]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/fs/locks.c | |
4 | * | |
5 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. | |
6 | * Doug Evans (dje@spiff.uucp), August 07, 1992 | |
7 | * | |
8 | * Deadlock detection added. | |
9 | * FIXME: one thing isn't handled yet: | |
10 | * - mandatory locks (requires lots of changes elsewhere) | |
11 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. | |
12 | * | |
13 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. | |
14 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 | |
7bbd1fc0 | 15 | * |
1da177e4 LT |
16 | * Converted file_lock_table to a linked list from an array, which eliminates |
17 | * the limits on how many active file locks are open. | |
18 | * Chad Page (pageone@netcom.com), November 27, 1994 | |
7bbd1fc0 | 19 | * |
1da177e4 LT |
20 | * Removed dependency on file descriptors. dup()'ed file descriptors now |
21 | * get the same locks as the original file descriptors, and a close() on | |
22 | * any file descriptor removes ALL the locks on the file for the current | |
23 | * process. Since locks still depend on the process id, locks are inherited | |
24 | * after an exec() but not after a fork(). This agrees with POSIX, and both | |
25 | * BSD and SVR4 practice. | |
26 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 | |
27 | * | |
28 | * Scrapped free list which is redundant now that we allocate locks | |
29 | * dynamically with kmalloc()/kfree(). | |
30 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 | |
31 | * | |
32 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. | |
33 | * | |
34 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the | |
35 | * fcntl() system call. They have the semantics described above. | |
36 | * | |
37 | * FL_FLOCK locks are created with calls to flock(), through the flock() | |
38 | * system call, which is new. Old C libraries implement flock() via fcntl() | |
39 | * and will continue to use the old, broken implementation. | |
40 | * | |
41 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated | |
42 | * with a file pointer (filp). As a result they can be shared by a parent | |
43 | * process and its children after a fork(). They are removed when the last | |
44 | * file descriptor referring to the file pointer is closed (unless explicitly | |
7bbd1fc0 | 45 | * unlocked). |
1da177e4 LT |
46 | * |
47 | * FL_FLOCK locks never deadlock, an existing lock is always removed before | |
48 | * upgrading from shared to exclusive (or vice versa). When this happens | |
49 | * any processes blocked by the current lock are woken up and allowed to | |
50 | * run before the new lock is applied. | |
51 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 | |
52 | * | |
53 | * Removed some race conditions in flock_lock_file(), marked other possible | |
7bbd1fc0 | 54 | * races. Just grep for FIXME to see them. |
1da177e4 LT |
55 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. |
56 | * | |
57 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. | |
58 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep | |
59 | * once we've checked for blocking and deadlocking. | |
60 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. | |
61 | * | |
62 | * Initial implementation of mandatory locks. SunOS turned out to be | |
63 | * a rotten model, so I implemented the "obvious" semantics. | |
395cf969 | 64 | * See 'Documentation/filesystems/mandatory-locking.txt' for details. |
1da177e4 LT |
65 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
66 | * | |
67 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to | |
68 | * check if a file has mandatory locks, used by mmap(), open() and creat() to | |
69 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference | |
70 | * Manual, Section 2. | |
71 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. | |
72 | * | |
73 | * Tidied up block list handling. Added '/proc/locks' interface. | |
74 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. | |
75 | * | |
76 | * Fixed deadlock condition for pathological code that mixes calls to | |
77 | * flock() and fcntl(). | |
78 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. | |
79 | * | |
80 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use | |
81 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to | |
82 | * guarantee sensible behaviour in the case where file system modules might | |
83 | * be compiled with different options than the kernel itself. | |
84 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
85 | * | |
86 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel | |
87 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. | |
88 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
89 | * | |
90 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK | |
91 | * locks. Changed process synchronisation to avoid dereferencing locks that | |
92 | * have already been freed. | |
93 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. | |
94 | * | |
95 | * Made the block list a circular list to minimise searching in the list. | |
96 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. | |
97 | * | |
98 | * Made mandatory locking a mount option. Default is not to allow mandatory | |
99 | * locking. | |
100 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. | |
101 | * | |
102 | * Some adaptations for NFS support. | |
103 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, | |
104 | * | |
105 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. | |
106 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. | |
107 | * | |
108 | * Use slab allocator instead of kmalloc/kfree. | |
109 | * Use generic list implementation from <linux/list.h>. | |
110 | * Sped up posix_locks_deadlock by only considering blocked locks. | |
111 | * Matthew Wilcox <willy@debian.org>, March, 2000. | |
112 | * | |
113 | * Leases and LOCK_MAND | |
114 | * Matthew Wilcox <willy@debian.org>, June, 2000. | |
115 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. | |
fd7732e0 N |
116 | * |
117 | * Locking conflicts and dependencies: | |
118 | * If multiple threads attempt to lock the same byte (or flock the same file) | |
119 | * only one can be granted the lock, and other must wait their turn. | |
120 | * The first lock has been "applied" or "granted", the others are "waiting" | |
121 | * and are "blocked" by the "applied" lock.. | |
122 | * | |
123 | * Waiting and applied locks are all kept in trees whose properties are: | |
124 | * | |
125 | * - the root of a tree may be an applied or waiting lock. | |
126 | * - every other node in the tree is a waiting lock that | |
127 | * conflicts with every ancestor of that node. | |
128 | * | |
129 | * Every such tree begins life as a waiting singleton which obviously | |
130 | * satisfies the above properties. | |
131 | * | |
132 | * The only ways we modify trees preserve these properties: | |
133 | * | |
134 | * 1. We may add a new leaf node, but only after first verifying that it | |
135 | * conflicts with all of its ancestors. | |
136 | * 2. We may remove the root of a tree, creating a new singleton | |
137 | * tree from the root and N new trees rooted in the immediate | |
138 | * children. | |
139 | * 3. If the root of a tree is not currently an applied lock, we may | |
140 | * apply it (if possible). | |
141 | * 4. We may upgrade the root of the tree (either extend its range, | |
142 | * or upgrade its entire range from read to write). | |
143 | * | |
144 | * When an applied lock is modified in a way that reduces or downgrades any | |
145 | * part of its range, we remove all its children (2 above). This particularly | |
146 | * happens when a lock is unlocked. | |
147 | * | |
148 | * For each of those child trees we "wake up" the thread which is | |
149 | * waiting for the lock so it can continue handling as follows: if the | |
150 | * root of the tree applies, we do so (3). If it doesn't, it must | |
151 | * conflict with some applied lock. We remove (wake up) all of its children | |
152 | * (2), and add it is a new leaf to the tree rooted in the applied | |
153 | * lock (1). We then repeat the process recursively with those | |
154 | * children. | |
155 | * | |
1da177e4 LT |
156 | */ |
157 | ||
158 | #include <linux/capability.h> | |
159 | #include <linux/file.h> | |
9f3acc31 | 160 | #include <linux/fdtable.h> |
1da177e4 LT |
161 | #include <linux/fs.h> |
162 | #include <linux/init.h> | |
1da177e4 LT |
163 | #include <linux/security.h> |
164 | #include <linux/slab.h> | |
1da177e4 LT |
165 | #include <linux/syscalls.h> |
166 | #include <linux/time.h> | |
4fb3a538 | 167 | #include <linux/rcupdate.h> |
ab1f1611 | 168 | #include <linux/pid_namespace.h> |
48f74186 | 169 | #include <linux/hashtable.h> |
7012b02a | 170 | #include <linux/percpu.h> |
1da177e4 | 171 | |
62af4f1f JL |
172 | #define CREATE_TRACE_POINTS |
173 | #include <trace/events/filelock.h> | |
174 | ||
7c0f6ba6 | 175 | #include <linux/uaccess.h> |
1da177e4 LT |
176 | |
177 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
178 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
11afe9f7 | 179 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) |
cff2fce5 | 180 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
9d5b86ac | 181 | #define IS_REMOTELCK(fl) (fl->fl_pid <= 0) |
1da177e4 | 182 | |
ab83fa4b BF |
183 | static bool lease_breaking(struct file_lock *fl) |
184 | { | |
778fc546 BF |
185 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
186 | } | |
187 | ||
188 | static int target_leasetype(struct file_lock *fl) | |
189 | { | |
190 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
191 | return F_UNLCK; | |
192 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
193 | return F_RDLCK; | |
194 | return fl->fl_type; | |
ab83fa4b BF |
195 | } |
196 | ||
1da177e4 LT |
197 | int leases_enable = 1; |
198 | int lease_break_time = 45; | |
199 | ||
1c8c601a | 200 | /* |
7012b02a | 201 | * The global file_lock_list is only used for displaying /proc/locks, so we |
7c3f654d PZ |
202 | * keep a list on each CPU, with each list protected by its own spinlock. |
203 | * Global serialization is done using file_rwsem. | |
204 | * | |
205 | * Note that alterations to the list also require that the relevant flc_lock is | |
206 | * held. | |
1c8c601a | 207 | */ |
7c3f654d PZ |
208 | struct file_lock_list_struct { |
209 | spinlock_t lock; | |
210 | struct hlist_head hlist; | |
211 | }; | |
212 | static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list); | |
aba37660 | 213 | DEFINE_STATIC_PERCPU_RWSEM(file_rwsem); |
88974691 | 214 | |
eb82dd39 | 215 | |
1c8c601a | 216 | /* |
48f74186 | 217 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 218 | * It is protected by blocked_lock_lock. |
48f74186 JL |
219 | * |
220 | * We hash locks by lockowner in order to optimize searching for the lock a | |
221 | * particular lockowner is waiting on. | |
222 | * | |
223 | * FIXME: make this value scale via some heuristic? We generally will want more | |
224 | * buckets when we have more lockowners holding locks, but that's a little | |
225 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 226 | */ |
48f74186 JL |
227 | #define BLOCKED_HASH_BITS 7 |
228 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 229 | |
1c8c601a | 230 | /* |
7b2296af JL |
231 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
232 | * want to be holding this lock. | |
1c8c601a | 233 | * |
ada5c1da N |
234 | * In addition, it also protects the fl->fl_blocked_requests list, and the |
235 | * fl->fl_blocker pointer for file_lock structures that are acting as lock | |
236 | * requests (in contrast to those that are acting as records of acquired locks). | |
1c8c601a JL |
237 | * |
238 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 239 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 240 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 241 | * flc_lock. |
1c8c601a | 242 | */ |
7b2296af | 243 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 244 | |
4a075e39 | 245 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 246 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 247 | |
4a075e39 | 248 | static struct file_lock_context * |
5c1c669a | 249 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 250 | { |
128a3785 | 251 | struct file_lock_context *ctx; |
4a075e39 | 252 | |
128a3785 DV |
