIST_HEAD(dispose); |
1da177e4 | 1734 | |
128a3785 | 1735 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1736 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
5f43086b | 1737 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1738 | spin_lock(&ctx->flc_lock); |
c568d683 | 1739 | time_out_leases(inode, &dispose); |
8634b51f JL |
1740 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
1741 | if (fl->fl_file != filp) | |
1742 | continue; | |
778fc546 | 1743 | type = target_leasetype(fl); |
1da177e4 LT |
1744 | break; |
1745 | } | |
6109c850 | 1746 | spin_unlock(&ctx->flc_lock); |
5f43086b PZ |
1747 | percpu_up_read_preempt_enable(&file_rwsem); |
1748 | ||
8634b51f | 1749 | locks_dispose_list(&dispose); |
1da177e4 | 1750 | } |
1da177e4 LT |
1751 | return type; |
1752 | } | |
1753 | ||
24cbe784 JL |
1754 | /** |
1755 | * check_conflicting_open - see if the given dentry points to a file that has | |
1756 | * an existing open that would conflict with the | |
1757 | * desired lease. | |
1758 | * @dentry: dentry to check | |
1759 | * @arg: type of lease that we're trying to acquire | |
7fadc59c | 1760 | * @flags: current lock flags |
24cbe784 JL |
1761 | * |
1762 | * Check to see if there's an existing open fd on this file that would | |
1763 | * conflict with the lease we're trying to set. | |
1764 | */ | |
1765 | static int | |
11afe9f7 | 1766 | check_conflicting_open(const struct dentry *dentry, const long arg, int flags) |
24cbe784 JL |
1767 | { |
1768 | int ret = 0; | |
1769 | struct inode *inode = dentry->d_inode; | |
1770 | ||
11afe9f7 CH |
1771 | if (flags & FL_LAYOUT) |
1772 | return 0; | |
1773 | ||
8cf9ee50 | 1774 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
24cbe784 JL |
1775 | return -EAGAIN; |
1776 | ||
1777 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1778 | (atomic_read(&inode->i_count) > 1))) | |
1779 | ret = -EAGAIN; | |
1780 | ||
1781 | return ret; | |
1782 | } | |
1783 | ||
e6f5c789 JL |
1784 | static int |
1785 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1786 | { |
8634b51f | 1787 | struct file_lock *fl, *my_fl = NULL, *lease; |
0f7fc9e4 | 1788 | struct dentry *dentry = filp->f_path.dentry; |
c568d683 | 1789 | struct inode *inode = dentry->d_inode; |
8634b51f | 1790 | struct file_lock_context *ctx; |
df4e8d2c | 1791 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1792 | int error; |
c45198ed | 1793 | LIST_HEAD(dispose); |
1da177e4 | 1794 | |
096657b6 | 1795 | lease = *flp; |
62af4f1f JL |
1796 | trace_generic_add_lease(inode, lease); |
1797 | ||
5c1c669a JL |
1798 | /* Note that arg is never F_UNLCK here */ |
1799 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1800 | if (!ctx) |
1801 | return -ENOMEM; | |
1802 | ||
df4e8d2c BF |
1803 | /* |
1804 | * In the delegation case we need mutual exclusion with | |
1805 | * a number of operations that take the i_mutex. We trylock | |
1806 | * because delegations are an optional optimization, and if | |
1807 | * there's some chance of a conflict--we'd rather not | |
1808 | * bother, maybe that's a sign this just isn't a good file to | |
1809 | * hand out a delegation on. | |
1810 | */ | |
5955102c | 1811 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1812 | return -EAGAIN; |
1813 | ||
1814 | if (is_deleg && arg == F_WRLCK) { | |
1815 | /* Write delegations are not currently supported: */ | |
5955102c | 1816 | inode_unlock(inode); |
df4e8d2c BF |
1817 | WARN_ON_ONCE(1); |
1818 | return -EINVAL; | |
1819 | } | |
096657b6 | 1820 | |
87709e28 | 1821 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1822 | spin_lock(&ctx->flc_lock); |
c45198ed | 1823 | time_out_leases(inode, &dispose); |
11afe9f7 | 1824 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
24cbe784 | 1825 | if (error) |
096657b6 | 1826 | goto out; |
6d5e8b05 | 1827 | |
1da177e4 LT |
1828 | /* |
1829 | * At this point, we know that if there is an exclusive | |
1830 | * lease on this file, then we hold it on this filp | |
1831 | * (otherwise our open of this file would have blocked). | |
1832 | * And if we are trying to acquire an exclusive lease, | |
1833 | * then the file is not open by anyone (including us) | |
1834 | * except for this filp. | |
1835 | */ | |
c1f24ef4 | 1836 | error = -EAGAIN; |
8634b51f | 1837 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1838 | if (fl->fl_file == filp && |
1839 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1840 | my_fl = fl; |
c1f24ef4 BF |
1841 | continue; |
1842 | } | |
8634b51f | 1843 | |
c1f24ef4 BF |
1844 | /* |
1845 | * No exclusive leases if someone else has a lease on | |
1846 | * this file: | |
1847 | */ | |
1848 | if (arg == F_WRLCK) | |
1849 | goto out; | |
1850 | /* | |
1851 | * Modifying our existing lease is OK, but no getting a | |
1852 | * new lease if someone else is opening for write: | |
1853 | */ | |
1854 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1855 | goto out; | |
1da177e4 LT |
1856 | } |
1857 | ||
8634b51f | 1858 | if (my_fl != NULL) { |
0164bf02 JL |
1859 | lease = my_fl; |
1860 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1861 | if (error) |
1862 | goto out; | |
1863 | goto out_setup; | |
1da177e4 LT |
1864 | } |
1865 | ||
1da177e4 LT |
1866 | error = -EINVAL; |
1867 | if (!leases_enable) | |
1868 | goto out; | |
1869 | ||
e084c1bd | 1870 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1871 | /* |
1872 | * The check in break_lease() is lockless. It's possible for another | |
1873 | * open to race in after we did the earlier check for a conflicting | |
1874 | * open but before the lease was inserted. Check again for a | |
