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 LT |
50 | */ |
51 | ||
52 | #include <linux/capability.h> | |
53 | #include <linux/file.h> | |
9f3acc31 | 54 | #include <linux/fdtable.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 LT |
71 | |
72 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
73 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
11afe9f7 | 74 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT)) |
cff2fce5 | 75 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
9d5b86ac | 76 | #define IS_REMOTELCK(fl) (fl->fl_pid <= 0) |
1da177e4 | 77 | |
ab83fa4b BF |
78 | static bool lease_breaking(struct file_lock *fl) |
79 | { | |
778fc546 BF |
80 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
81 | } | |
82 | ||
83 | static int target_leasetype(struct file_lock *fl) | |
84 | { | |
85 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
86 | return F_UNLCK; | |
87 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
88 | return F_RDLCK; | |
89 | return fl->fl_type; | |
ab83fa4b BF |
90 | } |
91 | ||
dd81faa8 LC |
92 | static int leases_enable = 1; |
93 | static int lease_break_time = 45; | |
94 | ||
95 | #ifdef CONFIG_SYSCTL | |
96 | static struct ctl_table locks_sysctls[] = { | |
97 | { | |
98 | .procname = "leases-enable", | |
99 | .data = &leases_enable, | |
100 | .maxlen = sizeof(int), | |
101 | .mode = 0644, | |
102 | .proc_handler = proc_dointvec, | |
103 | }, | |
104 | #ifdef CONFIG_MMU | |
105 | { | |
106 | .procname = "lease-break-time", | |
107 | .data = &lease_break_time, | |
108 | .maxlen = sizeof(int), | |
109 | .mode = 0644, | |
110 | .proc_handler = proc_dointvec, | |
111 | }, | |
112 | #endif /* CONFIG_MMU */ | |
113 | {} | |
114 | }; | |
115 | ||
116 | static int __init init_fs_locks_sysctls(void) | |
117 | { | |
118 | register_sysctl_init("fs", locks_sysctls); | |
119 | return 0; | |
120 | } | |
121 | early_initcall(init_fs_locks_sysctls); | |
122 | #endif /* CONFIG_SYSCTL */ | |
1da177e4 | 123 | |
1c8c601a | 124 | /* |
7012b02a | 125 | * The global file_lock_list is only used for displaying /proc/locks, so we |
7c3f654d PZ |
126 | * keep a list on each CPU, with each list protected by its own spinlock. |
127 | * Global serialization is done using file_rwsem. | |
128 | * | |
129 | * Note that alterations to the list also require that the relevant flc_lock is | |
130 | * held. | |
1c8c601a | 131 | */ |
7c3f654d PZ |
132 | struct file_lock_list_struct { |
133 | spinlock_t lock; | |
134 | struct hlist_head hlist; | |
135 | }; | |
136 | static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list); | |
aba37660 | 137 | DEFINE_STATIC_PERCPU_RWSEM(file_rwsem); |
88974691 | 138 | |
eb82dd39 | 139 | |
1c8c601a | 140 | /* |
48f74186 | 141 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 142 | * It is protected by blocked_lock_lock. |
48f74186 JL |
143 | * |
144 | * We hash locks by lockowner in order to optimize searching for the lock a | |
145 | * particular lockowner is waiting on. | |
146 | * | |
147 | * FIXME: make this value scale via some heuristic? We generally will want more | |
148 | * buckets when we have more lockowners holding locks, but that's a little | |
149 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 150 | */ |
48f74186 JL |
151 | #define BLOCKED_HASH_BITS 7 |
152 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 153 | |
1c8c601a | 154 | /* |
7b2296af JL |
155 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
156 | * want to be holding this lock. | |
1c8c601a | 157 | * |
ada5c1da N |
158 | * In addition, it also protects the fl->fl_blocked_requests list, and the |
159 | * fl->fl_blocker pointer for file_lock structures that are acting as lock | |
160 | * requests (in contrast to those that are acting as records of acquired locks). | |
1c8c601a JL |
161 | * |
162 | * Note that when we acquire this lock in order to change the above fields, | |
6109c850 | 163 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 164 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 165 | * flc_lock. |
1c8c601a | 166 | */ |
7b2296af | 167 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 168 | |
4a075e39 | 169 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 170 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 171 | |
4a075e39 | 172 | static struct file_lock_context * |
5c1c669a | 173 | locks_get_lock_context(struct inode *inode, int type) |
4a075e39 | 174 | { |
128a3785 | 175 | struct file_lock_context *ctx; |
4a075e39 | 176 | |
128a3785 DV |
177 | /* paired with cmpxchg() below */ |
178 | ctx = smp_load_acquire(&inode->i_flctx); | |
179 | if (likely(ctx) || type == F_UNLCK) | |
4a075e39 JL |
180 | goto out; |
181 | ||
128a3785 DV |
182 | ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL); |
183 | if (!ctx) | |
4a075e39 JL |
184 | goto out; |
185 | ||
128a3785 DV |
186 | spin_lock_init(&ctx->flc_lock); |
187 | INIT_LIST_HEAD(&ctx->flc_flock); | |
188 | INIT_LIST_HEAD(&ctx->flc_posix); | |
189 | INIT_LIST_HEAD(&ctx->flc_lease); | |
4a075e39 JL |
190 | |
191 | /* | |
192 | * Assign the pointer if it's not already assigned. If it is, then | |
193 | * free the context we just allocated. | |
194 | */ | |
128a3785 DV |
195 | if (cmpxchg(&inode->i_flctx, NULL, ctx)) { |
196 | kmem_cache_free(flctx_cache, ctx); | |
197 | ctx = smp_load_acquire(&inode->i_flctx); | |
198 | } | |
4a075e39 | 199 | out: |
1890910f | 200 | trace_locks_get_lock_context(inode, type, ctx); |
128a3785 | 201 | return ctx; |
4a075e39 JL |
202 | } |
203 | ||
e24dadab JL |
204 | static void |
205 | locks_dump_ctx_list(struct list_head *list, char *list_type) | |
206 | { | |
207 | struct file_lock *fl; | |
208 | ||
209 | list_for_each_entry(fl, list, fl_list) { | |
210 | pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type, fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
211 | } | |
212 | } | |
213 | ||
214 | static void | |
215 | locks_check_ctx_lists(struct inode *inode) | |
216 | { | |
217 | struct file_lock_context *ctx = inode->i_flctx; | |
218 | ||
219 | if (unlikely(!list_empty(&ctx->flc_flock) || | |
220 | !list_empty(&ctx->flc_posix) || | |
221 | !list_empty(&ctx->flc_lease))) { | |
222 | pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n", | |
223 | MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev), | |
224 | inode->i_ino); | |
225 | locks_dump_ctx_list(&ctx->flc_flock, "FLOCK"); | |
226 | locks_dump_ctx_list(&ctx->flc_posix, "POSIX"); | |
227 | locks_dump_ctx_list(&ctx->flc_lease, "LEASE"); | |
228 | } | |
229 | } | |
230 | ||
3953704f BC |
231 | static void |
232 | locks_check_ctx_file_list(struct file *filp, struct list_head *list, | |
233 | char *list_type) | |
234 | { | |
235 | struct file_lock *fl; | |
236 | struct inode *inode = locks_inode(filp); | |
237 | ||
238 | list_for_each_entry(fl, list, fl_list) | |
239 | if (fl->fl_file == filp) | |
240 | pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx " | |
241 | " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", | |
242 | list_type, MAJOR(inode->i_sb->s_dev), | |
243 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
244 | fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid); | |
245 | } | |
246 | ||
4a075e39 | 247 | void |
f27a0fe0 | 248 | locks_free_lock_context(struct inode *inode) |
4a075e39 | 249 | { |
f27a0fe0 JL |
250 | struct file_lock_context *ctx = inode->i_flctx; |
251 | ||
e24dadab JL |
252 | if (unlikely(ctx)) { |
253 | locks_check_ctx_lists(inode); | |
4a075e39 JL |
254 | kmem_cache_free(flctx_cache, ctx); |
255 | } | |
256 | } | |
257 | ||
ee19cc40 | 258 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 259 | { |
139ca04e | 260 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 261 | INIT_LIST_HEAD(&fl->fl_list); |
ada5c1da N |
262 | INIT_LIST_HEAD(&fl->fl_blocked_requests); |
263 | INIT_LIST_HEAD(&fl->fl_blocked_member); | |
ee19cc40 | 264 | init_waitqueue_head(&fl->fl_wait); |
a51cb91d MS |
265 | } |
266 | ||
1da177e4 | 267 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 268 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 269 | { |
ee19cc40 | 270 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
271 | |
272 | if (fl) | |
ee19cc40 | 273 | locks_init_lock_heads(fl); |
a51cb91d MS |
274 | |
275 | return fl; | |
1da177e4 | 276 | } |
c5b1f0d9 | 277 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 278 | |
a9e61e25 | 279 | void locks_release_private(struct file_lock *fl) |
47831f35 | 280 | { |
5926459e N |
281 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
282 | BUG_ON(!list_empty(&fl->fl_list)); | |
283 | BUG_ON(!list_empty(&fl->fl_blocked_requests)); | |
284 | BUG_ON(!list_empty(&fl->fl_blocked_member)); | |
285 | BUG_ON(!hlist_unhashed(&fl->fl_link)); | |
286 | ||
47831f35 TM |
287 | if (fl->fl_ops) { |
288 | if (fl->fl_ops->fl_release_private) | |
289 | fl->fl_ops->fl_release_private(fl); | |
290 | fl->fl_ops = NULL; | |
291 | } | |
47831f35 | 292 | |
5c97d7b1 | 293 | if (fl->fl_lmops) { |
cae80b30 JL |
294 | if (fl->fl_lmops->lm_put_owner) { |
295 | fl->fl_lmops->lm_put_owner(fl->fl_owner); | |
296 | fl->fl_owner = NULL; | |
297 | } | |
5c97d7b1 KM |
298 | fl->fl_lmops = NULL; |
299 | } | |
47831f35 | 300 | } |
a9e61e25 | 301 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 302 | |
591502c5 DN |
303 | /** |
304 | * locks_owner_has_blockers - Check for blocking lock requests | |
305 | * @flctx: file lock context | |
306 | * @owner: lock owner | |
307 | * | |
308 | * Return values: | |
309 | * %true: @owner has at least one blocker | |
310 | * %false: @owner has no blockers | |
311 | */ | |
312 | bool locks_owner_has_blockers(struct file_lock_context *flctx, | |
313 | fl_owner_t owner) | |
314 | { | |
315 | struct file_lock *fl; | |
316 | ||
317 | spin_lock(&flctx->flc_lock); | |
318 | list_for_each_entry(fl, &flctx->flc_posix, fl_list) { | |
319 | if (fl->fl_owner != owner) | |
320 | continue; | |
321 | if (!list_empty(&fl->fl_blocked_requests)) { | |
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)) { | |
6dee60f6 JL |
345 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
346 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
347 | locks_free_lock(fl); |
348 | } | |
349 | } | |
350 | ||
1da177e4 LT |
351 | void locks_init_lock(struct file_lock *fl) |
352 | { | |
ee19cc40 MS |
353 | memset(fl, 0, sizeof(struct file_lock)); |
354 | locks_init_lock_heads(fl); | |
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 LT |
362 | { |
363 | new->fl_owner = fl->fl_owner; | |
364 | new->fl_pid = fl->fl_pid; | |
0996905f | 365 | new->fl_file = NULL; |
1da177e4 LT |
366 | new->fl_flags = fl->fl_flags; |
367 | new->fl_type = fl->fl_type; | |
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) | |
cae80b30 | 375 | fl->fl_lmops->lm_get_owner(fl->fl_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 | |
0996905f | 387 | new->fl_file = fl->fl_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 | |
403 | * ->fl_blocked_requests, so we don't need a lock to check if it | |
404 | * is empty. | |
405 | */ | |
406 | if (list_empty(&fl->fl_blocked_requests)) | |
407 | return; | |
408 | spin_lock(&blocked_lock_lock); | |
409 | list_splice_init(&fl->fl_blocked_requests, &new->fl_blocked_requests); | |
bf77ae4c | 410 | list_for_each_entry(f, &new->fl_blocked_requests, fl_blocked_member) |
5946c431 N |
411 | f->fl_blocker = new; |
412 | spin_unlock(&blocked_lock_lock); | |
413 | } | |
414 | ||
1da177e4 | 415 | static inline int flock_translate_cmd(int cmd) { |
1da177e4 LT |
416 | switch (cmd) { |
417 | case LOCK_SH: | |
418 | return F_RDLCK; | |
419 | case LOCK_EX: | |
420 | return F_WRLCK; | |
421 | case LOCK_UN: | |
422 | return F_UNLCK; | |
423 | } | |
424 | return -EINVAL; | |
425 | } | |
426 | ||
427 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 | 428 | static struct file_lock * |
d6367d62 | 429 | flock_make_lock(struct file *filp, unsigned int cmd, struct file_lock *fl) |
1da177e4 | 430 | { |
1da177e4 | 431 | int type = flock_translate_cmd(cmd); |
6e129d00 | 432 | |
1da177e4 | 433 | if (type < 0) |
6e129d00 | 434 | return ERR_PTR(type); |
7bbd1fc0 | 435 | |
d6367d62 N |
436 | if (fl == NULL) { |
437 | fl = locks_alloc_lock(); | |
438 | if (fl == NULL) | |
439 | return ERR_PTR(-ENOMEM); | |
440 | } else { | |
441 | locks_init_lock(fl); | |
442 | } | |
1da177e4 LT |
443 | |
444 | fl->fl_file = filp; | |
73a8f5f7 | 445 | fl->fl_owner = filp; |
1da177e4 LT |
446 | fl->fl_pid = current->tgid; |
447 | fl->fl_flags = FL_FLOCK; | |
448 | fl->fl_type = type; | |
449 | fl->fl_end = OFFSET_MAX; | |
7bbd1fc0 | 450 | |
6e129d00 | 451 | return fl; |
1da177e4 LT |
452 | } |
453 | ||
0ec4f431 | 454 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
455 | { |
456 | switch (type) { | |
457 | case F_RDLCK: | |
458 | case F_WRLCK: | |
459 | case F_UNLCK: | |
460 | fl->fl_type = type; | |
461 | break; | |
462 | default: | |
463 | return -EINVAL; | |
464 | } | |
465 | return 0; | |
466 | } | |
467 | ||
ef12e72a BF |
468 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
469 | struct flock64 *l) | |
1da177e4 | 470 | { |
1da177e4 | 471 | switch (l->l_whence) { |
f5579f8c | 472 | case SEEK_SET: |
ef12e72a | 473 | fl->fl_start = 0; |
1da177e4 | 474 | break; |
f5579f8c | 475 | case SEEK_CUR: |
ef12e72a | 476 | fl->fl_start = filp->f_pos; |
1da177e4 | 477 | break; |
f5579f8c | 478 | case SEEK_END: |
ef12e72a | 479 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
480 | break; |
481 | default: | |
482 | return -EINVAL; | |
483 | } | |
ef12e72a BF |
484 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
485 | return -EOVERFLOW; | |
486 | fl->fl_start += l->l_start; | |