253 | /* paired with cmpxchg() below */ |
254 | ctx = smp_load_acquire(&inode->i_flctx); | |
255 | if (likely(ctx) || type == F_UNLCK) | |
4a075e39 JL |
256 | goto out; |
257 | ||
128a3785 DV |
258 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
259 | if (!ctx) | |
4a075e39 JL |
260 | goto out; |
261 | ||
128a3785 DV |
262 | spin_lock_init(&ctx->flc_lock); |
263 | INIT_LIST_HEAD(&ctx->flc_flock); | |
264 | INIT_LIST_HEAD(&ctx->flc_posix); | |
265 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
266 | |
267 | /* | |
268 | * Assign the pointer if it's not already assigned. If it is, then | |
269 | * free the context we just allocated. | |
270 | */ | |
128a3785 DV |
271 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
272 | kmem_cache_free(flctx_cache, ctx); | |
273 | ctx = smp_load_acquire(&inode->i_flctx); | |
274 | } | |
4a075e39 | 275 | out: |
1890910f | 276 | trace_locks_get_lock_context(inode, type, ctx); |
128a3785 | 277 | return ctx; |
4a075e39 JL |
278 | } |
279 | ||
e24dadab JL |
280 | static void |
281 | locks_dump_ctx_list(struct list_head *list, char *list_type) | |
282 | { | |
283 | struct file_lock *fl; | |
284 | ||
285 | list_for_each_entry(fl, list, fl_list) { | |
286 | 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); | |
287 | } | |
288 | } | |
289 | ||
290 | static void | |
291 | locks_check_ctx_lists(struct inode *inode) | |
292 | { | |
293 | struct file_lock_context *ctx = inode->i_flctx; | |
294 | ||
295 | if (unlikely(!list_empty(&ctx->flc_flock) || | |
296 | !list_empty(&ctx->flc_posix) || | |
297 | !list_empty(&ctx->flc_lease))) { | |
298 | pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n", | |
299 | MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), | |
300 | inode->i_ino); | |
301 | locks_dump_ctx_list(&ctx->flc_flock, "FLOCK"); | |
302 | locks_dump_ctx_list(&ctx->flc_posix, "POSIX"); | |
303 | locks_dump_ctx_list(&ctx->flc_lease, "LEASE"); | |
304 | } | |
305 | } | |
306 | ||
3953704f BC |
307 | static void |
308 | locks_check_ctx_file_list(struct file *filp, struct list_head *list, | |
309 | char *list_type) | |
310 | { | |
311 | struct file_lock *fl; | |
312 | struct inode *inode = locks_inode(filp); | |
313 | ||
314 | list_for_each_entry(fl, list, fl_list) | |
315 | if (fl->fl_file == filp) | |
316 | pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx " | |
317 | " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", | |
318 | list_type, MAJOR(inode->i_sb->s_dev), | |
319 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
320 | fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
321 | } | |
322 | ||
4a075e39 | 323 | void |
f27a0fe0 | 324 | locks_free_lock_context(struct inode *inode) |
4a075e39 | 325 | { |
f27a0fe0 JL |
326 | struct file_lock_context *ctx = inode->i_flctx; |
327 | ||
e24dadab JL |
328 | if (unlikely(ctx)) { |
329 | locks_check_ctx_lists(inode); | |
4a075e39 JL |
330 | kmem_cache_free(flctx_cache, ctx); |
331 | } | |
332 | } | |
333 | ||
ee19cc40 | 334 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 335 | { |
139ca04e | 336 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 337 | INIT_LIST_HEAD(&fl->fl_list); |
ada5c1da N |
338 | INIT_LIST_HEAD(&fl->fl_blocked_requests); |
339 | INIT_LIST_HEAD(&fl->fl_blocked_member); | |
ee19cc40 | 340 | init_waitqueue_head(&fl->fl_wait); |
a51cb91d MS |
341 | } |
342 | ||
1da177e4 | 343 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 344 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 345 | { |
ee19cc40 | 346 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
347 | |
348 | if (fl) | |
ee19cc40 | 349 | locks_init_lock_heads(fl); |
a51cb91d MS |
350 | |
351 | return fl; | |
1da177e4 | 352 | } |
c5b1f0d9 | 353 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 354 | |
a9e61e25 | 355 | void locks_release_private(struct file_lock *fl) |
47831f35 | 356 | { |
5926459e N |
357 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
358 | BUG_ON(!list_empty(&fl->fl_list)); | |
359 | BUG_ON(!list_empty(&fl->fl_blocked_requests)); | |
360 | BUG_ON(!list_empty(&fl->fl_blocked_member)); | |
361 | BUG_ON(!hlist_unhashed(&fl->fl_link)); | |
362 | ||
47831f35 TM |
363 | if (fl->fl_ops) { |
364 | if (fl->fl_ops->fl_release_private) | |
365 | fl->fl_ops->fl_release_private(fl); | |
366 | fl->fl_ops = NULL; | |
367 | } | |
47831f35 | 368 | |
5c97d7b1 | 369 | if (fl->fl_lmops) { |
cae80b30 JL |
370 | if (fl->fl_lmops->lm_put_owner) { |
371 | fl->fl_lmops->lm_put_owner(fl->fl_owner); | |
372 | fl->fl_owner = NULL; | |
373 | } | |
5c97d7b1 KM |
374 | fl->fl_lmops = NULL; |
375 | } | |
47831f35 | 376 | } |
a9e61e25 | 377 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 378 | |
1da177e4 | 379 | /* Free a lock which is not in use. */ |
05fa3135 | 380 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 381 | { |
47831f35 | 382 | locks_release_private(fl); |
1da177e4 LT |
383 | kmem_cache_free(filelock_cache, fl); |
384 | } | |
05fa3135 | 385 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 386 | |
ed9814d8 JL |
387 | static void |
388 | locks_dispose_list(struct list_head *dispose) | |
389 | { | |
390 | struct file_lock *fl; | |
391 | ||
392 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
393 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
394 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
395 | locks_free_lock(fl); |
396 | } | |
397 | } | |
398 | ||
1da177e4 LT |
399 | void locks_init_lock(struct file_lock *fl) |
400 | { | |
ee19cc40 MS |
401 | memset(fl, 0, sizeof(struct file_lock)); |
402 | locks_init_lock_heads(fl); | |
1da177e4 | 403 | } |
1da177e4 LT |
404 | EXPORT_SYMBOL(locks_init_lock); |
405 | ||
1da177e4 LT |
406 | /* |
407 | * Initialize a new lock from an existing file_lock structure. | |
408 | */ | |
3fe0fff1 | 409 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
410 | { |
411 | new->fl_owner = fl->fl_owner; | |
412 | new->fl_pid = fl->fl_pid; | |
0996905f | 413 | new->fl_file = NULL; |
1da177e4 LT |
414 | new->fl_flags = fl->fl_flags; |
415 | new->fl_type = fl->fl_type; | |
416 | new->fl_start = fl->fl_start; | |
417 | new->fl_end = fl->fl_end; | |
f328296e | 418 | new->fl_lmops = fl->fl_lmops; |
0996905f | 419 | new->fl_ops = NULL; |
f328296e KM |
420 | |
421 | if (fl->fl_lmops) { | |
422 | if (fl->fl_lmops->lm_get_owner) | |
cae80b30 | 423 | fl->fl_lmops->lm_get_owner(fl->fl_owner); |
f328296e | 424 | } |
0996905f | 425 | } |
3fe0fff1 | 426 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
427 | |
428 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
429 | { | |
566709bd JL |
430 | /* "new" must be a freshly-initialized lock */ |
431 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 432 | |
3fe0fff1 | 433 | locks_copy_conflock(new, fl); |
f328296e | 434 | |
0996905f | 435 | new->fl_file = fl->fl_file; |
1da177e4 | 436 | new->fl_ops = fl->fl_ops; |
47831f35 | 437 | |
f328296e KM |
438 | if (fl->fl_ops) { |
439 | if (fl->fl_ops->fl_copy_lock) | |
440 | fl->fl_ops->fl_copy_lock(new, fl); | |
441 | } | |
1da177e4 | 442 | } |
1da177e4 LT |
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); | |
bf77ae4c | 458 | list_for_each_entry(f, &new->fl_blocked_requests, fl_blocked_member) |
5946c431 N |
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); |
7bbd1fc0 | 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; | |
7bbd1fc0 | 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 | { |
1da177e4 LT |
662 | return fl1->fl_owner == fl2->fl_owner; |
663 | } | |
664 | ||
6109c850 | 665 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 666 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 667 | { |
7c3f654d PZ |
668 | struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list); |
669 | ||
aba37660 PZ |
670 | percpu_rwsem_assert_held(&file_rwsem); |
671 | ||
7c3f654d | 672 | spin_lock(&fll->lock); |
7012b02a | 673 | fl->fl_link_cpu = smp_processor_id(); |
7c3f654d PZ |
674 | hlist_add_head(&fl->fl_link, &fll->hlist); |
675 | spin_unlock(&fll->lock); | |
88974691 JL |
676 | } |
677 | ||
6109c850 | 678 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 679 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 680 | { |
7c3f654d PZ |
681 | struct file_lock_list_struct *fll; |
682 | ||
aba37660 PZ |
683 | percpu_rwsem_assert_held(&file_rwsem); |
684 | ||
7012b02a JL |
685 | /* |
686 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 687 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
688 | * also require that it be held. |
689 | */ | |
690 | if (hlist_unhashed(&fl->fl_link)) | |
691 | return; | |
7c3f654d PZ |
692 | |
693 | fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu); | |
694 | spin_lock(&fll->lock); | |
139ca04e | 695 | hlist_del_init(&fl->fl_link); |
7c3f654d | 696 | spin_unlock(&fll->lock); |
88974691 JL |
697 | } |
698 | ||
3999e493 JL |
699 | static unsigned long |
700 | posix_owner_key(struct file_lock *fl) | |
701 | { | |
3999e493 JL |
702 | return (unsigned long)fl->fl_owner; |
703 | } | |
704 | ||
6ca10ed8 | 705 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 706 | { |
663d5af7 DW |
707 | lockdep_assert_held(&blocked_lock_lock); |
708 | ||
3999e493 | 709 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
710 | } |
711 | ||
6ca10ed8 | 712 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 713 | { |
663d5af7 DW |
714 | lockdep_assert_held(&blocked_lock_lock); |
715 | ||
48f74186 | 716 | hash_del(&waiter->fl_link); |
88974691 JL |
717 | } |
718 | ||
1da177e4 LT |
719 | /* Remove waiter from blocker's block list. |
720 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 721 | * |
7b2296af | 722 | * Must be called with blocked_lock_lock held. |
1da177e4 | 723 | */ |
33443c42 | 724 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 725 | { |
88974691 | 726 | locks_delete_global_blocked(waiter); |
ada5c1da N |
727 | list_del_init(&waiter->fl_blocked_member); |
728 | waiter->fl_blocker = NULL; | |
1da177e4 LT |
729 | } |
730 | ||
ad6bbd8b N |
731 | static void __locks_wake_up_blocks(struct file_lock *blocker) |
732 | { | |
733 | while (!list_empty(&blocker->fl_blocked_requests)) { | |
734 | struct file_lock *waiter; | |
735 | ||
736 | waiter = list_first_entry(&blocker->fl_blocked_requests, | |
737 | struct file_lock, fl_blocked_member); | |
738 | __locks_delete_block(waiter); | |
739 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) | |
740 | waiter->fl_lmops->lm_notify(waiter); | |
741 | else | |
742 | wake_up(&waiter->fl_wait); | |
743 | } | |
744 | } | |
745 | ||
cb03f94f N |
746 | /** |
747 | * locks_delete_lock - stop waiting for a file lock | |
748 | * @waiter: the lock which was waiting | |
749 | * | |
750 | * lockd/nfsd need to disconnect the lock while working on it. | |
751 | */ | |
752 | int locks_delete_block(struct file_lock *waiter) | |
1da177e4 | 753 | { |
cb03f94f N |
754 | int status = -ENOENT; |
755 | ||
16306a61 N |
756 | /* |
757 | * If fl_blocker is NULL, it won't be set again as this thread | |
758 | * "owns" the lock and is the only one that might try to claim | |
759 | * the lock. So it is safe to test fl_blocker locklessly. | |
760 | * Also if fl_blocker is NULL, this waiter is not listed on | |
761 | * fl_blocked_requests for some lock, so no other request can | |
762 | * be added to the list of fl_blocked_requests for this | |
763 | * request. So if fl_blocker is NULL, it is safe to | |
764 | * locklessly check if fl_blocked_requests is empty. If both | |
765 | * of these checks succeed, there is no need to take the lock. | |
766 | */ | |
767 | if (waiter->fl_blocker == NULL && | |
768 | list_empty(&waiter->fl_blocked_requests)) | |
cb03f94f | 769 | return status; |
7b2296af | 770 | spin_lock(&blocked_lock_lock); |
cb03f94f N |
771 | if (waiter->fl_blocker) |
772 | status = 0; | |
5946c431 | 773 | __locks_wake_up_blocks(waiter); |
1da177e4 | 774 | __locks_delete_block(waiter); |
7b2296af | 775 | spin_unlock(&blocked_lock_lock); |
cb03f94f | 776 | return status; |
1da177e4 | 777 | } |
cb03f94f | 778 | EXPORT_SYMBOL(locks_delete_block); |
1da177e4 LT |
779 | |
780 | /* Insert waiter into blocker's block list. | |
781 | * We use a circular list so that processes can be easily woken up in | |
782 | * the order they blocked. The documentation doesn't require this but | |
783 | * it seems like the reasonable thing to do. | |
1c8c601a | 784 | * |