1875 | * conflicting open and cancel the lease if there is one. | |
1876 | * | |
1877 | * We also add a barrier here to ensure that the insertion of the lock | |
1878 | * precedes these checks. | |
1879 | */ | |
1880 | smp_mb(); | |
11afe9f7 | 1881 | error = check_conflicting_open(dentry, arg, lease->fl_flags); |
8634b51f | 1882 | if (error) { |
e084c1bd | 1883 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1884 | goto out; |
1885 | } | |
1c7dd2ff JL |
1886 | |
1887 | out_setup: | |
1888 | if (lease->fl_lmops->lm_setup) | |
1889 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1890 | out: |
6109c850 | 1891 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1892 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1893 | locks_dispose_list(&dispose); |
df4e8d2c | 1894 | if (is_deleg) |
5955102c | 1895 | inode_unlock(inode); |
8634b51f | 1896 | if (!error && !my_fl) |
1c7dd2ff | 1897 | *flp = NULL; |
1da177e4 LT |
1898 | return error; |
1899 | } | |
8335ebd9 | 1900 | |
2ab99ee1 | 1901 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1902 | { |
0efaa7e8 | 1903 | int error = -EAGAIN; |
8634b51f | 1904 | struct file_lock *fl, *victim = NULL; |
c568d683 | 1905 | struct inode *inode = locks_inode(filp); |
128a3785 | 1906 | struct file_lock_context *ctx; |
c45198ed | 1907 | LIST_HEAD(dispose); |
8335ebd9 | 1908 | |
128a3785 | 1909 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1910 | if (!ctx) { |
1911 | trace_generic_delete_lease(inode, NULL); | |
1912 | return error; | |
1913 | } | |
1914 | ||
87709e28 | 1915 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 1916 | spin_lock(&ctx->flc_lock); |
8634b51f | 1917 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1918 | if (fl->fl_file == filp && |
1919 | fl->fl_owner == owner) { | |
8634b51f | 1920 | victim = fl; |
0efaa7e8 | 1921 | break; |
8634b51f | 1922 | } |
8335ebd9 | 1923 | } |
a9b1b455 | 1924 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1925 | if (victim) |
7448cc37 | 1926 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1927 | spin_unlock(&ctx->flc_lock); |
87709e28 | 1928 | percpu_up_read_preempt_enable(&file_rwsem); |
c45198ed | 1929 | locks_dispose_list(&dispose); |
0efaa7e8 | 1930 | return error; |
8335ebd9 BF |
1931 | } |
1932 | ||
1933 | /** | |
1934 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1935 | * @filp: file pointer |
1936 | * @arg: type of lease to obtain | |
1937 | * @flp: input - file_lock to use, output - file_lock inserted | |
1938 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1939 | * doesn't require it) | |
8335ebd9 BF |
1940 | * |
1941 | * The (input) flp->fl_lmops->lm_break function is required | |
1942 | * by break_lease(). | |
8335ebd9 | 1943 | */ |
e6f5c789 JL |
1944 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1945 | void **priv) | |
8335ebd9 | 1946 | { |
c568d683 | 1947 | struct inode *inode = locks_inode(filp); |
8335ebd9 BF |
1948 | int error; |
1949 | ||
8e96e3b7 | 1950 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1951 | return -EACCES; |
1952 | if (!S_ISREG(inode->i_mode)) | |
1953 | return -EINVAL; | |
1954 | error = security_file_lock(filp, arg); | |
1955 | if (error) | |
1956 | return error; | |
1957 | ||
8335ebd9 BF |
1958 | switch (arg) { |
1959 | case F_UNLCK: | |
2ab99ee1 | 1960 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1961 | case F_RDLCK: |
1962 | case F_WRLCK: | |
0efaa7e8 JL |
1963 | if (!(*flp)->fl_lmops->lm_break) { |
1964 | WARN_ON_ONCE(1); | |
1965 | return -ENOLCK; | |
1966 | } | |
11afe9f7 | 1967 | |
e6f5c789 | 1968 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1969 | default: |
8d657eb3 | 1970 | return -EINVAL; |
8335ebd9 BF |
1971 | } |
1972 | } | |
0af1a450 | 1973 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1974 | |
b89f4321 | 1975 | /** |
e51673aa | 1976 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1977 | * @filp: file pointer |
1978 | * @arg: type of lease to obtain | |
1979 | * @lease: file_lock to use when adding a lease | |
1980 | * @priv: private info for lm_setup when adding a lease (may be | |
1981 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1982 | * |
1983 | * Call this to establish a lease on the file. The "lease" argument is not | |
1984 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
80b79dd0 MCC |
1985 | * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be |
1986 | * set; if not, this function will return -ENOLCK (and generate a scary-looking | |
e51673aa | 1987 | * stack trace). |
1c7dd2ff JL |
1988 | * |
1989 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1990 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1991 | */ |
e6f5c789 JL |
1992 | int |
1993 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1994 | { |
de2a4a50 | 1995 | if (filp->f_op->setlease) |
f82b4b67 | 1996 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1997 | else |
f82b4b67 | 1998 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1999 | } |
a9933cea | 2000 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 2001 | |
0ceaf6c7 | 2002 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 2003 | { |
1c7dd2ff | 2004 | struct file_lock *fl; |
f7347ce4 | 2005 | struct fasync_struct *new; |
1da177e4 LT |
2006 | int error; |
2007 | ||
c5b1f0d9 AB |
2008 | fl = lease_alloc(filp, arg); |
2009 | if (IS_ERR(fl)) | |
2010 | return PTR_ERR(fl); | |
1da177e4 | 2011 | |
f7347ce4 LT |
2012 | new = fasync_alloc(); |
2013 | if (!new) { | |
2014 | locks_free_lock(fl); | |
2015 | return -ENOMEM; | |
2016 | } | |
1c7dd2ff | 2017 | new->fa_fd = fd; |
f7347ce4 | 2018 | |
