orangefs: Remove redundant initialization of variable ret
[linux-2.6-block.git] / fs / locks.c
CommitLineData
457c8996 1// SPDX-License-Identifier: GPL-2.0-only
1da177e4
LT
2/*
3 * linux/fs/locks.c
4 *
5 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
6 * Doug Evans (dje@spiff.uucp), August 07, 1992
7 *
8 * Deadlock detection added.
9 * FIXME: one thing isn't handled yet:
10 * - mandatory locks (requires lots of changes elsewhere)
11 * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994.
12 *
13 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
14 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
7bbd1fc0 15 *
1da177e4
LT
16 * Converted file_lock_table to a linked list from an array, which eliminates
17 * the limits on how many active file locks are open.
18 * Chad Page (pageone@netcom.com), November 27, 1994
7bbd1fc0 19 *
1da177e4
LT
20 * Removed dependency on file descriptors. dup()'ed file descriptors now
21 * get the same locks as the original file descriptors, and a close() on
22 * any file descriptor removes ALL the locks on the file for the current
23 * process. Since locks still depend on the process id, locks are inherited
24 * after an exec() but not after a fork(). This agrees with POSIX, and both
25 * BSD and SVR4 practice.
26 * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995
27 *
28 * Scrapped free list which is redundant now that we allocate locks
29 * dynamically with kmalloc()/kfree().
30 * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995
31 *
32 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
33 *
34 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
35 * fcntl() system call. They have the semantics described above.
36 *
37 * FL_FLOCK locks are created with calls to flock(), through the flock()
38 * system call, which is new. Old C libraries implement flock() via fcntl()
39 * and will continue to use the old, broken implementation.
40 *
41 * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated
42 * with a file pointer (filp). As a result they can be shared by a parent
43 * process and its children after a fork(). They are removed when the last
44 * file descriptor referring to the file pointer is closed (unless explicitly
7bbd1fc0 45 * unlocked).
1da177e4
LT
46 *
47 * FL_FLOCK locks never deadlock, an existing lock is always removed before
48 * upgrading from shared to exclusive (or vice versa). When this happens
49 * any processes blocked by the current lock are woken up and allowed to
50 * run before the new lock is applied.
51 * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995
52 *
53 * Removed some race conditions in flock_lock_file(), marked other possible
7bbd1fc0 54 * races. Just grep for FIXME to see them.
1da177e4
LT
55 * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996.
56 *
57 * Addressed Dmitry's concerns. Deadlock checking no longer recursive.
58 * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep
59 * once we've checked for blocking and deadlocking.
60 * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996.
61 *
62 * Initial implementation of mandatory locks. SunOS turned out to be
63 * a rotten model, so I implemented the "obvious" semantics.
a02dcdf6 64 * See 'Documentation/filesystems/mandatory-locking.rst' for details.
1da177e4
LT
65 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
66 *
67 * Don't allow mandatory locks on mmap()'ed files. Added simple functions to
68 * check if a file has mandatory locks, used by mmap(), open() and creat() to
69 * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference
70 * Manual, Section 2.
71 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
72 *
73 * Tidied up block list handling. Added '/proc/locks' interface.
74 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
75 *
76 * Fixed deadlock condition for pathological code that mixes calls to
77 * flock() and fcntl().
78 * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996.
79 *
80 * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use
81 * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to
82 * guarantee sensible behaviour in the case where file system modules might
83 * be compiled with different options than the kernel itself.
84 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
85 *
86 * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel
87 * (Thomas.Meckel@mni.fh-giessen.de) for spotting this.
88 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
89 *
90 * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK
91 * locks. Changed process synchronisation to avoid dereferencing locks that
92 * have already been freed.
93 * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996.
94 *
95 * Made the block list a circular list to minimise searching in the list.
96 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
97 *
98 * Made mandatory locking a mount option. Default is not to allow mandatory
99 * locking.
100 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
101 *
102 * Some adaptations for NFS support.
103 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
104 *
105 * Fixed /proc/locks interface so that we can't overrun the buffer we are handed.
106 * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997.
107 *
108 * Use slab allocator instead of kmalloc/kfree.
109 * Use generic list implementation from <linux/list.h>.
110 * Sped up posix_locks_deadlock by only considering blocked locks.
111 * Matthew Wilcox <willy@debian.org>, March, 2000.
112 *
113 * Leases and LOCK_MAND
114 * Matthew Wilcox <willy@debian.org>, June, 2000.
