Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/fs/locks.c | |
4 | * | |
e9728cc7 BF |
5 | * We implement four types of file locks: BSD locks, posix locks, open |
6 | * file description locks, and leases. For details about BSD locks, | |
7 | * see the flock(2) man page; for details about the other three, see | |
8 | * fcntl(2). | |
1da177e4 | 9 | * |
fd7732e0 N |
10 | * |
11 | * Locking conflicts and dependencies: | |
12 | * If multiple threads attempt to lock the same byte (or flock the same file) | |
13 | * only one can be granted the lock, and other must wait their turn. | |
14 | * The first lock has been "applied" or "granted", the others are "waiting" | |
15 | * and are "blocked" by the "applied" lock.. | |
16 | * | |
17 | * Waiting and applied locks are all kept in trees whose properties are: | |
18 | * | |
19 | * - the root of a tree may be an applied or waiting lock. | |
20 | * - every other node in the tree is a waiting lock that | |
21 | * conflicts with every ancestor of that node. | |
22 | * | |
23 | * Every such tree begins life as a waiting singleton which obviously | |
24 | * satisfies the above properties. | |
25 | * | |
26 | * The only ways we modify trees preserve these properties: | |
27 | * | |
28 | * 1. We may add a new leaf node, but only after first verifying that it | |
29 | * conflicts with all of its ancestors. | |
30 | * 2. We may remove the root of a tree, creating a new singleton | |
31 | * tree from the root and N new trees rooted in the immediate | |
32 | * children. | |
33 | * 3. If the root of a tree is not currently an applied lock, we may | |
34 | * apply it (if possible). | |
35 | * 4. We may upgrade the root of the tree (either extend its range, | |
36 | * or upgrade its entire range from read to write). | |
37 | * | |
38 | * When an applied lock is modified in a way that reduces or downgrades any | |
39 | * part of its range, we remove all its children (2 above). This particularly | |
40 | * happens when a lock is unlocked. | |
41 | * | |
42 | * For each of those child trees we "wake up" the thread which is | |
43 | * waiting for the lock so it can continue handling as follows: if the | |
44 | * root of the tree applies, we do so (3). If it doesn't, it must | |
45 | * conflict with some applied lock. We remove (wake up) all of its children | |
46 | * (2), and add it is a new leaf to the tree rooted in the applied | |
47 | * lock (1). We then repeat the process recursively with those | |
48 | * children. | |
49 | * | |
1da177e4 | 50 | */ |
1da177e4 LT |
51 | #include <linux/capability.h> |
52 | #include <linux/file.h> | |
9f3acc31 | 53 | #include <linux/fdtable.h> |
5970e15d | 54 | #include <linux/filelock.h> |
1da177e4 LT |
55 | #include <linux/fs.h> |
56 | #include <linux/init.h> | |
1da177e4 LT |
57 | #include <linux/security.h> |
58 | #include <linux/slab.h> | |
1da177e4 LT |
59 | #include <linux/syscalls.h> |
60 | #include <linux/time.h> | |
4fb3a538 | 61 | #include <linux/rcupdate.h> |
ab1f1611 | 62 | #include <linux/pid_namespace.h> |
48f74186 | 63 | #include <linux/hashtable.h> |
7012b02a | 64 | #include <linux/percpu.h> |
dd81faa8 | 65 | #include <linux/sysctl.h> |
1da177e4 | 66 | |
62af4f1f JL |
67 | #define CREATE_TRACE_POINTS |
68 | #include <trace/events/filelock.h> | |
69 | ||
7c0f6ba6 | 70 | #include <linux/uaccess.h> |
1da177e4 | 71 | |
1a62c22a JL |
72 | static struct file_lock *file_lock(struct file_lock_core *flc) |
73 | { | |
74 | return container_of(flc, struct file_lock, c); | |
75 | } | |
76 | ||
ab83fa4b BF |
77 | static bool lease_breaking(struct file_lock *fl) |
78 | { | |
4ca52f53 | 79 | return fl->c.flc_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
778fc546 BF |
80 | } |
81 | ||
82 | static int target_leasetype(struct file_lock *fl) | |
83 | { | |
4ca52f53 | 84 | if (fl->c.flc_flags & FL_UNLOCK_PENDING) |
778fc546 | 85 | return F_UNLCK; |
4ca52f53 | 86 | if (fl->c.flc_flags & FL_DOWNGRADE_PENDING) |
778fc546 | 87 | return F_RDLCK; |
4ca52f53 | 88 | return fl->c.flc_type; |
ab83fa4b BF |
89 | } |
90 | ||
dd81faa8 LC |
91 | static int leases_enable = 1; |
92 | static int lease_break_time = 45; | |
93 | ||
94 | #ifdef CONFIG_SYSCTL | |
95 | static struct ctl_table locks_sysctls[] = { | |
96 | { | |
97 | .procname = "leases-enable", | |
98 | .data = &leases_enable, | |
99 | .maxlen = sizeof(int), | |
100 | .mode = 0644, | |
101 | .proc_handler = proc_dointvec, | |
102 | }, | |
103 | #ifdef CONFIG_MMU | |
104 | { | |
105 | .procname = "lease-break-time", | |
106 | .data = &lease_break_time, | |
107 | .maxlen = sizeof(int), | |
108 | .mode = 0644, | |
109 | .proc_handler = proc_dointvec, | |
110 | }, | |
111 | #endif /* CONFIG_MMU */ | |
dd81faa8 LC |
112 | }; |
113 | ||
114 | static int __init init_fs_locks_sysctls(void) | |
115 | { | |
116 | register_sysctl_init("fs", locks_sysctls); | |
117 | return 0; | |
118 | } | |
119 | early_initcall(init_fs_locks_sysctls); | |
120 | #endif /* CONFIG_SYSCTL */ | |
1da177e4 | 121 | |
1c8c601a | 122 | /* |
7012b02a | 123 | * The global file_lock_list is only used for displaying /proc/locks, so we |
7c3f654d PZ |
124 | * keep a list on each CPU, with each list protected by its own spinlock. |
125 | * Global serialization is done using file_rwsem. | |
126 | * | |
127 | * Note that alterations to the list also require that the relevant flc_lock is | |
128 | * held. | |
1c8c601a | 129 | */ |
7c3f654d PZ |
130 | struct file_lock_list_struct { |
131 | spinlock_t lock; | |
132 | struct hlist_head hlist; | |
133 | }; | |
134 | static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list); | |
aba37660 | 135 | DEFINE_STATIC_PERCPU_RWSEM(file_rwsem); |
88974691 | 136 | |
eb82dd39 | 137 | |
1c8c601a | 138 | /* |
48f74186 | 139 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 140 | * It is protected by blocked_lock_lock. |
48f74186 JL |
141 | * |
142 | * We hash locks by lockowner in order to optimize searching for the lock a | |
143 | * particular lockowner is waiting on. | |
144 | * | |
145 | * FIXME: make this value scale via some heuristic? We generally will want more | |
146 | * buckets when we have more lockowners holding locks, but that's a little | |
147 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 148 | */ |
48f74186 JL |
149 | #define BLOCKED_HASH_BITS 7 |
150 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 151 | |
1c8c601a | 152 | /* |
7b2296af JL |
153 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
154 | * want to be holding this lock. | |
1c8c601a | 155 | * |
ada5c1da N |
156 | * In addition, it also protects the fl->fl_blocked_requests list, and the |
157 | * fl->fl_blocker pointer for file_lock structures that are acting as lock | |
158 | * requests (in contrast to those that are acting as records of acquired locks). | |
1c8c601a JL |
159 | * |
160 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 161 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 162 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 163 | * flc_lock. |
1c8c601a | 164 | */ |
7b2296af | 165 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 166 | |
68279f9c AD |
167 | static struct kmem_cache *flctx_cache __ro_after_init; |
168 | static struct kmem_cache *filelock_cache __ro_after_init; | |
1da177e4 | 169 | |
4a075e39 | 170 | static struct file_lock_context * |
5c1c669a | 171 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 172 | { |
128a3785 | 173 | struct file_lock_context *ctx; |
4a075e39 | 174 | |
128a3785 | 175 | /* paired with cmpxchg() below */ |
401a8b8f | 176 | ctx = locks_inode_context(inode); |
128a3785 | 177 | if (likely(ctx) || type == F_UNLCK) |
4a075e39 JL |
178 | goto out; |
179 | ||
128a3785 DV |
180 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
181 | if (!ctx) | |
4a075e39 JL |
182 | goto out; |
183 | ||
128a3785 DV |
184 | spin_lock_init(&ctx->flc_lock); |
185 | INIT_LIST_HEAD(&ctx->flc_flock); | |
186 | INIT_LIST_HEAD(&ctx->flc_posix); | |
187 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
188 | |
189 | /* | |
190 | * Assign the pointer if it's not already assigned. If it is, then | |
191 | * free the context we just allocated. | |
192 | */ | |
128a3785 DV |
193 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
194 | kmem_cache_free(flctx_cache, ctx); | |
401a8b8f | 195 | ctx = locks_inode_context(inode); |
128a3785 | 196 | } |
4a075e39 | 197 | out: |
1890910f | 198 | trace_locks_get_lock_context(inode, type, ctx); |
128a3785 | 199 | return ctx; |
4a075e39 JL |
200 | } |
201 | ||
e24dadab JL |
202 | static void |
203 | locks_dump_ctx_list(struct list_head *list, char *list_type) | |
204 | { | |
fde49518 | 205 | struct file_lock_core *flc; |
e24dadab | 206 | |
fde49518 JL |
207 | list_for_each_entry(flc, list, flc_list) |
208 | pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", | |
209 | list_type, flc->flc_owner, flc->flc_flags, | |
210 | flc->flc_type, flc->flc_pid); | |
e24dadab JL |
211 | } |
212 | ||
213 | static void | |
214 | locks_check_ctx_lists(struct inode *inode) | |
215 | { | |
216 | struct file_lock_context *ctx = inode->i_flctx; | |
217 | ||
218 | if (unlikely(!list_empty(&ctx->flc_flock) || | |
219 | !list_empty(&ctx->flc_posix) || | |
220 | !list_empty(&ctx->flc_lease))) { | |
221 | pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n", | |
222 | MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), | |
223 | inode->i_ino); | |
224 | locks_dump_ctx_list(&ctx->flc_flock, "FLOCK"); | |
225 | locks_dump_ctx_list(&ctx->flc_posix, "POSIX"); | |
226 | locks_dump_ctx_list(&ctx->flc_lease, "LEASE"); | |
227 | } | |
228 | } | |
229 | ||
3953704f | 230 | static void |
fde49518 | 231 | locks_check_ctx_file_list(struct file *filp, struct list_head *list, char *list_type) |
3953704f | 232 | { |
fde49518 | 233 | struct file_lock_core *flc; |
c65454a9 | 234 | struct inode *inode = file_inode(filp); |
3953704f | 235 | |
fde49518 JL |
236 | list_for_each_entry(flc, list, flc_list) |
237 | if (flc->flc_file == filp) | |
3953704f BC |
238 | pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx " |
239 | " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", | |
240 | list_type, MAJOR(inode->i_sb->s_dev), | |
241 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
fde49518 JL |
242 | flc->flc_owner, flc->flc_flags, |
243 | flc->flc_type, flc->flc_pid); | |
3953704f BC |
244 | } |
245 | ||
4a075e39 | 246 | void |
f27a0fe0 | 247 | locks_free_lock_context(struct inode *inode) |
4a075e39 | 248 | { |
401a8b8f | 249 | struct file_lock_context *ctx = locks_inode_context(inode); |
f27a0fe0 | 250 | |
e24dadab JL |
251 | if (unlikely(ctx)) { |
252 | locks_check_ctx_lists(inode); | |
4a075e39 JL |
253 | kmem_cache_free(flctx_cache, ctx); |
254 | } | |
255 | } | |
256 | ||
4ca52f53 | 257 | static void locks_init_lock_heads(struct file_lock_core *flc) |
a51cb91d | 258 | { |
4ca52f53 JL |
259 | INIT_HLIST_NODE(&flc->flc_link); |
260 | INIT_LIST_HEAD(&flc->flc_list); | |
261 | INIT_LIST_HEAD(&flc->flc_blocked_requests); | |
262 | INIT_LIST_HEAD(&flc->flc_blocked_member); | |
263 | init_waitqueue_head(&flc->flc_wait); | |
a51cb91d MS |
264 | } |
265 | ||
1da177e4 | 266 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 267 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 268 | { |
ee19cc40 | 269 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
270 | |
271 | if (fl) | |
4ca52f53 | 272 | locks_init_lock_heads(&fl->c); |
a51cb91d MS |
273 | |
274 | return fl; | |
1da177e4 | 275 | } |
c5b1f0d9 | 276 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 277 | |
a9e61e25 | 278 | void locks_release_private(struct file_lock *fl) |
47831f35 | 279 | { |
fde49518 JL |
280 | struct file_lock_core *flc = &fl->c; |
281 | ||
282 | BUG_ON(waitqueue_active(&flc->flc_wait)); | |
283 | BUG_ON(!list_empty(&flc->flc_list)); | |
284 | BUG_ON(!list_empty(&flc->flc_blocked_requests)); | |
285 | BUG_ON(!list_empty(&flc->flc_blocked_member)); | |
286 | BUG_ON(!hlist_unhashed(&flc->flc_link)); | |
5926459e | 287 | |
47831f35 TM |
288 | if (fl->fl_ops) { |
289 | if (fl->fl_ops->fl_release_private) | |
290 | fl->fl_ops->fl_release_private(fl); | |
291 | fl->fl_ops = NULL; | |
292 | } | |
47831f35 | 293 | |
5c97d7b1 | 294 | if (fl->fl_lmops) { |
cae80b30 | 295 | if (fl->fl_lmops->lm_put_owner) { |
fde49518 JL |
296 | fl->fl_lmops->lm_put_owner(flc->flc_owner); |
297 | flc->flc_owner = NULL; | |
cae80b30 | 298 | } |
5c97d7b1 KM |
299 | fl->fl_lmops = NULL; |
300 | } | |
47831f35 | 301 | } |
a9e61e25 | 302 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 303 | |
591502c5 DN |
304 | /** |
305 | * locks_owner_has_blockers - Check for blocking lock requests | |
306 | * @flctx: file lock context | |
307 | * @owner: lock owner | |
308 | * | |
309 | * Return values: | |
310 | * %true: @owner has at least one blocker | |
311 | * %false: @owner has no blockers | |
312 | */ | |
fde49518 | 313 | bool locks_owner_has_blockers(struct file_lock_context *flctx, fl_owner_t owner) |
591502c5 | 314 | { |
fde49518 | 315 | struct file_lock_core *flc; |
591502c5 DN |
316 | |
317 | spin_lock(&flctx->flc_lock); | |
fde49518 JL |
318 | list_for_each_entry(flc, &flctx->flc_posix, flc_list) { |
319 | if (flc->flc_owner != owner) | |
591502c5 | 320 | continue; |
fde49518 | 321 | if (!list_empty(&flc->flc_blocked_requests)) { |
591502c5 DN |
322 | spin_unlock(&flctx->flc_lock); |
323 | return true; | |
324 | } | |
325 | } | |
326 | spin_unlock(&flctx->flc_lock); | |
327 | return false; | |
328 | } | |
329 | EXPORT_SYMBOL_GPL(locks_owner_has_blockers); | |
330 | ||
1da177e4 | 331 | /* Free a lock which is not in use. */ |
05fa3135 | 332 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 333 | { |
47831f35 | 334 | locks_release_private(fl); |
1da177e4 LT |
335 | kmem_cache_free(filelock_cache, fl); |
336 | } | |
05fa3135 | 337 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 338 | |
ed9814d8 JL |
339 | static void |
340 | locks_dispose_list(struct list_head *dispose) | |
341 | { | |
342 | struct file_lock *fl; | |
343 | ||
344 | while (!list_empty(dispose)) { | |
4ca52f53 JL |
345 | fl = list_first_entry(dispose, struct file_lock, c.flc_list); |
346 | list_del_init(&fl->c.flc_list); | |
ed9814d8 JL |
347 | locks_free_lock(fl); |
348 | } | |
349 | } | |
350 | ||
1da177e4 LT |
351 | void locks_init_lock(struct file_lock *fl) |
352 | { | |
ee19cc40 | 353 | memset(fl, 0, sizeof(struct file_lock)); |
4ca52f53 | 354 | locks_init_lock_heads(&fl->c); |
1da177e4 | 355 | } |
1da177e4 LT |
356 | EXPORT_SYMBOL(locks_init_lock); |
357 | ||
1da177e4 LT |
358 | /* |
359 | * Initialize a new lock from an existing file_lock structure. | |
360 | */ | |
3fe0fff1 | 361 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 | 362 | { |
4ca52f53 JL |
363 | new->c.flc_owner = fl->c.flc_owner; |
364 | new->c.flc_pid = fl->c.flc_pid; | |
365 | new->c.flc_file = NULL; | |