1c7dd2ff | 2019 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
2020 | if (fl) |
2021 | locks_free_lock(fl); | |
f7347ce4 LT |
2022 | if (new) |
2023 | fasync_free(new); | |
1da177e4 LT |
2024 | return error; |
2025 | } | |
2026 | ||
0ceaf6c7 BF |
2027 | /** |
2028 | * fcntl_setlease - sets a lease on an open file | |
2029 | * @fd: open file descriptor | |
2030 | * @filp: file pointer | |
2031 | * @arg: type of lease to obtain | |
2032 | * | |
2033 | * Call this fcntl to establish a lease on the file. | |
2034 | * Note that you also need to call %F_SETSIG to | |
2035 | * receive a signal when the lease is broken. | |
2036 | */ | |
2037 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
2038 | { | |
2039 | if (arg == F_UNLCK) | |
2ab99ee1 | 2040 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
2041 | return do_fcntl_add_lease(fd, filp, arg); |
2042 | } | |
2043 | ||
1da177e4 | 2044 | /** |
29d01b22 JL |
2045 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
2046 | * @inode: inode of the file to apply to | |
1da177e4 LT |
2047 | * @fl: The lock to be applied |
2048 | * | |
29d01b22 | 2049 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 2050 | */ |
616fb38f | 2051 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
2052 | { |
2053 | int error; | |
2054 | might_sleep(); | |
2055 | for (;;) { | |
29d01b22 | 2056 | error = flock_lock_inode(inode, fl); |
bde74e4b | 2057 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 2058 | break; |
ada5c1da | 2059 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
2060 | if (error) |
2061 | break; | |
1da177e4 | 2062 | } |
16306a61 | 2063 | locks_delete_block(fl); |
1da177e4 LT |
2064 | return error; |
2065 | } | |
2066 | ||
e55c34a6 BC |
2067 | /** |
2068 | * locks_lock_inode_wait - Apply a lock to an inode | |
2069 | * @inode: inode of the file to apply to | |
2070 | * @fl: The lock to be applied | |
2071 | * | |
2072 | * Apply a POSIX or FLOCK style lock request to an inode. | |
2073 | */ | |
2074 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
2075 | { | |
2076 | int res = 0; | |
2077 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
2078 | case FL_POSIX: | |
2079 | res = posix_lock_inode_wait(inode, fl); | |
2080 | break; | |
2081 | case FL_FLOCK: | |
2082 | res = flock_lock_inode_wait(inode, fl); | |
2083 | break; | |
2084 | default: | |
2085 | BUG(); | |
2086 | } | |
2087 | return res; | |
2088 | } | |
2089 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
2090 | ||
1da177e4 LT |
2091 | /** |
2092 | * sys_flock: - flock() system call. | |
2093 | * @fd: the file descriptor to lock. | |
2094 | * @cmd: the type of lock to apply. | |
2095 | * | |
2096 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
80b79dd0 | 2097 | * The @cmd can be one of: |
1da177e4 | 2098 | * |
80b79dd0 MCC |
2099 | * - %LOCK_SH -- a shared lock. |
2100 | * - %LOCK_EX -- an exclusive lock. | |
2101 | * - %LOCK_UN -- remove an existing lock. | |
2102 | * - %LOCK_MAND -- a 'mandatory' flock. | |
2103 | * This exists to emulate Windows Share Modes. | |
1da177e4 LT |
2104 | * |
2105 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
2106 | * processes read and write access respectively. | |
2107 | */ | |
002c8976 | 2108 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 2109 | { |
2903ff01 | 2110 | struct fd f = fdget(fd); |
1da177e4 LT |
2111 | struct file_lock *lock; |
2112 | int can_sleep, unlock; | |
2113 | int error; | |
2114 | ||
2115 | error = -EBADF; | |
2903ff01 | 2116 | if (!f.file) |
1da177e4 LT |
2117 | goto out; |
2118 | ||
2119 | can_sleep = !(cmd & LOCK_NB); | |
2120 | cmd &= ~LOCK_NB; | |
2121 | unlock = (cmd == LOCK_UN); | |
2122 | ||
aeb5d727 | 2123 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 2124 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
2125 | goto out_putf; |
2126 | ||
d6367d62 | 2127 | lock = flock_make_lock(f.file, cmd, NULL); |
6e129d00 JL |
2128 | if (IS_ERR(lock)) { |
2129 | error = PTR_ERR(lock); | |
1da177e4 | 2130 | goto out_putf; |
6e129d00 JL |
2131 | } |
2132 | ||
1da177e4 LT |
2133 | if (can_sleep) |
2134 | lock->fl_flags |= FL_SLEEP; | |
2135 | ||
2903ff01 | 2136 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
2137 | if (error) |
2138 | goto out_free; | |
2139 | ||
de2a4a50 | 2140 | if (f.file->f_op->flock) |
2903ff01 | 2141 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
2142 | (can_sleep) ? F_SETLKW : F_SETLK, |
2143 | lock); | |
2144 | else | |
4f656367 | 2145 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
2146 | |
2147 | out_free: | |
993dfa87 | 2148 | locks_free_lock(lock); |
1da177e4 LT |
2149 | |
2150 | out_putf: | |
2903ff01 | 2151 | fdput(f); |
1da177e4 LT |
2152 | out: |
2153 | return error; | |
2154 | } | |
2155 | ||
3ee17abd BF |
2156 | /** |
2157 | * vfs_test_lock - test file byte range lock | |
2158 | * @filp: The file to test lock for | |
6924c554 | 2159 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2160 | * |
2161 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2162 | * setting conf->fl_type to something other than F_UNLCK. | |
2163 | */ | |
2164 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2165 | { | |
de2a4a50 | 2166 | if (filp->f_op->lock) |
3ee17abd BF |
2167 | return filp->f_op->lock(filp, F_GETLK, fl); |
2168 | posix_test_lock(filp, fl); | |
2169 | return 0; | |
2170 | } | |
2171 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2172 | ||
9d5b86ac BC |
2173 | /** |
2174 | * locks_translate_pid - translate a file_lock's fl_pid number into a namespace | |
2175 | * @fl: The file_lock who's fl_pid should be translated | |