487 | if (fl->fl_start < 0) | |
488 | return -EINVAL; | |
1da177e4 LT |
489 | |
490 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
491 | POSIX-2001 defines it. */ | |
4c780a46 | 492 | if (l->l_len > 0) { |
ef12e72a BF |
493 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
494 | return -EOVERFLOW; | |
16238415 | 495 | fl->fl_end = fl->fl_start + (l->l_len - 1); |
ef12e72a | 496 | |
4c780a46 | 497 | } else if (l->l_len < 0) { |
ef12e72a | 498 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 499 | return -EINVAL; |
ef12e72a BF |
500 | fl->fl_end = fl->fl_start - 1; |
501 | fl->fl_start += l->l_len; | |
502 | } else | |
503 | fl->fl_end = OFFSET_MAX; | |
504 | ||
1da177e4 LT |
505 | fl->fl_owner = current->files; |
506 | fl->fl_pid = current->tgid; | |
507 | fl->fl_file = filp; | |
508 | fl->fl_flags = FL_POSIX; | |
509 | fl->fl_ops = NULL; | |
510 | fl->fl_lmops = NULL; | |
511 | ||
512 | return assign_type(fl, l->l_type); | |
513 | } | |
514 | ||
ef12e72a BF |
515 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
516 | * style lock. | |
517 | */ | |
518 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
519 | struct flock *l) | |
1da177e4 | 520 | { |
ef12e72a BF |
521 | struct flock64 ll = { |
522 | .l_type = l->l_type, | |
523 | .l_whence = l->l_whence, | |
524 | .l_start = l->l_start, | |
525 | .l_len = l->l_len, | |
526 | }; | |
527 | ||
528 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 529 | } |
1da177e4 LT |
530 | |
531 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
532 | static bool |
533 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
534 | { |
535 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 536 | return false; |
1da177e4 LT |
537 | } |
538 | ||
1c7dd2ff JL |
539 | static void |
540 | lease_setup(struct file_lock *fl, void **priv) | |
541 | { | |
542 | struct file *filp = fl->fl_file; | |
543 | struct fasync_struct *fa = *priv; | |
544 | ||
545 | /* | |
546 | * fasync_insert_entry() returns the old entry if any. If there was no | |
547 | * old entry, then it used "priv" and inserted it into the fasync list. | |
548 | * Clear the pointer to indicate that it shouldn't be freed. | |
549 | */ | |
550 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
551 | *priv = NULL; | |
552 | ||
01919134 | 553 | __f_setown(filp, task_pid(current), PIDTYPE_TGID, 0); |
1c7dd2ff JL |
554 | } |
555 | ||
7b021967 | 556 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 557 | .lm_break = lease_break_callback, |
8fb47a4f | 558 | .lm_change = lease_modify, |
1c7dd2ff | 559 | .lm_setup = lease_setup, |
1da177e4 LT |
560 | }; |
561 | ||
562 | /* | |
563 | * Initialize a lease, use the default lock manager operations | |
564 | */ | |
0ec4f431 | 565 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
447a5647 | 566 | { |
75dff55a TM |
567 | if (assign_type(fl, type) != 0) |
568 | return -EINVAL; | |
569 | ||
7ca76311 | 570 | fl->fl_owner = filp; |
1da177e4 LT |
571 | fl->fl_pid = current->tgid; |
572 | ||
573 | fl->fl_file = filp; | |
574 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
575 | fl->fl_start = 0; |
576 | fl->fl_end = OFFSET_MAX; | |
577 | fl->fl_ops = NULL; | |
578 | fl->fl_lmops = &lease_manager_ops; | |
579 | return 0; | |
580 | } | |
581 | ||
582 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 583 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
584 | { |
585 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 586 | int error = -ENOMEM; |
1da177e4 LT |
587 | |
588 | if (fl == NULL) | |
e32b8ee2 | 589 | return ERR_PTR(error); |
1da177e4 LT |
590 | |
591 | error = lease_init(filp, type, fl); | |
75dff55a TM |
592 | if (error) { |
593 | locks_free_lock(fl); | |
e32b8ee2 | 594 | return ERR_PTR(error); |
75dff55a | 595 | } |
e32b8ee2 | 596 | return fl; |
1da177e4 LT |
597 | } |
598 | ||
599 | /* Check if two locks overlap each other. | |
600 | */ | |
601 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
602 | { | |
603 | return ((fl1->fl_end >= fl2->fl_start) && | |
604 | (fl2->fl_end >= fl1->fl_start)); | |
605 | } | |
606 | ||
607 | /* | |
608 | * Check whether two locks have the same owner. | |
609 | */ | |
33443c42 | 610 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 611 | { |
1da177e4 LT |
612 | return fl1->fl_owner == fl2->fl_owner; |
613 | } | |
614 | ||
6109c850 | 615 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 616 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 617 | { |
7c3f654d PZ |
618 | struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list); |
619 | ||
aba37660 PZ |
620 | percpu_rwsem_assert_held(&file_rwsem); |
621 | ||
7c3f654d | 622 | spin_lock(&fll->lock); |
7012b02a | 623 | fl->fl_link_cpu = smp_processor_id(); |
7c3f654d PZ |
624 | hlist_add_head(&fl->fl_link, &fll->hlist); |
625 | spin_unlock(&fll->lock); | |
88974691 JL |
626 | } |
627 | ||
6109c850 | 628 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 629 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 630 | { |
7c3f654d PZ |
631 | struct file_lock_list_struct *fll; |
632 | ||
aba37660 PZ |
633 | percpu_rwsem_assert_held(&file_rwsem); |
634 | ||
7012b02a JL |
635 | /* |
636 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 637 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
638 | * also require that it be held. |
639 | */ | |
640 | if (hlist_unhashed(&fl->fl_link)) | |
641 | return; | |
7c3f654d PZ |
642 | |
643 | fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu); | |
644 | spin_lock(&fll->lock); | |
139ca04e | 645 | hlist_del_init(&fl->fl_link); |
7c3f654d | 646 | spin_unlock(&fll->lock); |
88974691 JL |
647 | } |
648 | ||
3999e493 JL |
649 | static unsigned long |
650 | posix_owner_key(struct file_lock *fl) | |
651 | { | |
3999e493 JL |
652 | return (unsigned long)fl->fl_owner; |
653 | } | |
654 | ||
6ca10ed8 | 655 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 656 | { |
663d5af7 DW |
657 | lockdep_assert_held(&blocked_lock_lock); |
658 | ||
3999e493 | 659 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
660 | } |
661 | ||
6ca10ed8 | 662 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 663 | { |
663d5af7 DW |
664 | lockdep_assert_held(&blocked_lock_lock); |
665 | ||
48f74186 | 666 | hash_del(&waiter->fl_link); |
88974691 JL |
667 | } |
668 | ||
1da177e4 LT |
669 | /* Remove waiter from blocker's block list. |
670 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 671 | * |
7b2296af | 672 | * Must be called with blocked_lock_lock held. |
1da177e4 | 673 | */ |
33443c42 | 674 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 675 | { |
88974691 | 676 | locks_delete_global_blocked(waiter); |
ada5c1da | 677 | list_del_init(&waiter->fl_blocked_member); |
1da177e4 LT |
678 | } |
679 | ||
ad6bbd8b N |
680 | static void __locks_wake_up_blocks(struct file_lock *blocker) |
681 | { | |
682 | while (!list_empty(&blocker->fl_blocked_requests)) { | |
683 | struct file_lock *waiter; | |
684 | ||
685 | waiter = list_first_entry(&blocker->fl_blocked_requests, | |
686 | struct file_lock, fl_blocked_member); | |
687 | __locks_delete_block(waiter); | |
688 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) | |
689 | waiter->fl_lmops->lm_notify(waiter); | |
690 | else | |
691 | wake_up(&waiter->fl_wait); | |
dcf23ac3 LT |
692 | |
693 | /* | |
694 | * The setting of fl_blocker to NULL marks the "done" | |
695 | * point in deleting a block. Paired with acquire at the top | |
696 | * of locks_delete_block(). | |
697 | */ | |
698 | smp_store_release(&waiter->fl_blocker, NULL); | |
ad6bbd8b N |
699 | } |
700 | } | |
701 | ||
cb03f94f | 702 | /** |
529adfe8 | 703 | * locks_delete_block - stop waiting for a file lock |
cb03f94f N |
704 | * @waiter: the lock which was waiting |
705 | * | |
706 | * lockd/nfsd need to disconnect the lock while working on it. | |
707 | */ | |
708 | int locks_delete_block(struct file_lock *waiter) | |
1da177e4 | 709 | { |
cb03f94f N |
710 | int status = -ENOENT; |
711 | ||
dcf23ac3 LT |
712 | /* |
713 | * If fl_blocker is NULL, it won't be set again as this thread "owns" | |
714 | * the lock and is the only one that might try to claim the lock. | |
715 | * | |
716 | * We use acquire/release to manage fl_blocker so that we can | |
717 | * optimize away taking the blocked_lock_lock in many cases. | |
718 | * | |
719 | * The smp_load_acquire guarantees two things: | |
720 | * | |
721 | * 1/ that fl_blocked_requests can be tested locklessly. If something | |
722 | * was recently added to that list it must have been in a locked region | |
723 | * *before* the locked region when fl_blocker was set to NULL. | |
724 | * | |
725 | * 2/ that no other thread is accessing 'waiter', so it is safe to free | |
726 | * it. __locks_wake_up_blocks is careful not to touch waiter after | |
727 | * fl_blocker is released. | |
728 | * | |
729 | * If a lockless check of fl_blocker shows it to be NULL, we know that | |
730 | * no new locks can be inserted into its fl_blocked_requests list, and | |
731 | * can avoid doing anything further if the list is empty. | |
732 | */ | |
733 | if (!smp_load_acquire(&waiter->fl_blocker) && | |
734 | list_empty(&waiter->fl_blocked_requests)) | |
735 | return status; | |
736 | ||
7b2296af | 737 | spin_lock(&blocked_lock_lock); |
cb03f94f N |
738 | if (waiter->fl_blocker) |
739 | status = 0; | |
5946c431 | 740 | __locks_wake_up_blocks(waiter); |
1da177e4 | 741 | __locks_delete_block(waiter); |
dcf23ac3 LT |
742 | |
743 | /* | |
744 | * The setting of fl_blocker to NULL marks the "done" point in deleting | |
745 | * a block. Paired with acquire at the top of this function. | |
746 | */ | |
747 | smp_store_release(&waiter->fl_blocker, NULL); | |
7b2296af | 748 | spin_unlock(&blocked_lock_lock); |
cb03f94f | 749 | return status; |
1da177e4 | 750 | } |
cb03f94f | 751 | EXPORT_SYMBOL(locks_delete_block); |
1da177e4 LT |
752 | |
753 | /* Insert waiter into blocker's block list. | |
754 | * We use a circular list so that processes can be easily woken up in | |
755 | * the order they blocked. The documentation doesn't require this but | |
756 | * it seems like the reasonable thing to do. | |
1c8c601a | 757 | * |
6109c850 | 758 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
ada5c1da N |
759 | * fl_blocked_requests list itself is protected by the blocked_lock_lock, |
760 | * but by ensuring that the flc_lock is also held on insertions we can avoid | |
761 | * taking the blocked_lock_lock in some cases when we see that the | |
762 | * fl_blocked_requests list is empty. | |
fd7732e0 N |
763 | * |
764 | * Rather than just adding to the list, we check for conflicts with any existing | |
765 | * waiters, and add beneath any waiter that blocks the new waiter. | |
766 | * Thus wakeups don't happen until needed. | |
1da177e4 | 767 | */ |
1c8c601a | 768 | static void __locks_insert_block(struct file_lock *blocker, |
fd7732e0 N |
769 | struct file_lock *waiter, |
770 | bool conflict(struct file_lock *, | |
771 | struct file_lock *)) | |
1da177e4 | 772 | { |
fd7732e0 | 773 | struct file_lock *fl; |
ada5c1da | 774 | BUG_ON(!list_empty(&waiter->fl_blocked_member)); |
fd7732e0 N |
775 | |
776 | new_blocker: | |
777 | list_for_each_entry(fl, &blocker->fl_blocked_requests, fl_blocked_member) | |
778 | if (conflict(fl, waiter)) { | |
779 | blocker = fl; | |
780 | goto new_blocker; | |
781 | } | |
ada5c1da N |
782 | waiter->fl_blocker = blocker; |
783 | list_add_tail(&waiter->fl_blocked_member, &blocker->fl_blocked_requests); | |
cff2fce5 | 784 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a | 785 | locks_insert_global_blocked(waiter); |
5946c431 N |
786 | |
787 | /* The requests in waiter->fl_blocked are known to conflict with | |
788 | * waiter, but might not conflict with blocker, or the requests | |
789 | * and lock which block it. So they all need to be woken. | |
790 | */ | |
791 | __locks_wake_up_blocks(waiter); | |
1c8c601a JL |
792 | } |
793 | ||
6109c850 | 794 | /* Must be called with flc_lock held. */ |
1c8c601a | 795 | static void locks_insert_block(struct file_lock *blocker, |
fd7732e0 N |
796 | struct file_lock *waiter, |
797 | bool conflict(struct file_lock *, | |
798 | struct file_lock *)) | |
1c8c601a | 799 | { |
7b2296af | 800 | spin_lock(&blocked_lock_lock); |
fd7732e0 | 801 | __locks_insert_block(blocker, waiter, conflict); |
7b2296af | 802 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
803 | } |
804 | ||
1cb36012 JL |
805 | /* |
806 | * Wake up processes blocked waiting for blocker. | |
807 | * | |
6109c850 | 808 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
809 | */ |
810 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
811 | { | |
4e8c765d JL |
812 | /* |
813 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 | 814 | * blocked requests are only added to the list under the flc_lock, and |
ada5c1da N |
815 | * the flc_lock is always held here. Note that removal from the |
816 | * fl_blocked_requests list does not require the flc_lock, so we must | |
817 | * recheck list_empty() after acquiring the blocked_lock_lock. | |
4e8c765d | 818 | */ |
ada5c1da | 819 | if (list_empty(&blocker->fl_blocked_requests)) |
4e8c765d JL |
820 | return; |
821 | ||
7b2296af | 822 | spin_lock(&blocked_lock_lock); |
ad6bbd8b | 823 | __locks_wake_up_blocks(blocker); |
7b2296af | 824 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
825 | } |
826 | ||
5263e31e | 827 | static void |
e084c1bd | 828 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) |
5263e31e | 829 | { |
5263e31e JL |
830 | list_add_tail(&fl->fl_list, before); |
831 | locks_insert_global_locks(fl); | |
832 | } | |
833 | ||