366 | new->c.flc_flags = fl->c.flc_flags; | |
367 | new->c.flc_type = fl->c.flc_type; | |
1da177e4 LT |
368 | new->fl_start = fl->fl_start; |
369 | new->fl_end = fl->fl_end; | |
f328296e | 370 | new->fl_lmops = fl->fl_lmops; |
0996905f | 371 | new->fl_ops = NULL; |
f328296e KM |
372 | |
373 | if (fl->fl_lmops) { | |
374 | if (fl->fl_lmops->lm_get_owner) | |
4ca52f53 | 375 | fl->fl_lmops->lm_get_owner(fl->c.flc_owner); |
f328296e | 376 | } |
0996905f | 377 | } |
3fe0fff1 | 378 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
379 | |
380 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
381 | { | |
566709bd JL |
382 | /* "new" must be a freshly-initialized lock */ |
383 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 384 | |
3fe0fff1 | 385 | locks_copy_conflock(new, fl); |
f328296e | 386 | |
4ca52f53 | 387 | new->c.flc_file = fl->c.flc_file; |
1da177e4 | 388 | new->fl_ops = fl->fl_ops; |
47831f35 | 389 | |
f328296e KM |
390 | if (fl->fl_ops) { |
391 | if (fl->fl_ops->fl_copy_lock) | |
392 | fl->fl_ops->fl_copy_lock(new, fl); | |
393 | } | |
1da177e4 | 394 | } |
1da177e4 LT |
395 | EXPORT_SYMBOL(locks_copy_lock); |
396 | ||
5946c431 N |
397 | static void locks_move_blocks(struct file_lock *new, struct file_lock *fl) |
398 | { | |
399 | struct file_lock *f; | |
400 | ||
401 | /* | |
402 | * As ctx->flc_lock is held, new requests cannot be added to | |
b6aaba5b | 403 | * ->flc_blocked_requests, so we don't need a lock to check if it |
5946c431 N |
404 | * is empty. |
405 | */ | |
4ca52f53 | 406 | if (list_empty(&fl->c.flc_blocked_requests)) |
5946c431 N |
407 | return; |
408 | spin_lock(&blocked_lock_lock); | |
4ca52f53 JL |
409 | list_splice_init(&fl->c.flc_blocked_requests, |
410 | &new->c.flc_blocked_requests); | |
411 | list_for_each_entry(f, &new->c.flc_blocked_requests, | |
412 | c.flc_blocked_member) | |
b6aaba5b | 413 | f->c.flc_blocker = &new->c; |
5946c431 N |
414 | spin_unlock(&blocked_lock_lock); |
415 | } | |
416 | ||
1da177e4 | 417 | static inline int flock_translate_cmd(int cmd) { |
1da177e4 LT |
418 | switch (cmd) { |
419 | case LOCK_SH: | |
420 | return F_RDLCK; | |
421 | case LOCK_EX: | |
422 | return F_WRLCK; | |
423 | case LOCK_UN: | |
424 | return F_UNLCK; | |
425 | } | |
426 | return -EINVAL; | |
427 | } | |
428 | ||
429 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
4149be7b | 430 | static void flock_make_lock(struct file *filp, struct file_lock *fl, int type) |
1da177e4 | 431 | { |
4149be7b | 432 | locks_init_lock(fl); |
1da177e4 | 433 | |
4ca52f53 JL |
434 | fl->c.flc_file = filp; |
435 | fl->c.flc_owner = filp; | |
436 | fl->c.flc_pid = current->tgid; | |
437 | fl->c.flc_flags = FL_FLOCK; | |
438 | fl->c.flc_type = type; | |
1da177e4 | 439 | fl->fl_end = OFFSET_MAX; |
1da177e4 LT |
440 | } |
441 | ||
d9077f7b | 442 | static int assign_type(struct file_lock_core *flc, int type) |
1da177e4 LT |
443 | { |
444 | switch (type) { | |
445 | case F_RDLCK: | |
446 | case F_WRLCK: | |
447 | case F_UNLCK: | |
d9077f7b | 448 | flc->flc_type = type; |
1da177e4 LT |
449 | break; |
450 | default: | |
451 | return -EINVAL; | |
452 | } | |
453 | return 0; | |
454 | } | |
455 | ||
ef12e72a BF |
456 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
457 | struct flock64 *l) | |
1da177e4 | 458 | { |
1da177e4 | 459 | switch (l->l_whence) { |
f5579f8c | 460 | case SEEK_SET: |
ef12e72a | 461 | fl->fl_start = 0; |
1da177e4 | 462 | break; |
f5579f8c | 463 | case SEEK_CUR: |
ef12e72a | 464 | fl->fl_start = filp->f_pos; |
1da177e4 | 465 | break; |
f5579f8c | 466 | case SEEK_END: |
ef12e72a | 467 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
468 | break; |
469 | default: | |
470 | return -EINVAL; | |
471 | } | |
ef12e72a BF |
472 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
473 | return -EOVERFLOW; | |
474 | fl->fl_start += l->l_start; | |
475 | if (fl->fl_start < 0) | |
476 | return -EINVAL; | |
1da177e4 LT |
477 | |
478 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
479 | POSIX-2001 defines it. */ | |
4c780a46 | 480 | if (l->l_len > 0) { |
ef12e72a BF |
481 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
482 | return -EOVERFLOW; | |
16238415 | 483 | fl->fl_end = fl->fl_start + (l->l_len - 1); |
ef12e72a | 484 | |
4c780a46 | 485 | } else if (l->l_len < 0) { |
ef12e72a | 486 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 487 | return -EINVAL; |
ef12e72a BF |
488 | fl->fl_end = fl->fl_start - 1; |
489 | fl->fl_start += l->l_len; | |
490 | } else | |
491 | fl->fl_end = OFFSET_MAX; | |
492 | ||
4ca52f53 JL |
493 | fl->c.flc_owner = current->files; |
494 | fl->c.flc_pid = current->tgid; | |
495 | fl->c.flc_file = filp; | |
496 | fl->c.flc_flags = FL_POSIX; | |
1da177e4 LT |
497 | fl->fl_ops = NULL; |
498 | fl->fl_lmops = NULL; | |
499 | ||
d9077f7b | 500 | return assign_type(&fl->c, l->l_type); |
1da177e4 LT |
501 | } |
502 | ||
ef12e72a BF |
503 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
504 | * style lock. | |
505 | */ | |
506 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
507 | struct flock *l) | |
1da177e4 | 508 | { |
ef12e72a BF |
509 | struct flock64 ll = { |
510 | .l_type = l->l_type, | |
511 | .l_whence = l->l_whence, | |
512 | .l_start = l->l_start, | |
513 | .l_len = l->l_len, | |
514 | }; | |
515 | ||
516 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 517 | } |
1da177e4 LT |
518 | |
519 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
520 | static bool |
521 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
522 | { |
523 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 524 | return false; |
1da177e4 LT |
525 | } |
526 | ||
1c7dd2ff JL |
527 | static void |
528 | lease_setup(struct file_lock *fl, void **priv) | |
529 | { | |
4ca52f53 | 530 | struct file *filp = fl->c.flc_file; |
1c7dd2ff JL |
531 | struct fasync_struct *fa = *priv; |
532 | ||
533 | /* | |
534 | * fasync_insert_entry() returns the old entry if any. If there was no | |
535 | * old entry, then it used "priv" and inserted it into the fasync list. | |
536 | * Clear the pointer to indicate that it shouldn't be freed. | |
537 | */ | |
538 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
539 | *priv = NULL; | |
540 | ||
01919134 | 541 | __f_setown(filp, task_pid(current), PIDTYPE_TGID, 0); |
1c7dd2ff JL |
542 | } |
543 | ||
7b021967 | 544 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 545 | .lm_break = lease_break_callback, |
8fb47a4f | 546 | .lm_change = lease_modify, |
1c7dd2ff | 547 | .lm_setup = lease_setup, |
1da177e4 LT |
548 | }; |
549 | ||
550 | /* | |
551 | * Initialize a lease, use the default lock manager operations | |
552 | */ | |
ed5f17f6 | 553 | static int lease_init(struct file *filp, int type, struct file_lock *fl) |
447a5647 | 554 | { |
d9077f7b | 555 | if (assign_type(&fl->c, type) != 0) |
75dff55a TM |
556 | return -EINVAL; |
557 | ||
4ca52f53 JL |
558 | fl->c.flc_owner = filp; |
559 | fl->c.flc_pid = current->tgid; | |
1da177e4 | 560 | |
4ca52f53 JL |
561 | fl->c.flc_file = filp; |
562 | fl->c.flc_flags = FL_LEASE; | |
1da177e4 LT |
563 | fl->fl_start = 0; |
564 | fl->fl_end = OFFSET_MAX; | |
565 | fl->fl_ops = NULL; | |
566 | fl->fl_lmops = &lease_manager_ops; | |
567 | return 0; | |
568 | } | |
569 | ||
570 | /* Allocate a file_lock initialised to this type of lease */ | |
ed5f17f6 | 571 | static struct file_lock *lease_alloc(struct file *filp, int type) |
1da177e4 LT |
572 | { |
573 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 574 | int error = -ENOMEM; |
1da177e4 LT |
575 | |
576 | if (fl == NULL) | |
e32b8ee2 | 577 | return ERR_PTR(error); |
1da177e4 LT |
578 | |
579 | error = lease_init(filp, type, fl); | |
75dff55a TM |
580 | if (error) { |
581 | locks_free_lock(fl); | |
e32b8ee2 | 582 | return ERR_PTR(error); |
75dff55a | 583 | } |
e32b8ee2 | 584 | return fl; |
1da177e4 LT |
585 | } |
586 | ||
587 | /* Check if two locks overlap each other. | |
588 | */ | |
589 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
590 | { | |
591 | return ((fl1->fl_end >= fl2->fl_start) && | |
592 | (fl2->fl_end >= fl1->fl_start)); | |
593 | } | |
594 | ||
595 | /* | |
596 | * Check whether two locks have the same owner. | |
597 | */ | |
9bb430a8 | 598 | static int posix_same_owner(struct file_lock_core *fl1, struct file_lock_core *fl2) |
1da177e4 | 599 | { |
9bb430a8 | 600 | return fl1->flc_owner == fl2->flc_owner; |
1da177e4 LT |
601 | } |
602 | ||
6109c850 | 603 | /* Must be called with the flc_lock held! */ |
4629172f | 604 | static void locks_insert_global_locks(struct file_lock_core *flc) |
88974691 | 605 | { |
7c3f654d PZ |
606 | struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list); |
607 | ||
aba37660 PZ |
608 | percpu_rwsem_assert_held(&file_rwsem); |
609 | ||
7c3f654d | 610 | spin_lock(&fll->lock); |
4629172f JL |
611 | flc->flc_link_cpu = smp_processor_id(); |
612 | hlist_add_head(&flc->flc_link, &fll->hlist); | |
7c3f654d | 613 | spin_unlock(&fll->lock); |
88974691 JL |
614 | } |
615 | ||
6109c850 | 616 | /* Must be called with the flc_lock held! */ |
4629172f | 617 | static void locks_delete_global_locks(struct file_lock_core *flc) |
88974691 | 618 | { |
7c3f654d PZ |
619 | struct file_lock_list_struct *fll; |
620 | ||
aba37660 PZ |
621 | percpu_rwsem_assert_held(&file_rwsem); |
622 | ||
7012b02a JL |
623 | /* |
624 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 625 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
626 | * also require that it be held. |
627 | */ | |
4629172f | 628 | if (hlist_unhashed(&flc->flc_link)) |
7012b02a | 629 | return; |
7c3f654d | 630 | |
4629172f | 631 | fll = per_cpu_ptr(&file_lock_list, flc->flc_link_cpu); |
7c3f654d | 632 | spin_lock(&fll->lock); |
4629172f | 633 | hlist_del_init(&flc->flc_link); |
7c3f654d | 634 | spin_unlock(&fll->lock); |
88974691 JL |
635 | } |
636 | ||
3999e493 | 637 | static unsigned long |
ad399740 | 638 | posix_owner_key(struct file_lock_core *flc) |
3999e493 | 639 | { |
ad399740 | 640 | return (unsigned long) flc->flc_owner; |
3999e493 JL |
641 | } |
642 | ||
1a6c75d4 | 643 | static void locks_insert_global_blocked(struct file_lock_core *waiter) |
88974691 | 644 | { |
663d5af7 DW |
645 | lockdep_assert_held(&blocked_lock_lock); |
646 | ||
1a6c75d4 | 647 | hash_add(blocked_hash, &waiter->flc_link, posix_owner_key(waiter)); |
88974691 JL |
648 | } |
649 | ||
1a6c75d4 | 650 | static void locks_delete_global_blocked(struct file_lock_core *waiter) |
88974691 | 651 | { |
663d5af7 DW |
652 | lockdep_assert_held(&blocked_lock_lock); |
653 | ||
1a6c75d4 | 654 | hash_del(&waiter->flc_link); |
88974691 JL |
655 | } |
656 | ||
1da177e4 LT |
657 | /* Remove waiter from blocker's block list. |
658 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 659 | * |
7b2296af | 660 | * Must be called with blocked_lock_lock held. |
1da177e4 | 661 | */ |
269a6194 | 662 | static void __locks_unlink_block(struct file_lock_core *waiter) |
1da177e4 | 663 | { |
1a62c22a JL |
664 | locks_delete_global_blocked(waiter); |
665 | list_del_init(&waiter->flc_blocked_member); | |
1da177e4 LT |
666 | } |
667 | ||
1a62c22a | 668 | static void __locks_wake_up_blocks(struct file_lock_core *blocker) |
ad6bbd8b | 669 | { |
1a62c22a JL |
670 | while (!list_empty(&blocker->flc_blocked_requests)) { |
671 | struct file_lock_core *waiter; | |
672 | struct file_lock *fl; | |
673 | ||
674 | waiter = list_first_entry(&blocker->flc_blocked_requests, | |
675 | struct file_lock_core, flc_blocked_member); | |
ad6bbd8b | 676 | |
1a62c22a | 677 | fl = file_lock(waiter); |
269a6194 | 678 | __locks_unlink_block(waiter); |
1a62c22a JL |
679 | if ((waiter->flc_flags & (FL_POSIX | FL_FLOCK)) && |
680 | fl->fl_lmops && fl->fl_lmops->lm_notify) | |
681 | fl->fl_lmops->lm_notify(fl); | |
ad6bbd8b | 682 | else |
1a62c22a | 683 | locks_wake_up(fl); |
dcf23ac3 LT |
684 | |
685 | /* | |
1a62c22a | 686 | * The setting of flc_blocker to NULL marks the "done" |
dcf23ac3 LT |
687 | * point in deleting a block. Paired with acquire at the top |
688 | * of locks_delete_block(). | |
689 | */ | |
1a62c22a | 690 | smp_store_release(&waiter->flc_blocker, NULL); |
ad6bbd8b N |
691 | } |
692 | } | |
693 | ||
269a6194 | 694 | static int __locks_delete_block(struct file_lock_core *waiter) |
1da177e4 | 695 | { |
cb03f94f N |
696 | int status = -ENOENT; |
697 | ||
dcf23ac3 LT |
698 | /* |
699 | * If fl_blocker is NULL, it won't be set again as this thread "owns" | |
700 | * the lock and is the only one that might try to claim the lock. | |
701 | * | |
702 | * We use acquire/release to manage fl_blocker so that we can | |
703 | * optimize away taking the blocked_lock_lock in many cases. | |
704 | * | |
705 | * The smp_load_acquire guarantees two things: | |
706 | * | |
707 | * 1/ that fl_blocked_requests can be tested locklessly. If something | |
708 | * was recently added to that list it must have been in a locked region | |
709 | * *before* the locked region when fl_blocker was set to NULL. | |
710 | * | |
711 | * 2/ that no other thread is accessing 'waiter', so it is safe to free | |
712 | * it. __locks_wake_up_blocks is careful not to touch waiter after | |
713 | * fl_blocker is released. | |
714 | * | |
715 | * If a lockless check of fl_blocker shows it to be NULL, we know that | |
716 | * no new locks can be inserted into its fl_blocked_requests list, and | |
717 | * can avoid doing anything further if the list is empty. | |
718 | */ | |
e8a166cf JL |
719 | if (!smp_load_acquire(&waiter->flc_blocker) && |
720 | list_empty(&waiter->flc_blocked_requests)) | |
dcf23ac3 LT |
721 | return status; |
722 | ||
7b2296af | 723 | spin_lock(&blocked_lock_lock); |
e8a166cf | 724 | if (waiter->flc_blocker) |
cb03f94f | 725 | status = 0; |
e8a166cf | 726 | __locks_wake_up_blocks(waiter); |
269a6194 | 727 | __locks_unlink_block(waiter); |
dcf23ac3 LT |
728 | |
729 | /* | |
730 | * The setting of fl_blocker to NULL marks the "done" point in deleting | |
731 | * a block. Paired with acquire at the top of this function. | |
732 | */ | |
e8a166cf | 733 | smp_store_release(&waiter->flc_blocker, NULL); |
7b2296af | 734 | spin_unlock(&blocked_lock_lock); |
cb03f94f | 735 | return status; |
1da177e4 | 736 | } |
269a6194 JL |
737 | |
738 | /** | |
739 | * locks_delete_block - stop waiting for a file lock | |
740 | * @waiter: the lock which was waiting | |
741 | * | |
742 | * lockd/nfsd need to disconnect the lock while working on it. | |
743 | */ | |
744 | int locks_delete_block(struct file_lock *waiter) | |
745 | { | |
746 | return __locks_delete_block(&waiter->c); | |
747 | } | |
cb03f94f | 748 | EXPORT_SYMBOL(locks_delete_block); |
1da177e4 LT |
749 | |
750 | /* Insert waiter into blocker's block list. | |
751 | * We use a circular list so that processes can be easily woken up in | |
752 | * the order they blocked. The documentation doesn't require this but | |
753 | * it seems like the reasonable thing to do. | |
1c8c601a | 754 | * |