2176 | * @ns: The namespace into which the pid should be translated | |
2177 | * | |
2178 | * Used to tranlate a fl_pid into a namespace virtual pid number | |
2179 | */ | |
2180 | static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns) | |
2181 | { | |
2182 | pid_t vnr; | |
2183 | struct pid *pid; | |
2184 | ||
2185 | if (IS_OFDLCK(fl)) | |
2186 | return -1; | |
2187 | if (IS_REMOTELCK(fl)) | |
2188 | return fl->fl_pid; | |
826d7bc9 KK |
2189 | /* |
2190 | * If the flock owner process is dead and its pid has been already | |
2191 | * freed, the translation below won't work, but we still want to show | |
2192 | * flock owner pid number in init pidns. | |
2193 | */ | |
2194 | if (ns == &init_pid_ns) | |
2195 | return (pid_t)fl->fl_pid; | |
9d5b86ac BC |
2196 | |
2197 | rcu_read_lock(); | |
2198 | pid = find_pid_ns(fl->fl_pid, &init_pid_ns); | |
2199 | vnr = pid_nr_ns(pid, ns); | |
2200 | rcu_read_unlock(); | |
2201 | return vnr; | |
2202 | } | |
2203 | ||
c2fa1b8a BF |
2204 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2205 | { | |
9d5b86ac | 2206 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2207 | #if BITS_PER_LONG == 32 |
2208 | /* | |
2209 | * Make sure we can represent the posix lock via | |
2210 | * legacy 32bit flock. | |
2211 | */ | |
2212 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2213 | return -EOVERFLOW; | |
2214 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2215 | return -EOVERFLOW; | |
2216 | #endif | |
2217 | flock->l_start = fl->fl_start; | |
2218 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2219 | fl->fl_end - fl->fl_start + 1; | |
2220 | flock->l_whence = 0; | |
129a84de | 2221 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2222 | return 0; |
2223 | } | |
2224 | ||
2225 | #if BITS_PER_LONG == 32 | |
2226 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2227 | { | |
9d5b86ac | 2228 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2229 | flock->l_start = fl->fl_start; |
2230 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2231 | fl->fl_end - fl->fl_start + 1; | |
2232 | flock->l_whence = 0; | |
2233 | flock->l_type = fl->fl_type; | |
2234 | } | |
2235 | #endif | |
2236 | ||
1da177e4 LT |
2237 | /* Report the first existing lock that would conflict with l. |
2238 | * This implements the F_GETLK command of fcntl(). | |
2239 | */ | |
a75d30c7 | 2240 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock) |
1da177e4 | 2241 | { |
52306e88 | 2242 | struct file_lock *fl; |
1da177e4 LT |
2243 | int error; |
2244 | ||
52306e88 BC |
2245 | fl = locks_alloc_lock(); |
2246 | if (fl == NULL) | |
2247 | return -ENOMEM; | |
1da177e4 | 2248 | error = -EINVAL; |
a75d30c7 | 2249 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2250 | goto out; |
2251 | ||
52306e88 | 2252 | error = flock_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2253 | if (error) |
2254 | goto out; | |
2255 | ||
0d3f7a2d | 2256 | if (cmd == F_OFD_GETLK) { |
90478939 | 2257 | error = -EINVAL; |
a75d30c7 | 2258 | if (flock->l_pid != 0) |
90478939 JL |
2259 | goto out; |
2260 | ||
5d50ffd7 | 2261 | cmd = F_GETLK; |
52306e88 BC |
2262 | fl->fl_flags |= FL_OFDLCK; |
2263 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2264 | } |
2265 | ||
52306e88 | 2266 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2267 | if (error) |
2268 | goto out; | |
1da177e4 | 2269 | |
52306e88 BC |
2270 | flock->l_type = fl->fl_type; |
2271 | if (fl->fl_type != F_UNLCK) { | |
2272 | error = posix_lock_to_flock(flock, fl); | |
c2fa1b8a | 2273 | if (error) |
52306e88 | 2274 | goto out; |
1da177e4 | 2275 | } |
1da177e4 | 2276 | out: |
52306e88 | 2277 | locks_free_lock(fl); |
1da177e4 LT |
2278 | return error; |
2279 | } | |
2280 | ||
7723ec97 ME |
2281 | /** |
2282 | * vfs_lock_file - file byte range lock | |
2283 | * @filp: The file to apply the lock to | |
2284 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2285 | * @fl: The lock to be applied | |
150b3934 ME |
2286 | * @conf: Place to return a copy of the conflicting lock, if found. |
2287 | * | |
2288 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2289 | * as the final argument. | |
2290 | * | |
2291 | * If the filesystem defines a private ->lock() method, then @conf will | |
2292 | * be left unchanged; so a caller that cares should initialize it to | |
2293 | * some acceptable default. | |
2beb6614 ME |
2294 | * |
2295 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2296 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2297 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2298 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2299 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2300 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2301 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2302 | * request completes. |
2303 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2304 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2305 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2306 | * nonzero return code and the file system should release the lock. The file |
2307 | * system is also responsible to keep a corresponding posix lock when it | |
2308 | * grants a lock so the VFS can find out which locks are locally held and do | |
2309 | * the correct lock cleanup when required. | |
2310 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2311 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2312 | * return code. |
7723ec97 | 2313 | */ |
150b3934 | 2314 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2315 | { |
de2a4a50 | 2316 | if (filp->f_op->lock) |
7723ec97 ME |