8634b51f | 834 | static void |
e084c1bd | 835 | locks_unlink_lock_ctx(struct file_lock *fl) |
1da177e4 | 836 | { |
88974691 | 837 | locks_delete_global_locks(fl); |
8634b51f | 838 | list_del_init(&fl->fl_list); |
1da177e4 | 839 | locks_wake_up_blocks(fl); |
24cbe784 JL |
840 | } |
841 | ||
8634b51f | 842 | static void |
e084c1bd | 843 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) |
24cbe784 | 844 | { |
e084c1bd | 845 | locks_unlink_lock_ctx(fl); |
ed9814d8 | 846 | if (dispose) |
6dee60f6 | 847 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
848 | else |
849 | locks_free_lock(fl); | |
1da177e4 LT |
850 | } |
851 | ||
852 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
853 | * checks for shared/exclusive status of overlapping locks. | |
854 | */ | |
c0e15908 N |
855 | static bool locks_conflict(struct file_lock *caller_fl, |
856 | struct file_lock *sys_fl) | |
1da177e4 LT |
857 | { |
858 | if (sys_fl->fl_type == F_WRLCK) | |
c0e15908 | 859 | return true; |
1da177e4 | 860 | if (caller_fl->fl_type == F_WRLCK) |
c0e15908 N |
861 | return true; |
862 | return false; | |
1da177e4 LT |
863 | } |
864 | ||
865 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
866 | * checking before calling the locks_conflict(). | |
867 | */ | |
c0e15908 N |
868 | static bool posix_locks_conflict(struct file_lock *caller_fl, |
869 | struct file_lock *sys_fl) | |
1da177e4 LT |
870 | { |
871 | /* POSIX locks owned by the same process do not conflict with | |
872 | * each other. | |
873 | */ | |
9b8c8695 | 874 | if (posix_same_owner(caller_fl, sys_fl)) |
c0e15908 | 875 | return false; |
1da177e4 LT |
876 | |
877 | /* Check whether they overlap */ | |
878 | if (!locks_overlap(caller_fl, sys_fl)) | |
c0e15908 | 879 | return false; |
1da177e4 | 880 | |
c0e15908 | 881 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
882 | } |
883 | ||
884 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
885 | * checking before calling the locks_conflict(). | |
886 | */ | |
c0e15908 N |
887 | static bool flock_locks_conflict(struct file_lock *caller_fl, |
888 | struct file_lock *sys_fl) | |
1da177e4 LT |
889 | { |
890 | /* FLOCK locks referring to the same filp do not conflict with | |
891 | * each other. | |
892 | */ | |
9b8c8695 | 893 | if (caller_fl->fl_file == sys_fl->fl_file) |
c0e15908 | 894 | return false; |
1da177e4 | 895 | |
c0e15908 | 896 | return locks_conflict(caller_fl, sys_fl); |
1da177e4 LT |
897 | } |
898 | ||
6d34ac19 | 899 | void |
9d6a8c5c | 900 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
901 | { |
902 | struct file_lock *cfl; | |
bd61e0a9 | 903 | struct file_lock_context *ctx; |
c568d683 | 904 | struct inode *inode = locks_inode(filp); |
2443da22 DN |
905 | void *owner; |
906 | void (*func)(void); | |
1da177e4 | 907 | |
128a3785 | 908 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 JL |
909 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { |
910 | fl->fl_type = F_UNLCK; | |
911 | return; | |
912 | } | |
913 | ||
2443da22 | 914 | retry: |
6109c850 | 915 | spin_lock(&ctx->flc_lock); |
bd61e0a9 | 916 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
2443da22 DN |
917 | if (!posix_locks_conflict(fl, cfl)) |
918 | continue; | |
919 | if (cfl->fl_lmops && cfl->fl_lmops->lm_lock_expirable | |
920 | && (*cfl->fl_lmops->lm_lock_expirable)(cfl)) { | |
921 | owner = cfl->fl_lmops->lm_mod_owner; | |
922 | func = cfl->fl_lmops->lm_expire_lock; | |
923 | __module_get(owner); | |
924 | spin_unlock(&ctx->flc_lock); | |
925 | (*func)(); | |
926 | module_put(owner); | |
927 | goto retry; | |
bd61e0a9 | 928 | } |
2443da22 DN |
929 | locks_copy_conflock(fl, cfl); |
930 | goto out; | |
1da177e4 | 931 | } |
bd61e0a9 JL |
932 | fl->fl_type = F_UNLCK; |
933 | out: | |
6109c850 | 934 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 935 | return; |
1da177e4 | 936 | } |
1da177e4 LT |
937 | EXPORT_SYMBOL(posix_test_lock); |
938 | ||
b533184f BF |
939 | /* |
940 | * Deadlock detection: | |
941 | * | |
942 | * We attempt to detect deadlocks that are due purely to posix file | |
943 | * locks. | |
1da177e4 | 944 | * |
b533184f BF |
945 | * We assume that a task can be waiting for at most one lock at a time. |
946 | * So for any acquired lock, the process holding that lock may be | |
947 | * waiting on at most one other lock. That lock in turns may be held by | |
948 | * someone waiting for at most one other lock. Given a requested lock | |
949 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
950 | * follow this chain of waiters to ensure we are not about to create a | |
951 | * cycle. | |
1da177e4 | 952 | * |
b533184f BF |
953 | * Since we do this before we ever put a process to sleep on a lock, we |
954 | * are ensured that there is never a cycle; that is what guarantees that | |
955 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 956 | * |
b533184f BF |
957 | * Note: the above assumption may not be true when handling lock |
958 | * requests from a broken NFS client. It may also fail in the presence | |
959 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 960 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 961 | * |
cff2fce5 | 962 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
963 | * Because the owner is not even nominally tied to a thread of |
964 | * execution, the deadlock detection below can't reasonably work well. Just | |
965 | * skip it for those. | |
966 | * | |
cff2fce5 | 967 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
968 | * locks that just checks for the case where two tasks are attempting to |
969 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 970 | */ |
97855b49 BF |
971 | |
972 | #define MAX_DEADLK_ITERATIONS 10 | |
973 | ||
b533184f BF |
974 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
975 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
976 | { | |
977 | struct file_lock *fl; | |
978 | ||
3999e493 | 979 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
5946c431 N |
980 | if (posix_same_owner(fl, block_fl)) { |
981 | while (fl->fl_blocker) | |
982 | fl = fl->fl_blocker; | |
983 | return fl; | |
984 | } | |
b533184f BF |
985 | } |
986 | return NULL; | |
987 | } | |
988 | ||
7b2296af | 989 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 990 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
991 | struct file_lock *block_fl) |
992 | { | |
97855b49 | 993 | int i = 0; |
1da177e4 | 994 | |
663d5af7 DW |
995 | lockdep_assert_held(&blocked_lock_lock); |
996 | ||
57b65325 JL |
997 | /* |
998 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 999 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 1000 | */ |
cff2fce5 | 1001 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
1002 | return 0; |
1003 | ||
b533184f BF |
1004 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
1005 | if (i++ > MAX_DEADLK_ITERATIONS) | |
1006 | return 0; | |
1007 | if (posix_same_owner(caller_fl, block_fl)) | |
1008 | return 1; | |
1da177e4 LT |
1009 | } |
1010 | return 0; | |
1011 | } | |
1012 | ||
1da177e4 | 1013 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 1014 | * after any leases, but before any posix locks. |
f475ae95 TM |
1015 | * |
1016 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1017 | * whether or not a lock was successfully freed by testing the return | |
1018 | * value for -ENOENT. | |
1da177e4 | 1019 | */ |
bcd7f78d | 1020 | static int flock_lock_inode(struct inode *inode, struct file_lock *request) |
1da177e4 | 1021 | { |
993dfa87 | 1022 | struct file_lock *new_fl = NULL; |
5263e31e JL |
1023 | struct file_lock *fl; |
1024 | struct file_lock_context *ctx; | |
1da177e4 | 1025 | int error = 0; |
5263e31e | 1026 | bool found = false; |
ed9814d8 | 1027 | LIST_HEAD(dispose); |
1da177e4 | 1028 | |
5c1c669a JL |
1029 | ctx = locks_get_lock_context(inode, request->fl_type); |
1030 | if (!ctx) { | |
1031 | if (request->fl_type != F_UNLCK) | |
1032 | return -ENOMEM; | |
1033 | return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0; | |
1034 | } | |
5263e31e | 1035 | |
b89f4321 | 1036 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 1037 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
1038 | if (!new_fl) |
1039 | return -ENOMEM; | |
84d535ad PE |
1040 | } |
1041 | ||
02e525b2 | 1042 | percpu_down_read(&file_rwsem); |
6109c850 | 1043 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
1044 | if (request->fl_flags & FL_ACCESS) |
1045 | goto find_conflict; | |
1046 | ||
5263e31e | 1047 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
bcd7f78d | 1048 | if (request->fl_file != fl->fl_file) |
1da177e4 | 1049 | continue; |
993dfa87 | 1050 | if (request->fl_type == fl->fl_type) |
1da177e4 | 1051 | goto out; |
5263e31e | 1052 | found = true; |
e084c1bd | 1053 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1054 | break; |
1055 | } | |
1da177e4 | 1056 | |
f475ae95 TM |
1057 | if (request->fl_type == F_UNLCK) { |
1058 | if ((request->fl_flags & FL_EXISTS) && !found) | |
1059 | error = -ENOENT; | |
993dfa87 | 1060 | goto out; |
f475ae95 | 1061 | } |
1da177e4 | 1062 | |
f07f18dd | 1063 | find_conflict: |
5263e31e | 1064 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 1065 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
1066 | continue; |
1067 | error = -EAGAIN; | |
bde74e4b MS |
1068 | if (!(request->fl_flags & FL_SLEEP)) |
1069 | goto out; | |
1070 | error = FILE_LOCK_DEFERRED; | |
fd7732e0 | 1071 | locks_insert_block(fl, request, flock_locks_conflict); |
1da177e4 LT |
1072 | goto out; |
1073 | } | |
f07f18dd TM |
1074 | if (request->fl_flags & FL_ACCESS) |
1075 | goto out; | |
993dfa87 | 1076 | locks_copy_lock(new_fl, request); |
5946c431 | 1077 | locks_move_blocks(new_fl, request); |
e084c1bd | 1078 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 1079 | new_fl = NULL; |
9cedc194 | 1080 | error = 0; |
1da177e4 LT |
1081 | |
1082 | out: | |
6109c850 | 1083 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1084 | percpu_up_read(&file_rwsem); |
993dfa87 TM |
1085 | if (new_fl) |
1086 | locks_free_lock(new_fl); | |
ed9814d8 | 1087 | locks_dispose_list(&dispose); |
c883da31 | 1088 | trace_flock_lock_inode(inode, request, error); |
1da177e4 LT |
1089 | return error; |
1090 | } | |
1091 | ||
b4d629a3 JL |
1092 | static int posix_lock_inode(struct inode *inode, struct file_lock *request, |
1093 | struct file_lock *conflock) | |
1da177e4 | 1094 | { |
bd61e0a9 | 1095 | struct file_lock *fl, *tmp; |
39005d02 MS |
1096 | struct file_lock *new_fl = NULL; |
1097 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
1098 | struct file_lock *left = NULL; |
1099 | struct file_lock *right = NULL; | |
bd61e0a9 | 1100 | struct file_lock_context *ctx; |
b9746ef8 JL |
1101 | int error; |
1102 | bool added = false; | |
ed9814d8 | 1103 | LIST_HEAD(dispose); |
2443da22 DN |
1104 | void *owner; |
1105 | void (*func)(void); | |
1da177e4 | 1106 | |
5c1c669a | 1107 | ctx = locks_get_lock_context(inode, request->fl_type); |
bd61e0a9 | 1108 | if (!ctx) |
5c1c669a | 1109 | return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM; |
bd61e0a9 | 1110 | |
1da177e4 LT |
1111 | /* |
1112 | * We may need two file_lock structures for this operation, | |
1113 | * so we get them in advance to avoid races. | |
39005d02 MS |
1114 | * |
1115 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1116 | */ |
39005d02 MS |
1117 | if (!(request->fl_flags & FL_ACCESS) && |
1118 | (request->fl_type != F_UNLCK || | |
1119 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1120 | new_fl = locks_alloc_lock(); | |
1121 | new_fl2 = locks_alloc_lock(); | |
1122 | } | |
1da177e4 | 1123 | |
2443da22 | 1124 | retry: |
02e525b2 | 1125 | percpu_down_read(&file_rwsem); |
6109c850 | 1126 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
1127 | /* |
1128 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1129 | * there are any, either return error or put the request on the | |
48f74186 | 1130 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1131 | */ |
1da177e4 | 1132 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1133 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1134 | if (!posix_locks_conflict(request, fl)) |
1135 | continue; | |
2443da22 DN |
1136 | if (fl->fl_lmops && fl->fl_lmops->lm_lock_expirable |
1137 | && (*fl->fl_lmops->lm_lock_expirable)(fl)) { | |
1138 | owner = fl->fl_lmops->lm_mod_owner; | |
1139 | func = fl->fl_lmops->lm_expire_lock; | |
1140 | __module_get(owner); | |
1141 | spin_unlock(&ctx->flc_lock); | |
1142 | percpu_up_read(&file_rwsem); | |
1143 | (*func)(); | |
1144 | module_put(owner); | |
1145 | goto retry; | |
1146 | } | |
5842add2 | 1147 | if (conflock) |
3fe0fff1 | 1148 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1149 | error = -EAGAIN; |
1150 | if (!(request->fl_flags & FL_SLEEP)) | |
1151 | goto out; | |
1c8c601a JL |
1152 | /* |
1153 | * Deadlock detection and insertion into the blocked | |
1154 | * locks list must be done while holding the same lock! | |
1155 | */ | |
1da177e4 | 1156 | error = -EDEADLK; |
7b2296af | 1157 | spin_lock(&blocked_lock_lock); |
945ab8f6 JL |
1158 | /* |
1159 | * Ensure that we don't find any locks blocked on this | |
1160 | * request during deadlock detection. | |
1161 | */ | |
1162 | __locks_wake_up_blocks(request); | |
1c8c601a JL |
1163 | if (likely(!posix_locks_deadlock(request, fl))) { |
1164 | error = FILE_LOCK_DEFERRED; | |
fd7732e0 N |
1165 | __locks_insert_block(fl, request, |
1166 | posix_locks_conflict); | |
1c8c601a | 1167 | } |
7b2296af | 1168 | spin_unlock(&blocked_lock_lock); |
1da177e4 | 1169 | goto out; |
7bbd1fc0 N |
1170 | } |
1171 | } | |
1da177e4 LT |
1172 | |