6109c850 | 785 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
ada5c1da N |
786 | * fl_blocked_requests list itself is protected by the blocked_lock_lock, |
787 | * but by ensuring that the flc_lock is also held on insertions we can avoid | |
788 | * taking the blocked_lock_lock in some cases when we see that the | |
789 | * fl_blocked_requests list is empty. | |
fd7732e0 N |
790 | * |
791 | * Rather than just adding to the list, we check for conflicts with any existing | |
792 | * waiters, and add beneath any waiter that blocks the new waiter. | |
793 | * Thus wakeups don't happen until needed. | |
1da177e4 | 794 | */ |
1c8c601a | 795 | static void __locks_insert_block(struct file_lock *blocker, |
fd7732e0 N |
796 | struct file_lock *waiter, |
797 | bool conflict(struct file_lock *, | |
798 | struct file_lock *)) | |
1da177e4 | 799 | { |
fd7732e0 | 800 | struct file_lock *fl; |
ada5c1da | 801 | BUG_ON(!list_empty(&waiter->fl_blocked_member)); |
fd7732e0 N |
802 | |
803 | new_blocker: | |
804 | list_for_each_entry(fl, &blocker->fl_blocked_requests, fl_blocked_member) | |
805 | if (conflict(fl, waiter)) { | |
806 | blocker = fl; | |
807 | goto new_blocker; | |
808 | } | |
ada5c1da N |
809 | waiter->fl_blocker = blocker; |
810 | list_add_tail(&waiter->fl_blocked_member, &blocker->fl_blocked_requests); | |
cff2fce5 | 811 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a | 812 | locks_insert_global_blocked(waiter); |
5946c431 N |
813 | |
814 | /* The requests in waiter->fl_blocked are known to conflict with | |
815 | * waiter, but might not conflict with blocker, or the requests | |
816 | * and lock which block it. So they all need to be woken. | |
817 | */ | |
818 | __locks_wake_up_blocks(waiter); | |
1c8c601a JL |
819 | } |
820 | ||
6109c850 | 821 | /* Must be called with flc_lock held. */ |
1c8c601a | 822 | static void locks_insert_block(struct file_lock *blocker, |
fd7732e0 N |
823 | struct file_lock *waiter, |
824 | bool conflict(struct file_lock *, | |
825 | struct file_lock *)) | |
1c8c601a | 826 | { |
7b2296af | 827 | spin_lock(&blocked_lock_lock); |
fd7732e0 | 828 | __locks_insert_block(blocker, waiter, conflict); |
7b2296af | 829 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
830 | } |
831 | ||
1cb36012 JL |
832 | /* |
833 | * Wake up processes blocked waiting for blocker. | |
834 | * | |
6109c850 | 835 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
836 | */ |
837 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
838 | { | |
4e8c765d JL |
839 | /* |
840 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 | 841 | * blocked requests are only added to the list under the flc_lock, and |
ada5c1da N |
842 | * the flc_lock is always held here. Note that removal from the |
843 | * fl_blocked_requests list does not require the flc_lock, so we must | |
844 | * recheck list_empty() after acquiring the blocked_lock_lock. | |
4e8c765d | 845 | */ |
ada5c1da | 846 | if (list_empty(&blocker->fl_blocked_requests)) |
4e8c765d JL |
847 | return; |
848 | ||
7b2296af | 849 | spin_lock(&blocked_lock_lock); |
ad6bbd8b | 850 | __locks_wake_up_blocks(blocker); |
7b2296af | 851 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
852 | } |
853 | ||
5263e31e | 854 | static void |
e084c1bd | 855 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e | 856 | { |
5263e31e JL |
857 | list_add_tail(&fl->fl_list, before); |
858 | locks_insert_global_locks(fl); | |
859 | } | |
860 | ||
8634b51f | 861 | static void |
e084c1bd | 862 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 863 | { |
88974691 | 864 | locks_delete_global_locks(fl); |
8634b51f | 865 | list_del_init(&fl->fl_list); |
1da177e4 | 866 | locks_wake_up_blocks(fl); |
24cbe784 JL |
867 | } |
868 | ||
8634b51f | 869 | static void |
e084c1bd | 870 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 871 | { |
e084c1bd | 872 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 873 | if (dispose) |
6dee60f6 | 874 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
875 | else |
876 | locks_free_lock(fl); | |
1da177e4 LT |
877 | } |
878 | ||
879 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
880 | * checks for shared/exclusive status of overlapping locks. | |
881 | */ | |
c0e15908 N |
882 | static bool locks_conflict(struct file_lock *caller_fl, |
883 | struct file_lock *sys_fl) | |
1da177e4 LT |
884 | { |
885 | if (sys_fl->fl_type == F_WRLCK) | |
c0e15908 | 886 | return true; |
1da177e4 | 887 | if (caller_fl->fl_type == F_WRLCK) |
c0e15908 N |
888 | return true; |
889 | return false; | |
1da177e4 LT |
890 | } |
891 | ||
892 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
893 | * checking before calling the locks_conflict(). | |
894 | */ | |
c0e15908 N |
895 | static bool posix_locks_conflict(struct file_lock *caller_fl, |
896 | struct file_lock *sys_fl) | |
1da177e4 LT |
897 | { |
898 | /* POSIX locks owned by the same process do not conflict with | |
899 | * each other. | |
900 | */ | |
9b8c8695 | 901 | if (posix_same_owner(caller_fl, sys_fl)) |
c0e15908 | 902 | return false; |
1da177e4 LT |
903 | |
904 | /* Check whether they overlap */ | |
905 | if (!locks_overlap(caller_fl, sys_fl)) | |
c0e15908 | 906 | return false; |
1da177e4 | 907 | |
c0e15908 | 908 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
909 | } |
910 | ||
911 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
912 | * checking before calling the locks_conflict(). | |
913 | */ | |
c0e15908 N |
914 | static bool flock_locks_conflict(struct file_lock *caller_fl, |
915 | struct file_lock *sys_fl) | |
1da177e4 LT |
916 | { |
917 | /* FLOCK locks referring to the same filp do not conflict with | |
918 | * each other. | |
919 | */ | |
9b8c8695 | 920 | if (caller_fl->fl_file == sys_fl->fl_file) |
c0e15908 | 921 | return false; |
1da177e4 | 922 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) |
c0e15908 | 923 | return false; |
1da177e4 | 924 | |
c0e15908 | 925 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
926 | } |
927 | ||
6d34ac19 | 928 | void |
9d6a8c5c | 929 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
930 | { |
931 | struct file_lock *cfl; | |
bd61e0a9 | 932 | struct file_lock_context *ctx; |
c568d683 | 933 | struct inode *inode = locks_inode(filp); |
1da177e4 | 934 | |
128a3785 | 935 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
936 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
937 | fl->fl_type = F_UNLCK; | |
938 | return; | |
939 | } | |
940 | ||
6109c850 | 941 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
942 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
943 | if (posix_locks_conflict(fl, cfl)) { | |
944 | locks_copy_conflock(fl, cfl); | |
bd61e0a9 JL |
945 | goto out; |
946 | } | |
1da177e4 | 947 | } |
bd61e0a9 JL |
948 | fl->fl_type = F_UNLCK; |
949 | out: | |
6109c850 | 950 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 951 | return; |
1da177e4 | 952 | } |
1da177e4 LT |
953 | EXPORT_SYMBOL(posix_test_lock); |
954 | ||
b533184f BF |
955 | /* |
956 | * Deadlock detection: | |
957 | * | |
958 | * We attempt to detect deadlocks that are due purely to posix file | |
959 | * locks. | |
1da177e4 | 960 | * |
b533184f BF |
961 | * We assume that a task can be waiting for at most one lock at a time. |
962 | * So for any acquired lock, the process holding that lock may be | |
963 | * waiting on at most one other lock. That lock in turns may be held by | |
964 | * someone waiting for at most one other lock. Given a requested lock | |
965 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
966 | * follow this chain of waiters to ensure we are not about to create a | |
967 | * cycle. | |
1da177e4 | 968 | * |
b533184f BF |
969 | * Since we do this before we ever put a process to sleep on a lock, we |
970 | * are ensured that there is never a cycle; that is what guarantees that | |
971 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 972 | * |
b533184f BF |
973 | * Note: the above assumption may not be true when handling lock |
974 | * requests from a broken NFS client. It may also fail in the presence | |
975 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 976 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 977 | * |
cff2fce5 | 978 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
979 | * Because the owner is not even nominally tied to a thread of |
980 | * execution, the deadlock detection below can't reasonably work well. Just | |
981 | * skip it for those. | |
982 | * | |
cff2fce5 | 983 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
984 | * locks that just checks for the case where two tasks are attempting to |
985 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 986 | */ |
97855b49 BF |
987 | |
988 | #define MAX_DEADLK_ITERATIONS 10 | |
989 | ||
b533184f BF |
990 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
991 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
992 | { | |
993 | struct file_lock *fl; | |
994 | ||
3999e493 | 995 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
5946c431 N |
996 | if (posix_same_owner(fl, block_fl)) { |
997 | while (fl->fl_blocker) | |
998 | fl = fl->fl_blocker; | |
999 | return fl; | |
1000 | } | |
b533184f BF |
1001 | } |
1002 | return NULL; | |
1003 | } | |
1004 | ||
7b2296af | 1005 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 1006 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
1007 | struct file_lock *block_fl) |
1008 | { | |
97855b49 | 1009 | int i = 0; |
1da177e4 | 1010 | |
663d5af7 DW |
1011 | lockdep_assert_held(&blocked_lock_lock); |
1012 | ||
57b65325 JL |
1013 | /* |
1014 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 1015 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 1016 | */ |
cff2fce5 | 1017 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
1018 | return 0; |
1019 | ||
b533184f BF |
1020 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
1021 | if (i++ > MAX_DEADLK_ITERATIONS) | |
1022 | return 0; | |
1023 | if (posix_same_owner(caller_fl, block_fl)) | |
1024 | return 1; | |
1da177e4 LT |
1025 | } |
1026 | return 0; | |
1027 | } | |
1028 | ||
1da177e4 | 1029 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 1030 | * after any leases, but before any posix locks. |
f475ae95 TM |
1031 | * |
1032 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1033 | * whether or not a lock was successfully freed by testing the return | |
1034 | * value for -ENOENT. | |
1da177e4 | 1035 | */ |
bcd7f78d | 1036 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 1037 | { |
993dfa87 | 1038 | struct file_lock *new_fl = NULL; |
5263e31e JL |
1039 | struct file_lock *fl; |
1040 | struct file_lock_context *ctx; | |
1da177e4 | 1041 | int error = 0; |
5263e31e | 1042 | bool found = false; |
ed9814d8 | 1043 | LIST_HEAD(dispose); |
1da177e4 | 1044 | |
5c1c669a JL |
1045 | ctx = locks_get_lock_context(inode, request->fl_type); |
1046 | if (!ctx) { | |
1047 | if (request->fl_type != F_UNLCK) | |
1048 | return -ENOMEM; | |
1049 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
1050 | } | |
5263e31e | 1051 | |
b89f4321 | 1052 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 1053 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
1054 | if (!new_fl) |
1055 | return -ENOMEM; | |
84d535ad PE |
1056 | } |
1057 | ||
02e525b2 | 1058 | percpu_down_read(&file_rwsem); |
6109c850 | 1059 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
1060 | if (request->fl_flags & FL_ACCESS) |
1061 | goto find_conflict; | |
1062 | ||
5263e31e | 1063 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 1064 | if (request->fl_file != fl->fl_file) |
1da177e4 | 1065 | continue; |
993dfa87 | 1066 | if (request->fl_type == fl->fl_type) |
1da177e4 | 1067 | goto out; |
5263e31e | 1068 | found = true; |
e084c1bd | 1069 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1070 | break; |
1071 | } | |
1da177e4 | 1072 | |
f475ae95 TM |
1073 | if (request->fl_type == F_UNLCK) { |
1074 | if ((request->fl_flags & FL_EXISTS) && !found) | |
1075 | error = -ENOENT; | |
993dfa87 | 1076 | goto out; |
f475ae95 | 1077 | } |
1da177e4 | 1078 | |
f07f18dd | 1079 | find_conflict: |
5263e31e | 1080 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 1081 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
1082 | continue; |
1083 | error = -EAGAIN; | |
bde74e4b MS |
1084 | if (!(request->fl_flags & FL_SLEEP)) |
1085 | goto out; | |