2317 | return filp->f_op->lock(filp, cmd, fl); |
2318 | else | |
150b3934 | 2319 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2320 | } |
2321 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2322 | ||
b648a6de MS |
2323 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2324 | struct file_lock *fl) | |
2325 | { | |
2326 | int error; | |
2327 | ||
2328 | error = security_file_lock(filp, fl->fl_type); | |
2329 | if (error) | |
2330 | return error; | |
2331 | ||
764c76b3 MS |
2332 | for (;;) { |
2333 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2334 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2335 | break; |
ada5c1da | 2336 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
2337 | if (error) |
2338 | break; | |
b648a6de | 2339 | } |
16306a61 | 2340 | locks_delete_block(fl); |
b648a6de MS |
2341 | |
2342 | return error; | |
2343 | } | |
2344 | ||
6ca7d910 | 2345 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2346 | static int |
2347 | check_fmode_for_setlk(struct file_lock *fl) | |
2348 | { | |
2349 | switch (fl->fl_type) { | |
2350 | case F_RDLCK: | |
2351 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2352 | return -EBADF; | |
2353 | break; | |
2354 | case F_WRLCK: | |
2355 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2356 | return -EBADF; | |
2357 | } | |
2358 | return 0; | |
2359 | } | |
2360 | ||
1da177e4 LT |
2361 | /* Apply the lock described by l to an open file descriptor. |
2362 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2363 | */ | |
c293621b | 2364 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2365 | struct flock *flock) |
1da177e4 LT |
2366 | { |
2367 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2368 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2369 | struct file *f; |
1da177e4 LT |
2370 | int error; |
2371 | ||
2372 | if (file_lock == NULL) | |
2373 | return -ENOLCK; | |
2374 | ||
1da177e4 LT |
2375 | /* Don't allow mandatory locks on files that may be memory mapped |
2376 | * and shared. | |
2377 | */ | |
a16877ca | 2378 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2379 | error = -EAGAIN; |
2380 | goto out; | |
2381 | } | |
2382 | ||
a75d30c7 | 2383 | error = flock_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2384 | if (error) |
2385 | goto out; | |
5d50ffd7 | 2386 | |
cf01f4ee JL |
2387 | error = check_fmode_for_setlk(file_lock); |
2388 | if (error) | |
2389 | goto out; | |
2390 | ||
5d50ffd7 JL |
2391 | /* |
2392 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2393 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2394 | */ |
2395 | switch (cmd) { | |
0d3f7a2d | 2396 | case F_OFD_SETLK: |
90478939 | 2397 | error = -EINVAL; |
a75d30c7 | 2398 | if (flock->l_pid != 0) |
90478939 JL |
2399 | goto out; |
2400 | ||
5d50ffd7 | 2401 | cmd = F_SETLK; |
cff2fce5 | 2402 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2403 | file_lock->fl_owner = filp; |
5d50ffd7 | 2404 | break; |
0d3f7a2d | 2405 | case F_OFD_SETLKW: |
90478939 | 2406 | error = -EINVAL; |
a75d30c7 | 2407 | if (flock->l_pid != 0) |
90478939 JL |
2408 | goto out; |
2409 | ||
5d50ffd7 | 2410 | cmd = F_SETLKW; |
cff2fce5 | 2411 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2412 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2413 | /* Fallthrough */ |
2414 | case F_SETLKW: | |
1da177e4 LT |
2415 | file_lock->fl_flags |= FL_SLEEP; |
2416 | } | |
5d50ffd7 | 2417 | |
b648a6de | 2418 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2419 | |
c293621b | 2420 | /* |
0752ba80 JL |
2421 | * Attempt to detect a close/fcntl race and recover by releasing the |
2422 | * lock that was just acquired. There is no need to do that when we're | |
2423 | * unlocking though, or for OFD locks. | |
c293621b | 2424 | */ |
0752ba80 JL |
2425 | if (!error && file_lock->fl_type != F_UNLCK && |
2426 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2427 | /* |
2428 | * We need that spin_lock here - it prevents reordering between | |
2429 | * update of i_flctx->flc_posix and check for it done in | |
2430 | * close(). rcu_read_lock() wouldn't do. | |
2431 | */ | |
2432 | spin_lock(¤t->files->file_lock); | |
2433 | f = fcheck(fd); | |
2434 | spin_unlock(¤t->files->file_lock); | |
2435 | if (f != filp) { | |
2436 | file_lock->fl_type = F_UNLCK; | |
2437 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2438 | WARN_ON_ONCE(error); | |
2439 | error = -EBADF; | |
2440 | } | |
1da177e4 | 2441 | } |
c293621b | 2442 | out: |
1890910f | 2443 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2444 | locks_free_lock(file_lock); |
2445 | return error; | |
2446 | } | |
2447 | ||
2448 | #if BITS_PER_LONG == 32 | |
2449 | /* Report the first existing lock that would conflict with l. | |
2450 | * This implements the F_GETLK command of fcntl(). | |
2451 | */ | |
a75d30c7 | 2452 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock) |
1da177e4 | 2453 | { |
52306e88 | 2454 | struct file_lock *fl; |
1da177e4 LT |
2455 | int error; |
2456 | ||
52306e88 BC |
2457 | fl = locks_alloc_lock(); |
2458 | if (fl == NULL) | |
2459 | return -ENOMEM; | |
2460 | ||
1da177e4 | 2461 | error = -EINVAL; |
a75d30c7 | 2462 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2463 | goto out; |
2464 | ||
52306e88 | 2465 | error = flock64_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2466 | if (error) |
2467 | goto out; | |
2468 | ||
0d3f7a2d | 2469 | if (cmd == F_OFD_GETLK) { |
90478939 | 2470 | error = -EINVAL; |
a75d30c7 | 2471 | if (flock->l_pid != 0) |
90478939 JL |
2472 | goto out; |
2473 | ||
5d50ffd7 | 2474 | cmd = F_GETLK64; |
52306e88 BC |
2475 | fl->fl_flags |= FL_OFDLCK; |