1173 | /* If we're just looking for a conflict, we're done. */ | |
1174 | error = 0; | |
1175 | if (request->fl_flags & FL_ACCESS) | |
1176 | goto out; | |
1177 | ||
bd61e0a9 JL |
1178 | /* Find the first old lock with the same owner as the new lock */ |
1179 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1180 | if (posix_same_owner(request, fl)) | |
1181 | break; | |
1da177e4 LT |
1182 | } |
1183 | ||
1cb36012 | 1184 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1185 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1186 | if (!posix_same_owner(request, fl)) | |
1187 | break; | |
1188 | ||
1189 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1190 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1191 | /* In all comparisons of start vs end, use |
1192 | * "start - 1" rather than "end + 1". If end | |
1193 | * is OFFSET_MAX, end + 1 will become negative. | |
1194 | */ | |
1da177e4 | 1195 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1196 | continue; |
1da177e4 LT |
1197 | /* If the next lock in the list has entirely bigger |
1198 | * addresses than the new one, insert the lock here. | |
1199 | */ | |
449231d6 | 1200 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1201 | break; |
1202 | ||
1203 | /* If we come here, the new and old lock are of the | |
1204 | * same type and adjacent or overlapping. Make one | |
1205 | * lock yielding from the lower start address of both | |
1206 | * locks to the higher end address. | |
1207 | */ | |
1208 | if (fl->fl_start > request->fl_start) | |
1209 | fl->fl_start = request->fl_start; | |
1210 | else | |
1211 | request->fl_start = fl->fl_start; | |
1212 | if (fl->fl_end < request->fl_end) | |
1213 | fl->fl_end = request->fl_end; | |
1214 | else | |
1215 | request->fl_end = fl->fl_end; | |
1216 | if (added) { | |
e084c1bd | 1217 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1218 | continue; |
1219 | } | |
1220 | request = fl; | |
b9746ef8 | 1221 | added = true; |
bd61e0a9 | 1222 | } else { |
1da177e4 LT |
1223 | /* Processing for different lock types is a bit |
1224 | * more complex. | |
1225 | */ | |
1226 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1227 | continue; |
1da177e4 LT |
1228 | if (fl->fl_start > request->fl_end) |
1229 | break; | |
1230 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1231 | added = true; |
1da177e4 LT |
1232 | if (fl->fl_start < request->fl_start) |
1233 | left = fl; | |
1234 | /* If the next lock in the list has a higher end | |
1235 | * address than the new one, insert the new one here. | |
1236 | */ | |
1237 | if (fl->fl_end > request->fl_end) { | |
1238 | right = fl; | |
1239 | break; | |
1240 | } | |
1241 | if (fl->fl_start >= request->fl_start) { | |
1242 | /* The new lock completely replaces an old | |
1243 | * one (This may happen several times). | |
1244 | */ | |
1245 | if (added) { | |
e084c1bd | 1246 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1247 | continue; |
1248 | } | |
b84d49f9 JL |
1249 | /* |
1250 | * Replace the old lock with new_fl, and | |
1251 | * remove the old one. It's safe to do the | |
1252 | * insert here since we know that we won't be | |
1253 | * using new_fl later, and that the lock is | |
1254 | * just replacing an existing lock. | |
1da177e4 | 1255 | */ |
b84d49f9 JL |
1256 | error = -ENOLCK; |
1257 | if (!new_fl) | |
1258 | goto out; | |
1259 | locks_copy_lock(new_fl, request); | |
5ef15968 | 1260 | locks_move_blocks(new_fl, request); |
b84d49f9 JL |
1261 | request = new_fl; |
1262 | new_fl = NULL; | |
e084c1bd JL |
1263 | locks_insert_lock_ctx(request, &fl->fl_list); |
1264 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1265 | added = true; |
1da177e4 LT |
1266 | } |
1267 | } | |
1da177e4 LT |
1268 | } |
1269 | ||
0d9a490a | 1270 | /* |
1cb36012 JL |
1271 | * The above code only modifies existing locks in case of merging or |
1272 | * replacing. If new lock(s) need to be inserted all modifications are | |
1273 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1274 | */ |
1275 | error = -ENOLCK; /* "no luck" */ | |
1276 | if (right && left == right && !new_fl2) | |
1277 | goto out; | |
1278 | ||
1da177e4 LT |
1279 | error = 0; |
1280 | if (!added) { | |
f475ae95 TM |
1281 | if (request->fl_type == F_UNLCK) { |
1282 | if (request->fl_flags & FL_EXISTS) | |
1283 | error = -ENOENT; | |
1da177e4 | 1284 | goto out; |
f475ae95 | 1285 | } |
0d9a490a MS |
1286 | |
1287 | if (!new_fl) { | |
1288 | error = -ENOLCK; | |
1289 | goto out; | |
1290 | } | |
1da177e4 | 1291 | locks_copy_lock(new_fl, request); |
5946c431 | 1292 | locks_move_blocks(new_fl, request); |
e084c1bd | 1293 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
2e2f756f | 1294 | fl = new_fl; |
1da177e4 LT |
1295 | new_fl = NULL; |
1296 | } | |
1297 | if (right) { | |
1298 | if (left == right) { | |
1299 | /* The new lock breaks the old one in two pieces, | |
1300 | * so we have to use the second new lock. | |
1301 | */ | |
1302 | left = new_fl2; | |
1303 | new_fl2 = NULL; | |
1304 | locks_copy_lock(left, right); | |
e084c1bd | 1305 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1306 | } |
1307 | right->fl_start = request->fl_end + 1; | |
1308 | locks_wake_up_blocks(right); | |
1309 | } | |
1310 | if (left) { | |
1311 | left->fl_end = request->fl_start - 1; | |
1312 | locks_wake_up_blocks(left); | |
1313 | } | |
1314 | out: | |
6109c850 | 1315 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1316 | percpu_up_read(&file_rwsem); |
1da177e4 LT |
1317 | /* |
1318 | * Free any unused locks. | |
1319 | */ | |
1320 | if (new_fl) | |
1321 | locks_free_lock(new_fl); | |
1322 | if (new_fl2) | |
1323 | locks_free_lock(new_fl2); | |
ed9814d8 | 1324 | locks_dispose_list(&dispose); |
1890910f JL |
1325 | trace_posix_lock_inode(inode, request, error); |
1326 | ||
1da177e4 LT |
1327 | return error; |
1328 | } | |
1329 | ||
1330 | /** | |
1331 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1332 | * @filp: The file to apply the lock to | |
1333 | * @fl: The lock to be applied | |
150b3934 | 1334 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1335 | * |
1336 | * Add a POSIX style lock to a file. | |
1337 | * We merge adjacent & overlapping locks whenever possible. | |
1338 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1339 | * |
1340 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1341 | * whether or not a lock was successfully freed by testing the return | |
1342 | * value for -ENOENT. | |
1da177e4 | 1343 | */ |
150b3934 | 1344 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1345 | struct file_lock *conflock) |
1346 | { | |
c568d683 | 1347 | return posix_lock_inode(locks_inode(filp), fl, conflock); |
1da177e4 | 1348 | } |
150b3934 | 1349 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1350 | |
1351 | /** | |
29d01b22 JL |
1352 | * posix_lock_inode_wait - Apply a POSIX-style lock to a file |
1353 | * @inode: inode of file to which lock request should be applied | |
1da177e4 LT |
1354 | * @fl: The lock to be applied |
1355 | * | |
616fb38f | 1356 | * Apply a POSIX style lock request to an inode. |
1da177e4 | 1357 | */ |
616fb38f | 1358 | static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
1359 | { |
1360 | int error; | |
1361 | might_sleep (); | |
1362 | for (;;) { | |
b4d629a3 | 1363 | error = posix_lock_inode(inode, fl, NULL); |
bde74e4b | 1364 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 1365 | break; |
dcf23ac3 LT |
1366 | error = wait_event_interruptible(fl->fl_wait, |
1367 | list_empty(&fl->fl_blocked_member)); | |
16306a61 N |
1368 | if (error) |
1369 | break; | |
1da177e4 | 1370 | } |
16306a61 | 1371 | locks_delete_block(fl); |
1da177e4 LT |
1372 | return error; |
1373 | } | |
29d01b22 | 1374 | |
778fc546 BF |
1375 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1376 | { | |
1377 | switch (arg) { | |
1378 | case F_UNLCK: | |
1379 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
df561f66 | 1380 | fallthrough; |
778fc546 BF |
1381 | case F_RDLCK: |
1382 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1383 | } | |
1384 | } | |
1385 | ||
1da177e4 | 1386 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1387 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1388 | { |
1da177e4 LT |
1389 | int error = assign_type(fl, arg); |
1390 | ||
1391 | if (error) | |
1392 | return error; | |
778fc546 | 1393 | lease_clear_pending(fl, arg); |
1da177e4 | 1394 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1395 | if (arg == F_UNLCK) { |
1396 | struct file *filp = fl->fl_file; | |
1397 | ||
1398 | f_delown(filp); | |
1399 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1400 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1401 | if (fl->fl_fasync != NULL) { | |
1402 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1403 | fl->fl_fasync = NULL; | |
1404 | } | |
e084c1bd | 1405 | locks_delete_lock_ctx(fl, dispose); |
3b6e2723 | 1406 | } |
1da177e4 LT |
1407 | return 0; |
1408 | } | |
1da177e4 LT |
1409 | EXPORT_SYMBOL(lease_modify); |
1410 | ||
778fc546 BF |
1411 | static bool past_time(unsigned long then) |
1412 | { | |
1413 | if (!then) | |
1414 | /* 0 is a special value meaning "this never expires": */ | |
1415 | return false; | |
1416 | return time_after(jiffies, then); | |
1417 | } | |
1418 | ||
c45198ed | 1419 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1420 | { |
8634b51f JL |
1421 | struct file_lock_context *ctx = inode->i_flctx; |
1422 | struct file_lock *fl, *tmp; | |
1da177e4 | 1423 | |
6109c850 | 1424 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1425 | |
8634b51f | 1426 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1427 | trace_time_out_leases(inode, fl); |
778fc546 | 1428 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1429 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1430 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1431 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1432 | } |
1433 | } | |
1434 | ||
df4e8d2c BF |
1435 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1436 | { | |
d51f527f IW |
1437 | bool rc; |
1438 | ||
28df3d15 BF |
1439 | if (lease->fl_lmops->lm_breaker_owns_lease |
1440 | && lease->fl_lmops->lm_breaker_owns_lease(lease)) | |
1441 | return false; | |
d51f527f IW |
1442 | if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) { |
1443 | rc = false; | |
1444 | goto trace; | |
1445 | } | |
1446 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) { | |
1447 | rc = false; | |
1448 | goto trace; | |
1449 | } | |
1450 | ||
1451 | rc = locks_conflict(breaker, lease); | |
1452 | trace: | |
1453 | trace_leases_conflict(rc, lease, breaker); | |
1454 | return rc; | |
df4e8d2c BF |
1455 | } |
1456 | ||
03d12ddf JL |
1457 | static bool |
1458 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1459 | { | |
8634b51f | 1460 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1461 | struct file_lock *fl; |
1462 | ||
6109c850 | 1463 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1464 | |
8634b51f | 1465 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1466 | if (leases_conflict(fl, breaker)) |
1467 | return true; | |
1468 | } | |
1469 | return false; | |
1470 | } | |
1471 | ||
1da177e4 LT |
1472 | /** |
1473 | * __break_lease - revoke all outstanding leases on file | |
1474 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1475 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1476 | * break all leases | |
1477 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1478 | * only delegations | |
1da177e4 | 1479 | * |
87250dd2 | 1480 | * break_lease (inlined for speed) has checked there already is at least |
1481 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1482 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1483 | * specified %O_NONBLOCK to your open(). |
1484 | */ | |
df4e8d2c | 1485 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1486 | { |
778fc546 | 1487 | int error = 0; |
128a3785 | 1488 | struct file_lock_context *ctx; |
a901125c | 1489 | struct file_lock *new_fl, *fl, *tmp; |
1da177e4 | 1490 | unsigned long break_time; |
8737c930 | 1491 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1492 | LIST_HEAD(dispose); |
1da177e4 | 1493 | |
8737c930 | 1494 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1495 | if (IS_ERR(new_fl)) |
1496 | return PTR_ERR(new_fl); | |
df4e8d2c | 1497 | new_fl->fl_flags = type; |
1da177e4 | 1498 | |
8634b51f | 1499 | /* typically we will check that ctx is non-NULL before calling */ |
128a3785 | 1500 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1501 | if (!ctx) { |
1502 | WARN_ON_ONCE(1); | |
cfddf9f4 | 1503 | goto free_lock; |
8634b51f JL |
1504 | } |
1505 | ||
02e525b2 | 1506 | percpu_down_read(&file_rwsem); |
6109c850 | 1507 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1508 | |
c45198ed | 1509 | time_out_leases(inode, &dispose); |
1da177e4 | 1510 | |
03d12ddf | 1511 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1512 | goto out; |
1513 | ||
1da177e4 LT |
1514 | break_time = 0; |
1515 | if (lease_break_time > 0) { | |
1516 | break_time = jiffies + lease_break_time * HZ; | |
1517 | if (break_time == 0) | |
1518 | break_time++; /* so that 0 means no break time */ | |
1519 | } | |
1520 | ||
a901125c | 1521 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1522 | if (!leases_conflict(fl, new_fl)) |
1523 | continue; | |
778fc546 BF |
1524 | if (want_write) { |
1525 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1526 | continue; | |
1527 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1528 | fl->fl_break_time = break_time; |
778fc546 | 1529 | } else { |
8634b51f | 1530 | if (lease_breaking(fl)) |
778fc546 BF |
1531 | continue; |