115 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
fd7732e0
N
116 *
117 * Locking conflicts and dependencies:
118 * If multiple threads attempt to lock the same byte (or flock the same file)
119 * only one can be granted the lock, and other must wait their turn.
120 * The first lock has been "applied" or "granted", the others are "waiting"
121 * and are "blocked" by the "applied" lock..
122 *
123 * Waiting and applied locks are all kept in trees whose properties are:
124 *
125 * - the root of a tree may be an applied or waiting lock.
126 * - every other node in the tree is a waiting lock that
127 * conflicts with every ancestor of that node.
128 *
129 * Every such tree begins life as a waiting singleton which obviously
130 * satisfies the above properties.
131 *
132 * The only ways we modify trees preserve these properties:
133 *
134 * 1. We may add a new leaf node, but only after first verifying that it
135 * conflicts with all of its ancestors.
136 * 2. We may remove the root of a tree, creating a new singleton
137 * tree from the root and N new trees rooted in the immediate
138 * children.
139 * 3. If the root of a tree is not currently an applied lock, we may
140 * apply it (if possible).
141 * 4. We may upgrade the root of the tree (either extend its range,
142 * or upgrade its entire range from read to write).
143 *
144 * When an applied lock is modified in a way that reduces or downgrades any
145 * part of its range, we remove all its children (2 above). This particularly
146 * happens when a lock is unlocked.
147 *
148 * For each of those child trees we "wake up" the thread which is
149 * waiting for the lock so it can continue handling as follows: if the
150 * root of the tree applies, we do so (3). If it doesn't, it must
151 * conflict with some applied lock. We remove (wake up) all of its children
152 * (2), and add it is a new leaf to the tree rooted in the applied
153 * lock (1). We then repeat the process recursively with those
154 * children.
155 *
1da177e4
LT
156 */
157
158#include <linux/capability.h>
159#include <linux/file.h>
9f3acc31 160#include <linux/fdtable.h>
1da177e4
LT
161#include <linux/fs.h>
162#include <linux/init.h>
1da177e4
LT
163#include <linux/security.h>
164#include <linux/slab.h>
1da177e4
LT
165#include <linux/syscalls.h>
166#include <linux/time.h>
4fb3a538 167#include <linux/rcupdate.h>
ab1f1611 168#include <linux/pid_namespace.h>
48f74186 169#include <linux/hashtable.h>
7012b02a 170#include <linux/percpu.h>
1da177e4 171
62af4f1f
JL
172#define CREATE_TRACE_POINTS
173#include <trace/events/filelock.h>
174
7c0f6ba6 175#include <linux/uaccess.h>
1da177e4
LT
176
177#define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
178#define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
11afe9f7 179#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
cff2fce5 180#define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
9d5b86ac 181#define IS_REMOTELCK(fl) (fl->fl_pid <= 0)
1da177e4 182
ab83fa4b
BF
183static bool lease_breaking(struct file_lock *fl)
184{
778fc546
BF
185 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
186}
187
188static int target_leasetype(struct file_lock *fl)
189{
190 if (fl->fl_flags & FL_UNLOCK_PENDING)
191 return F_UNLCK;
192 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
193 return F_RDLCK;
194 return fl->fl_type;
ab83fa4b
BF
195}
196
1da177e4
LT
197int leases_enable = 1;
198int lease_break_time = 45;
199
1c8c601a 200/*
7012b02a 201 * The global file_lock_list is only used for displaying /proc/locks, so we
7c3f654d
PZ
202 * keep a list on each CPU, with each list protected by its own spinlock.
203 * Global serialization is done using file_rwsem.
204 *
205 * Note that alterations to the list also require that the relevant flc_lock is
206 * held.
1c8c601a 207 */
7c3f654d
PZ
208struct file_lock_list_struct {
209 spinlock_t lock;
210 struct hlist_head hlist;
211};
212static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list);
aba37660 213DEFINE_STATIC_PERCPU_RWSEM(file_rwsem);
88974691 214
eb82dd39 215
1c8c601a 216/*
48f74186 217 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
7b2296af 218 * It is protected by blocked_lock_lock.
48f74186
JL
219 *
220 * We hash locks by lockowner in order to optimize searching for the lock a
221 * particular lockowner is waiting on.
222 *
223 * FIXME: make this value scale via some heuristic? We generally will want more
224 * buckets when we have more lockowners holding locks, but that's a little
225 * difficult to determine without knowing what the workload will look like.