6109c850 | 755 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
ada5c1da N |
756 | * fl_blocked_requests list itself is protected by the blocked_lock_lock, |
757 | * but by ensuring that the flc_lock is also held on insertions we can avoid | |
758 | * taking the blocked_lock_lock in some cases when we see that the | |
759 | * fl_blocked_requests list is empty. | |
fd7732e0 N |
760 | * |
761 | * Rather than just adding to the list, we check for conflicts with any existing | |
762 | * waiters, and add beneath any waiter that blocks the new waiter. | |
763 | * Thus wakeups don't happen until needed. | |
1da177e4 | 764 | */ |
269a6194 JL |
765 | static void __locks_insert_block(struct file_lock_core *blocker, |
766 | struct file_lock_core *waiter, | |
b6be3714 JL |
767 | bool conflict(struct file_lock_core *, |
768 | struct file_lock_core *)) | |
1da177e4 | 769 | { |
b6be3714 | 770 | struct file_lock_core *flc; |
fd7732e0 | 771 | |
b6be3714 | 772 | BUG_ON(!list_empty(&waiter->flc_blocked_member)); |
fd7732e0 | 773 | new_blocker: |
b6be3714 JL |
774 | list_for_each_entry(flc, &blocker->flc_blocked_requests, flc_blocked_member) |
775 | if (conflict(flc, waiter)) { | |
776 | blocker = flc; | |
fd7732e0 N |
777 | goto new_blocker; |
778 | } | |
b6aaba5b | 779 | waiter->flc_blocker = blocker; |
b6be3714 JL |
780 | list_add_tail(&waiter->flc_blocked_member, |
781 | &blocker->flc_blocked_requests); | |
782 | ||
783 | if ((blocker->flc_flags & (FL_POSIX|FL_OFDLCK)) == (FL_POSIX|FL_OFDLCK)) | |
784 | locks_insert_global_blocked(waiter); | |
5946c431 | 785 | |
b6be3714 | 786 | /* The requests in waiter->flc_blocked are known to conflict with |
5946c431 N |
787 | * waiter, but might not conflict with blocker, or the requests |
788 | * and lock which block it. So they all need to be woken. | |
789 | */ | |
b6be3714 | 790 | __locks_wake_up_blocks(waiter); |
1c8c601a JL |
791 | } |
792 | ||
6109c850 | 793 | /* Must be called with flc_lock held. */ |
269a6194 JL |
794 | static void locks_insert_block(struct file_lock_core *blocker, |
795 | struct file_lock_core *waiter, | |
b6be3714 JL |
796 | bool conflict(struct file_lock_core *, |
797 | struct file_lock_core *)) | |
1c8c601a | 798 | { |
7b2296af | 799 | spin_lock(&blocked_lock_lock); |
fd7732e0 | 800 | __locks_insert_block(blocker, waiter, conflict); |
7b2296af | 801 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
802 | } |
803 | ||
1cb36012 JL |
804 | /* |
805 | * Wake up processes blocked waiting for blocker. | |
806 | * | |
6109c850 | 807 | * Must be called with the inode->flc_lock held! |
1da177e4 | 808 | */ |
347d49fd | 809 | static void locks_wake_up_blocks(struct file_lock_core *blocker) |
1da177e4 | 810 | { |
4e8c765d JL |
811 | /* |
812 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 | 813 | * blocked requests are only added to the list under the flc_lock, and |
ada5c1da N |
814 | * the flc_lock is always held here. Note that removal from the |
815 | * fl_blocked_requests list does not require the flc_lock, so we must | |
816 | * recheck list_empty() after acquiring the blocked_lock_lock. | |
4e8c765d | 817 | */ |
347d49fd | 818 | if (list_empty(&blocker->flc_blocked_requests)) |
4e8c765d JL |
819 | return; |
820 | ||
7b2296af | 821 | spin_lock(&blocked_lock_lock); |
347d49fd | 822 | __locks_wake_up_blocks(blocker); |
7b2296af | 823 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
824 | } |
825 | ||
5263e31e | 826 | static void |
7c18509b | 827 | locks_insert_lock_ctx(struct file_lock_core *fl, struct list_head *before) |
5263e31e | 828 | { |
7c18509b JL |
829 | list_add_tail(&fl->flc_list, before); |
830 | locks_insert_global_locks(fl); | |
5263e31e JL |
831 | } |
832 | ||
8634b51f | 833 | static void |
7c18509b | 834 | locks_unlink_lock_ctx(struct file_lock_core *fl) |
1da177e4 | 835 | { |
7c18509b JL |
836 | locks_delete_global_locks(fl); |
837 | list_del_init(&fl->flc_list); | |
838 | locks_wake_up_blocks(fl); | |
24cbe784 JL |
839 | } |
840 | ||
8634b51f | 841 | static void |
7c18509b | 842 | locks_delete_lock_ctx(struct file_lock_core *fl, struct list_head *dispose) |
24cbe784 | 843 | { |
e084c1bd | 844 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 845 | if (dispose) |
7c18509b | 846 | list_add(&fl->flc_list, dispose); |
ed9814d8 | 847 | else |
7c18509b | 848 | locks_free_lock(file_lock(fl)); |
1da177e4 LT |
849 | } |
850 | ||
851 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
852 | * checks for shared/exclusive status of overlapping locks. | |
853 | */ | |
b6be3714 JL |
854 | static bool locks_conflict(struct file_lock_core *caller_flc, |
855 | struct file_lock_core *sys_flc) | |
1da177e4 | 856 | { |
b6be3714 | 857 | if (sys_flc->flc_type == F_WRLCK) |
c0e15908 | 858 | return true; |
b6be3714 | 859 | if (caller_flc->flc_type == F_WRLCK) |
c0e15908 N |
860 | return true; |
861 | return false; | |
1da177e4 LT |
862 | } |
863 | ||
864 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
865 | * checking before calling the locks_conflict(). | |
866 | */ | |
b6be3714 JL |
867 | static bool posix_locks_conflict(struct file_lock_core *caller_flc, |
868 | struct file_lock_core *sys_flc) | |
1da177e4 | 869 | { |
b6be3714 JL |
870 | struct file_lock *caller_fl = file_lock(caller_flc); |
871 | struct file_lock *sys_fl = file_lock(sys_flc); | |
872 | ||
1da177e4 LT |
873 | /* POSIX locks owned by the same process do not conflict with |
874 | * each other. | |
875 | */ | |
b6be3714 | 876 | if (posix_same_owner(caller_flc, sys_flc)) |
c0e15908 | 877 | return false; |
1da177e4 LT |
878 | |
879 | /* Check whether they overlap */ | |
880 | if (!locks_overlap(caller_fl, sys_fl)) | |
c0e15908 | 881 | return false; |
1da177e4 | 882 | |
b6be3714 | 883 | return locks_conflict(caller_flc, sys_flc); |
1da177e4 LT |
884 | } |
885 | ||
6c9007f6 SS |
886 | /* Determine if lock sys_fl blocks lock caller_fl. Used on xx_GETLK |
887 | * path so checks for additional GETLK-specific things like F_UNLCK. | |
888 | */ | |
889 | static bool posix_test_locks_conflict(struct file_lock *caller_fl, | |
890 | struct file_lock *sys_fl) | |
891 | { | |
b6be3714 JL |
892 | struct file_lock_core *caller = &caller_fl->c; |
893 | struct file_lock_core *sys = &sys_fl->c; | |
894 | ||
6c9007f6 | 895 | /* F_UNLCK checks any locks on the same fd. */ |
75cabec0 | 896 | if (lock_is_unlock(caller_fl)) { |
b6be3714 | 897 | if (!posix_same_owner(caller, sys)) |
6c9007f6 SS |
898 | return false; |
899 | return locks_overlap(caller_fl, sys_fl); | |
900 | } | |
b6be3714 | 901 | return posix_locks_conflict(caller, sys); |
6c9007f6 SS |
902 | } |
903 | ||
1da177e4 LT |
904 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific |
905 | * checking before calling the locks_conflict(). | |
906 | */ | |
b6be3714 JL |
907 | static bool flock_locks_conflict(struct file_lock_core *caller_flc, |
908 | struct file_lock_core *sys_flc) | |
1da177e4 LT |
909 | { |
910 | /* FLOCK locks referring to the same filp do not conflict with | |
911 | * each other. | |
912 | */ | |
b6be3714 | 913 | if (caller_flc->flc_file == sys_flc->flc_file) |
c0e15908 | 914 | return false; |
1da177e4 | 915 | |
b6be3714 | 916 | return locks_conflict(caller_flc, sys_flc); |
1da177e4 LT |
917 | } |
918 | ||
6d34ac19 | 919 | void |
9d6a8c5c | 920 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
921 | { |
922 | struct file_lock *cfl; | |
bd61e0a9 | 923 | struct file_lock_context *ctx; |
c65454a9 | 924 | struct inode *inode = file_inode(filp); |
2443da22 DN |
925 | void *owner; |
926 | void (*func)(void); | |
1da177e4 | 927 | |
401a8b8f | 928 | ctx = locks_inode_context(inode); |
bd61e0a9 | 929 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
4ca52f53 | 930 | fl->c.flc_type = F_UNLCK; |
bd61e0a9 JL |
931 | return; |
932 | } | |
933 | ||
2443da22 | 934 | retry: |
6109c850 | 935 | spin_lock(&ctx->flc_lock); |
4ca52f53 | 936 | list_for_each_entry(cfl, &ctx->flc_posix, c.flc_list) { |
6c9007f6 | 937 | if (!posix_test_locks_conflict(fl, cfl)) |
2443da22 DN |
938 | continue; |
939 | if (cfl->fl_lmops && cfl->fl_lmops->lm_lock_expirable | |
940 | && (*cfl->fl_lmops->lm_lock_expirable)(cfl)) { | |
941 | owner = cfl->fl_lmops->lm_mod_owner; | |
942 | func = cfl->fl_lmops->lm_expire_lock; | |
943 | __module_get(owner); | |
944 | spin_unlock(&ctx->flc_lock); | |
945 | (*func)(); | |
946 | module_put(owner); | |
947 | goto retry; | |
bd61e0a9 | 948 | } |
2443da22 DN |
949 | locks_copy_conflock(fl, cfl); |
950 | goto out; | |
1da177e4 | 951 | } |
4ca52f53 | 952 | fl->c.flc_type = F_UNLCK; |
bd61e0a9 | 953 | out: |
6109c850 | 954 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 955 | return; |
1da177e4 | 956 | } |
1da177e4 LT |
957 | EXPORT_SYMBOL(posix_test_lock); |
958 | ||
b533184f BF |
959 | /* |
960 | * Deadlock detection: | |
961 | * | |
962 | * We attempt to detect deadlocks that are due purely to posix file | |
963 | * locks. | |
1da177e4 | 964 | * |
b533184f BF |
965 | * We assume that a task can be waiting for at most one lock at a time. |
966 | * So for any acquired lock, the process holding that lock may be | |
967 | * waiting on at most one other lock. That lock in turns may be held by | |
968 | * someone waiting for at most one other lock. Given a requested lock | |
969 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
970 | * follow this chain of waiters to ensure we are not about to create a | |
971 | * cycle. | |
1da177e4 | 972 | * |
b533184f BF |
973 | * Since we do this before we ever put a process to sleep on a lock, we |
974 | * are ensured that there is never a cycle; that is what guarantees that | |
975 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 976 | * |
b533184f BF |
977 | * Note: the above assumption may not be true when handling lock |
978 | * requests from a broken NFS client. It may also fail in the presence | |
979 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 980 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 981 | * |
cff2fce5 | 982 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
983 | * Because the owner is not even nominally tied to a thread of |
984 | * execution, the deadlock detection below can't reasonably work well. Just | |
985 | * skip it for those. | |
986 | * | |
cff2fce5 | 987 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
988 | * locks that just checks for the case where two tasks are attempting to |
989 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 990 | */ |
97855b49 BF |
991 | |
992 | #define MAX_DEADLK_ITERATIONS 10 | |
993 | ||
b6be3714 JL |
994 | /* Find a lock that the owner of the given @blocker is blocking on. */ |
995 | static struct file_lock_core *what_owner_is_waiting_for(struct file_lock_core *blocker) | |
b533184f | 996 | { |
b6be3714 | 997 | struct file_lock_core *flc; |
b533184f | 998 | |
b6be3714 JL |
999 | hash_for_each_possible(blocked_hash, flc, flc_link, posix_owner_key(blocker)) { |
1000 | if (posix_same_owner(flc, blocker)) { | |
1001 | while (flc->flc_blocker) | |
b6aaba5b | 1002 | flc = flc->flc_blocker; |
b6be3714 | 1003 | return flc; |
5946c431 | 1004 | } |
b533184f BF |
1005 | } |
1006 | return NULL; | |
1007 | } | |
1008 | ||
7b2296af | 1009 | /* Must be called with the blocked_lock_lock held! */ |
b6be3714 JL |
1010 | static bool posix_locks_deadlock(struct file_lock *caller_fl, |
1011 | struct file_lock *block_fl) | |
1da177e4 | 1012 | { |
b6be3714 JL |
1013 | struct file_lock_core *caller = &caller_fl->c; |
1014 | struct file_lock_core *blocker = &block_fl->c; | |
97855b49 | 1015 | int i = 0; |
1da177e4 | 1016 | |
663d5af7 DW |
1017 | lockdep_assert_held(&blocked_lock_lock); |
1018 | ||
57b65325 JL |
1019 | /* |
1020 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 1021 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 1022 | */ |
b6be3714 JL |
1023 | if (caller->flc_flags & FL_OFDLCK) |
1024 | return false; | |
57b65325 | 1025 | |
b6be3714 | 1026 | while ((blocker = what_owner_is_waiting_for(blocker))) { |
b533184f | 1027 | if (i++ > MAX_DEADLK_ITERATIONS) |
b6be3714 JL |
1028 | return false; |
1029 | if (posix_same_owner(caller, blocker)) | |
1030 | return true; | |
1da177e4 | 1031 | } |
b6be3714 | 1032 | return false; |
1da177e4 LT |
1033 | } |
1034 | ||
1da177e4 | 1035 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 1036 | * after any leases, but before any posix locks. |
f475ae95 TM |
1037 | * |
1038 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1039 | * whether or not a lock was successfully freed by testing the return | |
1040 | * value for -ENOENT. | |
1da177e4 | 1041 | */ |
bcd7f78d | 1042 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 1043 | { |
993dfa87 | 1044 | struct file_lock *new_fl = NULL; |
5263e31e JL |
1045 | struct file_lock *fl; |
1046 | struct file_lock_context *ctx; | |
1da177e4 | 1047 | int error = 0; |
5263e31e | 1048 | bool found = false; |
ed9814d8 | 1049 | LIST_HEAD(dispose); |
1da177e4 | 1050 | |
4ca52f53 | 1051 | ctx = locks_get_lock_context(inode, request->c.flc_type); |
5c1c669a | 1052 | if (!ctx) { |
4ca52f53 | 1053 | if (request->c.flc_type != F_UNLCK) |
5c1c669a | 1054 | return -ENOMEM; |
4ca52f53 | 1055 | return (request->c.flc_flags & FL_EXISTS) ? -ENOENT : 0; |
5c1c669a | 1056 | } |
5263e31e | 1057 | |
4ca52f53 | 1058 | if (!(request->c.flc_flags & FL_ACCESS) && (request->c.flc_type != F_UNLCK)) { |
84d535ad | 1059 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
1060 | if (!new_fl) |
1061 | return -ENOMEM; | |
84d535ad PE |
1062 | } |
1063 | ||
02e525b2 | 1064 | percpu_down_read(&file_rwsem); |
6109c850 | 1065 | spin_lock(&ctx->flc_lock); |
4ca52f53 | 1066 | if (request->c.flc_flags & FL_ACCESS) |
b89f4321 AB |
1067 | goto find_conflict; |
1068 | ||
4ca52f53 JL |
1069 | list_for_each_entry(fl, &ctx->flc_flock, c.flc_list) { |
1070 | if (request->c.flc_file != fl->c.flc_file) | |
1da177e4 | 1071 | continue; |
4ca52f53 | 1072 | if (request->c.flc_type == fl->c.flc_type) |
1da177e4 | 1073 | goto out; |
5263e31e | 1074 | found = true; |
7c18509b | 1075 | locks_delete_lock_ctx(&fl->c, &dispose); |
1da177e4 LT |
1076 | break; |
1077 | } | |
1da177e4 | 1078 | |
75cabec0 | 1079 | if (lock_is_unlock(request)) { |
4ca52f53 | 1080 | if ((request->c.flc_flags & FL_EXISTS) && !found) |
f475ae95 | 1081 | error = -ENOENT; |
993dfa87 | 1082 | goto out; |
f475ae95 | 1083 | } |
1da177e4 | 1084 | |
f07f18dd | 1085 | find_conflict: |
4ca52f53 | 1086 | list_for_each_entry(fl, &ctx->flc_flock, c.flc_list) { |
b6be3714 | 1087 | if (!flock_locks_conflict(&request->c, &fl->c)) |
1da177e4 LT |
1088 | continue; |
1089 | error = -EAGAIN; | |
4ca52f53 | 1090 | if (!(request->c.flc_flags & FL_SLEEP)) |
bde74e4b MS |
1091 | goto out; |
1092 | error = FILE_LOCK_DEFERRED; | |
269a6194 | 1093 | locks_insert_block(&fl->c, &request->c, flock_locks_conflict); |
1da177e4 LT |
1094 | goto out; |
1095 | } | |
4ca52f53 | 1096 | if (request->c.flc_flags & FL_ACCESS) |