1532 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1533 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1534 | } |
4d01b7f5 | 1535 | if (fl->fl_lmops->lm_break(fl)) |
e084c1bd | 1536 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1537 | } |
1538 | ||
8634b51f | 1539 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1540 | goto out; |
1541 | ||
843c6b2f | 1542 | if (mode & O_NONBLOCK) { |
62af4f1f | 1543 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1544 | error = -EWOULDBLOCK; |
1545 | goto out; | |
1546 | } | |
1547 | ||
1548 | restart: | |
8634b51f JL |
1549 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1550 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1551 | if (break_time != 0) |
1da177e4 | 1552 | break_time -= jiffies; |
f1c6bb2c JL |
1553 | if (break_time == 0) |
1554 | break_time++; | |
fd7732e0 | 1555 | locks_insert_block(fl, new_fl, leases_conflict); |
62af4f1f | 1556 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1557 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1558 | percpu_up_read(&file_rwsem); |
aba37660 | 1559 | |
c45198ed | 1560 | locks_dispose_list(&dispose); |
4321e01e | 1561 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
dcf23ac3 LT |
1562 | list_empty(&new_fl->fl_blocked_member), |
1563 | break_time); | |
aba37660 | 1564 | |
02e525b2 | 1565 | percpu_down_read(&file_rwsem); |
6109c850 | 1566 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1567 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1568 | locks_delete_block(new_fl); |
1da177e4 | 1569 | if (error >= 0) { |
778fc546 BF |
1570 | /* |
1571 | * Wait for the next conflicting lease that has not been | |
1572 | * broken yet | |
1573 | */ | |
03d12ddf JL |
1574 | if (error == 0) |
1575 | time_out_leases(inode, &dispose); | |
1576 | if (any_leases_conflict(inode, new_fl)) | |
1577 | goto restart; | |
1da177e4 LT |
1578 | error = 0; |
1579 | } | |
1da177e4 | 1580 | out: |
6109c850 | 1581 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1582 | percpu_up_read(&file_rwsem); |
c45198ed | 1583 | locks_dispose_list(&dispose); |
cfddf9f4 | 1584 | free_lock: |
6d4b9e38 | 1585 | locks_free_lock(new_fl); |
1da177e4 LT |
1586 | return error; |
1587 | } | |
1da177e4 LT |
1588 | EXPORT_SYMBOL(__break_lease); |
1589 | ||
1590 | /** | |
76c47948 | 1591 | * lease_get_mtime - update modified time of an inode with exclusive lease |
1da177e4 | 1592 | * @inode: the inode |
76c47948 | 1593 | * @time: pointer to a timespec which contains the last modified time |
1da177e4 LT |
1594 | * |
1595 | * This is to force NFS clients to flush their caches for files with | |
1596 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1597 | * exclusive lease, then they could be modifying it. |
1da177e4 | 1598 | */ |
95582b00 | 1599 | void lease_get_mtime(struct inode *inode, struct timespec64 *time) |
1da177e4 | 1600 | { |
bfe86024 | 1601 | bool has_lease = false; |
128a3785 | 1602 | struct file_lock_context *ctx; |
8634b51f | 1603 | struct file_lock *fl; |
bfe86024 | 1604 | |
128a3785 | 1605 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1606 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1607 | spin_lock(&ctx->flc_lock); |
8ace5dfb GT |
1608 | fl = list_first_entry_or_null(&ctx->flc_lease, |
1609 | struct file_lock, fl_list); | |
1610 | if (fl && (fl->fl_type == F_WRLCK)) | |
1611 | has_lease = true; | |
6109c850 | 1612 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1613 | } |
1614 | ||
1615 | if (has_lease) | |
c2050a45 | 1616 | *time = current_time(inode); |
1da177e4 | 1617 | } |
1da177e4 LT |
1618 | EXPORT_SYMBOL(lease_get_mtime); |
1619 | ||
1620 | /** | |
1621 | * fcntl_getlease - Enquire what lease is currently active | |
1622 | * @filp: the file | |
1623 | * | |
1624 | * The value returned by this function will be one of | |
1625 | * (if no lease break is pending): | |
1626 | * | |
1627 | * %F_RDLCK to indicate a shared lease is held. | |
1628 | * | |
1629 | * %F_WRLCK to indicate an exclusive lease is held. | |
1630 | * | |
1631 | * %F_UNLCK to indicate no lease is held. | |
1632 | * | |
1633 | * (if a lease break is pending): | |
1634 | * | |
1635 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1636 | * changed to a shared lease (or removed). | |
1637 | * | |
1638 | * %F_UNLCK to indicate the lease needs to be removed. | |
1639 | * | |
1640 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1641 | * should be returned to userspace. | |
1642 | */ | |
1643 | int fcntl_getlease(struct file *filp) | |
1644 | { | |
1645 | struct file_lock *fl; | |
c568d683 | 1646 | struct inode *inode = locks_inode(filp); |
128a3785 | 1647 | struct file_lock_context *ctx; |
1da177e4 | 1648 | int type = F_UNLCK; |
c45198ed | 1649 | LIST_HEAD(dispose); |
1da177e4 | 1650 | |
128a3785 | 1651 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f | 1652 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
02e525b2 | 1653 | percpu_down_read(&file_rwsem); |
6109c850 | 1654 | spin_lock(&ctx->flc_lock); |
c568d683 | 1655 | time_out_leases(inode, &dispose); |
8634b51f JL |
1656 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
1657 | if (fl->fl_file != filp) | |
1658 | continue; | |
778fc546 | 1659 | type = target_leasetype(fl); |
1da177e4 LT |
1660 | break; |
1661 | } | |
6109c850 | 1662 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1663 | percpu_up_read(&file_rwsem); |
5f43086b | 1664 | |
8634b51f | 1665 | locks_dispose_list(&dispose); |
1da177e4 | 1666 | } |
1da177e4 LT |
1667 | return type; |
1668 | } | |
1669 | ||
24cbe784 | 1670 | /** |
387e3746 | 1671 | * check_conflicting_open - see if the given file points to an inode that has |
7bbd1fc0 N |
1672 | * an existing open that would conflict with the |
1673 | * desired lease. | |
387e3746 | 1674 | * @filp: file to check |
24cbe784 | 1675 | * @arg: type of lease that we're trying to acquire |
7fadc59c | 1676 | * @flags: current lock flags |
24cbe784 JL |
1677 | * |
1678 | * Check to see if there's an existing open fd on this file that would | |
1679 | * conflict with the lease we're trying to set. | |
1680 | */ | |
1681 | static int | |
387e3746 | 1682 | check_conflicting_open(struct file *filp, const long arg, int flags) |
24cbe784 | 1683 | { |
387e3746 AG |
1684 | struct inode *inode = locks_inode(filp); |
1685 | int self_wcount = 0, self_rcount = 0; | |
24cbe784 | 1686 | |
11afe9f7 CH |
1687 | if (flags & FL_LAYOUT) |
1688 | return 0; | |
aba2072f BF |
1689 | if (flags & FL_DELEG) |
1690 | /* We leave these checks to the caller */ | |
1691 | return 0; | |
11afe9f7 | 1692 | |
387e3746 AG |
1693 | if (arg == F_RDLCK) |
1694 | return inode_is_open_for_write(inode) ? -EAGAIN : 0; | |
1695 | else if (arg != F_WRLCK) | |
1696 | return 0; | |
1697 | ||
1698 | /* | |
1699 | * Make sure that only read/write count is from lease requestor. | |
1700 | * Note that this will result in denying write leases when i_writecount | |
1701 | * is negative, which is what we want. (We shouldn't grant write leases | |
1702 | * on files open for execution.) | |
1703 | */ | |
1704 | if (filp->f_mode & FMODE_WRITE) | |
1705 | self_wcount = 1; | |
1706 | else if (filp->f_mode & FMODE_READ) | |
1707 | self_rcount = 1; | |
24cbe784 | 1708 | |
387e3746 AG |
1709 | if (atomic_read(&inode->i_writecount) != self_wcount || |
1710 | atomic_read(&inode->i_readcount) != self_rcount) | |
1711 | return -EAGAIN; | |
24cbe784 | 1712 | |
387e3746 | 1713 | return 0; |
24cbe784 JL |
1714 | } |
1715 | ||
e6f5c789 JL |
1716 | static int |
1717 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1718 | { |
8634b51f | 1719 | struct file_lock *fl, *my_fl = NULL, *lease; |
387e3746 | 1720 | struct inode *inode = locks_inode(filp); |
8634b51f | 1721 | struct file_lock_context *ctx; |
df4e8d2c | 1722 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1723 | int error; |
c45198ed | 1724 | LIST_HEAD(dispose); |
1da177e4 | 1725 | |
096657b6 | 1726 | lease = *flp; |
62af4f1f JL |
1727 | trace_generic_add_lease(inode, lease); |
1728 | ||
5c1c669a JL |
1729 | /* Note that arg is never F_UNLCK here */ |
1730 | ctx = locks_get_lock_context(inode, arg); | |
8634b51f JL |
1731 | if (!ctx) |
1732 | return -ENOMEM; | |
1733 | ||
df4e8d2c BF |
1734 | /* |
1735 | * In the delegation case we need mutual exclusion with | |
1736 | * a number of operations that take the i_mutex. We trylock | |
1737 | * because delegations are an optional optimization, and if | |
1738 | * there's some chance of a conflict--we'd rather not | |
1739 | * bother, maybe that's a sign this just isn't a good file to | |
1740 | * hand out a delegation on. | |
1741 | */ | |
5955102c | 1742 | if (is_deleg && !inode_trylock(inode)) |
df4e8d2c BF |
1743 | return -EAGAIN; |
1744 | ||
1745 | if (is_deleg && arg == F_WRLCK) { | |
1746 | /* Write delegations are not currently supported: */ | |
5955102c | 1747 | inode_unlock(inode); |
df4e8d2c BF |
1748 | WARN_ON_ONCE(1); |
1749 | return -EINVAL; | |
1750 | } | |
096657b6 | 1751 | |
02e525b2 | 1752 | percpu_down_read(&file_rwsem); |
6109c850 | 1753 | spin_lock(&ctx->flc_lock); |
c45198ed | 1754 | time_out_leases(inode, &dispose); |
387e3746 | 1755 | error = check_conflicting_open(filp, arg, lease->fl_flags); |
24cbe784 | 1756 | if (error) |
096657b6 | 1757 | goto out; |
6d5e8b05 | 1758 | |
1da177e4 LT |
1759 | /* |
1760 | * At this point, we know that if there is an exclusive | |
1761 | * lease on this file, then we hold it on this filp | |
1762 | * (otherwise our open of this file would have blocked). | |
1763 | * And if we are trying to acquire an exclusive lease, | |
1764 | * then the file is not open by anyone (including us) | |
1765 | * except for this filp. | |
1766 | */ | |
c1f24ef4 | 1767 | error = -EAGAIN; |
8634b51f | 1768 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1769 | if (fl->fl_file == filp && |
1770 | fl->fl_owner == lease->fl_owner) { | |
8634b51f | 1771 | my_fl = fl; |
c1f24ef4 BF |
1772 | continue; |
1773 | } | |
8634b51f | 1774 | |
c1f24ef4 BF |
1775 | /* |
1776 | * No exclusive leases if someone else has a lease on | |
1777 | * this file: | |
1778 | */ | |
1779 | if (arg == F_WRLCK) | |
1780 | goto out; | |
1781 | /* | |
1782 | * Modifying our existing lease is OK, but no getting a | |
1783 | * new lease if someone else is opening for write: | |
1784 | */ | |
1785 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1786 | goto out; | |
1da177e4 LT |
1787 | } |
1788 | ||
8634b51f | 1789 | if (my_fl != NULL) { |
0164bf02 JL |
1790 | lease = my_fl; |
1791 | error = lease->fl_lmops->lm_change(lease, arg, &dispose); | |
1c7dd2ff JL |
1792 | if (error) |
1793 | goto out; | |
1794 | goto out_setup; | |
1da177e4 LT |
1795 | } |
1796 | ||
1da177e4 LT |
1797 | error = -EINVAL; |
1798 | if (!leases_enable) | |
1799 | goto out; | |
1800 | ||
e084c1bd | 1801 | locks_insert_lock_ctx(lease, &ctx->flc_lease); |
24cbe784 JL |
1802 | /* |
1803 | * The check in break_lease() is lockless. It's possible for another | |
1804 | * open to race in after we did the earlier check for a conflicting | |
1805 | * open but before the lease was inserted. Check again for a | |
1806 | * conflicting open and cancel the lease if there is one. | |
1807 | * | |
1808 | * We also add a barrier here to ensure that the insertion of the lock | |
1809 | * precedes these checks. | |
1810 | */ | |
1811 | smp_mb(); | |
387e3746 | 1812 | error = check_conflicting_open(filp, arg, lease->fl_flags); |
8634b51f | 1813 | if (error) { |
e084c1bd | 1814 | locks_unlink_lock_ctx(lease); |
8634b51f JL |
1815 | goto out; |
1816 | } | |
1c7dd2ff JL |
1817 | |
1818 | out_setup: | |
1819 | if (lease->fl_lmops->lm_setup) | |
1820 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1821 | out: |
6109c850 | 1822 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1823 | percpu_up_read(&file_rwsem); |
c45198ed | 1824 | locks_dispose_list(&dispose); |
df4e8d2c | 1825 | if (is_deleg) |
5955102c | 1826 | inode_unlock(inode); |
8634b51f | 1827 | if (!error && !my_fl) |
1c7dd2ff | 1828 | *flp = NULL; |
1da177e4 LT |
1829 | return error; |
1830 | } | |
8335ebd9 | 1831 | |
2ab99ee1 | 1832 | static int generic_delete_lease(struct file *filp, void *owner) |
8335ebd9 | 1833 | { |
0efaa7e8 | 1834 | int error = -EAGAIN; |
8634b51f | 1835 | struct file_lock *fl, *victim = NULL; |
c568d683 | 1836 | struct inode *inode = locks_inode(filp); |
128a3785 | 1837 | struct file_lock_context *ctx; |
c45198ed | 1838 | LIST_HEAD(dispose); |
8335ebd9 | 1839 | |
128a3785 | 1840 | ctx = smp_load_acquire(&inode->i_flctx); |
8634b51f JL |
1841 | if (!ctx) { |
1842 | trace_generic_delete_lease(inode, NULL); | |
1843 | return error; | |
1844 | } | |
1845 | ||
02e525b2 | 1846 | percpu_down_read(&file_rwsem); |
6109c850 | 1847 | spin_lock(&ctx->flc_lock); |
8634b51f | 1848 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
2ab99ee1 CH |
1849 | if (fl->fl_file == filp && |
1850 | fl->fl_owner == owner) { | |
8634b51f | 1851 | victim = fl; |
0efaa7e8 | 1852 | break; |
8634b51f | 1853 | } |
8335ebd9 | 1854 | } |
a9b1b455 | 1855 | trace_generic_delete_lease(inode, victim); |
8634b51f | 1856 | if (victim) |
7448cc37 | 1857 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1858 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 1859 | percpu_up_read(&file_rwsem); |
c45198ed | 1860 | locks_dispose_list(&dispose); |
0efaa7e8 | 1861 | return error; |
8335ebd9 BF |
1862 | } |
1863 | ||
1864 | /** | |
1865 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1866 | * @filp: file pointer |
1867 | * @arg: type of lease to obtain | |
1868 | * @flp: input - file_lock to use, output - file_lock inserted | |