2476 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2477 | } |
2478 | ||
52306e88 | 2479 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2480 | if (error) |
2481 | goto out; | |
2482 | ||
52306e88 BC |
2483 | flock->l_type = fl->fl_type; |
2484 | if (fl->fl_type != F_UNLCK) | |
2485 | posix_lock_to_flock64(flock, fl); | |
f328296e | 2486 | |
1da177e4 | 2487 | out: |
52306e88 | 2488 | locks_free_lock(fl); |
1da177e4 LT |
2489 | return error; |
2490 | } | |
2491 | ||
2492 | /* Apply the lock described by l to an open file descriptor. | |
2493 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2494 | */ | |
c293621b | 2495 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2496 | struct flock64 *flock) |
1da177e4 LT |
2497 | { |
2498 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2499 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2500 | struct file *f; |
1da177e4 LT |
2501 | int error; |
2502 | ||
2503 | if (file_lock == NULL) | |
2504 | return -ENOLCK; | |
2505 | ||
1da177e4 LT |
2506 | /* Don't allow mandatory locks on files that may be memory mapped |
2507 | * and shared. | |
2508 | */ | |
a16877ca | 2509 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2510 | error = -EAGAIN; |
2511 | goto out; | |
2512 | } | |
2513 | ||
a75d30c7 | 2514 | error = flock64_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2515 | if (error) |
2516 | goto out; | |
5d50ffd7 | 2517 | |
cf01f4ee JL |
2518 | error = check_fmode_for_setlk(file_lock); |
2519 | if (error) | |
2520 | goto out; | |
2521 | ||
5d50ffd7 JL |
2522 | /* |
2523 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2524 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2525 | */ |
2526 | switch (cmd) { | |
0d3f7a2d | 2527 | case F_OFD_SETLK: |
90478939 | 2528 | error = -EINVAL; |
a75d30c7 | 2529 | if (flock->l_pid != 0) |
90478939 JL |
2530 | goto out; |
2531 | ||
5d50ffd7 | 2532 | cmd = F_SETLK64; |
cff2fce5 | 2533 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2534 | file_lock->fl_owner = filp; |
5d50ffd7 | 2535 | break; |
0d3f7a2d | 2536 | case F_OFD_SETLKW: |
90478939 | 2537 | error = -EINVAL; |
a75d30c7 | 2538 | if (flock->l_pid != 0) |
90478939 JL |
2539 | goto out; |
2540 | ||
5d50ffd7 | 2541 | cmd = F_SETLKW64; |
cff2fce5 | 2542 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2543 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2544 | /* Fallthrough */ |
2545 | case F_SETLKW64: | |
1da177e4 LT |
2546 | file_lock->fl_flags |= FL_SLEEP; |
2547 | } | |
5d50ffd7 | 2548 | |
b648a6de | 2549 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2550 | |
c293621b | 2551 | /* |
0752ba80 JL |
2552 | * Attempt to detect a close/fcntl race and recover by releasing the |
2553 | * lock that was just acquired. There is no need to do that when we're | |
2554 | * unlocking though, or for OFD locks. | |
c293621b | 2555 | */ |
0752ba80 JL |
2556 | if (!error && file_lock->fl_type != F_UNLCK && |
2557 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2558 | /* |
2559 | * We need that spin_lock here - it prevents reordering between | |
2560 | * update of i_flctx->flc_posix and check for it done in | |
2561 | * close(). rcu_read_lock() wouldn't do. | |
2562 | */ | |
2563 | spin_lock(¤t->files->file_lock); | |
2564 | f = fcheck(fd); | |
2565 | spin_unlock(¤t->files->file_lock); | |
2566 | if (f != filp) { | |
2567 | file_lock->fl_type = F_UNLCK; | |
2568 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2569 | WARN_ON_ONCE(error); | |
2570 | error = -EBADF; | |
2571 | } | |
1da177e4 | 2572 | } |
1da177e4 LT |
2573 | out: |
2574 | locks_free_lock(file_lock); | |
2575 | return error; | |
2576 | } | |
2577 | #endif /* BITS_PER_LONG == 32 */ | |
2578 | ||
2579 | /* | |
2580 | * This function is called when the file is being removed | |
2581 | * from the task's fd array. POSIX locks belonging to this task | |
2582 | * are deleted at this time. | |
2583 | */ | |
2584 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2585 | { | |
1890910f | 2586 | int error; |
c568d683 | 2587 | struct inode *inode = locks_inode(filp); |
ff7b86b8 | 2588 | struct file_lock lock; |
128a3785 | 2589 | struct file_lock_context *ctx; |
1da177e4 LT |
2590 | |
2591 | /* | |
2592 | * If there are no locks held on this file, we don't need to call | |
2593 | * posix_lock_file(). Another process could be setting a lock on this | |
2594 | * file at the same time, but we wouldn't remove that lock anyway. | |
2595 | */ | |
c568d683 | 2596 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 | 2597 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2598 | return; |
2599 | ||
d6367d62 | 2600 | locks_init_lock(&lock); |
1da177e4 | 2601 | lock.fl_type = F_UNLCK; |
75e1fcc0 | 2602 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2603 | lock.fl_start = 0; |
2604 | lock.fl_end = OFFSET_MAX; | |
2605 | lock.fl_owner = owner; | |
2606 | lock.fl_pid = current->tgid; | |
2607 | lock.fl_file = filp; | |
2608 | lock.fl_ops = NULL; | |
2609 | lock.fl_lmops = NULL; | |
2610 | ||
1890910f | 2611 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2612 | |
1da177e4 LT |
2613 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2614 | lock.fl_ops->fl_release_private(&lock); | |
c568d683 | 2615 | trace_locks_remove_posix(inode, &lock, error); |
1da177e4 LT |
2616 | } |
2617 | ||
2618 | EXPORT_SYMBOL(locks_remove_posix); | |
2619 | ||
3d8e560d | 2620 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2621 | static void |
128a3785 | 2622 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 | 2623 | { |
d6367d62 | 2624 | struct file_lock fl; |