f07f18dd | 1097 | goto out; |
993dfa87 | 1098 | locks_copy_lock(new_fl, request); |
5946c431 | 1099 | locks_move_blocks(new_fl, request); |
7c18509b | 1100 | locks_insert_lock_ctx(&new_fl->c, &ctx->flc_flock); |
993dfa87 | 1101 | new_fl = NULL; |
9cedc194 | 1102 | error = 0; |
1da177e4 LT |
1103 | |
1104 | out: | |
6109c850 | 1105 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1106 | percpu_up_read(&file_rwsem); |
993dfa87 TM |
1107 | if (new_fl) |
1108 | locks_free_lock(new_fl); | |
ed9814d8 | 1109 | locks_dispose_list(&dispose); |
c883da31 | 1110 | trace_flock_lock_inode(inode, request, error); |
1da177e4 LT |
1111 | return error; |
1112 | } | |
1113 | ||
b4d629a3 JL |
1114 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
1115 | struct file_lock *conflock) | |
1da177e4 | 1116 | { |
bd61e0a9 | 1117 | struct file_lock *fl, *tmp; |
39005d02 MS |
1118 | struct file_lock *new_fl = NULL; |
1119 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
1120 | struct file_lock *left = NULL; |
1121 | struct file_lock *right = NULL; | |
bd61e0a9 | 1122 | struct file_lock_context *ctx; |
b9746ef8 JL |
1123 | int error; |
1124 | bool added = false; | |
ed9814d8 | 1125 | LIST_HEAD(dispose); |
2443da22 DN |
1126 | void *owner; |
1127 | void (*func)(void); | |
1da177e4 | 1128 | |
4ca52f53 | 1129 | ctx = locks_get_lock_context(inode, request->c.flc_type); |
bd61e0a9 | 1130 | if (!ctx) |
75cabec0 | 1131 | return lock_is_unlock(request) ? 0 : -ENOMEM; |
bd61e0a9 | 1132 | |
1da177e4 LT |
1133 | /* |
1134 | * We may need two file_lock structures for this operation, | |
1135 | * so we get them in advance to avoid races. | |
39005d02 MS |
1136 | * |
1137 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1138 | */ |
4ca52f53 JL |
1139 | if (!(request->c.flc_flags & FL_ACCESS) && |
1140 | (request->c.flc_type != F_UNLCK || | |
39005d02 MS |
1141 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { |
1142 | new_fl = locks_alloc_lock(); | |
1143 | new_fl2 = locks_alloc_lock(); | |
1144 | } | |
1da177e4 | 1145 | |
2443da22 | 1146 | retry: |
02e525b2 | 1147 | percpu_down_read(&file_rwsem); |
6109c850 | 1148 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1149 | /* |
1150 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1151 | * there are any, either return error or put the request on the | |
48f74186 | 1152 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1153 | */ |
4ca52f53 JL |
1154 | if (request->c.flc_type != F_UNLCK) { |
1155 | list_for_each_entry(fl, &ctx->flc_posix, c.flc_list) { | |
b6be3714 | 1156 | if (!posix_locks_conflict(&request->c, &fl->c)) |
1da177e4 | 1157 | continue; |
2443da22 DN |
1158 | if (fl->fl_lmops && fl->fl_lmops->lm_lock_expirable |
1159 | && (*fl->fl_lmops->lm_lock_expirable)(fl)) { | |
1160 | owner = fl->fl_lmops->lm_mod_owner; | |
1161 | func = fl->fl_lmops->lm_expire_lock; | |
1162 | __module_get(owner); | |
1163 | spin_unlock(&ctx->flc_lock); | |
1164 | percpu_up_read(&file_rwsem); | |
1165 | (*func)(); | |
1166 | module_put(owner); | |
1167 | goto retry; | |
1168 | } | |
5842add2 | 1169 | if (conflock) |
3fe0fff1 | 1170 | locks_copy_conflock(conflock, fl); |
1da177e4 | 1171 | error = -EAGAIN; |
4ca52f53 | 1172 | if (!(request->c.flc_flags & FL_SLEEP)) |
1da177e4 | 1173 | goto out; |
1c8c601a JL |
1174 | /* |
1175 | * Deadlock detection and insertion into the blocked | |
1176 | * locks list must be done while holding the same lock! | |
1177 | */ | |
1da177e4 | 1178 | error = -EDEADLK; |
7b2296af | 1179 | spin_lock(&blocked_lock_lock); |
945ab8f6 JL |
1180 | /* |
1181 | * Ensure that we don't find any locks blocked on this | |
1182 | * request during deadlock detection. | |
1183 | */ | |
1a62c22a | 1184 | __locks_wake_up_blocks(&request->c); |
1c8c601a JL |
1185 | if (likely(!posix_locks_deadlock(request, fl))) { |
1186 | error = FILE_LOCK_DEFERRED; | |
269a6194 | 1187 | __locks_insert_block(&fl->c, &request->c, |
fd7732e0 | 1188 | posix_locks_conflict); |
1c8c601a | 1189 | } |
7b2296af | 1190 | spin_unlock(&blocked_lock_lock); |
1da177e4 | 1191 | goto out; |
7bbd1fc0 N |
1192 | } |
1193 | } | |
1da177e4 LT |
1194 | |
1195 | /* If we're just looking for a conflict, we're done. */ | |
1196 | error = 0; | |
4ca52f53 | 1197 | if (request->c.flc_flags & FL_ACCESS) |
1da177e4 LT |
1198 | goto out; |
1199 | ||
bd61e0a9 | 1200 | /* Find the first old lock with the same owner as the new lock */ |
4ca52f53 | 1201 | list_for_each_entry(fl, &ctx->flc_posix, c.flc_list) { |
9bb430a8 | 1202 | if (posix_same_owner(&request->c, &fl->c)) |
bd61e0a9 | 1203 | break; |
1da177e4 LT |
1204 | } |
1205 | ||
1cb36012 | 1206 | /* Process locks with this owner. */ |
4ca52f53 | 1207 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, c.flc_list) { |
9bb430a8 | 1208 | if (!posix_same_owner(&request->c, &fl->c)) |
bd61e0a9 JL |
1209 | break; |
1210 | ||
1211 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
4ca52f53 | 1212 | if (request->c.flc_type == fl->c.flc_type) { |
449231d6 OK |
1213 | /* In all comparisons of start vs end, use |
1214 | * "start - 1" rather than "end + 1". If end | |
1215 | * is OFFSET_MAX, end + 1 will become negative. | |
1216 | */ | |
1da177e4 | 1217 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1218 | continue; |
1da177e4 LT |
1219 | /* If the next lock in the list has entirely bigger |
1220 | * addresses than the new one, insert the lock here. | |
1221 | */ | |
449231d6 | 1222 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1223 | break; |
1224 | ||
1225 | /* If we come here, the new and old lock are of the | |
1226 | * same type and adjacent or overlapping. Make one | |
1227 | * lock yielding from the lower start address of both | |
1228 | * locks to the higher end address. | |
1229 | */ | |
1230 | if (fl->fl_start > request->fl_start) | |
1231 | fl->fl_start = request->fl_start; | |
1232 | else | |
1233 | request->fl_start = fl->fl_start; | |
1234 | if (fl->fl_end < request->fl_end) | |
1235 | fl->fl_end = request->fl_end; | |
1236 | else | |
1237 | request->fl_end = fl->fl_end; | |
1238 | if (added) { | |
7c18509b | 1239 | locks_delete_lock_ctx(&fl->c, &dispose); |
1da177e4 LT |
1240 | continue; |
1241 | } | |
1242 | request = fl; | |
b9746ef8 | 1243 | added = true; |
bd61e0a9 | 1244 | } else { |
1da177e4 LT |
1245 | /* Processing for different lock types is a bit |
1246 | * more complex. | |
1247 | */ | |
1248 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1249 | continue; |
1da177e4 LT |
1250 | if (fl->fl_start > request->fl_end) |
1251 | break; | |
75cabec0 | 1252 | if (lock_is_unlock(request)) |
b9746ef8 | 1253 | added = true; |
1da177e4 LT |
1254 | if (fl->fl_start < request->fl_start) |
1255 | left = fl; | |
1256 | /* If the next lock in the list has a higher end | |
1257 | * address than the new one, insert the new one here. | |
1258 | */ | |
1259 | if (fl->fl_end > request->fl_end) { | |
1260 | right = fl; | |
1261 | break; | |
1262 | } | |
1263 | if (fl->fl_start >= request->fl_start) { | |
1264 | /* The new lock completely replaces an old | |
1265 | * one (This may happen several times). | |
1266 | */ | |
1267 | if (added) { | |
7c18509b | 1268 | locks_delete_lock_ctx(&fl->c, &dispose); |
1da177e4 LT |
1269 | continue; |
1270 | } | |
b84d49f9 JL |
1271 | /* |
1272 | * Replace the old lock with new_fl, and | |
1273 | * remove the old one. It's safe to do the | |
1274 | * insert here since we know that we won't be | |
1275 | * using new_fl later, and that the lock is | |
1276 | * just replacing an existing lock. | |
1da177e4 | 1277 | */ |
b84d49f9 JL |
1278 | error = -ENOLCK; |
1279 | if (!new_fl) | |
1280 | goto out; | |
1281 | locks_copy_lock(new_fl, request); | |
5ef15968 | 1282 | locks_move_blocks(new_fl, request); |
b84d49f9 JL |
1283 | request = new_fl; |
1284 | new_fl = NULL; | |
7c18509b | 1285 | locks_insert_lock_ctx(&request->c, |
4ca52f53 | 1286 | &fl->c.flc_list); |
7c18509b | 1287 | locks_delete_lock_ctx(&fl->c, &dispose); |
b9746ef8 | 1288 | added = true; |
1da177e4 LT |
1289 | } |
1290 | } | |
1da177e4 LT |
1291 | } |
1292 | ||
0d9a490a | 1293 | /* |
1cb36012 JL |
1294 | * The above code only modifies existing locks in case of merging or |
1295 | * replacing. If new lock(s) need to be inserted all modifications are | |
1296 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1297 | */ |
1298 | error = -ENOLCK; /* "no luck" */ | |
1299 | if (right && left == right && !new_fl2) | |
1300 | goto out; | |
1301 | ||
1da177e4 LT |
1302 | error = 0; |
1303 | if (!added) { | |
75cabec0 | 1304 | if (lock_is_unlock(request)) { |
4ca52f53 | 1305 | if (request->c.flc_flags & FL_EXISTS) |
f475ae95 | 1306 | error = -ENOENT; |
1da177e4 | 1307 | goto out; |
f475ae95 | 1308 | } |
0d9a490a MS |
1309 | |
1310 | if (!new_fl) { | |
1311 | error = -ENOLCK; | |
1312 | goto out; | |
1313 | } | |
1da177e4 | 1314 | locks_copy_lock(new_fl, request); |
5946c431 | 1315 | locks_move_blocks(new_fl, request); |
7c18509b | 1316 | locks_insert_lock_ctx(&new_fl->c, &fl->c.flc_list); |
2e2f756f | 1317 | fl = new_fl; |
1da177e4 LT |
1318 | new_fl = NULL; |
1319 | } | |
1320 | if (right) { | |
1321 | if (left == right) { | |
1322 | /* The new lock breaks the old one in two pieces, | |
1323 | * so we have to use the second new lock. | |
1324 | */ | |
1325 | left = new_fl2; | |
1326 | new_fl2 = NULL; | |
1327 | locks_copy_lock(left, right); | |
7c18509b | 1328 | locks_insert_lock_ctx(&left->c, &fl->c.flc_list); |
1da177e4 LT |
1329 | } |
1330 | right->fl_start = request->fl_end + 1; | |
347d49fd | 1331 | locks_wake_up_blocks(&right->c); |
1da177e4 LT |
1332 | } |
1333 | if (left) { | |
1334 | left->fl_end = request->fl_start - 1; | |
347d49fd | 1335 | locks_wake_up_blocks(&left->c); |
1da177e4 LT |
1336 | } |
1337 | out: | |
6109c850 | 1338 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1339 | percpu_up_read(&file_rwsem); |
74f6f591 | 1340 | trace_posix_lock_inode(inode, request, error); |
1da177e4 LT |
1341 | /* |
1342 | * Free any unused locks. | |
1343 | */ | |
1344 | if (new_fl) | |
1345 | locks_free_lock(new_fl); | |
1346 | if (new_fl2) | |
1347 | locks_free_lock(new_fl2); | |
ed9814d8 | 1348 | locks_dispose_list(&dispose); |
1890910f | 1349 | |
1da177e4 LT |
1350 | return error; |
1351 | } | |
1352 | ||
1353 | /** | |
1354 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1355 | * @filp: The file to apply the lock to | |
1356 | * @fl: The lock to be applied | |
150b3934 | 1357 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1358 | * |
1359 | * Add a POSIX style lock to a file. | |
1360 | * We merge adjacent & overlapping locks whenever possible. | |
1361 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1362 | * |
1363 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1364 | * whether or not a lock was successfully freed by testing the return | |
1365 | * value for -ENOENT. | |
1da177e4 | 1366 | */ |
150b3934 | 1367 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1368 | struct file_lock *conflock) |
1369 | { | |
c65454a9 | 1370 | return posix_lock_inode(file_inode(filp), fl, conflock); |
1da177e4 | 1371 | } |
150b3934 | 1372 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1373 | |
1374 | /** | |
29d01b22 JL |
1375 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1376 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1377 | * @fl: The lock to be applied |
1378 | * | |
616fb38f | 1379 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1380 | */ |
616fb38f | 1381 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1382 | { |
1383 | int error; | |
1384 | might_sleep (); | |
1385 | for (;;) { | |
b4d629a3 | 1386 | error = posix_lock_inode(inode, fl, NULL); |
bde74e4b | 1387 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1388 | break; |
4ca52f53 JL |
1389 | error = wait_event_interruptible(fl->c.flc_wait, |
1390 | list_empty(&fl->c.flc_blocked_member)); | |
16306a61 N |
1391 | if (error) |
1392 | break; | |
1da177e4 | 1393 | } |
16306a61 | 1394 | locks_delete_block(fl); |
1da177e4 LT |
1395 | return error; |
1396 | } | |
29d01b22 | 1397 | |
778fc546 BF |
1398 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1399 | { | |
1400 | switch (arg) { | |
1401 | case F_UNLCK: | |
4ca52f53 | 1402 | fl->c.flc_flags &= ~FL_UNLOCK_PENDING; |
df561f66 | 1403 | fallthrough; |
778fc546 | 1404 | case F_RDLCK: |
4ca52f53 | 1405 | fl->c.flc_flags &= ~FL_DOWNGRADE_PENDING; |
778fc546 BF |
1406 | } |
1407 | } | |
1408 | ||
1da177e4 | 1409 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1410 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1411 | { |
d9077f7b | 1412 | int error = assign_type(&fl->c, arg); |
1da177e4 LT |
1413 | |
1414 | if (error) | |
1415 | return error; | |
778fc546 | 1416 | lease_clear_pending(fl, arg); |
347d49fd | 1417 | locks_wake_up_blocks(&fl->c); |
3b6e2723 | 1418 | if (arg == F_UNLCK) { |
4ca52f53 | 1419 | struct file *filp = fl->c.flc_file; |
3b6e2723 FB |
1420 | |
1421 | f_delown(filp); | |
1422 | filp->f_owner.signum = 0; | |
4ca52f53 | 1423 | fasync_helper(0, fl->c.flc_file, 0, &fl->fl_fasync); |
96d6d59c BF |
1424 | if (fl->fl_fasync != NULL) { |
1425 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1426 | fl->fl_fasync = NULL; | |
1427 | } | |
7c18509b | 1428 | locks_delete_lock_ctx(&fl->c, dispose); |
3b6e2723 | 1429 | } |
1da177e4 LT |
1430 | return 0; |
1431 | } | |
1da177e4 LT |
1432 | EXPORT_SYMBOL(lease_modify); |
1433 | ||
778fc546 BF |
1434 | static bool past_time(unsigned long then) |
1435 | { | |
1436 | if (!then) | |
1437 | /* 0 is a special value meaning "this never expires": */ | |
1438 | return false; | |
1439 | return time_after(jiffies, then); | |
1440 | } | |
1441 | ||
c45198ed | 1442 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1443 | { |
8634b51f JL |
1444 | struct file_lock_context *ctx = inode->i_flctx; |
1445 | struct file_lock *fl, *tmp; | |
1da177e4 | 1446 | |
6109c850 | 1447 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1448 | |
4ca52f53 | 1449 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, c.flc_list) { |
62af4f1f | 1450 | trace_time_out_leases(inode, fl); |
778fc546 | 1451 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1452 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1453 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1454 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1455 | } |
1456 | } | |
1457 | ||
b6be3714 | 1458 | static bool leases_conflict(struct file_lock_core *lc, struct file_lock_core *bc) |
df4e8d2c | 1459 | { |
d51f527f | 1460 | bool rc; |
b6be3714 JL |
1461 | struct file_lock *lease = file_lock(lc); |
1462 | struct file_lock *breaker = file_lock(bc); | |
d51f527f | 1463 | |
28df3d15 BF |
1464 | if (lease->fl_lmops->lm_breaker_owns_lease |
1465 | && lease->fl_lmops->lm_breaker_owns_lease(lease)) | |
1466 | return false; | |
b6be3714 | 1467 | if ((bc->flc_flags & FL_LAYOUT) != (lc->flc_flags & FL_LAYOUT)) { |