1869 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1870 | * doesn't require it) | |
8335ebd9 BF |
1871 | * |
1872 | * The (input) flp->fl_lmops->lm_break function is required | |
1873 | * by break_lease(). | |
8335ebd9 | 1874 | */ |
e6f5c789 JL |
1875 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1876 | void **priv) | |
8335ebd9 | 1877 | { |
c568d683 | 1878 | struct inode *inode = locks_inode(filp); |
8335ebd9 BF |
1879 | int error; |
1880 | ||
8e96e3b7 | 1881 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1882 | return -EACCES; |
1883 | if (!S_ISREG(inode->i_mode)) | |
1884 | return -EINVAL; | |
1885 | error = security_file_lock(filp, arg); | |
1886 | if (error) | |
1887 | return error; | |
1888 | ||
8335ebd9 BF |
1889 | switch (arg) { |
1890 | case F_UNLCK: | |
2ab99ee1 | 1891 | return generic_delete_lease(filp, *priv); |
8335ebd9 BF |
1892 | case F_RDLCK: |
1893 | case F_WRLCK: | |
0efaa7e8 JL |
1894 | if (!(*flp)->fl_lmops->lm_break) { |
1895 | WARN_ON_ONCE(1); | |
1896 | return -ENOLCK; | |
1897 | } | |
11afe9f7 | 1898 | |
e6f5c789 | 1899 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1900 | default: |
8d657eb3 | 1901 | return -EINVAL; |
8335ebd9 BF |
1902 | } |
1903 | } | |
0af1a450 | 1904 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1905 | |
18f6622e JL |
1906 | #if IS_ENABLED(CONFIG_SRCU) |
1907 | /* | |
1908 | * Kernel subsystems can register to be notified on any attempt to set | |
1909 | * a new lease with the lease_notifier_chain. This is used by (e.g.) nfsd | |
1910 | * to close files that it may have cached when there is an attempt to set a | |
1911 | * conflicting lease. | |
1912 | */ | |
1913 | static struct srcu_notifier_head lease_notifier_chain; | |
1914 | ||
1915 | static inline void | |
1916 | lease_notifier_chain_init(void) | |
1917 | { | |
1918 | srcu_init_notifier_head(&lease_notifier_chain); | |
1919 | } | |
1920 | ||
1921 | static inline void | |
1922 | setlease_notifier(long arg, struct file_lock *lease) | |
1923 | { | |
1924 | if (arg != F_UNLCK) | |
1925 | srcu_notifier_call_chain(&lease_notifier_chain, arg, lease); | |
1926 | } | |
1927 | ||
1928 | int lease_register_notifier(struct notifier_block *nb) | |
1929 | { | |
1930 | return srcu_notifier_chain_register(&lease_notifier_chain, nb); | |
1931 | } | |
1932 | EXPORT_SYMBOL_GPL(lease_register_notifier); | |
1933 | ||
1934 | void lease_unregister_notifier(struct notifier_block *nb) | |
1935 | { | |
1936 | srcu_notifier_chain_unregister(&lease_notifier_chain, nb); | |
1937 | } | |
1938 | EXPORT_SYMBOL_GPL(lease_unregister_notifier); | |
1939 | ||
1940 | #else /* !IS_ENABLED(CONFIG_SRCU) */ | |
1941 | static inline void | |
1942 | lease_notifier_chain_init(void) | |
1943 | { | |
1944 | } | |
1945 | ||
1946 | static inline void | |
1947 | setlease_notifier(long arg, struct file_lock *lease) | |
1948 | { | |
1949 | } | |
1950 | ||
1951 | int lease_register_notifier(struct notifier_block *nb) | |
1952 | { | |
1953 | return 0; | |
1954 | } | |
1955 | EXPORT_SYMBOL_GPL(lease_register_notifier); | |
1956 | ||
1957 | void lease_unregister_notifier(struct notifier_block *nb) | |
1958 | { | |
1959 | } | |
1960 | EXPORT_SYMBOL_GPL(lease_unregister_notifier); | |
1961 | ||
1962 | #endif /* IS_ENABLED(CONFIG_SRCU) */ | |
1963 | ||
b89f4321 | 1964 | /** |
e51673aa | 1965 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1966 | * @filp: file pointer |
1967 | * @arg: type of lease to obtain | |
1968 | * @lease: file_lock to use when adding a lease | |
1969 | * @priv: private info for lm_setup when adding a lease (may be | |
7bbd1fc0 | 1970 | * NULL if lm_setup doesn't require it) |
e51673aa JL |
1971 | * |
1972 | * Call this to establish a lease on the file. The "lease" argument is not | |
1973 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
80b79dd0 MCC |
1974 | * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be |
1975 | * set; if not, this function will return -ENOLCK (and generate a scary-looking | |
e51673aa | 1976 | * stack trace). |
1c7dd2ff JL |
1977 | * |
1978 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1979 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1980 | */ |
e6f5c789 JL |
1981 | int |
1982 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1983 | { |
18f6622e JL |
1984 | if (lease) |
1985 | setlease_notifier(arg, *lease); | |
de2a4a50 | 1986 | if (filp->f_op->setlease) |
f82b4b67 | 1987 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1988 | else |
f82b4b67 | 1989 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1990 | } |
a9933cea | 1991 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1992 | |
0ceaf6c7 | 1993 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1994 | { |
1c7dd2ff | 1995 | struct file_lock *fl; |
f7347ce4 | 1996 | struct fasync_struct *new; |
1da177e4 LT |
1997 | int error; |
1998 | ||
c5b1f0d9 AB |
1999 | fl = lease_alloc(filp, arg); |
2000 | if (IS_ERR(fl)) | |
2001 | return PTR_ERR(fl); | |
1da177e4 | 2002 | |
f7347ce4 LT |
2003 | new = fasync_alloc(); |
2004 | if (!new) { | |
2005 | locks_free_lock(fl); | |
2006 | return -ENOMEM; | |
2007 | } | |
1c7dd2ff | 2008 | new->fa_fd = fd; |
f7347ce4 | 2009 | |
1c7dd2ff | 2010 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
2011 | if (fl) |
2012 | locks_free_lock(fl); | |
f7347ce4 LT |
2013 | if (new) |
2014 | fasync_free(new); | |
1da177e4 LT |
2015 | return error; |
2016 | } | |
2017 | ||
0ceaf6c7 BF |
2018 | /** |
2019 | * fcntl_setlease - sets a lease on an open file | |
2020 | * @fd: open file descriptor | |
2021 | * @filp: file pointer | |
2022 | * @arg: type of lease to obtain | |
2023 | * | |
2024 | * Call this fcntl to establish a lease on the file. | |
2025 | * Note that you also need to call %F_SETSIG to | |
2026 | * receive a signal when the lease is broken. | |
2027 | */ | |
2028 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
2029 | { | |
2030 | if (arg == F_UNLCK) | |
2ab99ee1 | 2031 | return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp); |
0ceaf6c7 BF |
2032 | return do_fcntl_add_lease(fd, filp, arg); |
2033 | } | |
2034 | ||
1da177e4 | 2035 | /** |
29d01b22 JL |
2036 | * flock_lock_inode_wait - Apply a FLOCK-style lock to a file |
2037 | * @inode: inode of the file to apply to | |
1da177e4 LT |
2038 | * @fl: The lock to be applied |
2039 | * | |
29d01b22 | 2040 | * Apply a FLOCK style lock request to an inode. |
1da177e4 | 2041 | */ |
616fb38f | 2042 | static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl) |
1da177e4 LT |
2043 | { |
2044 | int error; | |
2045 | might_sleep(); | |
2046 | for (;;) { | |
29d01b22 | 2047 | error = flock_lock_inode(inode, fl); |
bde74e4b | 2048 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 | 2049 | break; |
dcf23ac3 LT |
2050 | error = wait_event_interruptible(fl->fl_wait, |
2051 | list_empty(&fl->fl_blocked_member)); | |
16306a61 N |
2052 | if (error) |
2053 | break; | |
1da177e4 | 2054 | } |
16306a61 | 2055 | locks_delete_block(fl); |
1da177e4 LT |
2056 | return error; |
2057 | } | |
2058 | ||
e55c34a6 BC |
2059 | /** |
2060 | * locks_lock_inode_wait - Apply a lock to an inode | |
2061 | * @inode: inode of the file to apply to | |
2062 | * @fl: The lock to be applied | |
2063 | * | |
2064 | * Apply a POSIX or FLOCK style lock request to an inode. | |
2065 | */ | |
2066 | int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) | |
2067 | { | |
2068 | int res = 0; | |
2069 | switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) { | |
2070 | case FL_POSIX: | |
2071 | res = posix_lock_inode_wait(inode, fl); | |
2072 | break; | |
2073 | case FL_FLOCK: | |
2074 | res = flock_lock_inode_wait(inode, fl); | |
2075 | break; | |
2076 | default: | |
2077 | BUG(); | |
2078 | } | |
2079 | return res; | |
2080 | } | |
2081 | EXPORT_SYMBOL(locks_lock_inode_wait); | |
2082 | ||
1da177e4 LT |
2083 | /** |
2084 | * sys_flock: - flock() system call. | |
2085 | * @fd: the file descriptor to lock. | |
2086 | * @cmd: the type of lock to apply. | |
2087 | * | |
2088 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
80b79dd0 | 2089 | * The @cmd can be one of: |
1da177e4 | 2090 | * |
80b79dd0 MCC |
2091 | * - %LOCK_SH -- a shared lock. |
2092 | * - %LOCK_EX -- an exclusive lock. | |
2093 | * - %LOCK_UN -- remove an existing lock. | |
90f7d7a0 | 2094 | * - %LOCK_MAND -- a 'mandatory' flock. (DEPRECATED) |
1da177e4 | 2095 | * |
90f7d7a0 | 2096 | * %LOCK_MAND support has been removed from the kernel. |
1da177e4 | 2097 | */ |
002c8976 | 2098 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 2099 | { |
2903ff01 | 2100 | struct fd f = fdget(fd); |
1da177e4 LT |
2101 | struct file_lock *lock; |
2102 | int can_sleep, unlock; | |
2103 | int error; | |
2104 | ||
2105 | error = -EBADF; | |
2903ff01 | 2106 | if (!f.file) |
1da177e4 LT |
2107 | goto out; |
2108 | ||
2109 | can_sleep = !(cmd & LOCK_NB); | |
2110 | cmd &= ~LOCK_NB; | |
2111 | unlock = (cmd == LOCK_UN); | |
2112 | ||
90f7d7a0 JL |
2113 | if (!unlock && !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
2114 | goto out_putf; | |
2115 | ||
2116 | /* | |
2117 | * LOCK_MAND locks were broken for a long time in that they never | |
2118 | * conflicted with one another and didn't prevent any sort of open, | |
2119 | * read or write activity. | |
2120 | * | |
2121 | * Just ignore these requests now, to preserve legacy behavior, but | |
2122 | * throw a warning to let people know that they don't actually work. | |
2123 | */ | |
2124 | if (cmd & LOCK_MAND) { | |
2125 | pr_warn_once("Attempt to set a LOCK_MAND lock via flock(2). This support has been removed and the request ignored.\n"); | |
2126 | error = 0; | |
1da177e4 | 2127 | goto out_putf; |
90f7d7a0 | 2128 | } |
1da177e4 | 2129 | |
d6367d62 | 2130 | lock = flock_make_lock(f.file, cmd, NULL); |
6e129d00 JL |
2131 | if (IS_ERR(lock)) { |
2132 | error = PTR_ERR(lock); | |
1da177e4 | 2133 | goto out_putf; |
6e129d00 JL |
2134 | } |
2135 | ||
1da177e4 LT |
2136 | if (can_sleep) |
2137 | lock->fl_flags |= FL_SLEEP; | |
2138 | ||
2903ff01 | 2139 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
2140 | if (error) |
2141 | goto out_free; | |
2142 | ||
de2a4a50 | 2143 | if (f.file->f_op->flock) |
2903ff01 | 2144 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
2145 | (can_sleep) ? F_SETLKW : F_SETLK, |
2146 | lock); | |
2147 | else | |
4f656367 | 2148 | error = locks_lock_file_wait(f.file, lock); |
1da177e4 LT |
2149 | |
2150 | out_free: | |
993dfa87 | 2151 | locks_free_lock(lock); |
1da177e4 LT |
2152 | |
2153 | out_putf: | |
2903ff01 | 2154 | fdput(f); |
1da177e4 LT |
2155 | out: |
2156 | return error; | |
2157 | } | |
2158 | ||
3ee17abd BF |
2159 | /** |
2160 | * vfs_test_lock - test file byte range lock | |
2161 | * @filp: The file to test lock for | |
6924c554 | 2162 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
2163 | * |
2164 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
2165 | * setting conf->fl_type to something other than F_UNLCK. | |
2166 | */ | |
2167 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
2168 | { | |
de2a4a50 | 2169 | if (filp->f_op->lock) |
3ee17abd BF |
2170 | return filp->f_op->lock(filp, F_GETLK, fl); |
2171 | posix_test_lock(filp, fl); | |
2172 | return 0; | |
2173 | } | |
2174 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
2175 | ||
9d5b86ac BC |
2176 | /** |
2177 | * locks_translate_pid - translate a file_lock's fl_pid number into a namespace | |
2178 | * @fl: The file_lock who's fl_pid should be translated | |
2179 | * @ns: The namespace into which the pid should be translated | |
2180 | * | |
2181 | * Used to tranlate a fl_pid into a namespace virtual pid number | |
2182 | */ | |
2183 | static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns) | |
2184 | { | |
2185 | pid_t vnr; | |
2186 | struct pid *pid; | |
2187 | ||
2188 | if (IS_OFDLCK(fl)) | |
2189 | return -1; | |
2190 | if (IS_REMOTELCK(fl)) | |
2191 | return fl->fl_pid; | |
826d7bc9 KK |
2192 | /* |
2193 | * If the flock owner process is dead and its pid has been already | |
2194 | * freed, the translation below won't work, but we still want to show | |
2195 | * flock owner pid number in init pidns. | |
2196 | */ | |
2197 | if (ns == &init_pid_ns) | |
2198 | return (pid_t)fl->fl_pid; | |
9d5b86ac BC |
2199 | |
2200 | rcu_read_lock(); | |
2201 | pid = find_pid_ns(fl->fl_pid, &init_pid_ns); | |
2202 | vnr = pid_nr_ns(pid, ns); | |
2203 | rcu_read_unlock(); | |
2204 | return vnr; | |
2205 | } | |
2206 | ||
c2fa1b8a BF |
2207 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
2208 | { | |
9d5b86ac | 2209 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2210 | #if BITS_PER_LONG == 32 |
2211 | /* | |
2212 | * Make sure we can represent the posix lock via | |
2213 | * legacy 32bit flock. | |
2214 | */ | |
2215 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2216 | return -EOVERFLOW; | |
2217 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2218 | return -EOVERFLOW; | |
2219 | #endif | |
2220 | flock->l_start = fl->fl_start; | |
2221 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2222 | fl->fl_end - fl->fl_start + 1; | |
2223 | flock->l_whence = 0; | |
129a84de | 2224 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2225 | return 0; |
2226 | } | |
2227 | ||
2228 | #if BITS_PER_LONG == 32 | |
2229 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2230 | { | |
9d5b86ac | 2231 | flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current)); |
c2fa1b8a BF |
2232 | flock->l_start = fl->fl_start; |
2233 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2234 | fl->fl_end - fl->fl_start + 1; | |
2235 | flock->l_whence = 0; | |