1086 | error = FILE_LOCK_DEFERRED; | |
fd7732e0 | 1087 | locks_insert_block(fl, request, flock_locks_conflict); |
1da177e4 LT |
1088 | goto out; |
1089 | } | |
f07f18dd TM |
1090 | if (request->fl_flags & FL_ACCESS) |
1091 | goto out; | |
993dfa87 | 1092 | locks_copy_lock(new_fl, request); |
5946c431 | 1093 | locks_move_blocks(new_fl, request); |
e084c1bd | 1094 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 1095 | new_fl = NULL; |
9cedc194 | 1096 | error = 0; |
1da177e4 LT |
1097 | |
1098 | out: | |
6109c850 | 1099 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1100 | percpu_up_read(&file_rwsem); |
993dfa87 TM |
1101 | if (new_fl) |
1102 | locks_free_lock(new_fl); | |
ed9814d8 | 1103 | locks_dispose_list(&dispose); |
c883da31 | 1104 | trace_flock_lock_inode(inode, request, error); |
1da177e4 LT |
1105 | return error; |
1106 | } | |
1107 | ||
b4d629a3 JL |
1108 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
1109 | struct file_lock *conflock) | |
1da177e4 | 1110 | { |
bd61e0a9 | 1111 | struct file_lock *fl, *tmp; |
39005d02 MS |
1112 | struct file_lock *new_fl = NULL; |
1113 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
1114 | struct file_lock *left = NULL; |
1115 | struct file_lock *right = NULL; | |
bd61e0a9 | 1116 | struct file_lock_context *ctx; |
b9746ef8 JL |
1117 | int error; |
1118 | bool added = false; | |
ed9814d8 | 1119 | LIST_HEAD(dispose); |
1da177e4 | 1120 | |
5c1c669a | 1121 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 1122 | if (!ctx) |
5c1c669a | 1123 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 1124 | |
1da177e4 LT |
1125 | /* |
1126 | * We may need two file_lock structures for this operation, | |
1127 | * so we get them in advance to avoid races. | |
39005d02 MS |
1128 | * |
1129 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1130 | */ |
39005d02 MS |
1131 | if (!(request->fl_flags & FL_ACCESS) && |
1132 | (request->fl_type != F_UNLCK || | |
1133 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1134 | new_fl = locks_alloc_lock(); | |
1135 | new_fl2 = locks_alloc_lock(); | |
1136 | } | |
1da177e4 | 1137 | |
02e525b2 | 1138 | percpu_down_read(&file_rwsem); |
6109c850 | 1139 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1140 | /* |
1141 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1142 | * there are any, either return error or put the request on the | |
48f74186 | 1143 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1144 | */ |
1da177e4 | 1145 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1146 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1147 | if (!posix_locks_conflict(request, fl)) |
1148 | continue; | |
5842add2 | 1149 | if (conflock) |
3fe0fff1 | 1150 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1151 | error = -EAGAIN; |
1152 | if (!(request->fl_flags & FL_SLEEP)) | |
1153 | goto out; | |
1c8c601a JL |
1154 | /* |
1155 | * Deadlock detection and insertion into the blocked | |
1156 | * locks list must be done while holding the same lock! | |
1157 | */ | |
1da177e4 | 1158 | error = -EDEADLK; |
7b2296af | 1159 | spin_lock(&blocked_lock_lock); |
945ab8f6 JL |
1160 | /* |
1161 | * Ensure that we don't find any locks blocked on this | |
1162 | * request during deadlock detection. | |
1163 | */ | |
1164 | __locks_wake_up_blocks(request); | |
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 | 1171 | goto out; |
7bbd1fc0 N |
1172 | } |
1173 | } | |
1da177e4 LT |
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); |
02e525b2 | 1317 | percpu_up_read(&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 | } | |
1da177e4 | 1468 | EXPORT_SYMBOL(locks_mandatory_area); |
9e8925b6 | 1469 | #endif /* CONFIG_MANDATORY_FILE_LOCKING */ |
1da177e4 | 1470 | |
778fc546 BF |
1471 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1472 | { | |
1473 | switch (arg) { | |
1474 | case F_UNLCK: | |
1475 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
0a4c9265 | 1476 | /* fall through */ |
778fc546 BF |
1477 | case F_RDLCK: |
1478 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1479 | } | |
1480 | } | |
1481 | ||
1da177e4 | 1482 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1483 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1484 | { |
1da177e4 LT |
1485 | int error = assign_type(fl, arg); |
1486 | ||
1487 | if (error) | |
1488 | return error; | |
778fc546 | 1489 | lease_clear_pending(fl, arg); |
1da177e4 | 1490 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1491 | if (arg == F_UNLCK) { |
1492 | struct file *filp = fl->fl_file; | |
1493 | ||
1494 | f_delown(filp); | |
1495 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1496 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1497 | if (fl->fl_fasync != NULL) { | |
1498 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1499 | fl->fl_fasync = NULL; | |
1500 | } | |
e084c1bd | 1501 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1502 | } |
1da177e4 LT |
1503 | return 0; |
1504 | } | |
1da177e4 LT |
1505 | EXPORT_SYMBOL(lease_modify); |
1506 | ||
778fc546 BF |
1507 | static bool past_time(unsigned long then) |
1508 | { | |
1509 | if (!then) | |
1510 | /* 0 is a special value meaning "this never expires": */ | |
1511 | return false; | |
1512 | return time_after(jiffies, then); | |
1513 | } | |
1514 | ||
c45198ed | 1515 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1516 | { |
8634b51f JL |
1517 | struct file_lock_context *ctx = inode->i_flctx; |
1518 | struct file_lock *fl, *tmp; | |
1da177e4 | 1519 | |
6109c850 | 1520 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1521 | |
8634b51f | 1522 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1523 | trace_time_out_leases(inode, fl); |
778fc546 | 1524 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1525 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1526 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1527 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1528 | } |
1529 | } | |
1530 | ||
df4e8d2c BF |
1531 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1532 | { | |
d51f527f IW |
1533 | bool rc; |
1534 | ||
1535 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) { | |
1536 | rc = false; | |
1537 | goto trace; | |
1538 | } | |
1539 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) { | |
1540 | rc = false; | |
1541 | goto trace; | |
1542 | } | |
1543 | ||
1544 | rc = locks_conflict(breaker, lease); | |
1545 | trace: | |
1546 | trace_leases_conflict(rc, lease, breaker); | |
1547 | return rc; | |
df4e8d2c BF |
1548 | } |
1549 | ||
03d12ddf JL |
1550 | static bool |
1551 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1552 | { | |
8634b51f | 1553 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1554 | struct file_lock *fl; |
1555 | ||
6109c850 | 1556 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1557 | |
8634b51f | 1558 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1559 | if (leases_conflict(fl, breaker)) |
1560 | return true; | |
1561 | } | |
1562 | return false; | |
1563 | } | |
1564 | ||
1da177e4 LT |
1565 | /** |
1566 | * __break_lease - revoke all outstanding leases on file | |
1567 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1568 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1569 | * break all leases | |
1570 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1571 | * only delegations | |
1da177e4 | 1572 | * |
87250dd2 | 1573 | * break_lease (inlined for speed) has checked there already is at least |
1574 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1575 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1576 | * specified %O_NONBLOCK to your open(). |
1577 | */ | |
df4e8d2c | 1578 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1579 | { |
778fc546 | 1580 | int error = 0; |
128a3785 | 1581 | struct file_lock_context *ctx; |
a901125c | 1582 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1583 | unsigned long break_time; |
8737c930 | 1584 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1585 | LIST_HEAD(dispose); |
1da177e4 | 1586 | |
8737c930 | 1587 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1588 | if (IS_ERR(new_fl)) |
1589 | return PTR_ERR(new_fl); | |
df4e8d2c | 1590 | new_fl->fl_flags = type; |
1da177e4 | 1591 | |
8634b51f | 1592 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1593 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1594 | if (!ctx) { |
1595 | WARN_ON_ONCE(1); | |
cfddf9f4 | 1596 | goto free_lock; |
8634b51f JL |
1597 | } |
1598 | ||
02e525b2 | 1599 | percpu_down_read(&file_rwsem); |
6109c850 | 1600 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1601 | |
c45198ed | 1602 | time_out_leases(inode, &dispose); |
1da177e4 | 1603 | |
03d12ddf | 1604 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1605 | goto out; |
1606 | ||
1da177e4 LT |
1607 | break_time = 0; |
1608 | if (lease_break_time > 0) { | |
1609 | break_time = jiffies + lease_break_time * HZ; | |
1610 | if (break_time == 0) | |
1611 | break_time++; /* so that 0 means no break time */ | |
1612 | } | |
1613 | ||
a901125c | 1614 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1615 | if (!leases_conflict(fl, new_fl)) |
1616 | continue; | |
778fc546 BF |
1617 | if (want_write) { |
1618 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1619 | continue; | |
1620 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1621 | fl->fl_break_time = break_time; |
778fc546 | 1622 | } else { |
8634b51f | 1623 | if (lease_breaking(fl)) |
778fc546 BF |
1624 | continue; |
1625 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1626 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1627 | } |
4d01b7f5 | 1628 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1629 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1630 | } |
1631 | ||
8634b51f | 1632 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1633 | goto out; |
1634 | ||
843c6b2f | 1635 | if (mode & O_NONBLOCK) { |
62af4f1f | 1636 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1637 | error = -EWOULDBLOCK; |
1638 | goto out; | |
1639 | } | |
1640 | ||
1641 | restart: | |
8634b51f JL |
1642 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1643 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1644 | if (break_time != 0) |
1da177e4 | 1645 | break_time -= jiffies; |
f1c6bb2c JL |
1646 | if (break_time == 0) |
1647 | break_time++; | |
fd7732e0 | 1648 | locks_insert_block(fl, new_fl, leases_conflict); |
62af4f1f | 1649 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1650 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1651 | percpu_up_read(&file_rwsem); |
aba37660 | 1652 | |
c45198ed | 1653 | locks_dispose_list(&dispose); |
4321e01e | 1654 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
ada5c1da | 1655 | !new_fl->fl_blocker, break_time); |
aba37660 | 1656 | |
02e525b2 | 1657 | percpu_down_read(&file_rwsem); |
6109c850 | 1658 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1659 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1660 | locks_delete_block(new_fl); |
1da177e4 | 1661 | if (error >= 0) { |
778fc546 BF |
1662 | /* |
1663 | * Wait for the next conflicting lease that has not been | |
1664 | * broken yet | |
1665 | */ | |
03d12ddf JL |
1666 | if (error == 0) |
1667 | time_out_leases(inode, &dispose); | |
1668 | if (any_leases_conflict(inode, new_fl)) | |
1669 | goto restart; | |
1da177e4 LT |
1670 | error = 0; |
1671 | } | |
1da177e4 | 1672 | out: |
6109c850 | 1673 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1674 | percpu_up_read(&file_rwsem); |
c45198ed | 1675 | locks_dispose_list(&dispose); |
cfddf9f4 | 1676 | free_lock: |
6d4b9e38 | 1677 | locks_free_lock(new_fl); |
1da177e4 LT |
1678 | return error; |
1679 | } | |
1da177e4 LT |
1680 | EXPORT_SYMBOL(__break_lease); |
1681 | ||
1682 | /** | |
76c47948 | 1683 | * lease_get_mtime - update modified time of an inode with exclusive lease |
1da177e4 | 1684 | * @inode: the inode |
76c47948 | 1685 | * @time: pointer to a timespec which contains the last modified time |
1da177e4 LT |
1686 | * |
1687 | * This is to force NFS clients to flush their caches for files with | |
1688 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1689 | * exclusive lease, then they could be modifying it. |
1da177e4 | 1690 | */ |
95582b00 | 1691 | void lease_get_mtime(struct inode *inode, struct timespec64 *time) |
1da177e4 | 1692 | { |