d51f527f IW |
1468 | rc = false; |
1469 | goto trace; | |
1470 | } | |
b6be3714 | 1471 | if ((bc->flc_flags & FL_DELEG) && (lc->flc_flags & FL_LEASE)) { |
d51f527f IW |
1472 | rc = false; |
1473 | goto trace; | |
1474 | } | |
1475 | ||
b6be3714 | 1476 | rc = locks_conflict(bc, lc); |
d51f527f IW |
1477 | trace: |
1478 | trace_leases_conflict(rc, lease, breaker); | |
1479 | return rc; | |
df4e8d2c BF |
1480 | } |
1481 | ||
03d12ddf JL |
1482 | static bool |
1483 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1484 | { | |
8634b51f | 1485 | struct file_lock_context *ctx = inode->i_flctx; |
b6be3714 | 1486 | struct file_lock_core *flc; |
03d12ddf | 1487 | |
6109c850 | 1488 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1489 | |
b6be3714 JL |
1490 | list_for_each_entry(flc, &ctx->flc_lease, flc_list) { |
1491 | if (leases_conflict(flc, &breaker->c)) | |
03d12ddf JL |
1492 | return true; |
1493 | } | |
1494 | return false; | |
1495 | } | |
1496 | ||
1da177e4 LT |
1497 | /** |
1498 | * __break_lease - revoke all outstanding leases on file | |
1499 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1500 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1501 | * break all leases | |
1502 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1503 | * only delegations | |
1da177e4 | 1504 | * |
87250dd2 | 1505 | * break_lease (inlined for speed) has checked there already is at least |
1506 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1507 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1508 | * specified %O_NONBLOCK to your open(). |
1509 | */ | |
df4e8d2c | 1510 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1511 | { |
778fc546 | 1512 | int error = 0; |
128a3785 | 1513 | struct file_lock_context *ctx; |
a901125c | 1514 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1515 | unsigned long break_time; |
8737c930 | 1516 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1517 | LIST_HEAD(dispose); |
1da177e4 | 1518 | |
8737c930 | 1519 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1520 | if (IS_ERR(new_fl)) |
1521 | return PTR_ERR(new_fl); | |
4ca52f53 | 1522 | new_fl->c.flc_flags = type; |
1da177e4 | 1523 | |
8634b51f | 1524 | /* typically we will check that ctx is non-NULL before calling */ |
401a8b8f | 1525 | ctx = locks_inode_context(inode); |
8634b51f JL |
1526 | if (!ctx) { |
1527 | WARN_ON_ONCE(1); | |
cfddf9f4 | 1528 | goto free_lock; |
8634b51f JL |
1529 | } |
1530 | ||
02e525b2 | 1531 | percpu_down_read(&file_rwsem); |
6109c850 | 1532 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1533 | |
c45198ed | 1534 | time_out_leases(inode, &dispose); |
1da177e4 | 1535 | |
03d12ddf | 1536 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1537 | goto out; |
1538 | ||
1da177e4 LT |
1539 | break_time = 0; |
1540 | if (lease_break_time > 0) { | |
1541 | break_time = jiffies + lease_break_time * HZ; | |
1542 | if (break_time == 0) | |
1543 | break_time++; /* so that 0 means no break time */ | |
1544 | } | |
1545 | ||
4ca52f53 | 1546 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, c.flc_list) { |
b6be3714 | 1547 | if (!leases_conflict(&fl->c, &new_fl->c)) |
df4e8d2c | 1548 | continue; |
778fc546 | 1549 | if (want_write) { |
4ca52f53 | 1550 | if (fl->c.flc_flags & FL_UNLOCK_PENDING) |
778fc546 | 1551 | continue; |
4ca52f53 | 1552 | fl->c.flc_flags |= FL_UNLOCK_PENDING; |
1da177e4 | 1553 | fl->fl_break_time = break_time; |
778fc546 | 1554 | } else { |
8634b51f | 1555 | if (lease_breaking(fl)) |
778fc546 | 1556 | continue; |
4ca52f53 | 1557 | fl->c.flc_flags |= FL_DOWNGRADE_PENDING; |
778fc546 | 1558 | fl->fl_downgrade_time = break_time; |
1da177e4 | 1559 | } |
4d01b7f5 | 1560 | if (fl->fl_lmops->lm_break(fl)) |
7c18509b | 1561 | locks_delete_lock_ctx(&fl->c, &dispose); |
1da177e4 LT |
1562 | } |
1563 | ||
8634b51f | 1564 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1565 | goto out; |
1566 | ||
843c6b2f | 1567 | if (mode & O_NONBLOCK) { |
62af4f1f | 1568 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1569 | error = -EWOULDBLOCK; |
1570 | goto out; | |
1571 | } | |
1572 | ||
1573 | restart: | |
4ca52f53 | 1574 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, c.flc_list); |
8634b51f | 1575 | break_time = fl->fl_break_time; |
f1c6bb2c | 1576 | if (break_time != 0) |
1da177e4 | 1577 | break_time -= jiffies; |
f1c6bb2c JL |
1578 | if (break_time == 0) |
1579 | break_time++; | |
269a6194 | 1580 | locks_insert_block(&fl->c, &new_fl->c, leases_conflict); |
62af4f1f | 1581 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1582 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1583 | percpu_up_read(&file_rwsem); |
aba37660 | 1584 | |
c45198ed | 1585 | locks_dispose_list(&dispose); |
4ca52f53 JL |
1586 | error = wait_event_interruptible_timeout(new_fl->c.flc_wait, |
1587 | list_empty(&new_fl->c.flc_blocked_member), | |
1588 | break_time); | |
aba37660 | 1589 | |
02e525b2 | 1590 | percpu_down_read(&file_rwsem); |
6109c850 | 1591 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1592 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1593 | locks_delete_block(new_fl); |
1da177e4 | 1594 | if (error >= 0) { |
778fc546 BF |
1595 | /* |
1596 | * Wait for the next conflicting lease that has not been | |
1597 | * broken yet | |
1598 | */ | |
03d12ddf JL |
1599 | if (error == 0) |
1600 | time_out_leases(inode, &dispose); | |
1601 | if (any_leases_conflict(inode, new_fl)) | |
1602 | goto restart; | |
1da177e4 LT |
1603 | error = 0; |
1604 | } | |
1da177e4 | 1605 | out: |
6109c850 | 1606 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1607 | percpu_up_read(&file_rwsem); |
c45198ed | 1608 | locks_dispose_list(&dispose); |
cfddf9f4 | 1609 | free_lock: |
6d4b9e38 | 1610 | locks_free_lock(new_fl); |
1da177e4 LT |
1611 | return error; |
1612 | } | |
1da177e4 LT |
1613 | EXPORT_SYMBOL(__break_lease); |
1614 | ||
1615 | /** | |
76c47948 | 1616 | * lease_get_mtime - update modified time of an inode with exclusive lease |
1da177e4 | 1617 | * @inode: the inode |
76c47948 | 1618 | * @time: pointer to a timespec which contains the last modified time |
1da177e4 LT |
1619 | * |
1620 | * This is to force NFS clients to flush their caches for files with | |
1621 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1622 | * exclusive lease, then they could be modifying it. |
1da177e4 | 1623 | */ |
95582b00 | 1624 | void lease_get_mtime(struct inode *inode, struct timespec64 *time) |
1da177e4 | 1625 | { |
bfe86024 | 1626 | bool has_lease = false; |
128a3785 | 1627 | struct file_lock_context *ctx; |
8634b51f | 1628 | struct file_lock *fl; |
bfe86024 | 1629 | |
401a8b8f | 1630 | ctx = locks_inode_context(inode); |
8634b51f | 1631 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1632 | spin_lock(&ctx->flc_lock); |
8ace5dfb | 1633 | fl = list_first_entry_or_null(&ctx->flc_lease, |
4ca52f53 | 1634 | struct file_lock, c.flc_list); |
75cabec0 | 1635 | if (fl && lock_is_write(fl)) |
8ace5dfb | 1636 | has_lease = true; |
6109c850 | 1637 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1638 | } |
1639 | ||
1640 | if (has_lease) | |
c2050a45 | 1641 | *time = current_time(inode); |
1da177e4 | 1642 | } |
1da177e4 LT |
1643 | EXPORT_SYMBOL(lease_get_mtime); |
1644 | ||
1645 | /** | |
1646 | * fcntl_getlease - Enquire what lease is currently active | |
1647 | * @filp: the file | |
1648 | * | |
1649 | * The value returned by this function will be one of | |
1650 | * (if no lease break is pending): | |
1651 | * | |
1652 | * %F_RDLCK to indicate a shared lease is held. | |
1653 | * | |
1654 | * %F_WRLCK to indicate an exclusive lease is held. | |
1655 | * | |
1656 | * %F_UNLCK to indicate no lease is held. | |
1657 | * | |
1658 | * (if a lease break is pending): | |
1659 | * | |
1660 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1661 | * changed to a shared lease (or removed). | |
1662 | * | |
1663 | * %F_UNLCK to indicate the lease needs to be removed. | |
1664 | * | |
1665 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1666 | * should be returned to userspace. | |
1667 | */ | |
1668 | int fcntl_getlease(struct file *filp) | |
1669 | { | |
1670 | struct file_lock *fl; | |
c65454a9 | 1671 | struct inode *inode = file_inode(filp); |
128a3785 | 1672 | struct file_lock_context *ctx; |
1da177e4 | 1673 | int type = F_UNLCK; |
c45198ed | 1674 | LIST_HEAD(dispose); |
1da177e4 | 1675 | |
401a8b8f | 1676 | ctx = locks_inode_context(inode); |
8634b51f | 1677 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
02e525b2 | 1678 | percpu_down_read(&file_rwsem); |
6109c850 | 1679 | spin_lock(&ctx->flc_lock); |
c568d683 | 1680 | time_out_leases(inode, &dispose); |
4ca52f53 JL |
1681 | list_for_each_entry(fl, &ctx->flc_lease, c.flc_list) { |
1682 | if (fl->c.flc_file != filp) | |
8634b51f | 1683 | continue; |
778fc546 | 1684 | type = target_leasetype(fl); |
1da177e4 LT |
1685 | break; |
1686 | } | |
6109c850 | 1687 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1688 | percpu_up_read(&file_rwsem); |
5f43086b | 1689 | |
8634b51f | 1690 | locks_dispose_list(&dispose); |
1da177e4 | 1691 | } |
1da177e4 LT |
1692 | return type; |
1693 | } | |
1694 | ||
24cbe784 | 1695 | /** |
387e3746 | 1696 | * check_conflicting_open - see if the given file points to an inode that has |
7bbd1fc0 N |
1697 | * an existing open that would conflict with the |
1698 | * desired lease. | |
387e3746 | 1699 | * @filp: file to check |
24cbe784 | 1700 | * @arg: type of lease that we're trying to acquire |
7fadc59c | 1701 | * @flags: current lock flags |
24cbe784 JL |
1702 | * |
1703 | * Check to see if there's an existing open fd on this file that would | |
1704 | * conflict with the lease we're trying to set. | |
1705 | */ | |
1706 | static int | |
ed5f17f6 | 1707 | check_conflicting_open(struct file *filp, const int arg, int flags) |
24cbe784 | 1708 | { |
c65454a9 | 1709 | struct inode *inode = file_inode(filp); |
387e3746 | 1710 | int self_wcount = 0, self_rcount = 0; |
24cbe784 | 1711 | |
11afe9f7 CH |
1712 | if (flags & FL_LAYOUT) |
1713 | return 0; | |
aba2072f BF |
1714 | if (flags & FL_DELEG) |
1715 | /* We leave these checks to the caller */ | |
1716 | return 0; | |
11afe9f7 | 1717 | |
387e3746 AG |
1718 | if (arg == F_RDLCK) |
1719 | return inode_is_open_for_write(inode) ? -EAGAIN : 0; | |
1720 | else if (arg != F_WRLCK) | |
1721 | return 0; | |
1722 | ||
1723 | /* | |
1724 | * Make sure that only read/write count is from lease requestor. | |
1725 | * Note that this will result in denying write leases when i_writecount | |
1726 | * is negative, which is what we want. (We shouldn't grant write leases | |
1727 | * on files open for execution.) | |
1728 | */ | |
1729 | if (filp->f_mode & FMODE_WRITE) | |
1730 | self_wcount = 1; | |
1731 | else if (filp->f_mode & FMODE_READ) | |
1732 | self_rcount = 1; | |
24cbe784 | 1733 | |
387e3746 AG |
1734 | if (atomic_read(&inode->i_writecount) != self_wcount || |
1735 | atomic_read(&inode->i_readcount) != self_rcount) | |
1736 | return -EAGAIN; | |
24cbe784 | 1737 | |
387e3746 | 1738 | return 0; |
24cbe784 JL |
1739 | } |
1740 | ||
e6f5c789 | 1741 | static int |
ed5f17f6 | 1742 | generic_add_lease(struct file *filp, int arg, struct file_lock **flp, void **priv) |
1da177e4 | 1743 | { |
8634b51f | 1744 | struct file_lock *fl, *my_fl = NULL, *lease; |
c65454a9 | 1745 | struct inode *inode = file_inode(filp); |
8634b51f | 1746 | struct file_lock_context *ctx; |
4ca52f53 | 1747 | bool is_deleg = (*flp)->c.flc_flags & FL_DELEG; |
c1f24ef4 | 1748 | int error; |
c45198ed | 1749 | LIST_HEAD(dispose); |
1da177e4 | 1750 | |
096657b6 | 1751 | lease = *flp; |
62af4f1f JL |
1752 | trace_generic_add_lease(inode, lease); |
1753 | ||
5c1c669a JL |
1754 | /* Note that arg is never F_UNLCK here */ |
1755 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1756 | if (!ctx) |
1757 | return -ENOMEM; | |
1758 | ||
df4e8d2c BF |
1759 | /* |
1760 | * In the delegation case we need mutual exclusion with | |
1761 | * a number of operations that take the i_mutex. We trylock | |
1762 | * because delegations are an optional optimization, and if | |
1763 | * there's some chance of a conflict--we'd rather not | |
1764 | * bother, maybe that's a sign this just isn't a good file to | |
1765 | * hand out a delegation on. | |
1766 | */ | |
5955102c | 1767 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1768 | return -EAGAIN; |
1769 | ||
02e525b2 | 1770 | percpu_down_read(&file_rwsem); |
6109c850 | 1771 | spin_lock(&ctx->flc_lock); |
c45198ed | 1772 | time_out_leases(inode, &dispose); |
4ca52f53 | 1773 | error = check_conflicting_open(filp, arg, lease->c.flc_flags); |
24cbe784 | 1774 | if (error) |
096657b6 | 1775 | goto out; |
6d5e8b05 | 1776 | |
1da177e4 LT |
1777 | /* |
1778 | * At this point, we know that if there is an exclusive | |
1779 | * lease on this file, then we hold it on this filp | |
1780 | * (otherwise our open of this file would have blocked). | |
1781 | * And if we are trying to acquire an exclusive lease, | |
1782 | * then the file is not open by anyone (including us) | |
1783 | * except for this filp. | |
1784 | */ | |
c1f24ef4 | 1785 | error = -EAGAIN; |
4ca52f53 JL |
1786 | list_for_each_entry(fl, &ctx->flc_lease, c.flc_list) { |
1787 | if (fl->c.flc_file == filp && | |
1788 | fl->c.flc_owner == lease->c.flc_owner) { | |
8634b51f | 1789 | my_fl = fl; |
c1f24ef4 BF |
1790 | continue; |
1791 | } | |
8634b51f | 1792 | |
c1f24ef4 BF |
1793 | /* |
1794 | * No exclusive leases if someone else has a lease on | |
1795 | * this file: | |
1796 | */ | |
1797 | if (arg == F_WRLCK) | |
1798 | goto out; | |
1799 | /* | |
1800 | * Modifying our existing lease is OK, but no getting a | |
1801 | * new lease if someone else is opening for write: | |
1802 | */ | |
4ca52f53 | 1803 | if (fl->c.flc_flags & FL_UNLOCK_PENDING) |
c1f24ef4 | 1804 | goto out; |
1da177e4 LT |
1805 | } |
1806 | ||
8634b51f | 1807 | if (my_fl != NULL) { |
0164bf02 JL |
1808 | lease = my_fl; |
1809 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1810 | if (error) |
1811 | goto out; | |
1812 | goto out_setup; | |
1da177e4 LT |
1813 | } |
1814 | ||
1da177e4 LT |
1815 | error = -EINVAL; |
1816 | if (!leases_enable) | |
1817 | goto out; | |
1818 | ||
7c18509b | 1819 | locks_insert_lock_ctx(&lease->c, &ctx->flc_lease); |
24cbe784 JL |
1820 | /* |
1821 | * The check in break_lease() is lockless. It's possible for another | |
1822 | * open to race in after we did the earlier check for a conflicting | |
1823 | * open but before the lease was inserted. Check again for a | |
1824 | * conflicting open and cancel the lease if there is one. | |
1825 | * | |
1826 | * We also add a barrier here to ensure that the insertion of the lock | |
1827 | * precedes these checks. | |
1828 | */ | |
1829 | smp_mb(); | |
4ca52f53 | 1830 | error = check_conflicting_open(filp, arg, lease->c.flc_flags); |