2236 | flock->l_type = fl->fl_type; | |
2237 | } | |
2238 | #endif | |
2239 | ||
1da177e4 LT |
2240 | /* Report the first existing lock that would conflict with l. |
2241 | * This implements the F_GETLK command of fcntl(). | |
2242 | */ | |
a75d30c7 | 2243 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock) |
1da177e4 | 2244 | { |
52306e88 | 2245 | struct file_lock *fl; |
1da177e4 LT |
2246 | int error; |
2247 | ||
52306e88 BC |
2248 | fl = locks_alloc_lock(); |
2249 | if (fl == NULL) | |
2250 | return -ENOMEM; | |
1da177e4 | 2251 | error = -EINVAL; |
a75d30c7 | 2252 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2253 | goto out; |
2254 | ||
52306e88 | 2255 | error = flock_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2256 | if (error) |
2257 | goto out; | |
2258 | ||
0d3f7a2d | 2259 | if (cmd == F_OFD_GETLK) { |
90478939 | 2260 | error = -EINVAL; |
a75d30c7 | 2261 | if (flock->l_pid != 0) |
90478939 JL |
2262 | goto out; |
2263 | ||
52306e88 BC |
2264 | fl->fl_flags |= FL_OFDLCK; |
2265 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2266 | } |
2267 | ||
52306e88 | 2268 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2269 | if (error) |
2270 | goto out; | |
7bbd1fc0 | 2271 | |
52306e88 BC |
2272 | flock->l_type = fl->fl_type; |
2273 | if (fl->fl_type != F_UNLCK) { | |
2274 | error = posix_lock_to_flock(flock, fl); | |
c2fa1b8a | 2275 | if (error) |
52306e88 | 2276 | goto out; |
1da177e4 | 2277 | } |
1da177e4 | 2278 | out: |
52306e88 | 2279 | locks_free_lock(fl); |
1da177e4 LT |
2280 | return error; |
2281 | } | |
2282 | ||
7723ec97 ME |
2283 | /** |
2284 | * vfs_lock_file - file byte range lock | |
2285 | * @filp: The file to apply the lock to | |
2286 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2287 | * @fl: The lock to be applied | |
150b3934 ME |
2288 | * @conf: Place to return a copy of the conflicting lock, if found. |
2289 | * | |
2290 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2291 | * as the final argument. | |
2292 | * | |
2293 | * If the filesystem defines a private ->lock() method, then @conf will | |
2294 | * be left unchanged; so a caller that cares should initialize it to | |
2295 | * some acceptable default. | |
2beb6614 ME |
2296 | * |
2297 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2298 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2299 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2300 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2301 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2302 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2303 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2304 | * request completes. |
2305 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2306 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2307 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2308 | * nonzero return code and the file system should release the lock. The file |
2309 | * system is also responsible to keep a corresponding posix lock when it | |
2310 | * grants a lock so the VFS can find out which locks are locally held and do | |
2311 | * the correct lock cleanup when required. | |
2312 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2313 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2314 | * return code. |
7723ec97 | 2315 | */ |
150b3934 | 2316 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2317 | { |
de2a4a50 | 2318 | if (filp->f_op->lock) |
7723ec97 ME |
2319 | return filp->f_op->lock(filp, cmd, fl); |
2320 | else | |
150b3934 | 2321 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2322 | } |
2323 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2324 | ||
b648a6de MS |
2325 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2326 | struct file_lock *fl) | |
2327 | { | |
2328 | int error; | |
2329 | ||
2330 | error = security_file_lock(filp, fl->fl_type); | |
2331 | if (error) | |
2332 | return error; | |
2333 | ||
764c76b3 MS |
2334 | for (;;) { |
2335 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2336 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2337 | break; |
dcf23ac3 LT |
2338 | error = wait_event_interruptible(fl->fl_wait, |
2339 | list_empty(&fl->fl_blocked_member)); | |
16306a61 N |
2340 | if (error) |
2341 | break; | |
b648a6de | 2342 | } |
16306a61 | 2343 | locks_delete_block(fl); |
b648a6de MS |
2344 | |
2345 | return error; | |
2346 | } | |
2347 | ||
6ca7d910 | 2348 | /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */ |
cf01f4ee JL |
2349 | static int |
2350 | check_fmode_for_setlk(struct file_lock *fl) | |
2351 | { | |
2352 | switch (fl->fl_type) { | |
2353 | case F_RDLCK: | |
2354 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2355 | return -EBADF; | |
2356 | break; | |
2357 | case F_WRLCK: | |
2358 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2359 | return -EBADF; | |
2360 | } | |
2361 | return 0; | |
2362 | } | |
2363 | ||
1da177e4 LT |
2364 | /* Apply the lock described by l to an open file descriptor. |
2365 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2366 | */ | |
c293621b | 2367 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2368 | struct flock *flock) |
1da177e4 LT |
2369 | { |
2370 | struct file_lock *file_lock = locks_alloc_lock(); | |
a75d30c7 | 2371 | struct inode *inode = locks_inode(filp); |
0b2bac2f | 2372 | struct file *f; |
1da177e4 LT |
2373 | int error; |
2374 | ||
2375 | if (file_lock == NULL) | |
2376 | return -ENOLCK; | |
2377 | ||
a75d30c7 | 2378 | error = flock_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2379 | if (error) |
2380 | goto out; | |
5d50ffd7 | 2381 | |
cf01f4ee JL |
2382 | error = check_fmode_for_setlk(file_lock); |
2383 | if (error) | |
2384 | goto out; | |
2385 | ||
5d50ffd7 JL |
2386 | /* |
2387 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2388 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2389 | */ |
2390 | switch (cmd) { | |
0d3f7a2d | 2391 | case F_OFD_SETLK: |
90478939 | 2392 | error = -EINVAL; |
a75d30c7 | 2393 | if (flock->l_pid != 0) |
90478939 JL |
2394 | goto out; |
2395 | ||
5d50ffd7 | 2396 | cmd = F_SETLK; |
cff2fce5 | 2397 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2398 | file_lock->fl_owner = filp; |
5d50ffd7 | 2399 | break; |
0d3f7a2d | 2400 | case F_OFD_SETLKW: |
90478939 | 2401 | error = -EINVAL; |
a75d30c7 | 2402 | if (flock->l_pid != 0) |
90478939 JL |
2403 | goto out; |
2404 | ||
5d50ffd7 | 2405 | cmd = F_SETLKW; |
cff2fce5 | 2406 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2407 | file_lock->fl_owner = filp; |
df561f66 | 2408 | fallthrough; |
5d50ffd7 | 2409 | case F_SETLKW: |
1da177e4 LT |
2410 | file_lock->fl_flags |= FL_SLEEP; |
2411 | } | |
5d50ffd7 | 2412 | |
b648a6de | 2413 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2414 | |
c293621b | 2415 | /* |
0752ba80 JL |
2416 | * Attempt to detect a close/fcntl race and recover by releasing the |
2417 | * lock that was just acquired. There is no need to do that when we're | |
2418 | * unlocking though, or for OFD locks. | |
c293621b | 2419 | */ |
0752ba80 JL |
2420 | if (!error && file_lock->fl_type != F_UNLCK && |
2421 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
120ce2b0 | 2422 | struct files_struct *files = current->files; |
7f3697e2 JL |
2423 | /* |
2424 | * We need that spin_lock here - it prevents reordering between | |
2425 | * update of i_flctx->flc_posix and check for it done in | |
2426 | * close(). rcu_read_lock() wouldn't do. | |
2427 | */ | |
120ce2b0 EB |
2428 | spin_lock(&files->file_lock); |
2429 | f = files_lookup_fd_locked(files, fd); | |
2430 | spin_unlock(&files->file_lock); | |
7f3697e2 JL |
2431 | if (f != filp) { |
2432 | file_lock->fl_type = F_UNLCK; | |
2433 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2434 | WARN_ON_ONCE(error); | |
2435 | error = -EBADF; | |
2436 | } | |
1da177e4 | 2437 | } |
c293621b | 2438 | out: |
1890910f | 2439 | trace_fcntl_setlk(inode, file_lock, error); |
1da177e4 LT |
2440 | locks_free_lock(file_lock); |
2441 | return error; | |
2442 | } | |
2443 | ||
2444 | #if BITS_PER_LONG == 32 | |
2445 | /* Report the first existing lock that would conflict with l. | |
2446 | * This implements the F_GETLK command of fcntl(). | |
2447 | */ | |
a75d30c7 | 2448 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock) |
1da177e4 | 2449 | { |
52306e88 | 2450 | struct file_lock *fl; |
1da177e4 LT |
2451 | int error; |
2452 | ||
52306e88 BC |
2453 | fl = locks_alloc_lock(); |
2454 | if (fl == NULL) | |
2455 | return -ENOMEM; | |
2456 | ||
1da177e4 | 2457 | error = -EINVAL; |
a75d30c7 | 2458 | if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK) |
1da177e4 LT |
2459 | goto out; |
2460 | ||
52306e88 | 2461 | error = flock64_to_posix_lock(filp, fl, flock); |
1da177e4 LT |
2462 | if (error) |
2463 | goto out; | |
2464 | ||
0d3f7a2d | 2465 | if (cmd == F_OFD_GETLK) { |
90478939 | 2466 | error = -EINVAL; |
a75d30c7 | 2467 | if (flock->l_pid != 0) |
90478939 JL |
2468 | goto out; |
2469 | ||
5d50ffd7 | 2470 | cmd = F_GETLK64; |
52306e88 BC |
2471 | fl->fl_flags |= FL_OFDLCK; |
2472 | fl->fl_owner = filp; | |
5d50ffd7 JL |
2473 | } |
2474 | ||
52306e88 | 2475 | error = vfs_test_lock(filp, fl); |
3ee17abd BF |
2476 | if (error) |
2477 | goto out; | |
2478 | ||
52306e88 BC |
2479 | flock->l_type = fl->fl_type; |
2480 | if (fl->fl_type != F_UNLCK) | |
2481 | posix_lock_to_flock64(flock, fl); | |
f328296e | 2482 | |
1da177e4 | 2483 | out: |
52306e88 | 2484 | locks_free_lock(fl); |
1da177e4 LT |
2485 | return error; |
2486 | } | |
2487 | ||
2488 | /* Apply the lock described by l to an open file descriptor. | |
2489 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2490 | */ | |
c293621b | 2491 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
a75d30c7 | 2492 | struct flock64 *flock) |
1da177e4 LT |
2493 | { |
2494 | struct file_lock *file_lock = locks_alloc_lock(); | |
0b2bac2f | 2495 | struct file *f; |
1da177e4 LT |
2496 | int error; |
2497 | ||
2498 | if (file_lock == NULL) | |
2499 | return -ENOLCK; | |
2500 | ||
a75d30c7 | 2501 | error = flock64_to_posix_lock(filp, file_lock, flock); |
1da177e4 LT |
2502 | if (error) |
2503 | goto out; | |
5d50ffd7 | 2504 | |
cf01f4ee JL |
2505 | error = check_fmode_for_setlk(file_lock); |
2506 | if (error) | |
2507 | goto out; | |
2508 | ||
5d50ffd7 JL |
2509 | /* |
2510 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2511 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2512 | */ |
2513 | switch (cmd) { | |
0d3f7a2d | 2514 | case F_OFD_SETLK: |
90478939 | 2515 | error = -EINVAL; |
a75d30c7 | 2516 | if (flock->l_pid != 0) |
90478939 JL |
2517 | goto out; |
2518 | ||
5d50ffd7 | 2519 | cmd = F_SETLK64; |
cff2fce5 | 2520 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2521 | file_lock->fl_owner = filp; |
5d50ffd7 | 2522 | break; |
0d3f7a2d | 2523 | case F_OFD_SETLKW: |
90478939 | 2524 | error = -EINVAL; |
a75d30c7 | 2525 | if (flock->l_pid != 0) |
90478939 JL |
2526 | goto out; |
2527 | ||
5d50ffd7 | 2528 | cmd = F_SETLKW64; |
cff2fce5 | 2529 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2530 | file_lock->fl_owner = filp; |
df561f66 | 2531 | fallthrough; |
5d50ffd7 | 2532 | case F_SETLKW64: |
1da177e4 LT |
2533 | file_lock->fl_flags |= FL_SLEEP; |
2534 | } | |
5d50ffd7 | 2535 | |
b648a6de | 2536 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2537 | |
c293621b | 2538 | /* |
0752ba80 JL |
2539 | * Attempt to detect a close/fcntl race and recover by releasing the |
2540 | * lock that was just acquired. There is no need to do that when we're | |
2541 | * unlocking though, or for OFD locks. | |
c293621b | 2542 | */ |
0752ba80 JL |
2543 | if (!error && file_lock->fl_type != F_UNLCK && |
2544 | !(file_lock->fl_flags & FL_OFDLCK)) { | |
120ce2b0 | 2545 | struct files_struct *files = current->files; |
7f3697e2 JL |
2546 | /* |
2547 | * We need that spin_lock here - it prevents reordering between | |
2548 | * update of i_flctx->flc_posix and check for it done in | |
2549 | * close(). rcu_read_lock() wouldn't do. | |
2550 | */ | |
120ce2b0 EB |
2551 | spin_lock(&files->file_lock); |
2552 | f = files_lookup_fd_locked(files, fd); | |
2553 | spin_unlock(&files->file_lock); | |
7f3697e2 JL |
2554 | if (f != filp) { |
2555 | file_lock->fl_type = F_UNLCK; | |
2556 | error = do_lock_file_wait(filp, cmd, file_lock); | |
2557 | WARN_ON_ONCE(error); | |
2558 | error = -EBADF; | |
2559 | } | |
1da177e4 | 2560 | } |
1da177e4 LT |
2561 | out: |
2562 | locks_free_lock(file_lock); | |
2563 | return error; | |
2564 | } | |
2565 | #endif /* BITS_PER_LONG == 32 */ | |
2566 | ||
2567 | /* | |
2568 | * This function is called when the file is being removed | |
2569 | * from the task's fd array. POSIX locks belonging to this task | |
2570 | * are deleted at this time. | |
2571 | */ | |
2572 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2573 | { | |
1890910f | 2574 | int error; |
c568d683 | 2575 | struct inode *inode = locks_inode(filp); |
ff7b86b8 | 2576 | struct file_lock lock; |
128a3785 | 2577 | struct file_lock_context *ctx; |
1da177e4 LT |
2578 | |
2579 | /* | |
2580 | * If there are no locks held on this file, we don't need to call | |
2581 | * posix_lock_file(). Another process could be setting a lock on this | |
2582 | * file at the same time, but we wouldn't remove that lock anyway. | |
2583 | */ | |
c568d683 | 2584 | ctx = smp_load_acquire(&inode->i_flctx); |
bd61e0a9 | 2585 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2586 | return; |
2587 | ||
d6367d62 | 2588 | locks_init_lock(&lock); |
1da177e4 | 2589 | lock.fl_type = F_UNLCK; |
75e1fcc0 | 2590 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2591 | lock.fl_start = 0; |