bfe86024 | 1693 | bool has_lease = false; |
128a3785 | 1694 | struct file_lock_context *ctx; |
8634b51f | 1695 | struct file_lock *fl; |
bfe86024 | 1696 | |
128a3785 | 1697 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1698 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1699 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1700 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1701 | struct file_lock, fl_list); | |
1702 | if (fl && (fl->fl_type == F_WRLCK)) | |
1703 | has_lease = true; | |
6109c850 | 1704 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1705 | } |
1706 | ||
1707 | if (has_lease) | |
c2050a45 | 1708 | *time = current_time(inode); |
1da177e4 | 1709 | } |
1da177e4 LT |
1710 | EXPORT_SYMBOL(lease_get_mtime); |
1711 | ||
1712 | /** | |
1713 | * fcntl_getlease - Enquire what lease is currently active | |
1714 | * @filp: the file | |
1715 | * | |
1716 | * The value returned by this function will be one of | |
1717 | * (if no lease break is pending): | |
1718 | * | |
1719 | * %F_RDLCK to indicate a shared lease is held. | |
1720 | * | |
1721 | * %F_WRLCK to indicate an exclusive lease is held. | |
1722 | * | |
1723 | * %F_UNLCK to indicate no lease is held. | |
1724 | * | |
1725 | * (if a lease break is pending): | |
1726 | * | |
1727 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1728 | * changed to a shared lease (or removed). | |
1729 | * | |
1730 | * %F_UNLCK to indicate the lease needs to be removed. | |
1731 | * | |
1732 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1733 | * should be returned to userspace. | |
1734 | */ | |
1735 | int fcntl_getlease(struct file *filp) | |
1736 | { | |
1737 | struct file_lock *fl; | |
c568d683 | 1738 | struct inode *inode = locks_inode(filp); |
128a3785 | 1739 | struct file_lock_context *ctx; |
1da177e4 | 1740 | int type = F_UNLCK; |
c45198ed | 1741 | LIST_HEAD(dispose); |
1da177e4 | 1742 | |
128a3785 | 1743 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1744 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
02e525b2 | 1745 | percpu_down_read(&file_rwsem); |
6109c850 | 1746 | spin_lock(&ctx->flc_lock); |
c568d683 | 1747 | time_out_leases(inode, &dispose); |
8634b51f JL |
1748 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
1749 | if (fl->fl_file != filp) | |
1750 | continue; | |
778fc546 | 1751 | type = target_leasetype(fl); |
1da177e4 LT |
1752 | break; |
1753 | } | |
6109c850 | 1754 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1755 | percpu_up_read(&file_rwsem); |
5f43086b | 1756 | |
8634b51f | 1757 | locks_dispose_list(&dispose); |
1da177e4 | 1758 | } |
1da177e4 LT |
1759 | return type; |
1760 | } | |
1761 | ||
24cbe784 | 1762 | /** |
387e3746 | 1763 | * check_conflicting_open - see if the given file points to an inode that has |
7bbd1fc0 N |
1764 | * an existing open that would conflict with the |
1765 | * desired lease. | |
387e3746 | 1766 | * @filp: file to check |
24cbe784 | 1767 | * @arg: type of lease that we're trying to acquire |
7fadc59c | 1768 | * @flags: current lock flags |
24cbe784 JL |
1769 | * |
1770 | * Check to see if there's an existing open fd on this file that would | |
1771 | * conflict with the lease we're trying to set. | |
1772 | */ | |
1773 | static int | |
387e3746 | 1774 | check_conflicting_open(struct file *filp, const long arg, int flags) |
24cbe784 | 1775 | { |
387e3746 AG |
1776 | struct inode *inode = locks_inode(filp); |
1777 | int self_wcount = 0, self_rcount = 0; | |
24cbe784 | 1778 | |
11afe9f7 CH |
1779 | if (flags & FL_LAYOUT) |
1780 | return 0; | |
1781 | ||
387e3746 AG |
1782 | if (arg == F_RDLCK) |
1783 | return inode_is_open_for_write(inode) ? -EAGAIN : 0; | |
1784 | else if (arg != F_WRLCK) | |
1785 | return 0; | |
1786 | ||
1787 | /* | |
1788 | * Make sure that only read/write count is from lease requestor. | |
1789 | * Note that this will result in denying write leases when i_writecount | |
1790 | * is negative, which is what we want. (We shouldn't grant write leases | |
1791 | * on files open for execution.) | |
1792 | */ | |
1793 | if (filp->f_mode & FMODE_WRITE) | |
1794 | self_wcount = 1; | |
1795 | else if (filp->f_mode & FMODE_READ) | |
1796 | self_rcount = 1; | |
24cbe784 | 1797 | |
387e3746 AG |
1798 | if (atomic_read(&inode->i_writecount) != self_wcount || |
1799 | atomic_read(&inode->i_readcount) != self_rcount) | |
1800 | return -EAGAIN; | |
24cbe784 | 1801 | |
387e3746 | 1802 | return 0; |
24cbe784 JL |
1803 | } |
1804 | ||
e6f5c789 JL |
1805 | static int |
1806 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1807 | { |
8634b51f | 1808 | struct file_lock *fl, *my_fl = NULL, *lease; |
387e3746 | 1809 | struct inode *inode = locks_inode(filp); |
8634b51f | 1810 | struct file_lock_context *ctx; |
df4e8d2c | 1811 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1812 | int error; |
c45198ed | 1813 | LIST_HEAD(dispose); |
1da177e4 | 1814 | |
096657b6 | 1815 | lease = *flp; |
62af4f1f JL |
1816 | trace_generic_add_lease(inode, lease); |
1817 | ||
5c1c669a JL |
1818 | /* Note that arg is never F_UNLCK here */ |
1819 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1820 | if (!ctx) |
1821 | return -ENOMEM; | |
1822 | ||
df4e8d2c BF |
1823 | /* |
1824 | * In the delegation case we need mutual exclusion with | |
1825 | * a number of operations that take the i_mutex. We trylock | |
1826 | * because delegations are an optional optimization, and if | |
1827 | * there's some chance of a conflict--we'd rather not | |
1828 | * bother, maybe that's a sign this just isn't a good file to | |
1829 | * hand out a delegation on. | |
1830 | */ | |
5955102c | 1831 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1832 | return -EAGAIN; |
1833 | ||
1834 | if (is_deleg && arg == F_WRLCK) { | |
1835 | /* Write delegations are not currently supported: */ | |
5955102c | 1836 | inode_unlock(inode); |
df4e8d2c BF |
1837 | WARN_ON_ONCE(1); |
1838 | return -EINVAL; | |
1839 | } | |
096657b6 | 1840 | |
02e525b2 | 1841 | percpu_down_read(&file_rwsem); |
6109c850 | 1842 | spin_lock(&ctx->flc_lock); |
c45198ed | 1843 | time_out_leases(inode, &dispose); |
387e3746 | 1844 | error = check_conflicting_open(filp, arg, lease->fl_flags); |
24cbe784 | 1845 | if (error) |
096657b6 | 1846 | goto out; |
6d5e8b05 | 1847 | |
1da177e4 LT |
1848 | /* |
1849 | * At this point, we know that if there is an exclusive | |
1850 | * lease on this file, then we hold it on this filp | |
1851 | * (otherwise our open of this file would have blocked). | |
1852 | * And if we are trying to acquire an exclusive lease, | |
1853 | * then the file is not open by anyone (including us) | |
1854 | * except for this filp. | |
1855 | */ | |
c1f24ef4 | 1856 | error = -EAGAIN; |
8634b51f | 1857 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1858 | if (fl->fl_file == filp && |
1859 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1860 | my_fl = fl; |
c1f24ef4 BF |
1861 | continue; |
1862 | } | |
8634b51f | 1863 | |
c1f24ef4 BF |
1864 | /* |
1865 | * No exclusive leases if someone else has a lease on | |
1866 | * this file: | |
1867 | */ | |
1868 | if (arg == F_WRLCK) | |
1869 | goto out; | |
1870 | /* | |
1871 | * Modifying our existing lease is OK, but no getting a | |
1872 | * new lease if someone else is opening for write: | |
1873 | */ | |
1874 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1875 | goto out; | |
1da177e4 LT |
1876 | } |
1877 | ||
8634b51f | 1878 | if (my_fl != NULL) { |
0164bf02 JL |
1879 | lease = my_fl; |
1880 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1881 | if (error) |
1882 | goto out; | |
1883 | goto out_setup; | |
1da177e4 LT |
1884 | } |
1885 | ||
1da177e4 LT |
1886 | error = -EINVAL; |
1887 | if (!leases_enable) | |
1888 | goto out; | |
1889 | ||
e084c1bd | 1890 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1891 | /* |
1892 | * The check in break_lease() is lockless. It's possible for another | |
1893 | * open to race in after we did the earlier check for a conflicting | |
1894 | * open but before the lease was inserted. Check again for a | |
1895 | * conflicting open and cancel the lease if there is one. | |
1896 | * | |
1897 | * We also add a barrier here to ensure that the insertion of the lock | |
1898 | * precedes these checks. | |
1899 | */ | |
1900 | smp_mb(); | |
387e3746 | 1901 | error = check_conflicting_open(filp, arg, lease->fl_flags); |
8634b51f | 1902 | if (error) { |
e084c1bd | 1903 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1904 | goto out; |
1905 | } | |
1c7dd2ff JL |
1906 | |
1907 | out_setup: | |
1908 | if (lease->fl_lmops->lm_setup) | |
1909 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1910 | out: |
6109c850 | 1911 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1912 | percpu_up_read(&file_rwsem); |
c45198ed | 1913 | locks_dispose_list(&dispose); |
df4e8d2c | 1914 | if (is_deleg) |
5955102c | 1915 | inode_unlock(inode); |
8634b51f | 1916 | if (!error && !my_fl) |
1c7dd2ff | 1917 | *flp = NULL; |
1da177e4 LT |
1918 | return error; |
1919 | } | |
8335ebd9 | 1920 | |
2ab99ee1 | 1921 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1922 | { |
0efaa7e8 | 1923 | int error = -EAGAIN; |
8634b51f | 1924 | struct file_lock *fl, *victim = NULL; |
c568d683 | 1925 | struct inode *inode = locks_inode(filp); |
128a3785 | 1926 | struct file_lock_context *ctx; |
c45198ed | 1927 | LIST_HEAD(dispose); |
8335ebd9 | 1928 | |
128a3785 | 1929 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1930 | if (!ctx) { |
1931 | trace_generic_delete_lease(inode, NULL); | |
1932 | return error; | |
1933 | } | |
1934 | ||
02e525b2 | 1935 | percpu_down_read(&file_rwsem); |
6109c850 | 1936 | spin_lock(&ctx->flc_lock); |
8634b51f | 1937 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1938 | if (fl->fl_file == filp && |
1939 | fl->fl_owner == owner) { | |
8634b51f | 1940 | victim = fl; |
0efaa7e8 | 1941 | break; |
8634b51f | 1942 | } |
8335ebd9 | 1943 | } |
a9b1b455 | 1944 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1945 | if (victim) |
7448cc37 | 1946 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1947 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1948 | percpu_up_read(&file_rwsem); |
c45198ed | 1949 | locks_dispose_list(&dispose); |
0efaa7e8 | 1950 | return error; |
8335ebd9 BF |
1951 | } |
1952 | ||
1953 | /** | |
1954 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1955 | * @filp: file pointer |
1956 | * @arg: type of lease to obtain | |
1957 | * @flp: input - file_lock to use, output - file_lock inserted | |
1958 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1959 | * doesn't require it) | |
8335ebd9 BF |
1960 | * |
1961 | * The (input) flp->fl_lmops->lm_break function is required | |
1962 | * by break_lease(). | |
8335ebd9 | 1963 | */ |
e6f5c789 JL |
1964 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1965 | void **priv) | |
8335ebd9 | 1966 | { |
c568d683 | 1967 | struct inode *inode = locks_inode(filp); |
8335ebd9 BF |
1968 | int error; |
1969 | ||
8e96e3b7 | 1970 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1971 | return -EACCES; |
1972 | if (!S_ISREG(inode->i_mode)) | |
1973 | return -EINVAL; | |
1974 | error = security_file_lock(filp, arg); | |
1975 | if (error) | |
1976 | return error; | |
1977 | ||
8335ebd9 BF |
1978 | switch (arg) { |
1979 | case F_UNLCK: | |
2ab99ee1 | 1980 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1981 | case F_RDLCK: |
1982 | case F_WRLCK: | |
0efaa7e8 JL |
1983 | if (!(*flp)->fl_lmops->lm_break) { |
1984 | WARN_ON_ONCE(1); | |
1985 | return -ENOLCK; | |
1986 | } | |
11afe9f7 | 1987 | |
e6f5c789 | 1988 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1989 | default: |
8d657eb3 | 1990 | return -EINVAL; |
8335ebd9 BF |
1991 | } |
1992 | } | |
0af1a450 | 1993 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1994 | |
18f6622e JL |
1995 | #if IS_ENABLED(CONFIG_SRCU) |
1996 | /* | |
1997 | * Kernel subsystems can register to be notified on any attempt to set | |
1998 | * a new lease with the lease_notifier_chain. This is used by (e.g.) nfsd | |
1999 | * to close files that it may have cached when there is an attempt to set a | |
2000 | * conflicting lease. | |
2001 | */ | |
2002 | static struct srcu_notifier_head lease_notifier_chain; | |
2003 | ||
2004 | static inline void | |
2005 | lease_notifier_chain_init(void) | |
2006 | { | |