8634b51f | 1831 | if (error) { |
7c18509b | 1832 | locks_unlink_lock_ctx(&lease->c); |
8634b51f JL |
1833 | goto out; |
1834 | } | |
1c7dd2ff JL |
1835 | |
1836 | out_setup: | |
1837 | if (lease->fl_lmops->lm_setup) | |
1838 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1839 | out: |
6109c850 | 1840 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1841 | percpu_up_read(&file_rwsem); |
c45198ed | 1842 | locks_dispose_list(&dispose); |
df4e8d2c | 1843 | if (is_deleg) |
5955102c | 1844 | inode_unlock(inode); |
8634b51f | 1845 | if (!error && !my_fl) |
1c7dd2ff | 1846 | *flp = NULL; |
1da177e4 LT |
1847 | return error; |
1848 | } | |
8335ebd9 | 1849 | |
2ab99ee1 | 1850 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1851 | { |
0efaa7e8 | 1852 | int error = -EAGAIN; |
8634b51f | 1853 | struct file_lock *fl, *victim = NULL; |
c65454a9 | 1854 | struct inode *inode = file_inode(filp); |
128a3785 | 1855 | struct file_lock_context *ctx; |
c45198ed | 1856 | LIST_HEAD(dispose); |
8335ebd9 | 1857 | |
401a8b8f | 1858 | ctx = locks_inode_context(inode); |
8634b51f JL |
1859 | if (!ctx) { |
1860 | trace_generic_delete_lease(inode, NULL); | |
1861 | return error; | |
1862 | } | |
1863 | ||
02e525b2 | 1864 | percpu_down_read(&file_rwsem); |
6109c850 | 1865 | spin_lock(&ctx->flc_lock); |
4ca52f53 JL |
1866 | list_for_each_entry(fl, &ctx->flc_lease, c.flc_list) { |
1867 | if (fl->c.flc_file == filp && | |
1868 | fl->c.flc_owner == owner) { | |
8634b51f | 1869 | victim = fl; |
0efaa7e8 | 1870 | break; |
8634b51f | 1871 | } |
8335ebd9 | 1872 | } |
a9b1b455 | 1873 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1874 | if (victim) |
7448cc37 | 1875 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1876 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1877 | percpu_up_read(&file_rwsem); |
c45198ed | 1878 | locks_dispose_list(&dispose); |
0efaa7e8 | 1879 | return error; |
8335ebd9 BF |
1880 | } |
1881 | ||
1882 | /** | |
1883 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1884 | * @filp: file pointer |
1885 | * @arg: type of lease to obtain | |
1886 | * @flp: input - file_lock to use, output - file_lock inserted | |
1887 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1888 | * doesn't require it) | |
8335ebd9 BF |
1889 | * |
1890 | * The (input) flp->fl_lmops->lm_break function is required | |
1891 | * by break_lease(). | |
8335ebd9 | 1892 | */ |
ed5f17f6 | 1893 | int generic_setlease(struct file *filp, int arg, struct file_lock **flp, |
e6f5c789 | 1894 | void **priv) |
8335ebd9 | 1895 | { |
c65454a9 | 1896 | struct inode *inode = file_inode(filp); |
42d0c4bd | 1897 | vfsuid_t vfsuid = i_uid_into_vfsuid(file_mnt_idmap(filp), inode); |
8335ebd9 BF |
1898 | int error; |
1899 | ||
42d0c4bd | 1900 | if ((!vfsuid_eq_kuid(vfsuid, current_fsuid())) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1901 | return -EACCES; |
1902 | if (!S_ISREG(inode->i_mode)) | |
1903 | return -EINVAL; | |
1904 | error = security_file_lock(filp, arg); | |
1905 | if (error) | |
1906 | return error; | |
1907 | ||
8335ebd9 BF |
1908 | switch (arg) { |
1909 | case F_UNLCK: | |
2ab99ee1 | 1910 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1911 | case F_RDLCK: |
1912 | case F_WRLCK: | |
0efaa7e8 JL |
1913 | if (!(*flp)->fl_lmops->lm_break) { |
1914 | WARN_ON_ONCE(1); | |
1915 | return -ENOLCK; | |
1916 | } | |
11afe9f7 | 1917 | |
e6f5c789 | 1918 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1919 | default: |
8d657eb3 | 1920 | return -EINVAL; |
8335ebd9 BF |
1921 | } |
1922 | } | |
0af1a450 | 1923 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1924 | |
18f6622e JL |
1925 | /* |
1926 | * Kernel subsystems can register to be notified on any attempt to set | |
1927 | * a new lease with the lease_notifier_chain. This is used by (e.g.) nfsd | |
1928 | * to close files that it may have cached when there is an attempt to set a | |
1929 | * conflicting lease. | |
1930 | */ | |
1931 | static struct srcu_notifier_head lease_notifier_chain; | |
1932 | ||
1933 | static inline void | |
1934 | lease_notifier_chain_init(void) | |
1935 | { | |
1936 | srcu_init_notifier_head(&lease_notifier_chain); | |
1937 | } | |
1938 | ||
1939 | static inline void | |
ed5f17f6 | 1940 | setlease_notifier(int arg, struct file_lock *lease) |
18f6622e JL |
1941 | { |
1942 | if (arg != F_UNLCK) | |
1943 | srcu_notifier_call_chain(&lease_notifier_chain, arg, lease); | |
1944 | } | |
1945 | ||
1946 | int lease_register_notifier(struct notifier_block *nb) | |
1947 | { | |
1948 | return srcu_notifier_chain_register(&lease_notifier_chain, nb); | |
1949 | } | |
1950 | EXPORT_SYMBOL_GPL(lease_register_notifier); | |
1951 | ||
1952 | void lease_unregister_notifier(struct notifier_block *nb) | |
1953 | { | |
1954 | srcu_notifier_chain_unregister(&lease_notifier_chain, nb); | |
1955 | } | |
1956 | EXPORT_SYMBOL_GPL(lease_unregister_notifier); | |
1957 | ||
b89f4321 | 1958 | /** |
e51673aa | 1959 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1960 | * @filp: file pointer |
1961 | * @arg: type of lease to obtain | |
1962 | * @lease: file_lock to use when adding a lease | |
1963 | * @priv: private info for lm_setup when adding a lease (may be | |
7bbd1fc0 | 1964 | * NULL if lm_setup doesn't require it) |
e51673aa JL |
1965 | * |
1966 | * Call this to establish a lease on the file. The "lease" argument is not | |
1967 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
80b79dd0 MCC |
1968 | * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be |
1969 | * set; if not, this function will return -ENOLCK (and generate a scary-looking | |
e51673aa | 1970 | * stack trace). |
1c7dd2ff JL |
1971 | * |
1972 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1973 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1974 | */ |
e6f5c789 | 1975 | int |
ed5f17f6 | 1976 | vfs_setlease(struct file *filp, int arg, struct file_lock **lease, void **priv) |
1da177e4 | 1977 | { |
18f6622e JL |
1978 | if (lease) |
1979 | setlease_notifier(arg, *lease); | |
de2a4a50 | 1980 | if (filp->f_op->setlease) |
f82b4b67 | 1981 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1982 | else |
f82b4b67 | 1983 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1984 | } |
a9933cea | 1985 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1986 | |
ed5f17f6 | 1987 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, int arg) |
1da177e4 | 1988 | { |
1c7dd2ff | 1989 | struct file_lock *fl; |
f7347ce4 | 1990 | struct fasync_struct *new; |
1da177e4 LT |
1991 | int error; |
1992 | ||
c5b1f0d9 AB |
1993 | fl = lease_alloc(filp, arg); |
1994 | if (IS_ERR(fl)) | |
1995 | return PTR_ERR(fl); | |
1da177e4 | 1996 | |
f7347ce4 LT |
1997 | new = fasync_alloc(); |
1998 | if (!new) { | |
1999 | locks_free_lock(fl); | |
2000 | return -ENOMEM; | |
2001 | } | |
1c7dd2ff | 2002 | new->fa_fd = fd; |
f7347ce4 | 2003 | |
1c7dd2ff | 2004 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
2005 | if (fl) |
2006 | locks_free_lock(fl); | |
f7347ce4 LT |
2007 | if (new) |
2008 | fasync_free(new); | |
1da177e4 LT |
2009 | return error; |
2010 | } | |
2011 | ||
0ceaf6c7 BF |
2012 | /** |
2013 | * fcntl_setlease - sets a lease on an open file | |
2014 | * @fd: open file descriptor | |
2015 | * @filp: file pointer | |
2016 | * @arg: type of lease to obtain | |
2017 | * | |
2018 | * Call this fcntl to establish a lease on the file. | |
2019 | * Note that you also need to call %F_SETSIG to | |
2020 | * receive a signal when the lease is broken. | |
2021 | */ | |
ed5f17f6 | 2022 | int fcntl_setlease(unsigned int fd, struct file *filp, int arg) |
0ceaf6c7 BF |
2023 | { |
2024 | if (arg == F_UNLCK) | |
2ab99ee1 | 2025 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
2026 | return do_fcntl_add_lease(fd, filp, arg); |
2027 | } | |
2028 | ||
1da177e4 | 2029 | /** |
29d01b22 JL |
2030 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
2031 | * @inode: inode of the file to apply to | |
1da177e4 LT |
2032 | * @fl: The lock to be applied |
2033 | * | |
29d01b22 | 2034 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 2035 | */ |
616fb38f | 2036 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
2037 | { |
2038 | int error; | |
2039 | might_sleep(); | |
2040 | for (;;) { | |
29d01b22 | 2041 | error = flock_lock_inode(inode, fl); |
bde74e4b | 2042 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 2043 | break; |
4ca52f53 JL |
2044 | error = wait_event_interruptible(fl->c.flc_wait, |
2045 | list_empty(&fl->c.flc_blocked_member)); | |
16306a61 N |
2046 | if (error) |
2047 | break; | |
1da177e4 | 2048 | } |
16306a61 | 2049 | locks_delete_block(fl); |
1da177e4 LT |
2050 | return error; |
2051 | } | |
2052 | ||
e55c34a6 BC |
2053 | /** |
2054 | * locks_lock_inode_wait - Apply a lock to an inode | |
2055 | * @inode: inode of the file to apply to | |
2056 | * @fl: The lock to be applied | |
2057 | * | |
2058 | * Apply a POSIX or FLOCK style lock request to an inode. | |
2059 | */ | |
2060 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
2061 | { | |
2062 | int res = 0; | |
4ca52f53 | 2063 | switch (fl->c.flc_flags & (FL_POSIX|FL_FLOCK)) { |
e55c34a6 BC |
2064 | case FL_POSIX: |
2065 | res = posix_lock_inode_wait(inode, fl); | |
2066 | break; | |
2067 | case FL_FLOCK: | |
2068 | res = flock_lock_inode_wait(inode, fl); | |
2069 | break; | |
2070 | default: | |
2071 | BUG(); | |
2072 | } | |
2073 | return res; | |
2074 | } | |
2075 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
2076 | ||
1da177e4 LT |
2077 | /** |
2078 | * sys_flock: - flock() system call. | |
2079 | * @fd: the file descriptor to lock. | |
2080 | * @cmd: the type of lock to apply. | |
2081 | * | |
2082 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
80b79dd0 | 2083 | * The @cmd can be one of: |
1da177e4 | 2084 | * |
80b79dd0 MCC |
2085 | * - %LOCK_SH -- a shared lock. |
2086 | * - %LOCK_EX -- an exclusive lock. | |
2087 | * - %LOCK_UN -- remove an existing lock. | |
90f7d7a0 | 2088 | * - %LOCK_MAND -- a 'mandatory' flock. (DEPRECATED) |
1da177e4 | 2089 | * |
90f7d7a0 | 2090 | * %LOCK_MAND support has been removed from the kernel. |
1da177e4 | 2091 | */ |
002c8976 | 2092 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 2093 | { |
db4abb4a | 2094 | int can_sleep, error, type; |
4149be7b | 2095 | struct file_lock fl; |
db4abb4a | 2096 | struct fd f; |
90f7d7a0 JL |
2097 | |
2098 | /* | |
2099 | * LOCK_MAND locks were broken for a long time in that they never | |
2100 | * conflicted with one another and didn't prevent any sort of open, | |
2101 | * read or write activity. | |
2102 | * | |
2103 | * Just ignore these requests now, to preserve legacy behavior, but | |
2104 | * throw a warning to let people know that they don't actually work. | |
2105 | */ | |
2106 | if (cmd & LOCK_MAND) { | |
f2f2494c | 2107 | pr_warn_once("%s(%d): Attempt to set a LOCK_MAND lock via flock(2). This support has been removed and the request ignored.\n", current->comm, current->pid); |
db4abb4a | 2108 | return 0; |
90f7d7a0 | 2109 | } |
1da177e4 | 2110 | |
db4abb4a KI |
2111 | type = flock_translate_cmd(cmd & ~LOCK_NB); |
2112 | if (type < 0) | |
2113 | return type; | |
2114 | ||
2115 | error = -EBADF; | |
2116 | f = fdget(fd); | |
2117 | if (!f.file) | |
2118 | return error; | |
2119 | ||
2120 | if (type != F_UNLCK && !(f.file->f_mode & (FMODE_READ | FMODE_WRITE))) | |
1da177e4 | 2121 | goto out_putf; |
6e129d00 | 2122 | |
4149be7b KI |
2123 | flock_make_lock(f.file, &fl, type); |
2124 | ||
4ca52f53 | 2125 | error = security_file_lock(f.file, fl.c.flc_type); |
1da177e4 | 2126 | if (error) |
4149be7b | 2127 | goto out_putf; |
1da177e4 | 2128 | |
db4abb4a KI |
2129 | can_sleep = !(cmd & LOCK_NB); |
2130 | if (can_sleep) | |
4ca52f53 | 2131 | fl.c.flc_flags |= FL_SLEEP; |
db4abb4a | 2132 | |
de2a4a50 | 2133 | if (f.file->f_op->flock) |
2903ff01 | 2134 | error = f.file->f_op->flock(f.file, |
db4abb4a | 2135 | (can_sleep) ? F_SETLKW : F_SETLK, |
4149be7b | 2136 | &fl); |
1da177e4 | 2137 | else |
4149be7b | 2138 | error = locks_lock_file_wait(f.file, &fl); |
1da177e4 | 2139 | |
932c29a1 | 2140 | locks_release_private(&fl); |
1da177e4 | 2141 | out_putf: |
2903ff01 | 2142 | fdput(f); |
db4abb4a | 2143 | |
1da177e4 LT |
2144 | return error; |
2145 | } | |
2146 | ||
3ee17abd BF |
2147 | /** |
2148 | * vfs_test_lock - test file byte range lock | |
2149 | * @filp: The file to test lock for | |
6924c554 | 2150 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2151 | * |
2152 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2153 | * setting conf->fl_type to something other than F_UNLCK. | |
2154 | */ | |
2155 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2156 | { | |
4ca52f53 | 2157 | WARN_ON_ONCE(filp != fl->c.flc_file); |
de2a4a50 | 2158 | if (filp->f_op->lock) |
3ee17abd BF |
2159 | return filp->f_op->lock(filp, F_GETLK, fl); |
2160 | posix_test_lock(filp, fl); | |
2161 | return 0; | |
2162 | } | |
2163 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2164 | ||
9d5b86ac BC |
2165 | /** |
2166 | * locks_translate_pid - translate a file_lock's fl_pid number into a namespace | |
2167 | * @fl: The file_lock who's fl_pid should be translated | |
2168 | * @ns: The namespace into which the pid should be translated | |
2169 | * | |
bd4c4680 | 2170 | * Used to translate a fl_pid into a namespace virtual pid number |
9d5b86ac | 2171 | */ |
ae7eb16e | 2172 | static pid_t locks_translate_pid(struct file_lock_core *fl, struct pid_namespace *ns) |
9d5b86ac BC |
2173 | { |
2174 | pid_t vnr; | |
2175 | struct pid *pid; | |
2176 | ||
ae7eb16e | 2177 | if (fl->flc_flags & FL_OFDLCK) |
9d5b86ac | 2178 | return -1; |
3d40f781 JL |
2179 | |
2180 | /* Remote locks report a negative pid value */ | |
ae7eb16e JL |
2181 | if (fl->flc_pid <= 0) |
2182 | return fl->flc_pid; | |
3d40f781 | 2183 | |
826d7bc9 KK |
2184 | /* |
2185 | * If the flock owner process is dead and its pid has been already | |
2186 | * freed, the translation below won't work, but we still want to show | |
2187 | * flock owner pid number in init pidns. | |
2188 | */ | |
2189 | if (ns == &init_pid_ns) | |
ae7eb16e | 2190 | return (pid_t) fl->flc_pid; |
9d5b86ac BC |
2191 | |
2192 | rcu_read_lock(); | |
ae7eb16e | 2193 | pid = find_pid_ns(fl->flc_pid, &init_pid_ns); |
9d5b86ac BC |
2194 | vnr = pid_nr_ns(pid, ns); |
2195 | rcu_read_unlock(); | |
2196 | return vnr; | |
2197 | } | |
2198 | ||
c2fa1b8a BF |
2199 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2200 | { | |
ae7eb16e | 2201 | flock->l_pid = locks_translate_pid(&fl->c, task_active_pid_ns(current)); |
c2fa1b8a BF |
2202 | #if BITS_PER_LONG == 32 |
2203 | /* | |
2204 | * Make sure we can represent the posix lock via | |
2205 | * legacy 32bit flock. | |
2206 | */ | |
2207 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2208 | return -EOVERFLOW; | |
2209 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2210 | return -EOVERFLOW; | |