c568d683 | 2625 | struct inode *inode = locks_inode(filp); |
dd459bb1 | 2626 | |
3d8e560d | 2627 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2628 | return; |
2629 | ||
d6367d62 N |
2630 | flock_make_lock(filp, LOCK_UN, &fl); |
2631 | fl.fl_flags |= FL_CLOSE; | |
2632 | ||
de2a4a50 | 2633 | if (filp->f_op->flock) |
dd459bb1 JL |
2634 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2635 | else | |
bcd7f78d | 2636 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2637 | |
2638 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2639 | fl.fl_ops->fl_release_private(&fl); | |
2640 | } | |
2641 | ||
3d8e560d | 2642 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2643 | static void |
128a3785 | 2644 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2645 | { |
8634b51f JL |
2646 | struct file_lock *fl, *tmp; |
2647 | LIST_HEAD(dispose); | |
2648 | ||
3d8e560d | 2649 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2650 | return; |
2651 | ||
5f43086b | 2652 | percpu_down_read_preempt_disable(&file_rwsem); |
6109c850 | 2653 | spin_lock(&ctx->flc_lock); |
8634b51f | 2654 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2655 | if (filp == fl->fl_file) |
2656 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2657 | spin_unlock(&ctx->flc_lock); |
5f43086b PZ |
2658 | percpu_up_read_preempt_enable(&file_rwsem); |
2659 | ||
8634b51f JL |
2660 | locks_dispose_list(&dispose); |
2661 | } | |
2662 | ||
1da177e4 LT |
2663 | /* |
2664 | * This function is called on the last close of an open file. | |
2665 | */ | |
78ed8a13 | 2666 | void locks_remove_file(struct file *filp) |
1da177e4 | 2667 | { |
128a3785 DV |
2668 | struct file_lock_context *ctx; |
2669 | ||
c568d683 | 2670 | ctx = smp_load_acquire(&locks_inode(filp)->i_flctx); |
128a3785 | 2671 | if (!ctx) |
3d8e560d JL |
2672 | return; |
2673 | ||
dd459bb1 | 2674 | /* remove any OFD locks */ |
73a8f5f7 | 2675 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2676 | |
dd459bb1 | 2677 | /* remove flock locks */ |
128a3785 | 2678 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2679 | |
8634b51f | 2680 | /* remove any leases */ |
128a3785 | 2681 | locks_remove_lease(filp, ctx); |
3953704f BC |
2682 | |
2683 | spin_lock(&ctx->flc_lock); | |
2684 | locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX"); | |
2685 | locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK"); | |
2686 | locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE"); | |
2687 | spin_unlock(&ctx->flc_lock); | |
1da177e4 LT |
2688 | } |
2689 | ||
9b9d2ab4 ME |
2690 | /** |
2691 | * vfs_cancel_lock - file byte range unblock lock | |
2692 | * @filp: The file to apply the unblock to | |
2693 | * @fl: The lock to be unblocked | |
2694 | * | |
2695 | * Used by lock managers to cancel blocked requests | |
2696 | */ | |
2697 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2698 | { | |
de2a4a50 | 2699 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2700 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2701 | return 0; | |
2702 | } | |
2703 | ||
2704 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2705 | ||
7f8ada98 | 2706 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2707 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2708 | #include <linux/seq_file.h> |
2709 | ||
7012b02a JL |
2710 | struct locks_iterator { |
2711 | int li_cpu; | |
2712 | loff_t li_pos; | |
2713 | }; | |
2714 | ||
7f8ada98 | 2715 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2716 | loff_t id, char *pfx) |
1da177e4 LT |
2717 | { |
2718 | struct inode *inode = NULL; | |
ab1f1611 | 2719 | unsigned int fl_pid; |
9d5b86ac | 2720 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
ab1f1611 | 2721 | |
9d5b86ac BC |
2722 | fl_pid = locks_translate_pid(fl, proc_pidns); |
2723 | /* | |
1cf8e5de KK |
2724 | * If lock owner is dead (and pid is freed) or not visible in current |
2725 | * pidns, zero is shown as a pid value. Check lock info from | |
2726 | * init_pid_ns to get saved lock pid value. | |
9d5b86ac | 2727 | */ |
1da177e4 LT |
2728 | |
2729 | if (fl->fl_file != NULL) | |
c568d683 | 2730 | inode = locks_inode(fl->fl_file); |
1da177e4 | 2731 | |
99dc8292 | 2732 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2733 | if (IS_POSIX(fl)) { |
c918d42a | 2734 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2735 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2736 | else if (IS_OFDLCK(fl)) |
5315c26a | 2737 | seq_puts(f, "OFDLCK"); |
c918d42a | 2738 | else |
5315c26a | 2739 | seq_puts(f, "POSIX "); |
c918d42a JL |
2740 | |
2741 | seq_printf(f, " %s ", | |
1da177e4 | 2742 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2743 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2744 | } else if (IS_FLOCK(fl)) { |
2745 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2746 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2747 | } else { |
5315c26a | 2748 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2749 | } |
2750 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2751 | if (fl->fl_flags & FL_DELEG) |
2752 | seq_puts(f, "DELEG "); | |
2753 | else | |
2754 | seq_puts(f, "LEASE "); | |
2755 | ||
ab83fa4b | 2756 | if (lease_breaking(fl)) |
5315c26a | 2757 | seq_puts(f, "BREAKING "); |
1da177e4 | 2758 | else if (fl->fl_file) |
5315c26a | 2759 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2760 | else |
5315c26a | 2761 | seq_puts(f, "BREAKER "); |
1da177e4 | 2762 | } else { |
5315c26a | 2763 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2764 | } |
2765 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2766 | seq_printf(f, "%s ", |