2007 | srcu_init_notifier_head(&lease_notifier_chain); | |
2008 | } | |
2009 | ||
2010 | static inline void | |
2011 | setlease_notifier(long arg, struct file_lock *lease) | |
2012 | { | |
2013 | if (arg != F_UNLCK) | |
2014 | srcu_notifier_call_chain(&lease_notifier_chain, arg, lease); | |
2015 | } | |
2016 | ||
2017 | int lease_register_notifier(struct notifier_block *nb) | |
2018 | { | |
2019 | return srcu_notifier_chain_register(&lease_notifier_chain, nb); | |
2020 | } | |
2021 | EXPORT_SYMBOL_GPL(lease_register_notifier); | |
2022 | ||
2023 | void lease_unregister_notifier(struct notifier_block *nb) | |
2024 | { | |
2025 | srcu_notifier_chain_unregister(&lease_notifier_chain, nb); | |
2026 | } | |
2027 | EXPORT_SYMBOL_GPL(lease_unregister_notifier); | |
2028 | ||
2029 | #else /* !IS_ENABLED(CONFIG_SRCU) */ | |
2030 | static inline void | |
2031 | lease_notifier_chain_init(void) | |
2032 | { | |
2033 | } | |
2034 | ||
2035 | static inline void | |
2036 | setlease_notifier(long arg, struct file_lock *lease) | |
2037 | { | |
2038 | } | |
2039 | ||
2040 | int lease_register_notifier(struct notifier_block *nb) | |
2041 | { | |
2042 | return 0; | |
2043 | } | |
2044 | EXPORT_SYMBOL_GPL(lease_register_notifier); | |
2045 | ||
2046 | void lease_unregister_notifier(struct notifier_block *nb) | |
2047 | { | |
2048 | } | |
2049 | EXPORT_SYMBOL_GPL(lease_unregister_notifier); | |
2050 | ||
2051 | #endif /* IS_ENABLED(CONFIG_SRCU) */ | |
2052 | ||
b89f4321 | 2053 | /** |
e51673aa | 2054 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
2055 | * @filp: file pointer |
2056 | * @arg: type of lease to obtain | |
2057 | * @lease: file_lock to use when adding a lease | |
2058 | * @priv: private info for lm_setup when adding a lease (may be | |
7bbd1fc0 | 2059 | * NULL if lm_setup doesn't require it) |
e51673aa JL |
2060 | * |
2061 | * Call this to establish a lease on the file. The "lease" argument is not | |
2062 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
80b79dd0 MCC |
2063 | * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be |
2064 | * set; if not, this function will return -ENOLCK (and generate a scary-looking | |
e51673aa | 2065 | * stack trace). |
1c7dd2ff JL |
2066 | * |
2067 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
2068 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 2069 | */ |
e6f5c789 JL |
2070 | int |
2071 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 2072 | { |
18f6622e JL |
2073 | if (lease) |
2074 | setlease_notifier(arg, *lease); | |
de2a4a50 | 2075 | if (filp->f_op->setlease) |
f82b4b67 | 2076 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 2077 | else |
f82b4b67 | 2078 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 2079 | } |
a9933cea | 2080 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 2081 | |
0ceaf6c7 | 2082 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 2083 | { |
1c7dd2ff | 2084 | struct file_lock *fl; |
f7347ce4 | 2085 | struct fasync_struct *new; |
1da177e4 LT |
2086 | int error; |
2087 | ||
c5b1f0d9 AB |
2088 | fl = lease_alloc(filp, arg); |
2089 | if (IS_ERR(fl)) | |
2090 | return PTR_ERR(fl); | |
1da177e4 | 2091 | |
f7347ce4 LT |
2092 | new = fasync_alloc(); |
2093 | if (!new) { | |
2094 | locks_free_lock(fl); | |
2095 | return -ENOMEM; | |
2096 | } | |
1c7dd2ff | 2097 | new->fa_fd = fd; |
f7347ce4 | 2098 | |
1c7dd2ff | 2099 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
2100 | if (fl) |
2101 | locks_free_lock(fl); | |
f7347ce4 LT |
2102 | if (new) |
2103 | fasync_free(new); | |
1da177e4 LT |
2104 | return error; |
2105 | } | |
2106 | ||
0ceaf6c7 BF |
2107 | /** |
2108 | * fcntl_setlease - sets a lease on an open file | |
2109 | * @fd: open file descriptor | |
2110 | * @filp: file pointer | |
2111 | * @arg: type of lease to obtain | |
2112 | * | |
2113 | * Call this fcntl to establish a lease on the file. | |
2114 | * Note that you also need to call %F_SETSIG to | |
2115 | * receive a signal when the lease is broken. | |
2116 | */ | |
2117 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
2118 | { | |
2119 | if (arg == F_UNLCK) | |
2ab99ee1 | 2120 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
2121 | return do_fcntl_add_lease(fd, filp, arg); |
2122 | } | |
2123 | ||
1da177e4 | 2124 | /** |
29d01b22 JL |
2125 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
2126 | * @inode: inode of the file to apply to | |
1da177e4 LT |
2127 | * @fl: The lock to be applied |
2128 | * | |
29d01b22 | 2129 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 2130 | */ |
616fb38f | 2131 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
2132 | { |
2133 | int error; | |
2134 | might_sleep(); | |
2135 | for (;;) { | |
29d01b22 | 2136 | error = flock_lock_inode(inode, fl); |
bde74e4b | 2137 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 2138 | break; |
ada5c1da | 2139 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
2140 | if (error) |
2141 | break; | |
1da177e4 | 2142 | } |
16306a61 | 2143 | locks_delete_block(fl); |
1da177e4 LT |
2144 | return error; |
2145 | } | |
2146 | ||
e55c34a6 BC |
2147 | /** |
2148 | * locks_lock_inode_wait - Apply a lock to an inode | |
2149 | * @inode: inode of the file to apply to | |
2150 | * @fl: The lock to be applied | |
2151 | * | |
2152 | * Apply a POSIX or FLOCK style lock request to an inode. | |
2153 | */ | |
2154 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
2155 | { | |
2156 | int res = 0; | |
2157 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
2158 | case FL_POSIX: | |
2159 | res = posix_lock_inode_wait(inode, fl); | |
2160 | break; | |
2161 | case FL_FLOCK: | |
2162 | res = flock_lock_inode_wait(inode, fl); | |
2163 | break; | |
2164 | default: | |
2165 | BUG(); | |
2166 | } | |
2167 | return res; | |
2168 | } | |
2169 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
2170 | ||
1da177e4 LT |
2171 | /** |
2172 | * sys_flock: - flock() system call. | |
2173 | * @fd: the file descriptor to lock. | |
2174 | * @cmd: the type of lock to apply. | |
2175 | * | |
2176 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
80b79dd0 | 2177 | * The @cmd can be one of: |
1da177e4 | 2178 | * |
80b79dd0 MCC |
2179 | * - %LOCK_SH -- a shared lock. |
2180 | * - %LOCK_EX -- an exclusive lock. | |
2181 | * - %LOCK_UN -- remove an existing lock. | |
2182 | * - %LOCK_MAND -- a 'mandatory' flock. | |
2183 | * This exists to emulate Windows Share Modes. | |
1da177e4 LT |
2184 | * |
2185 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
2186 | * processes read and write access respectively. | |
2187 | */ | |
002c8976 | 2188 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 2189 | { |
2903ff01 | 2190 | struct fd f = fdget(fd); |
1da177e4 LT |
2191 | struct file_lock *lock; |
2192 | int can_sleep, unlock; | |
2193 | int error; | |
2194 | ||
2195 | error = -EBADF; | |
2903ff01 | 2196 | if (!f.file) |
1da177e4 LT |
2197 | goto out; |
2198 | ||
2199 | can_sleep = !(cmd & LOCK_NB); | |
2200 | cmd &= ~LOCK_NB; | |
2201 | unlock = (cmd == LOCK_UN); | |
2202 | ||
aeb5d727 | 2203 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 2204 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
2205 | goto out_putf; |
2206 | ||
d6367d62 | 2207 | lock = flock_make_lock(f.file, cmd, NULL); |
6e129d00 JL |
2208 | if (IS_ERR(lock)) { |
2209 | error = PTR_ERR(lock); | |
1da177e4 | 2210 | goto out_putf; |
6e129d00 JL |
2211 | } |
2212 | ||
1da177e4 LT |
2213 | if (can_sleep) |
2214 | lock->fl_flags |= FL_SLEEP; | |
2215 | ||
2903ff01 | 2216 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
2217 | if (error) |
2218 | goto out_free; | |
2219 | ||
de2a4a50 | 2220 | if (f.file->f_op->flock) |
2903ff01 | 2221 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
2222 | (can_sleep) ? F_SETLKW : F_SETLK, |
2223 | lock); | |
2224 | else | |
4f656367 | 2225 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
2226 | |
2227 | out_free: | |
993dfa87 | 2228 | locks_free_lock(lock); |
1da177e4 LT |
2229 | |
2230 | out_putf: | |
2903ff01 | 2231 | fdput(f); |
1da177e4 LT |
2232 | out: |
2233 | return error; | |
2234 | } | |
2235 | ||
3ee17abd BF |
2236 | /** |
2237 | * vfs_test_lock - test file byte range lock | |
2238 | * @filp: The file to test lock for | |
6924c554 | 2239 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2240 | * |
2241 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2242 | * setting conf->fl_type to something other than F_UNLCK. | |
2243 | */ | |
2244 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2245 | { | |
de2a4a50 | 2246 | if (filp->f_op->lock) |
3ee17abd BF |
2247 | return filp->f_op->lock(filp, F_GETLK, fl); |
2248 | posix_test_lock(filp, fl); | |
2249 | return 0; | |
2250 | } | |
2251 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2252 | ||
9d5b86ac BC |
2253 | /** |
2254 | * locks_translate_pid - translate a file_lock's fl_pid number into a namespace | |
2255 | * @fl: The file_lock who's fl_pid should be translated | |
2256 | * @ns: The namespace into which the pid should be translated | |
2257 | * | |
2258 | * Used to tranlate a fl_pid into a namespace virtual pid number | |
2259 | */ | |
2260 | static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns) | |
2261 | { | |
2262 | pid_t vnr; | |
2263 | struct pid *pid; | |
2264 | ||
2265 | if (IS_OFDLCK(fl)) | |
2266 | return -1; | |
2267 | if (IS_REMOTELCK(fl)) | |
2268 | return fl->fl_pid; | |
826d7bc9 KK |
2269 | /* |
2270 | * If the flock owner process is dead and its pid has been already | |
2271 | * freed, the translation below won't work, but we still want to show | |
2272 | * flock owner pid number in init pidns. | |
2273 | */ | |
2274 | if (ns == &init_pid_ns) | |
2275 | return (pid_t)fl->fl_pid; | |
9d5b86ac BC |
2276 | |
2277 | rcu_read_lock(); | |
2278 | pid = find_pid_ns(fl->fl_pid, &init_pid_ns); | |
2279 | vnr = pid_nr_ns(pid, ns); | |
2280 | rcu_read_unlock(); | |
2281 | return vnr; | |
2282 | } | |
2283 | ||
c2fa1b8a BF |
2284 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2285 | { | |
9d5b86ac | 2286 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2287 | #if BITS_PER_LONG == 32 |
2288 | /* | |
2289 | * Make sure we can represent the posix lock via | |
2290 | * legacy 32bit flock. | |
2291 | */ | |
2292 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2293 | return -EOVERFLOW; | |
2294 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2295 | return -EOVERFLOW; | |
2296 | #endif | |
2297 | flock->l_start = fl->fl_start; | |
2298 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2299 | fl->fl_end - fl->fl_start + 1; | |
2300 | flock->l_whence = 0; | |
129a84de | 2301 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2302 | return 0; |
2303 | } | |
2304 | ||
2305 | #if BITS_PER_LONG == 32 | |
2306 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2307 | { | |
9d5b86ac | 2308 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2309 | flock->l_start = fl->fl_start; |
2310 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2311 | fl->fl_end - fl->fl_start + 1; | |
2312 | flock->l_whence = 0; | |
2313 | flock->l_type = fl->fl_type; | |
2314 | } | |
2315 | #endif | |
2316 | ||
1da177e4 LT |
2317 | /* Report the first existing lock that would conflict with l. |
2318 | * This implements the F_GETLK command of fcntl(). | |
2319 | */ | |
a75d30c7 | 2320 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock) |
1da177e4 | 2321 | { |
52306e88 | 2322 | struct file_lock *fl; |
1da177e4 LT |
2323 | int error; |
2324 | ||
52306e88 BC |
2325 | fl = locks_alloc_lock(); |
2326 | if (fl == NULL) | |
2327 | return -ENOMEM; | |
1da177e4 | 2328 | error = -EINVAL; |
a75d30c7 | 2329 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2330 | goto out; |
2331 | ||
52306e88 | 2332 | error = flock_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2333 | if (error) |
2334 | goto out; | |
2335 | ||
0d3f7a2d | 2336 | if (cmd == F_OFD_GETLK) { |
90478939 | 2337 | error = -EINVAL; |
a75d30c7 | 2338 | if (flock->l_pid != 0) |
90478939 JL |
2339 | goto out; |
2340 | ||
5d50ffd7 | 2341 | cmd = F_GETLK; |
52306e88 BC |
2342 | fl->fl_flags |= FL_OFDLCK; |
2343 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2344 | } |
2345 | ||
52306e88 | 2346 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2347 | if (error) |
2348 | goto out; | |
7bbd1fc0 | 2349 | |
52306e88 BC |