2211 | #endif | |
2212 | flock->l_start = fl->fl_start; | |
2213 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2214 | fl->fl_end - fl->fl_start + 1; | |
2215 | flock->l_whence = 0; | |
4ca52f53 | 2216 | flock->l_type = fl->c.flc_type; |
c2fa1b8a BF |
2217 | return 0; |
2218 | } | |
2219 | ||
2220 | #if BITS_PER_LONG == 32 | |
2221 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2222 | { | |
ae7eb16e | 2223 | flock->l_pid = locks_translate_pid(&fl->c, task_active_pid_ns(current)); |
c2fa1b8a BF |
2224 | flock->l_start = fl->fl_start; |
2225 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2226 | fl->fl_end - fl->fl_start + 1; | |
2227 | flock->l_whence = 0; | |
4ca52f53 | 2228 | flock->l_type = fl->c.flc_type; |
c2fa1b8a BF |
2229 | } |
2230 | #endif | |
2231 | ||
1da177e4 LT |
2232 | /* Report the first existing lock that would conflict with l. |
2233 | * This implements the F_GETLK command of fcntl(). | |
2234 | */ | |
a75d30c7 | 2235 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock) |
1da177e4 | 2236 | { |
52306e88 | 2237 | struct file_lock *fl; |
1da177e4 LT |
2238 | int error; |
2239 | ||
52306e88 BC |
2240 | fl = locks_alloc_lock(); |
2241 | if (fl == NULL) | |
2242 | return -ENOMEM; | |
1da177e4 | 2243 | error = -EINVAL; |
6c9007f6 SS |
2244 | if (cmd != F_OFD_GETLK && flock->l_type != F_RDLCK |
2245 | && flock->l_type != F_WRLCK) | |
1da177e4 LT |
2246 | goto out; |
2247 | ||
52306e88 | 2248 | error = flock_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2249 | if (error) |
2250 | goto out; | |
2251 | ||
0d3f7a2d | 2252 | if (cmd == F_OFD_GETLK) { |
90478939 | 2253 | error = -EINVAL; |
a75d30c7 | 2254 | if (flock->l_pid != 0) |
90478939 JL |
2255 | goto out; |
2256 | ||
4ca52f53 JL |
2257 | fl->c.flc_flags |= FL_OFDLCK; |
2258 | fl->c.flc_owner = filp; | |
5d50ffd7 JL |
2259 | } |
2260 | ||
52306e88 | 2261 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2262 | if (error) |
2263 | goto out; | |
7bbd1fc0 | 2264 | |
4ca52f53 JL |
2265 | flock->l_type = fl->c.flc_type; |
2266 | if (fl->c.flc_type != F_UNLCK) { | |
52306e88 | 2267 | error = posix_lock_to_flock(flock, fl); |
c2fa1b8a | 2268 | if (error) |
52306e88 | 2269 | goto out; |
1da177e4 | 2270 | } |
1da177e4 | 2271 | out: |
52306e88 | 2272 | locks_free_lock(fl); |
1da177e4 LT |
2273 | return error; |
2274 | } | |
2275 | ||
7723ec97 ME |
2276 | /** |
2277 | * vfs_lock_file - file byte range lock | |
2278 | * @filp: The file to apply the lock to | |
2279 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2280 | * @fl: The lock to be applied | |
150b3934 ME |
2281 | * @conf: Place to return a copy of the conflicting lock, if found. |
2282 | * | |
2283 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2284 | * as the final argument. | |
2285 | * | |
2286 | * If the filesystem defines a private ->lock() method, then @conf will | |
2287 | * be left unchanged; so a caller that cares should initialize it to | |
2288 | * some acceptable default. | |
2beb6614 ME |
2289 | * |
2290 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2291 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2292 | * been granted or denied by the underlying filesystem, if (and only if) | |
e70da176 AA |
2293 | * lm_grant is set. Additionally EXPORT_OP_ASYNC_LOCK in export_operations |
2294 | * flags need to be set. | |
2295 | * | |
2296 | * Callers expecting ->lock() to return asynchronously will only use F_SETLK, | |
2297 | * not F_SETLKW; they will set FL_SLEEP if (and only if) the request is for a | |
2298 | * blocking lock. When ->lock() does return asynchronously, it must return | |
2299 | * FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock request completes. | |
2beb6614 | 2300 | * If the request is for non-blocking lock the file system should return |
bde74e4b MS |
2301 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2302 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2303 | * nonzero return code and the file system should release the lock. The file |
2304 | * system is also responsible to keep a corresponding posix lock when it | |
2305 | * grants a lock so the VFS can find out which locks are locally held and do | |
2306 | * the correct lock cleanup when required. | |
2307 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2308 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2309 | * return code. |
7723ec97 | 2310 | */ |
150b3934 | 2311 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2312 | { |
4ca52f53 | 2313 | WARN_ON_ONCE(filp != fl->c.flc_file); |
de2a4a50 | 2314 | if (filp->f_op->lock) |
7723ec97 ME |
2315 | return filp->f_op->lock(filp, cmd, fl); |
2316 | else | |
150b3934 | 2317 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2318 | } |
2319 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2320 | ||
b648a6de MS |
2321 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2322 | struct file_lock *fl) | |
2323 | { | |
2324 | int error; | |
2325 | ||
4ca52f53 | 2326 | error = security_file_lock(filp, fl->c.flc_type); |
b648a6de MS |
2327 | if (error) |
2328 | return error; | |
2329 | ||
764c76b3 MS |
2330 | for (;;) { |
2331 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2332 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2333 | break; |
4ca52f53 JL |
2334 | error = wait_event_interruptible(fl->c.flc_wait, |
2335 | list_empty(&fl->c.flc_blocked_member)); | |
16306a61 N |
2336 | if (error) |
2337 | break; | |
b648a6de | 2338 | } |
16306a61 | 2339 | locks_delete_block(fl); |
b648a6de MS |
2340 | |
2341 | return error; | |
2342 | } | |
2343 | ||
6ca7d910 | 2344 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2345 | static int |
2346 | check_fmode_for_setlk(struct file_lock *fl) | |
2347 | { | |
4ca52f53 | 2348 | switch (fl->c.flc_type) { |
cf01f4ee | 2349 | case F_RDLCK: |
4ca52f53 | 2350 | if (!(fl->c.flc_file->f_mode & FMODE_READ)) |
cf01f4ee JL |
2351 | return -EBADF; |
2352 | break; | |
2353 | case F_WRLCK: | |
4ca52f53 | 2354 | if (!(fl->c.flc_file->f_mode & FMODE_WRITE)) |
cf01f4ee JL |
2355 | return -EBADF; |
2356 | } | |
2357 | return 0; | |
2358 | } | |
2359 | ||
1da177e4 LT |
2360 | /* Apply the lock described by l to an open file descriptor. |
2361 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2362 | */ | |
c293621b | 2363 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2364 | struct flock *flock) |
1da177e4 LT |
2365 | { |
2366 | struct file_lock *file_lock = locks_alloc_lock(); | |
c65454a9 | 2367 | struct inode *inode = file_inode(filp); |
0b2bac2f | 2368 | struct file *f; |
1da177e4 LT |
2369 | int error; |
2370 | ||
2371 | if (file_lock == NULL) | |
2372 | return -ENOLCK; | |
2373 | ||
a75d30c7 | 2374 | error = flock_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2375 | if (error) |
2376 | goto out; | |
5d50ffd7 | 2377 | |
cf01f4ee JL |
2378 | error = check_fmode_for_setlk(file_lock); |
2379 | if (error) | |
2380 | goto out; | |
2381 | ||
5d50ffd7 JL |
2382 | /* |
2383 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2384 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2385 | */ |
2386 | switch (cmd) { | |
0d3f7a2d | 2387 | case F_OFD_SETLK: |
90478939 | 2388 | error = -EINVAL; |
a75d30c7 | 2389 | if (flock->l_pid != 0) |
90478939 JL |
2390 | goto out; |
2391 | ||
5d50ffd7 | 2392 | cmd = F_SETLK; |
4ca52f53 JL |
2393 | file_lock->c.flc_flags |= FL_OFDLCK; |
2394 | file_lock->c.flc_owner = filp; | |
5d50ffd7 | 2395 | break; |
0d3f7a2d | 2396 | case F_OFD_SETLKW: |
90478939 | 2397 | error = -EINVAL; |
a75d30c7 | 2398 | if (flock->l_pid != 0) |
90478939 JL |
2399 | goto out; |
2400 | ||
5d50ffd7 | 2401 | cmd = F_SETLKW; |
4ca52f53 JL |
2402 | file_lock->c.flc_flags |= FL_OFDLCK; |
2403 | file_lock->c.flc_owner = filp; | |
df561f66 | 2404 | fallthrough; |
5d50ffd7 | 2405 | case F_SETLKW: |
4ca52f53 | 2406 | file_lock->c.flc_flags |= FL_SLEEP; |
1da177e4 | 2407 | } |
5d50ffd7 | 2408 | |
b648a6de | 2409 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2410 | |
c293621b | 2411 | /* |
0752ba80 JL |
2412 | * Attempt to detect a close/fcntl race and recover by releasing the |
2413 | * lock that was just acquired. There is no need to do that when we're | |
2414 | * unlocking though, or for OFD locks. | |
c293621b | 2415 | */ |
4ca52f53 JL |
2416 | if (!error && file_lock->c.flc_type != F_UNLCK && |
2417 | !(file_lock->c.flc_flags & FL_OFDLCK)) { | |
120ce2b0 | 2418 | struct files_struct *files = current->files; |
7f3697e2 JL |
2419 | /* |
2420 | * We need that spin_lock here - it prevents reordering between | |
2421 | * update of i_flctx->flc_posix and check for it done in | |
2422 | * close(). rcu_read_lock() wouldn't do. | |
2423 | */ | |
120ce2b0 EB |
2424 | spin_lock(&files->file_lock); |
2425 | f = files_lookup_fd_locked(files, fd); | |
2426 | spin_unlock(&files->file_lock); | |
7f3697e2 | 2427 | if (f != filp) { |
4ca52f53 | 2428 | file_lock->c.flc_type = F_UNLCK; |
7f3697e2 JL |
2429 | error = do_lock_file_wait(filp, cmd, file_lock); |
2430 | WARN_ON_ONCE(error); | |
2431 | error = -EBADF; | |
2432 | } | |
1da177e4 | 2433 | } |
c293621b | 2434 | out: |
1890910f | 2435 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2436 | locks_free_lock(file_lock); |
2437 | return error; | |
2438 | } | |
2439 | ||
2440 | #if BITS_PER_LONG == 32 | |
2441 | /* Report the first existing lock that would conflict with l. | |
2442 | * This implements the F_GETLK command of fcntl(). | |
2443 | */ | |
a75d30c7 | 2444 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock) |
1da177e4 | 2445 | { |
52306e88 | 2446 | struct file_lock *fl; |
1da177e4 LT |
2447 | int error; |
2448 | ||
52306e88 BC |
2449 | fl = locks_alloc_lock(); |
2450 | if (fl == NULL) | |
2451 | return -ENOMEM; | |
2452 | ||
1da177e4 | 2453 | error = -EINVAL; |
6c9007f6 SS |
2454 | if (cmd != F_OFD_GETLK && flock->l_type != F_RDLCK |
2455 | && flock->l_type != F_WRLCK) | |
1da177e4 LT |
2456 | goto out; |
2457 | ||
52306e88 | 2458 | error = flock64_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2459 | if (error) |
2460 | goto out; | |
2461 | ||
0d3f7a2d | 2462 | if (cmd == F_OFD_GETLK) { |
90478939 | 2463 | error = -EINVAL; |
a75d30c7 | 2464 | if (flock->l_pid != 0) |
90478939 JL |
2465 | goto out; |
2466 | ||
4ca52f53 JL |
2467 | fl->c.flc_flags |= FL_OFDLCK; |
2468 | fl->c.flc_owner = filp; | |
5d50ffd7 JL |
2469 | } |
2470 | ||
52306e88 | 2471 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2472 | if (error) |
2473 | goto out; | |
2474 | ||
4ca52f53 JL |
2475 | flock->l_type = fl->c.flc_type; |
2476 | if (fl->c.flc_type != F_UNLCK) | |
52306e88 | 2477 | posix_lock_to_flock64(flock, fl); |
f328296e | 2478 | |
1da177e4 | 2479 | out: |
52306e88 | 2480 | locks_free_lock(fl); |
1da177e4 LT |
2481 | return error; |
2482 | } | |
2483 | ||
2484 | /* Apply the lock described by l to an open file descriptor. | |
2485 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2486 | */ | |
c293621b | 2487 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2488 | struct flock64 *flock) |
1da177e4 LT |
2489 | { |
2490 | struct file_lock *file_lock = locks_alloc_lock(); | |
0b2bac2f | 2491 | struct file *f; |
1da177e4 LT |
2492 | int error; |
2493 | ||
2494 | if (file_lock == NULL) | |
2495 | return -ENOLCK; | |
2496 | ||
a75d30c7 | 2497 | error = flock64_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2498 | if (error) |
2499 | goto out; | |
5d50ffd7 | 2500 | |
cf01f4ee JL |
2501 | error = check_fmode_for_setlk(file_lock); |
2502 | if (error) | |
2503 | goto out; | |
2504 | ||
5d50ffd7 JL |
2505 | /* |
2506 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2507 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2508 | */ |
2509 | switch (cmd) { | |
0d3f7a2d | 2510 | case F_OFD_SETLK: |
90478939 | 2511 | error = -EINVAL; |
a75d30c7 | 2512 | if (flock->l_pid != 0) |
90478939 JL |
2513 | goto out; |
2514 | ||
5d50ffd7 | 2515 | cmd = F_SETLK64; |
4ca52f53 JL |
2516 | file_lock->c.flc_flags |= FL_OFDLCK; |
2517 | file_lock->c.flc_owner = filp; | |
5d50ffd7 | 2518 | break; |
0d3f7a2d | 2519 | case F_OFD_SETLKW: |
90478939 | 2520 | error = -EINVAL; |
a75d30c7 | 2521 | if (flock->l_pid != 0) |
90478939 JL |
2522 | goto out; |
2523 | ||
5d50ffd7 | 2524 | cmd = F_SETLKW64; |
4ca52f53 JL |
2525 | file_lock->c.flc_flags |= FL_OFDLCK; |
2526 | file_lock->c.flc_owner = filp; | |
df561f66 | 2527 | fallthrough; |
5d50ffd7 | 2528 | case F_SETLKW64: |
4ca52f53 | 2529 | file_lock->c.flc_flags |= FL_SLEEP; |
1da177e4 | 2530 | } |
5d50ffd7 | 2531 | |
b648a6de | 2532 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2533 | |
c293621b | 2534 | /* |
0752ba80 JL |
2535 | * Attempt to detect a close/fcntl race and recover by releasing the |
2536 | * lock that was just acquired. There is no need to do that when we're | |
2537 | * unlocking though, or for OFD locks. | |
c293621b | 2538 | */ |
4ca52f53 JL |
2539 | if (!error && file_lock->c.flc_type != F_UNLCK && |
2540 | !(file_lock->c.flc_flags & FL_OFDLCK)) { | |
120ce2b0 | 2541 | struct files_struct *files = current->files; |
7f3697e2 JL |
2542 | /* |
2543 | * We need that spin_lock here - it prevents reordering between | |
2544 | * update of i_flctx->flc_posix and check for it done in | |
2545 | * close(). rcu_read_lock() wouldn't do. | |
2546 | */ | |
120ce2b0 EB |
2547 | spin_lock(&files->file_lock); |
2548 | f = files_lookup_fd_locked(files, fd); | |
2549 | spin_unlock(&files->file_lock); | |
7f3697e2 | 2550 | if (f != filp) { |
4ca52f53 | 2551 | file_lock->c.flc_type = F_UNLCK; |
7f3697e2 JL |
2552 | error = do_lock_file_wait(filp, cmd, file_lock); |
2553 | WARN_ON_ONCE(error); | |
2554 | error = -EBADF; | |
2555 | } | |
1da177e4 | 2556 | } |
1da177e4 LT |
2557 | out: |
2558 | locks_free_lock(file_lock); | |
2559 | return error; | |
2560 | } | |
2561 | #endif /* BITS_PER_LONG == 32 */ | |
2562 | ||
2563 | /* | |
2564 | * This function is called when the file is being removed | |
2565 | * from the task's fd array. POSIX locks belonging to this task | |
2566 | * are deleted at this time. | |
2567 | */ | |
2568 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2569 | { | |
1890910f | 2570 | int error; |
c65454a9 | 2571 | struct inode *inode = file_inode(filp); |
ff7b86b8 | 2572 | struct file_lock lock; |
128a3785 | 2573 | struct file_lock_context *ctx; |
1da177e4 LT |
2574 | |
2575 | /* | |
2576 | * If there are no locks held on this file, we don't need to call | |
2577 | * posix_lock_file(). Another process could be setting a lock on this | |
2578 | * file at the same time, but we wouldn't remove that lock anyway. | |
2579 | */ | |
401a8b8f | 2580 | ctx = locks_inode_context(inode); |
bd61e0a9 | 2581 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2582 | return; |
2583 | ||
d6367d62 | 2584 | locks_init_lock(&lock); |
4ca52f53 JL |
2585 | lock.c.flc_type = F_UNLCK; |
2586 | lock.c.flc_flags = FL_POSIX | FL_CLOSE; | |