2592 | lock.fl_end = OFFSET_MAX; | |
2593 | lock.fl_owner = owner; | |
2594 | lock.fl_pid = current->tgid; | |
2595 | lock.fl_file = filp; | |
2596 | lock.fl_ops = NULL; | |
2597 | lock.fl_lmops = NULL; | |
2598 | ||
1890910f | 2599 | error = vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2600 | |
1da177e4 LT |
2601 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2602 | lock.fl_ops->fl_release_private(&lock); | |
c568d683 | 2603 | trace_locks_remove_posix(inode, &lock, error); |
1da177e4 | 2604 | } |
1da177e4 LT |
2605 | EXPORT_SYMBOL(locks_remove_posix); |
2606 | ||
3d8e560d | 2607 | /* The i_flctx must be valid when calling into here */ |
dd459bb1 | 2608 | static void |
128a3785 | 2609 | locks_remove_flock(struct file *filp, struct file_lock_context *flctx) |
dd459bb1 | 2610 | { |
d6367d62 | 2611 | struct file_lock fl; |
c568d683 | 2612 | struct inode *inode = locks_inode(filp); |
dd459bb1 | 2613 | |
3d8e560d | 2614 | if (list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2615 | return; |
2616 | ||
d6367d62 N |
2617 | flock_make_lock(filp, LOCK_UN, &fl); |
2618 | fl.fl_flags |= FL_CLOSE; | |
2619 | ||
de2a4a50 | 2620 | if (filp->f_op->flock) |
dd459bb1 JL |
2621 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2622 | else | |
bcd7f78d | 2623 | flock_lock_inode(inode, &fl); |
dd459bb1 JL |
2624 | |
2625 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2626 | fl.fl_ops->fl_release_private(&fl); | |
2627 | } | |
2628 | ||
3d8e560d | 2629 | /* The i_flctx must be valid when calling into here */ |
8634b51f | 2630 | static void |
128a3785 | 2631 | locks_remove_lease(struct file *filp, struct file_lock_context *ctx) |
8634b51f | 2632 | { |
8634b51f JL |
2633 | struct file_lock *fl, *tmp; |
2634 | LIST_HEAD(dispose); | |
2635 | ||
3d8e560d | 2636 | if (list_empty(&ctx->flc_lease)) |
8634b51f JL |
2637 | return; |
2638 | ||
02e525b2 | 2639 | percpu_down_read(&file_rwsem); |
6109c850 | 2640 | spin_lock(&ctx->flc_lock); |
8634b51f | 2641 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
c4e136cd JL |
2642 | if (filp == fl->fl_file) |
2643 | lease_modify(fl, F_UNLCK, &dispose); | |
6109c850 | 2644 | spin_unlock(&ctx->flc_lock); |
02e525b2 | 2645 | percpu_up_read(&file_rwsem); |
5f43086b | 2646 | |
8634b51f JL |
2647 | locks_dispose_list(&dispose); |
2648 | } | |
2649 | ||
1da177e4 LT |
2650 | /* |
2651 | * This function is called on the last close of an open file. | |
2652 | */ | |
78ed8a13 | 2653 | void locks_remove_file(struct file *filp) |
1da177e4 | 2654 | { |
128a3785 DV |
2655 | struct file_lock_context *ctx; |
2656 | ||
c568d683 | 2657 | ctx = smp_load_acquire(&locks_inode(filp)->i_flctx); |
128a3785 | 2658 | if (!ctx) |
3d8e560d JL |
2659 | return; |
2660 | ||
dd459bb1 | 2661 | /* remove any OFD locks */ |
73a8f5f7 | 2662 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2663 | |
dd459bb1 | 2664 | /* remove flock locks */ |
128a3785 | 2665 | locks_remove_flock(filp, ctx); |
dd459bb1 | 2666 | |
8634b51f | 2667 | /* remove any leases */ |
128a3785 | 2668 | locks_remove_lease(filp, ctx); |
3953704f BC |
2669 | |
2670 | spin_lock(&ctx->flc_lock); | |
2671 | locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX"); | |
2672 | locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK"); | |
2673 | locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE"); | |
2674 | spin_unlock(&ctx->flc_lock); | |
1da177e4 LT |
2675 | } |
2676 | ||
9b9d2ab4 ME |
2677 | /** |
2678 | * vfs_cancel_lock - file byte range unblock lock | |
2679 | * @filp: The file to apply the unblock to | |
2680 | * @fl: The lock to be unblocked | |
2681 | * | |
2682 | * Used by lock managers to cancel blocked requests | |
2683 | */ | |
2684 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2685 | { | |
de2a4a50 | 2686 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2687 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2688 | return 0; | |
2689 | } | |
9b9d2ab4 ME |
2690 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); |
2691 | ||
7f8ada98 | 2692 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2693 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2694 | #include <linux/seq_file.h> |
2695 | ||
7012b02a JL |
2696 | struct locks_iterator { |
2697 | int li_cpu; | |
2698 | loff_t li_pos; | |
2699 | }; | |
2700 | ||
7f8ada98 | 2701 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
b8da9b10 | 2702 | loff_t id, char *pfx, int repeat) |
1da177e4 LT |
2703 | { |
2704 | struct inode *inode = NULL; | |
ab1f1611 | 2705 | unsigned int fl_pid; |
9d78edea | 2706 | struct pid_namespace *proc_pidns = proc_pid_ns(file_inode(f->file)->i_sb); |
90f7d7a0 | 2707 | int type; |
ab1f1611 | 2708 | |
9d5b86ac BC |
2709 | fl_pid = locks_translate_pid(fl, proc_pidns); |
2710 | /* | |
1cf8e5de KK |
2711 | * If lock owner is dead (and pid is freed) or not visible in current |
2712 | * pidns, zero is shown as a pid value. Check lock info from | |
2713 | * init_pid_ns to get saved lock pid value. | |
9d5b86ac | 2714 | */ |
1da177e4 LT |
2715 | |
2716 | if (fl->fl_file != NULL) | |
c568d683 | 2717 | inode = locks_inode(fl->fl_file); |
1da177e4 | 2718 | |
b8da9b10 LL |
2719 | seq_printf(f, "%lld: ", id); |
2720 | ||
2721 | if (repeat) | |
2722 | seq_printf(f, "%*s", repeat - 1 + (int)strlen(pfx), pfx); | |
2723 | ||
1da177e4 | 2724 | if (IS_POSIX(fl)) { |
c918d42a | 2725 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2726 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2727 | else if (IS_OFDLCK(fl)) |
5315c26a | 2728 | seq_puts(f, "OFDLCK"); |
c918d42a | 2729 | else |
5315c26a | 2730 | seq_puts(f, "POSIX "); |
c918d42a JL |
2731 | |
2732 | seq_printf(f, " %s ", | |
f7e33bdb | 2733 | (inode == NULL) ? "*NOINODE*" : "ADVISORY "); |
1da177e4 | 2734 | } else if (IS_FLOCK(fl)) { |
90f7d7a0 | 2735 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 | 2736 | } else if (IS_LEASE(fl)) { |
8144f1f6 JL |
2737 | if (fl->fl_flags & FL_DELEG) |
2738 | seq_puts(f, "DELEG "); | |
2739 | else | |
2740 | seq_puts(f, "LEASE "); | |
2741 | ||
ab83fa4b | 2742 | if (lease_breaking(fl)) |
5315c26a | 2743 | seq_puts(f, "BREAKING "); |
1da177e4 | 2744 | else if (fl->fl_file) |
5315c26a | 2745 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2746 | else |
5315c26a | 2747 | seq_puts(f, "BREAKER "); |
1da177e4 | 2748 | } else { |
5315c26a | 2749 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 | 2750 | } |
90f7d7a0 | 2751 | type = IS_LEASE(fl) ? target_leasetype(fl) : fl->fl_type; |
43e4cb94 | 2752 | |
90f7d7a0 JL |
2753 | seq_printf(f, "%s ", (type == F_WRLCK) ? "WRITE" : |
2754 | (type == F_RDLCK) ? "READ" : "UNLCK"); | |
1da177e4 | 2755 | if (inode) { |
3648888e | 2756 | /* userspace relies on this representation of dev_t */ |
98ca480a | 2757 | seq_printf(f, "%d %02x:%02x:%lu ", fl_pid, |
1da177e4 LT |
2758 | MAJOR(inode->i_sb->s_dev), |
2759 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
1da177e4 | 2760 | } else { |
ab1f1611 | 2761 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2762 | } |
2763 | if (IS_POSIX(fl)) { | |
2764 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2765 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2766 | else |
7f8ada98 | 2767 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2768 | } else { |
5315c26a | 2769 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2770 | } |
2771 | } | |
2772 | ||
b8da9b10 LL |
2773 | static struct file_lock *get_next_blocked_member(struct file_lock *node) |
2774 | { | |
2775 | struct file_lock *tmp; | |
2776 | ||
2777 | /* NULL node or root node */ | |
2778 | if (node == NULL || node->fl_blocker == NULL) | |
2779 | return NULL; | |
2780 | ||
2781 | /* Next member in the linked list could be itself */ | |
2782 | tmp = list_next_entry(node, fl_blocked_member); | |
2783 | if (list_entry_is_head(tmp, &node->fl_blocker->fl_blocked_requests, fl_blocked_member) | |
2784 | || tmp == node) { | |
2785 | return NULL; | |
2786 | } | |
2787 | ||
2788 | return tmp; | |
2789 | } | |
2790 | ||
7f8ada98 | 2791 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2792 | { |
7012b02a | 2793 | struct locks_iterator *iter = f->private; |
b8da9b10 | 2794 | struct file_lock *cur, *tmp; |
9d78edea | 2795 | struct pid_namespace *proc_pidns = proc_pid_ns(file_inode(f->file)->i_sb); |
b8da9b10 | 2796 | int level = 0; |
1da177e4 | 2797 | |
b8da9b10 | 2798 | cur = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2799 | |
b8da9b10 | 2800 | if (locks_translate_pid(cur, proc_pidns) == 0) |
d67fd44f NB |
2801 | return 0; |
2802 | ||
b8da9b10 LL |
2803 | /* View this crossed linked list as a binary tree, the first member of fl_blocked_requests |
2804 | * is the left child of current node, the next silibing in fl_blocked_member is the | |
2805 | * right child, we can alse get the parent of current node from fl_blocker, so this | |
2806 | * question becomes traversal of a binary tree | |
2807 | */ | |
2808 | while (cur != NULL) { | |
2809 | if (level) | |
2810 | lock_get_status(f, cur, iter->li_pos, "-> ", level); | |
2811 | else | |
2812 | lock_get_status(f, cur, iter->li_pos, "", level); | |
1da177e4 | 2813 | |
b8da9b10 LL |
2814 | if (!list_empty(&cur->fl_blocked_requests)) { |
2815 | /* Turn left */ | |
2816 | cur = list_first_entry_or_null(&cur->fl_blocked_requests, | |
2817 | struct file_lock, fl_blocked_member); | |
2818 | level++; | |
2819 | } else { | |
2820 | /* Turn right */ | |
2821 | tmp = get_next_blocked_member(cur); | |
2822 | /* Fall back to parent node */ | |
2823 | while (tmp == NULL && cur->fl_blocker != NULL) { | |
2824 | cur = cur->fl_blocker; | |
2825 | level--; | |
2826 | tmp = get_next_blocked_member(cur); | |
2827 | } | |
2828 | cur = tmp; | |
2829 | } | |
2830 | } | |
094f2825 | 2831 | |
7f8ada98 PE |
2832 | return 0; |
2833 | } | |
1da177e4 | 2834 | |
6c8c9031 AV |
2835 | static void __show_fd_locks(struct seq_file *f, |
2836 | struct list_head *head, int *id, | |
2837 | struct file *filp, struct files_struct *files) | |
2838 | { | |
2839 | struct file_lock *fl; | |
2840 | ||
2841 | list_for_each_entry(fl, head, fl_list) { | |
2842 | ||
2843 | if (filp != fl->fl_file) | |
2844 | continue; | |
2845 | if (fl->fl_owner != files && | |
2846 | fl->fl_owner != filp) | |
2847 | continue; | |
2848 | ||
2849 | (*id)++; | |
2850 | seq_puts(f, "lock:\t"); | |
b8da9b10 | 2851 | lock_get_status(f, fl, *id, "", 0); |
6c8c9031 AV |
2852 | } |
2853 | } | |
2854 | ||
2855 | void show_fd_locks(struct seq_file *f, | |
2856 | struct file *filp, struct files_struct *files) | |
2857 | { | |
c568d683 | 2858 | struct inode *inode = locks_inode(filp); |
6c8c9031 AV |
2859 | struct file_lock_context *ctx; |
2860 | int id = 0; | |
2861 | ||
128a3785 | 2862 | ctx = smp_load_acquire(&inode->i_flctx); |
6c8c9031 AV |
2863 | if (!ctx) |
2864 | return; | |
2865 | ||
2866 | spin_lock(&ctx->flc_lock); | |
2867 | __show_fd_locks(f, &ctx->flc_flock, &id, filp, files); | |
2868 | __show_fd_locks(f, &ctx->flc_posix, &id, filp, files); | |
2869 | __show_fd_locks(f, &ctx->flc_lease, &id, filp, files); | |
2870 | spin_unlock(&ctx->flc_lock); | |
2871 | } | |
2872 | ||
7f8ada98 | 2873 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2874 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2875 | { |
7012b02a | 2876 | struct locks_iterator *iter = f->private; |
99dc8292 | 2877 | |
7012b02a | 2878 | iter->li_pos = *pos + 1; |
aba37660 | 2879 | percpu_down_write(&file_rwsem); |
7b2296af | 2880 | spin_lock(&blocked_lock_lock); |
7c3f654d | 2881 | return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos); |
7f8ada98 | 2882 | } |
1da177e4 | 2883 | |
7f8ada98 PE |
2884 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2885 | { | |
7012b02a JL |
2886 | struct locks_iterator *iter = f->private; |
2887 | ||
2888 | ++iter->li_pos; | |
7c3f654d | 2889 | return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos); |
7f8ada98 | 2890 | } |
1da177e4 | 2891 | |
7f8ada98 | 2892 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2893 | __releases(&blocked_lock_lock) |
7f8ada98 | 2894 | { |
7b2296af | 2895 | spin_unlock(&blocked_lock_lock); |
aba37660 | 2896 | percpu_up_write(&file_rwsem); |
1da177e4 LT |
2897 | } |
2898 | ||
d8ba7a36 | 2899 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2900 | .start = locks_start, |
2901 | .next = locks_next, | |
2902 | .stop = locks_stop, | |
2903 | .show = locks_show, | |
2904 | }; | |
d8ba7a36 | 2905 | |
d8ba7a36 AD |
2906 | static int __init proc_locks_init(void) |
2907 | { | |
44414d82 CH |
2908 | proc_create_seq_private("locks", 0, NULL, &locks_seq_operations, |
2909 | sizeof(struct locks_iterator), NULL); | |
d8ba7a36 AD |
2910 | return 0; |
2911 | } | |
91899226 | 2912 | fs_initcall(proc_locks_init); |
7f8ada98 PE |
2913 | #endif |
2914 | ||
1da177e4 LT |
2915 | static int __init filelock_init(void) |
2916 | { | |
7012b02a JL |
2917 | int i; |
2918 | ||
4a075e39 | 2919 | flctx_cache = kmem_cache_create("file_lock_ctx", |
3754707b | 2920 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); |
4a075e39 | 2921 | |
1da177e4 | 2922 | filelock_cache = kmem_cache_create("file_lock_cache", |
3754707b | 2923 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
ee19cc40 | 2924 | |
7c3f654d PZ |
2925 | for_each_possible_cpu(i) { |
2926 | struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i); | |
2927 | ||
2928 | spin_lock_init(&fll->lock); | |
2929 | INIT_HLIST_HEAD(&fll->hlist); | |
2930 | } | |
7012b02a | 2931 | |
18f6622e | 2932 | lease_notifier_chain_init(); |
1da177e4 LT |
2933 | return 0; |
2934 | } | |
1da177e4 | 2935 | core_initcall(filelock_init); |