1da177e4 LT |
2767 | (fl->fl_type & LOCK_READ) |
2768 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2769 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2770 | } else { | |
7f8ada98 | 2771 | seq_printf(f, "%s ", |
ab83fa4b | 2772 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2773 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2774 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2775 | } |
2776 | if (inode) { | |
3648888e | 2777 | /* userspace relies on this representation of dev_t */ |
ab1f1611 | 2778 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2779 | MAJOR(inode->i_sb->s_dev), |
2780 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2781 | } else { |
ab1f1611 | 2782 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2783 | } |
2784 | if (IS_POSIX(fl)) { | |
2785 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2786 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2787 | else |
7f8ada98 | 2788 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2789 | } else { |
5315c26a | 2790 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2791 | } |
2792 | } | |
2793 | ||
7f8ada98 | 2794 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2795 | { |
7012b02a | 2796 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2797 | struct file_lock *fl, *bfl; |
d67fd44f | 2798 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
1da177e4 | 2799 | |
139ca04e | 2800 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2801 | |
9d5b86ac | 2802 | if (locks_translate_pid(fl, proc_pidns) == 0) |
d67fd44f NB |
2803 | return 0; |
2804 | ||
7012b02a | 2805 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2806 | |
ada5c1da | 2807 | list_for_each_entry(bfl, &fl->fl_blocked_requests, fl_blocked_member) |
7012b02a | 2808 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2809 | |
7f8ada98 PE |
2810 | return 0; |
2811 | } | |
1da177e4 | 2812 | |
6c8c9031 AV |
2813 | static void __show_fd_locks(struct seq_file *f, |
2814 | struct list_head *head, int *id, | |
2815 | struct file *filp, struct files_struct *files) | |
2816 | { | |
2817 | struct file_lock *fl; | |
2818 | ||
2819 | list_for_each_entry(fl, head, fl_list) { | |
2820 | ||
2821 | if (filp != fl->fl_file) | |
2822 | continue; | |
2823 | if (fl->fl_owner != files && | |
2824 | fl->fl_owner != filp) | |
2825 | continue; | |
2826 | ||
2827 | (*id)++; | |
2828 | seq_puts(f, "lock:\t"); | |
2829 | lock_get_status(f, fl, *id, ""); | |
2830 | } | |
2831 | } | |
2832 | ||
2833 | void show_fd_locks(struct seq_file *f, | |
2834 | struct file *filp, struct files_struct *files) | |
2835 | { | |
c568d683 | 2836 | struct inode *inode = locks_inode(filp); |
6c8c9031 AV |
2837 | struct file_lock_context *ctx; |
2838 | int id = 0; | |
2839 | ||
128a3785 | 2840 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2841 | if (!ctx) |
2842 | return; | |
2843 | ||
2844 | spin_lock(&ctx->flc_lock); | |
2845 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2846 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2847 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2848 | spin_unlock(&ctx->flc_lock); | |
2849 | } | |
2850 | ||
7f8ada98 | 2851 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2852 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2853 | { |
7012b02a | 2854 | struct locks_iterator *iter = f->private; |
99dc8292 | 2855 | |
7012b02a | 2856 | iter->li_pos = *pos + 1; |
aba37660 | 2857 | percpu_down_write(&file_rwsem); |
7b2296af | 2858 | spin_lock(&blocked_lock_lock); |
7c3f654d | 2859 | return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos); |
7f8ada98 | 2860 | } |
1da177e4 | 2861 | |
7f8ada98 PE |
2862 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2863 | { | |
7012b02a JL |
2864 | struct locks_iterator *iter = f->private; |
2865 | ||
2866 | ++iter->li_pos; | |
7c3f654d | 2867 | return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos); |
7f8ada98 | 2868 | } |
1da177e4 | 2869 | |
7f8ada98 | 2870 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2871 | __releases(&blocked_lock_lock) |
7f8ada98 | 2872 | { |
7b2296af | 2873 | spin_unlock(&blocked_lock_lock); |
aba37660 | 2874 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2875 | } |
2876 | ||
d8ba7a36 | 2877 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2878 | .start = locks_start, |
2879 | .next = locks_next, | |
2880 | .stop = locks_stop, | |
2881 | .show = locks_show, | |
2882 | }; | |
d8ba7a36 | 2883 | |
d8ba7a36 AD |
2884 | static int __init proc_locks_init(void) |
2885 | { | |
44414d82 CH |
2886 | proc_create_seq_private("locks", 0, NULL, &locks_seq_operations, |
2887 | sizeof(struct locks_iterator), NULL); | |
d8ba7a36 AD |
2888 | return 0; |
2889 | } | |
91899226 | 2890 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2891 | #endif |
2892 | ||
1da177e4 LT |
2893 | static int __init filelock_init(void) |
2894 | { | |
7012b02a JL |
2895 | int i; |
2896 | ||
4a075e39 JL |
2897 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2898 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2899 | ||
1da177e4 | 2900 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2901 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2902 | ||
7012b02a | 2903 | |
7c3f654d PZ |
2904 | for_each_possible_cpu(i) { |
2905 | struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i); | |
2906 | ||
2907 | spin_lock_init(&fll->lock); | |
2908 | INIT_HLIST_HEAD(&fll->hlist); | |
2909 | } | |
7012b02a | 2910 | |
1da177e4 LT |
2911 | return 0; |
2912 | } | |
2913 | ||
2914 | core_initcall(filelock_init); |