1da177e4 LT |
2587 | lock.fl_start = 0; |
2588 | lock.fl_end = OFFSET_MAX; | |
4ca52f53 JL |
2589 | lock.c.flc_owner = owner; |
2590 | lock.c.flc_pid = current->tgid; | |
2591 | lock.c.flc_file = filp; | |
1da177e4 LT |
2592 | lock.fl_ops = NULL; |
2593 | lock.fl_lmops = NULL; | |
2594 | ||
1890910f | 2595 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2596 | |
1da177e4 LT |
2597 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2598 | lock.fl_ops->fl_release_private(&lock); | |
c568d683 | 2599 | trace_locks_remove_posix(inode, &lock, error); |
1da177e4 | 2600 | } |
1da177e4 LT |
2601 | EXPORT_SYMBOL(locks_remove_posix); |
2602 | ||
3d8e560d | 2603 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2604 | static void |
128a3785 | 2605 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 | 2606 | { |
d6367d62 | 2607 | struct file_lock fl; |
c65454a9 | 2608 | struct inode *inode = file_inode(filp); |
dd459bb1 | 2609 | |
3d8e560d | 2610 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2611 | return; |
2612 | ||
4149be7b | 2613 | flock_make_lock(filp, &fl, F_UNLCK); |
4ca52f53 | 2614 | fl.c.flc_flags |= FL_CLOSE; |
d6367d62 | 2615 | |
de2a4a50 | 2616 | if (filp->f_op->flock) |
dd459bb1 JL |
2617 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2618 | else | |
bcd7f78d | 2619 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2620 | |
2621 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2622 | fl.fl_ops->fl_release_private(&fl); | |
2623 | } | |
2624 | ||
3d8e560d | 2625 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2626 | static void |
128a3785 | 2627 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2628 | { |
8634b51f JL |
2629 | struct file_lock *fl, *tmp; |
2630 | LIST_HEAD(dispose); | |
2631 | ||
3d8e560d | 2632 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2633 | return; |
2634 | ||
02e525b2 | 2635 | percpu_down_read(&file_rwsem); |
6109c850 | 2636 | spin_lock(&ctx->flc_lock); |
4ca52f53 JL |
2637 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, c.flc_list) |
2638 | if (filp == fl->c.flc_file) | |
c4e136cd | 2639 | lease_modify(fl, F_UNLCK, &dispose); |
6109c850 | 2640 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 2641 | percpu_up_read(&file_rwsem); |
5f43086b | 2642 | |
8634b51f JL |
2643 | locks_dispose_list(&dispose); |
2644 | } | |
2645 | ||
1da177e4 LT |
2646 | /* |
2647 | * This function is called on the last close of an open file. | |
2648 | */ | |
78ed8a13 | 2649 | void locks_remove_file(struct file *filp) |
1da177e4 | 2650 | { |
128a3785 DV |
2651 | struct file_lock_context *ctx; |
2652 | ||
c65454a9 | 2653 | ctx = locks_inode_context(file_inode(filp)); |
128a3785 | 2654 | if (!ctx) |
3d8e560d JL |
2655 | return; |
2656 | ||
dd459bb1 | 2657 | /* remove any OFD locks */ |
73a8f5f7 | 2658 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2659 | |
dd459bb1 | 2660 | /* remove flock locks */ |
128a3785 | 2661 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2662 | |
8634b51f | 2663 | /* remove any leases */ |
128a3785 | 2664 | locks_remove_lease(filp, ctx); |
3953704f BC |
2665 | |
2666 | spin_lock(&ctx->flc_lock); | |
2667 | locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX"); | |
2668 | locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK"); | |
2669 | locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE"); | |
2670 | spin_unlock(&ctx->flc_lock); | |
1da177e4 LT |
2671 | } |
2672 | ||
9b9d2ab4 ME |
2673 | /** |
2674 | * vfs_cancel_lock - file byte range unblock lock | |
2675 | * @filp: The file to apply the unblock to | |
2676 | * @fl: The lock to be unblocked | |
2677 | * | |
2678 | * Used by lock managers to cancel blocked requests | |
2679 | */ | |
2680 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2681 | { | |
4ca52f53 | 2682 | WARN_ON_ONCE(filp != fl->c.flc_file); |
de2a4a50 | 2683 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2684 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2685 | return 0; | |
2686 | } | |
9b9d2ab4 ME |
2687 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); |
2688 | ||
ab1ddef9 JL |
2689 | /** |
2690 | * vfs_inode_has_locks - are any file locks held on @inode? | |
2691 | * @inode: inode to check for locks | |
2692 | * | |
2693 | * Return true if there are any FL_POSIX or FL_FLOCK locks currently | |
2694 | * set on @inode. | |
2695 | */ | |
2696 | bool vfs_inode_has_locks(struct inode *inode) | |
2697 | { | |
2698 | struct file_lock_context *ctx; | |
2699 | bool ret; | |
2700 | ||
401a8b8f | 2701 | ctx = locks_inode_context(inode); |
ab1ddef9 JL |
2702 | if (!ctx) |
2703 | return false; | |
2704 | ||
2705 | spin_lock(&ctx->flc_lock); | |
2706 | ret = !list_empty(&ctx->flc_posix) || !list_empty(&ctx->flc_flock); | |
2707 | spin_unlock(&ctx->flc_lock); | |
2708 | return ret; | |
2709 | } | |
2710 | EXPORT_SYMBOL_GPL(vfs_inode_has_locks); | |
2711 | ||
7f8ada98 | 2712 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2713 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2714 | #include <linux/seq_file.h> |
2715 | ||
7012b02a JL |
2716 | struct locks_iterator { |
2717 | int li_cpu; | |
2718 | loff_t li_pos; | |
2719 | }; | |
2720 | ||
a1c2af32 | 2721 | static void lock_get_status(struct seq_file *f, struct file_lock_core *flc, |
b8da9b10 | 2722 | loff_t id, char *pfx, int repeat) |
1da177e4 LT |
2723 | { |
2724 | struct inode *inode = NULL; | |
6021d62c | 2725 | unsigned int pid; |
9d78edea | 2726 | struct pid_namespace *proc_pidns = proc_pid_ns(file_inode(f->file)->i_sb); |
a1c2af32 JL |
2727 | int type = flc->flc_type; |
2728 | struct file_lock *fl = file_lock(flc); | |
2729 | ||
2730 | pid = locks_translate_pid(flc, proc_pidns); | |
ab1f1611 | 2731 | |
9d5b86ac | 2732 | /* |
1cf8e5de KK |
2733 | * If lock owner is dead (and pid is freed) or not visible in current |
2734 | * pidns, zero is shown as a pid value. Check lock info from | |
2735 | * init_pid_ns to get saved lock pid value. | |
9d5b86ac | 2736 | */ |
a1c2af32 JL |
2737 | if (flc->flc_file != NULL) |
2738 | inode = file_inode(flc->flc_file); | |
1da177e4 | 2739 | |
b8da9b10 LL |
2740 | seq_printf(f, "%lld: ", id); |
2741 | ||
2742 | if (repeat) | |
2743 | seq_printf(f, "%*s", repeat - 1 + (int)strlen(pfx), pfx); | |
2744 | ||
a1c2af32 JL |
2745 | if (flc->flc_flags & FL_POSIX) { |
2746 | if (flc->flc_flags & FL_ACCESS) | |
5315c26a | 2747 | seq_puts(f, "ACCESS"); |
a1c2af32 | 2748 | else if (flc->flc_flags & FL_OFDLCK) |
5315c26a | 2749 | seq_puts(f, "OFDLCK"); |
c918d42a | 2750 | else |
5315c26a | 2751 | seq_puts(f, "POSIX "); |
c918d42a JL |
2752 | |
2753 | seq_printf(f, " %s ", | |
f7e33bdb | 2754 | (inode == NULL) ? "*NOINODE*" : "ADVISORY "); |
a1c2af32 | 2755 | } else if (flc->flc_flags & FL_FLOCK) { |
90f7d7a0 | 2756 | seq_puts(f, "FLOCK ADVISORY "); |
a1c2af32 | 2757 | } else if (flc->flc_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) { |
3d40f781 JL |
2758 | type = target_leasetype(fl); |
2759 | ||
a1c2af32 | 2760 | if (flc->flc_flags & FL_DELEG) |
8144f1f6 JL |
2761 | seq_puts(f, "DELEG "); |
2762 | else | |
2763 | seq_puts(f, "LEASE "); | |
2764 | ||
ab83fa4b | 2765 | if (lease_breaking(fl)) |
5315c26a | 2766 | seq_puts(f, "BREAKING "); |
a1c2af32 | 2767 | else if (flc->flc_file) |
5315c26a | 2768 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2769 | else |
5315c26a | 2770 | seq_puts(f, "BREAKER "); |
1da177e4 | 2771 | } else { |
5315c26a | 2772 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 | 2773 | } |
43e4cb94 | 2774 | |
90f7d7a0 JL |
2775 | seq_printf(f, "%s ", (type == F_WRLCK) ? "WRITE" : |
2776 | (type == F_RDLCK) ? "READ" : "UNLCK"); | |
1da177e4 | 2777 | if (inode) { |
3648888e | 2778 | /* userspace relies on this representation of dev_t */ |
6021d62c | 2779 | seq_printf(f, "%d %02x:%02x:%lu ", pid, |
1da177e4 LT |
2780 | MAJOR(inode->i_sb->s_dev), |
2781 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2782 | } else { |
6021d62c | 2783 | seq_printf(f, "%d <none>:0 ", pid); |
1da177e4 | 2784 | } |
a1c2af32 | 2785 | if (flc->flc_flags & FL_POSIX) { |
1da177e4 | 2786 | if (fl->fl_end == OFFSET_MAX) |
7f8ada98 | 2787 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2788 | else |
7f8ada98 | 2789 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2790 | } else { |
5315c26a | 2791 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2792 | } |
2793 | } | |
2794 | ||
a1c2af32 | 2795 | static struct file_lock_core *get_next_blocked_member(struct file_lock_core *node) |
b8da9b10 | 2796 | { |
a1c2af32 | 2797 | struct file_lock_core *tmp; |
b8da9b10 LL |
2798 | |
2799 | /* NULL node or root node */ | |
a1c2af32 | 2800 | if (node == NULL || node->flc_blocker == NULL) |
b8da9b10 LL |
2801 | return NULL; |
2802 | ||
2803 | /* Next member in the linked list could be itself */ | |
a1c2af32 JL |
2804 | tmp = list_next_entry(node, flc_blocked_member); |
2805 | if (list_entry_is_head(tmp, &node->flc_blocker->flc_blocked_requests, | |
2806 | flc_blocked_member) | |
b6aaba5b | 2807 | || tmp == node) { |
b8da9b10 LL |
2808 | return NULL; |
2809 | } | |
2810 | ||
2811 | return tmp; | |
2812 | } | |
2813 | ||
7f8ada98 | 2814 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2815 | { |
7012b02a | 2816 | struct locks_iterator *iter = f->private; |
a1c2af32 | 2817 | struct file_lock_core *cur, *tmp; |
9d78edea | 2818 | struct pid_namespace *proc_pidns = proc_pid_ns(file_inode(f->file)->i_sb); |
b8da9b10 | 2819 | int level = 0; |
1da177e4 | 2820 | |
a1c2af32 | 2821 | cur = hlist_entry(v, struct file_lock_core, flc_link); |
1da177e4 | 2822 | |
a1c2af32 | 2823 | if (locks_translate_pid(cur, proc_pidns) == 0) |
d67fd44f NB |
2824 | return 0; |
2825 | ||
a1c2af32 | 2826 | /* View this crossed linked list as a binary tree, the first member of flc_blocked_requests |
4ca52f53 | 2827 | * is the left child of current node, the next silibing in flc_blocked_member is the |
a1c2af32 | 2828 | * right child, we can alse get the parent of current node from flc_blocker, so this |
b8da9b10 LL |
2829 | * question becomes traversal of a binary tree |
2830 | */ | |
2831 | while (cur != NULL) { | |
2832 | if (level) | |
2833 | lock_get_status(f, cur, iter->li_pos, "-> ", level); | |
2834 | else | |
2835 | lock_get_status(f, cur, iter->li_pos, "", level); | |
1da177e4 | 2836 | |
a1c2af32 | 2837 | if (!list_empty(&cur->flc_blocked_requests)) { |
b8da9b10 | 2838 | /* Turn left */ |
a1c2af32 JL |
2839 | cur = list_first_entry_or_null(&cur->flc_blocked_requests, |
2840 | struct file_lock_core, | |
2841 | flc_blocked_member); | |
b8da9b10 LL |
2842 | level++; |
2843 | } else { | |
2844 | /* Turn right */ | |
2845 | tmp = get_next_blocked_member(cur); | |
2846 | /* Fall back to parent node */ | |
a1c2af32 JL |
2847 | while (tmp == NULL && cur->flc_blocker != NULL) { |
2848 | cur = cur->flc_blocker; | |
b8da9b10 LL |
2849 | level--; |
2850 | tmp = get_next_blocked_member(cur); | |
2851 | } | |
2852 | cur = tmp; | |
2853 | } | |
2854 | } | |
094f2825 | 2855 | |
7f8ada98 PE |
2856 | return 0; |
2857 | } | |
1da177e4 | 2858 | |
6c8c9031 AV |
2859 | static void __show_fd_locks(struct seq_file *f, |
2860 | struct list_head *head, int *id, | |
2861 | struct file *filp, struct files_struct *files) | |
2862 | { | |
a1c2af32 | 2863 | struct file_lock_core *fl; |
6c8c9031 | 2864 | |
a1c2af32 | 2865 | list_for_each_entry(fl, head, flc_list) { |
6c8c9031 | 2866 | |
a1c2af32 | 2867 | if (filp != fl->flc_file) |
6c8c9031 | 2868 | continue; |
a1c2af32 | 2869 | if (fl->flc_owner != files && fl->flc_owner != filp) |
6c8c9031 AV |
2870 | continue; |
2871 | ||
2872 | (*id)++; | |
2873 | seq_puts(f, "lock:\t"); | |
b8da9b10 | 2874 | lock_get_status(f, fl, *id, "", 0); |
6c8c9031 AV |
2875 | } |
2876 | } | |
2877 | ||
2878 | void show_fd_locks(struct seq_file *f, | |
2879 | struct file *filp, struct files_struct *files) | |
2880 | { | |
c65454a9 | 2881 | struct inode *inode = file_inode(filp); |
6c8c9031 AV |
2882 | struct file_lock_context *ctx; |
2883 | int id = 0; | |
2884 | ||
401a8b8f | 2885 | ctx = locks_inode_context(inode); |
6c8c9031 AV |
2886 | if (!ctx) |
2887 | return; | |
2888 | ||
2889 | spin_lock(&ctx->flc_lock); | |
2890 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2891 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2892 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2893 | spin_unlock(&ctx->flc_lock); | |
2894 | } | |
2895 | ||
7f8ada98 | 2896 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2897 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2898 | { |
7012b02a | 2899 | struct locks_iterator *iter = f->private; |
99dc8292 | 2900 | |
7012b02a | 2901 | iter->li_pos = *pos + 1; |
aba37660 | 2902 | percpu_down_write(&file_rwsem); |
7b2296af | 2903 | spin_lock(&blocked_lock_lock); |
7c3f654d | 2904 | return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos); |
7f8ada98 | 2905 | } |
1da177e4 | 2906 | |
7f8ada98 PE |
2907 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2908 | { | |
7012b02a JL |
2909 | struct locks_iterator *iter = f->private; |
2910 | ||
2911 | ++iter->li_pos; | |
7c3f654d | 2912 | return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos); |
7f8ada98 | 2913 | } |
1da177e4 | 2914 | |
7f8ada98 | 2915 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2916 | __releases(&blocked_lock_lock) |
7f8ada98 | 2917 | { |
7b2296af | 2918 | spin_unlock(&blocked_lock_lock); |
aba37660 | 2919 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2920 | } |
2921 | ||
d8ba7a36 | 2922 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2923 | .start = locks_start, |
2924 | .next = locks_next, | |
2925 | .stop = locks_stop, | |
2926 | .show = locks_show, | |
2927 | }; | |
d8ba7a36 | 2928 | |
d8ba7a36 AD |
2929 | static int __init proc_locks_init(void) |
2930 | { | |
44414d82 CH |
2931 | proc_create_seq_private("locks", 0, NULL, &locks_seq_operations, |
2932 | sizeof(struct locks_iterator), NULL); | |
d8ba7a36 AD |
2933 | return 0; |
2934 | } | |
91899226 | 2935 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2936 | #endif |
2937 | ||
1da177e4 LT |
2938 | static int __init filelock_init(void) |
2939 | { | |
7012b02a JL |
2940 | int i; |
2941 | ||
4a075e39 | 2942 | flctx_cache = kmem_cache_create("file_lock_ctx", |
3754707b | 2943 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); |
4a075e39 | 2944 | |
1da177e4 | 2945 | filelock_cache = kmem_cache_create("file_lock_cache", |
3754707b | 2946 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
ee19cc40 | 2947 | |
7c3f654d PZ |
2948 | for_each_possible_cpu(i) { |
2949 | struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i); | |
2950 | ||
2951 | spin_lock_init(&fll->lock); | |
2952 | INIT_HLIST_HEAD(&fll->hlist); | |
2953 | } | |
7012b02a | 2954 | |
18f6622e | 2955 | lease_notifier_chain_init(); |
1da177e4 LT |
2956 | return 0; |
2957 | } | |
1da177e4 | 2958 | core_initcall(filelock_init); |