2350 | flock->l_type = fl->fl_type; |
2351 | if (fl->fl_type != F_UNLCK) { | |
2352 | error = posix_lock_to_flock(flock, fl); | |
c2fa1b8a | 2353 | if (error) |
52306e88 | 2354 | goto out; |
1da177e4 | 2355 | } |
1da177e4 | 2356 | out: |
52306e88 | 2357 | locks_free_lock(fl); |
1da177e4 LT |
2358 | return error; |
2359 | } | |
2360 | ||
7723ec97 ME |
2361 | /** |
2362 | * vfs_lock_file - file byte range lock | |
2363 | * @filp: The file to apply the lock to | |
2364 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2365 | * @fl: The lock to be applied | |
150b3934 ME |
2366 | * @conf: Place to return a copy of the conflicting lock, if found. |
2367 | * | |
2368 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2369 | * as the final argument. | |
2370 | * | |
2371 | * If the filesystem defines a private ->lock() method, then @conf will | |
2372 | * be left unchanged; so a caller that cares should initialize it to | |
2373 | * some acceptable default. | |
2beb6614 ME |
2374 | * |
2375 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2376 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2377 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2378 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2379 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2380 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2381 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2382 | * request completes. |
2383 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2384 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2385 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2386 | * nonzero return code and the file system should release the lock. The file |
2387 | * system is also responsible to keep a corresponding posix lock when it | |
2388 | * grants a lock so the VFS can find out which locks are locally held and do | |
2389 | * the correct lock cleanup when required. | |
2390 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2391 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2392 | * return code. |
7723ec97 | 2393 | */ |
150b3934 | 2394 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2395 | { |
de2a4a50 | 2396 | if (filp->f_op->lock) |
7723ec97 ME |
2397 | return filp->f_op->lock(filp, cmd, fl); |
2398 | else | |
150b3934 | 2399 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2400 | } |
2401 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2402 | ||
b648a6de MS |
2403 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2404 | struct file_lock *fl) | |
2405 | { | |
2406 | int error; | |
2407 | ||
2408 | error = security_file_lock(filp, fl->fl_type); | |
2409 | if (error) | |
2410 | return error; | |
2411 | ||
764c76b3 MS |
2412 | for (;;) { |
2413 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2414 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2415 | break; |
ada5c1da | 2416 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_blocker); |
16306a61 N |
2417 | if (error) |
2418 | break; | |
b648a6de | 2419 | } |
16306a61 | 2420 | locks_delete_block(fl); |
b648a6de MS |
2421 | |
2422 | return error; | |
2423 | } | |
2424 | ||
6ca7d910 | 2425 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2426 | static int |
2427 | check_fmode_for_setlk(struct file_lock *fl) | |
2428 | { | |
2429 | switch (fl->fl_type) { | |
2430 | case F_RDLCK: | |
2431 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2432 | return -EBADF; | |
2433 | break; | |
2434 | case F_WRLCK: | |
2435 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2436 | return -EBADF; | |
2437 | } | |
2438 | return 0; | |
2439 | } | |
2440 | ||
1da177e4 LT |
2441 | /* Apply the lock described by l to an open file descriptor. |
2442 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2443 | */ | |
c293621b | 2444 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2445 | struct flock *flock) |
1da177e4 LT |
2446 | { |
2447 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2448 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2449 | struct file *f; |
1da177e4 LT |
2450 | int error; |
2451 | ||
2452 | if (file_lock == NULL) | |
2453 | return -ENOLCK; | |
2454 | ||
1da177e4 LT |
2455 | /* Don't allow mandatory locks on files that may be memory mapped |
2456 | * and shared. | |
2457 | */ | |
a16877ca | 2458 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2459 | error = -EAGAIN; |
2460 | goto out; | |
2461 | } | |
2462 | ||
a75d30c7 | 2463 | error = flock_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2464 | if (error) |
2465 | goto out; | |
5d50ffd7 | 2466 | |
cf01f4ee JL |
2467 | error = check_fmode_for_setlk(file_lock); |
2468 | if (error) | |
2469 | goto out; | |
2470 | ||
5d50ffd7 JL |
2471 | /* |
2472 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2473 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2474 | */ |
2475 | switch (cmd) { | |
0d3f7a2d | 2476 | case F_OFD_SETLK: |
90478939 | 2477 | error = -EINVAL; |
a75d30c7 | 2478 | if (flock->l_pid != 0) |
90478939 JL |
2479 | goto out; |
2480 | ||
5d50ffd7 | 2481 | cmd = F_SETLK; |
cff2fce5 | 2482 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2483 | file_lock->fl_owner = filp; |
5d50ffd7 | 2484 | break; |
0d3f7a2d | 2485 | case F_OFD_SETLKW: |
90478939 | 2486 | error = -EINVAL; |
a75d30c7 | 2487 | if (flock->l_pid != 0) |
90478939 JL |
2488 | goto out; |
2489 | ||
5d50ffd7 | 2490 | cmd = F_SETLKW; |
cff2fce5 | 2491 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2492 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2493 | /* Fallthrough */ |
2494 | case F_SETLKW: | |
1da177e4 LT |
2495 | file_lock->fl_flags |= FL_SLEEP; |
2496 | } | |
5d50ffd7 | 2497 | |
b648a6de | 2498 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2499 | |
c293621b | 2500 | /* |
0752ba80 JL |
2501 | * Attempt to detect a close/fcntl race and recover by releasing the |
2502 | * lock that was just acquired. There is no need to do that when we're | |
2503 | * unlocking though, or for OFD locks. | |
c293621b | 2504 | */ |
0752ba80 JL |
2505 | if (!error && file_lock->fl_type != F_UNLCK && |
2506 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2507 | /* |
2508 | * We need that spin_lock here - it prevents reordering between | |
2509 | * update of i_flctx->flc_posix and check for it done in | |
2510 | * close(). rcu_read_lock() wouldn't do. | |
2511 | */ | |
2512 | spin_lock(¤t->files->file_lock); | |
2513 | f = fcheck(fd); | |
2514 | spin_unlock(¤t->files->file_lock); | |
2515 | if (f != filp) { | |
2516 | file_lock->fl_type = F_UNLCK; | |
2517 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2518 | WARN_ON_ONCE(error); | |
2519 | error = -EBADF; | |
2520 | } | |
1da177e4 | 2521 | } |
c293621b | 2522 | out: |
1890910f | 2523 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2524 | locks_free_lock(file_lock); |
2525 | return error; | |
2526 | } | |
2527 | ||
2528 | #if BITS_PER_LONG == 32 | |
2529 | /* Report the first existing lock that would conflict with l. | |
2530 | * This implements the F_GETLK command of fcntl(). | |
2531 | */ | |
a75d30c7 | 2532 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock) |
1da177e4 | 2533 | { |
52306e88 | 2534 | struct file_lock *fl; |
1da177e4 LT |
2535 | int error; |
2536 | ||
52306e88 BC |
2537 | fl = locks_alloc_lock(); |
2538 | if (fl == NULL) | |
2539 | return -ENOMEM; | |
2540 | ||
1da177e4 | 2541 | error = -EINVAL; |
a75d30c7 | 2542 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2543 | goto out; |
2544 | ||
52306e88 | 2545 | error = flock64_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2546 | if (error) |
2547 | goto out; | |
2548 | ||
0d3f7a2d | 2549 | if (cmd == F_OFD_GETLK) { |
90478939 | 2550 | error = -EINVAL; |
a75d30c7 | 2551 | if (flock->l_pid != 0) |
90478939 JL |
2552 | goto out; |
2553 | ||
5d50ffd7 | 2554 | cmd = F_GETLK64; |
52306e88 BC |
2555 | fl->fl_flags |= FL_OFDLCK; |
2556 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2557 | } |
2558 | ||
52306e88 | 2559 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2560 | if (error) |
2561 | goto out; | |
2562 | ||
52306e88 BC |
2563 | flock->l_type = fl->fl_type; |
2564 | if (fl->fl_type != F_UNLCK) | |
2565 | posix_lock_to_flock64(flock, fl); | |
f328296e | 2566 | |
1da177e4 | 2567 | out: |
52306e88 | 2568 | locks_free_lock(fl); |
1da177e4 LT |
2569 | return error; |
2570 | } | |
2571 | ||
2572 | /* Apply the lock described by l to an open file descriptor. | |
2573 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2574 | */ | |
c293621b | 2575 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2576 | struct flock64 *flock) |
1da177e4 LT |
2577 | { |
2578 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2579 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2580 | struct file *f; |
1da177e4 LT |
2581 | int error; |
2582 | ||
2583 | if (file_lock == NULL) | |
2584 | return -ENOLCK; | |
2585 | ||
1da177e4 LT |
2586 | /* Don't allow mandatory locks on files that may be memory mapped |
2587 | * and shared. | |
2588 | */ | |
a16877ca | 2589 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2590 | error = -EAGAIN; |
2591 | goto out; | |
2592 | } | |
2593 | ||
a75d30c7 | 2594 | error = flock64_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2595 | if (error) |
2596 | goto out; | |
5d50ffd7 | 2597 | |
cf01f4ee JL |
2598 | error = check_fmode_for_setlk(file_lock); |
2599 | if (error) | |
2600 | goto out; | |
2601 | ||
5d50ffd7 JL |
2602 | /* |
2603 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2604 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2605 | */ |
2606 | switch (cmd) { | |
0d3f7a2d | 2607 | case F_OFD_SETLK: |
90478939 | 2608 | error = -EINVAL; |
a75d30c7 | 2609 | if (flock->l_pid != 0) |
90478939 JL |
2610 | goto out; |
2611 | ||
5d50ffd7 | 2612 | cmd = F_SETLK64; |
cff2fce5 | 2613 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2614 | file_lock->fl_owner = filp; |
5d50ffd7 | 2615 | break; |
0d3f7a2d | 2616 | case F_OFD_SETLKW: |
90478939 | 2617 | error = -EINVAL; |
a75d30c7 | 2618 | if (flock->l_pid != 0) |
90478939 JL |
2619 | goto out; |
2620 | ||
5d50ffd7 | 2621 | cmd = F_SETLKW64; |
cff2fce5 | 2622 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2623 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2624 | /* Fallthrough */ |
2625 | case F_SETLKW64: | |
1da177e4 LT |
2626 | file_lock->fl_flags |= FL_SLEEP; |
2627 | } | |
5d50ffd7 | 2628 | |
b648a6de | 2629 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2630 | |
c293621b | 2631 | /* |
0752ba80 JL |
2632 | * Attempt to detect a close/fcntl race and recover by releasing the |
2633 | * lock that was just acquired. There is no need to do that when we're | |
2634 | * unlocking though, or for OFD locks. | |
c293621b | 2635 | */ |
0752ba80 JL |
2636 | if (!error && file_lock->fl_type != F_UNLCK && |
2637 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
7f3697e2 JL |
2638 | /* |
2639 | * We need that spin_lock here - it prevents reordering between | |
2640 | * update of i_flctx->flc_posix and check for it done in | |
2641 | * close(). rcu_read_lock() wouldn't do. | |
2642 | */ | |
2643 | spin_lock(¤t->files->file_lock); | |
2644 | f = fcheck(fd); | |
2645 | spin_unlock(¤t->files->file_lock); | |
2646 | if (f != filp) { | |
2647 | file_lock->fl_type = F_UNLCK; | |
2648 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2649 | WARN_ON_ONCE(error); | |
2650 | error = -EBADF; | |
2651 | } | |
1da177e4 | 2652 | } |
1da177e4 LT |
2653 | out: |
2654 | locks_free_lock(file_lock); | |
2655 | return error; | |
2656 | } | |
2657 | #endif /* BITS_PER_LONG == 32 */ | |
2658 | ||
2659 | /* | |
2660 | * This function is called when the file is being removed | |
2661 | * from the task's fd array. POSIX locks belonging to this task | |
2662 | * are deleted at this time. | |
2663 | */ | |
2664 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2665 | { | |
1890910f | 2666 | int error; |
c568d683 | 2667 | struct inode *inode = locks_inode(filp); |
ff7b86b8 | 2668 | struct file_lock lock; |
128a3785 | 2669 | struct file_lock_context *ctx; |
1da177e4 LT |
2670 | |
2671 | /* | |
2672 | * If there are no locks held on this file, we don't need to call | |
2673 | * posix_lock_file(). Another process could be setting a lock on this | |
2674 | * file at the same time, but we wouldn't remove that lock anyway. | |
2675 | */ | |
c568d683 | 2676 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 | 2677 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2678 | return; |
2679 | ||
d6367d62 | 2680 | locks_init_lock(&lock); |
1da177e4 | 2681 | lock.fl_type = F_UNLCK; |
75e1fcc0 | 2682 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2683 | lock.fl_start = 0; |
2684 | lock.fl_end = OFFSET_MAX; | |
2685 | lock.fl_owner = owner; | |
2686 | lock.fl_pid = current->tgid; | |
2687 | lock.fl_file = filp; | |
2688 | lock.fl_ops = NULL; | |
2689 | lock.fl_lmops = NULL; | |
2690 | ||
1890910f | 2691 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2692 | |
1da177e4 LT |
2693 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2694 | lock.fl_ops->fl_release_private(&lock); | |
c568d683 | 2695 | trace_locks_remove_posix(inode, &lock, error); |
1da177e4 | 2696 | } |
1da177e4 LT |
2697 | EXPORT_SYMBOL(locks_remove_posix); |
2698 | ||
3d8e560d | 2699 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2700 | static void |
128a3785 | 2701 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 | 2702 | { |
d6367d62 | 2703 | struct file_lock fl; |
c568d683 | 2704 | struct inode *inode = locks_inode(filp); |
dd459bb1 | 2705 | |
3d8e560d | 2706 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2707 | return; |
2708 | ||
d6367d62 N |
2709 | flock_make_lock(filp, LOCK_UN, &fl); |
2710 | fl.fl_flags |= FL_CLOSE; | |
2711 | ||
de2a4a50 | 2712 | if (filp->f_op->flock) |
dd459bb1 JL |
2713 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2714 | else | |
bcd7f78d | 2715 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2716 | |
2717 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2718 | fl.fl_ops->fl_release_private(&fl); | |
2719 | } | |
2720 | ||
3d8e560d | 2721 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2722 | static void |
128a3785 | 2723 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2724 | { |
8634b51f JL |
2725 | struct file_lock *fl, *tmp; |
2726 | LIST_HEAD(dispose); | |
2727 | ||
3d8e560d | 2728 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2729 | return; |
2730 | ||
02e525b2 | 2731 | percpu_down_read(&file_rwsem); |
6109c850 | 2732 | spin_lock(&ctx->flc_lock); |
8634b51f | 2733 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2734 | if (filp == fl->fl_file) |
2735 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2736 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 2737 | percpu_up_read(&file_rwsem); |
5f43086b | 2738 | |
8634b51f JL |
2739 | locks_dispose_list(&dispose); |
2740 | } | |
2741 | ||
1da177e4 LT |
2742 | /* |
2743 | * This function is called on the last close of an open file. | |
2744 | */ | |
78ed8a13 | 2745 | void locks_remove_file(struct file *filp) |
1da177e4 | 2746 | { |
128a3785 DV |
2747 | struct file_lock_context *ctx; |
2748 | ||
c568d683 | 2749 | ctx = smp_load_acquire(&locks_inode(filp)->i_flctx); |
128a3785 | 2750 | if (!ctx) |
3d8e560d JL |
2751 | return; |
2752 | ||
dd459bb1 | 2753 | /* remove any OFD locks */ |
73a8f5f7 | 2754 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2755 | |
dd459bb1 | 2756 | /* remove flock locks */ |
128a3785 | 2757 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2758 | |
8634b51f | 2759 | /* remove any leases */ |
128a3785 | 2760 | locks_remove_lease(filp, ctx); |
3953704f BC |
2761 | |
2762 | spin_lock(&ctx->flc_lock); | |
2763 | locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX"); | |
2764 | locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK"); | |
2765 | locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE"); | |
2766 | spin_unlock(&ctx->flc_lock); | |
1da177e4 LT |
2767 | } |
2768 | ||
9b9d2ab4 ME |
2769 | /** |
2770 | * vfs_cancel_lock - file byte range unblock lock | |
2771 | * @filp: The file to apply the unblock to | |
2772 | * @fl: The lock to be unblocked | |
2773 | * | |
2774 | * Used by lock managers to cancel blocked requests | |
2775 | */ | |
2776 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2777 | { | |
de2a4a50 | 2778 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2779 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2780 | return 0; | |
2781 | } | |
9b9d2ab4 ME |
2782 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); |
2783 | ||
7f8ada98 | 2784 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2785 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2786 | #include <linux/seq_file.h> |
2787 | ||
7012b02a JL |
2788 | struct locks_iterator { |
2789 | int li_cpu; | |
2790 | loff_t li_pos; | |
2791 | }; | |
2792 | ||
7f8ada98 | 2793 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2794 | loff_t id, char *pfx) |
1da177e4 LT |
2795 | { |
2796 | struct inode *inode = NULL; | |
ab1f1611 | 2797 | unsigned int fl_pid; |
9d5b86ac | 2798 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
ab1f1611 | 2799 | |
9d5b86ac BC |
2800 | fl_pid = locks_translate_pid(fl, proc_pidns); |
2801 | /* | |
1cf8e5de KK |
2802 | * If lock owner is dead (and pid is freed) or not visible in current |
2803 | * pidns, zero is shown as a pid value. Check lock info from | |
2804 | * init_pid_ns to get saved lock pid value. | |
9d5b86ac | 2805 | */ |
1da177e4 LT |
2806 | |
2807 | if (fl->fl_file != NULL) | |
c568d683 | 2808 | inode = locks_inode(fl->fl_file); |
1da177e4 | 2809 | |
99dc8292 | 2810 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2811 | if (IS_POSIX(fl)) { |
c918d42a | 2812 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2813 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2814 | else if (IS_OFDLCK(fl)) |
5315c26a | 2815 | seq_puts(f, "OFDLCK"); |
c918d42a | 2816 | else |
5315c26a | 2817 | seq_puts(f, "POSIX "); |
c918d42a JL |
2818 | |
2819 | seq_printf(f, " %s ", | |
1da177e4 | 2820 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2821 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2822 | } else if (IS_FLOCK(fl)) { |
2823 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2824 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2825 | } else { |
5315c26a | 2826 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2827 | } |
2828 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2829 | if (fl->fl_flags & FL_DELEG) |
2830 | seq_puts(f, "DELEG "); | |
2831 | else | |
2832 | seq_puts(f, "LEASE "); | |
2833 | ||
ab83fa4b | 2834 | if (lease_breaking(fl)) |
5315c26a | 2835 | seq_puts(f, "BREAKING "); |
1da177e4 | 2836 | else if (fl->fl_file) |
5315c26a | 2837 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2838 | else |
5315c26a | 2839 | seq_puts(f, "BREAKER "); |
1da177e4 | 2840 | } else { |
5315c26a | 2841 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2842 | } |
2843 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2844 | seq_printf(f, "%s ", |
1da177e4 LT |
2845 | (fl->fl_type & LOCK_READ) |
2846 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2847 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2848 | } else { | |
43e4cb94 PB |
2849 | int type = IS_LEASE(fl) ? target_leasetype(fl) : fl->fl_type; |
2850 | ||
2851 | seq_printf(f, "%s ", (type == F_WRLCK) ? "WRITE" : | |
2852 | (type == F_RDLCK) ? "READ" : "UNLCK"); | |
1da177e4 LT |
2853 | } |
2854 | if (inode) { | |
3648888e | 2855 | /* userspace relies on this representation of dev_t */ |
ab1f1611 | 2856 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2857 | MAJOR(inode->i_sb->s_dev), |
2858 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2859 | } else { |
ab1f1611 | 2860 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2861 | } |
2862 | if (IS_POSIX(fl)) { | |
2863 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2864 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2865 | else |
7f8ada98 | 2866 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2867 | } else { |
5315c26a | 2868 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2869 | } |
2870 | } | |
2871 | ||
7f8ada98 | 2872 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2873 | { |
7012b02a | 2874 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2875 | struct file_lock *fl, *bfl; |
d67fd44f | 2876 | struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info; |
1da177e4 | 2877 | |
139ca04e | 2878 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2879 | |
9d5b86ac | 2880 | if (locks_translate_pid(fl, proc_pidns) == 0) |
d67fd44f NB |
2881 | return 0; |
2882 | ||
7012b02a | 2883 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2884 | |
ada5c1da | 2885 | list_for_each_entry(bfl, &fl->fl_blocked_requests, fl_blocked_member) |
7012b02a | 2886 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2887 | |
7f8ada98 PE |
2888 | return 0; |
2889 | } | |
1da177e4 | 2890 | |
6c8c9031 AV |
2891 | static void __show_fd_locks(struct seq_file *f, |
2892 | struct list_head *head, int *id, | |
2893 | struct file *filp, struct files_struct *files) | |
2894 | { | |
2895 | struct file_lock *fl; | |
2896 | ||
2897 | list_for_each_entry(fl, head, fl_list) { | |
2898 | ||
2899 | if (filp != fl->fl_file) | |
2900 | continue; | |
2901 | if (fl->fl_owner != files && | |
2902 | fl->fl_owner != filp) | |
2903 | continue; | |
2904 | ||
2905 | (*id)++; | |
2906 | seq_puts(f, "lock:\t"); | |
2907 | lock_get_status(f, fl, *id, ""); | |
2908 | } | |
2909 | } | |
2910 | ||
2911 | void show_fd_locks(struct seq_file *f, | |
2912 | struct file *filp, struct files_struct *files) | |
2913 | { | |
c568d683 | 2914 | struct inode *inode = locks_inode(filp); |
6c8c9031 AV |
2915 | struct file_lock_context *ctx; |
2916 | int id = 0; | |
2917 | ||
128a3785 | 2918 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2919 | if (!ctx) |
2920 | return; | |
2921 | ||
2922 | spin_lock(&ctx->flc_lock); | |
2923 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2924 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2925 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2926 | spin_unlock(&ctx->flc_lock); | |
2927 | } | |
2928 | ||
7f8ada98 | 2929 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2930 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2931 | { |
7012b02a | 2932 | struct locks_iterator *iter = f->private; |
99dc8292 | 2933 | |
7012b02a | 2934 | iter->li_pos = *pos + 1; |
aba37660 | 2935 | percpu_down_write(&file_rwsem); |
7b2296af | 2936 | spin_lock(&blocked_lock_lock); |
7c3f654d | 2937 | return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos); |
7f8ada98 | 2938 | } |
1da177e4 | 2939 | |
7f8ada98 PE |
2940 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2941 | { | |
7012b02a JL |
2942 | struct locks_iterator *iter = f->private; |
2943 | ||
2944 | ++iter->li_pos; | |
7c3f654d | 2945 | return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos); |
7f8ada98 | 2946 | } |
1da177e4 | 2947 | |
7f8ada98 | 2948 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2949 | __releases(&blocked_lock_lock) |
7f8ada98 | 2950 | { |
7b2296af | 2951 | spin_unlock(&blocked_lock_lock); |
aba37660 | 2952 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2953 | } |
2954 | ||
d8ba7a36 | 2955 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2956 | .start = locks_start, |
2957 | .next = locks_next, | |
2958 | .stop = locks_stop, | |
2959 | .show = locks_show, | |
2960 | }; | |
d8ba7a36 | 2961 | |
d8ba7a36 AD |
2962 | static int __init proc_locks_init(void) |
2963 | { | |
44414d82 CH |
2964 | proc_create_seq_private("locks", 0, NULL, &locks_seq_operations, |
2965 | sizeof(struct locks_iterator), NULL); | |
d8ba7a36 AD |
2966 | return 0; |
2967 | } | |
91899226 | 2968 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2969 | #endif |
2970 | ||
1da177e4 LT |
2971 | static int __init filelock_init(void) |
2972 | { | |
7012b02a JL |
2973 | int i; |
2974 | ||
4a075e39 JL |
2975 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2976 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2977 | ||
1da177e4 | 2978 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2979 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2980 | ||
7c3f654d PZ |
2981 | for_each_possible_cpu(i) { |
2982 | struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i); | |
2983 | ||
2984 | spin_lock_init(&fll->lock); | |
2985 | INIT_HLIST_HEAD(&fll->hlist); | |
2986 | } | |
7012b02a | 2987 | |
18f6622e | 2988 | lease_notifier_chain_init(); |
1da177e4 LT |
2989 | return 0; |
2990 | } | |
1da177e4 | 2991 | core_initcall(filelock_init); |