1c8c601a 226 */
48f74186
JL
227#define BLOCKED_HASH_BITS 7
228static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
88974691 229
1c8c601a 230/*
7b2296af
JL
231 * This lock protects the blocked_hash. Generally, if you're accessing it, you
232 * want to be holding this lock.
1c8c601a 233 *
ada5c1da
N
234 * In addition, it also protects the fl->fl_blocked_requests list, and the
235 * fl->fl_blocker pointer for file_lock structures that are acting as lock
236 * requests (in contrast to those that are acting as records of acquired locks).
1c8c601a
JL
237 *
238 * Note that when we acquire this lock in order to change the above fields,
6109c850 239 * we often hold the flc_lock as well. In certain cases, when reading the fields
1c8c601a 240 * protected by this lock, we can skip acquiring it iff we already hold the
6109c850 241 * flc_lock.
1c8c601a 242 */
7b2296af 243static DEFINE_SPINLOCK(blocked_lock_lock);
1da177e4 244
4a075e39 245static struct kmem_cache *flctx_cache __read_mostly;
e18b890b 246static struct kmem_cache *filelock_cache __read_mostly;
1da177e4 247
4a075e39 248static struct file_lock_context *
5c1c669a 249locks_get_lock_context(struct inode *inode, int type)
4a075e39 250{
128a3785 251 struct file_lock_context *ctx;
4a075e39 252
128a3785
DV
253 /* paired with cmpxchg() below */
254 ctx = smp_load_acquire(&inode->i_flctx);
255 if (likely(ctx) || type == F_UNLCK)
4a075e39
JL
256 goto out;
257
128a3785
DV
258 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
259 if (!ctx)
4a075e39
JL
260 goto out;
261
128a3785
DV
262 spin_lock_init(&ctx->flc_lock);
263 INIT_LIST_HEAD(&ctx->flc_flock);
264 INIT_LIST_HEAD(&ctx->flc_posix);
265 INIT_LIST_HEAD(&ctx->flc_lease);
4a075e39
JL
266
267 /*
268 * Assign the pointer if it's not already assigned. If it is, then
269 * free the context we just allocated.
270 */
128a3785
DV
271 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
272 kmem_cache_free(flctx_cache, ctx);
273 ctx = smp_load_acquire(&inode->i_flctx);
274 }
4a075e39 275out:
1890910f 276 trace_locks_get_lock_context(inode, type, ctx);
128a3785 277 return ctx;
4a075e39
JL
278}
279
e24dadab
JL
280static void
281locks_dump_ctx_list(struct list_head *list, char *list_type)
282{
283 struct file_lock *fl;
284
285 list_for_each_entry(fl, list, fl_list) {
286 pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type, fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
287 }
288}
289
290static void
291locks_check_ctx_lists(struct inode *inode)
292{
293 struct file_lock_context *ctx = inode->i_flctx;
294
295 if (unlikely(!list_empty(&ctx->flc_flock) ||
296 !list_empty(&ctx->flc_posix) ||
297 !list_empty(&ctx->flc_lease))) {
298 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
299 MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev),
300 inode->i_ino);
301 locks_dump_ctx_list(&ctx->flc_flock, "FLOCK");
302 locks_dump_ctx_list(&ctx->flc_posix, "POSIX");
303 locks_dump_ctx_list(&ctx->flc_lease, "LEASE");
304 }
305}
306
3953704f
BC
307static void
308locks_check_ctx_file_list(struct file *filp, struct list_head *list,
309 char *list_type)
310{
311 struct file_lock *fl;
312 struct inode *inode = locks_inode(filp);
313
314 list_for_each_entry(fl, list, fl_list)
315 if (fl->fl_file == filp)
316 pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx "
317 " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n",
318 list_type, MAJOR(inode->i_sb->s_dev),
319 MINOR(inode->i_sb->s_dev), inode->i_ino,
320 fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
321}
322
4a075e39 323void
f27a0fe0 324locks_free_lock_context(struct inode *inode)
4a075e39 325{
f27a0fe0
JL
326 struct file_lock_context *ctx = inode->i_flctx;
327
e24dadab
JL
328 if (unlikely(ctx)) {
329 locks_check_ctx_lists(inode);
4a075e39
JL
330 kmem_cache_free(flctx_cache, ctx);
331 }
332}
333
ee19cc40 334static void locks_init_lock_heads(struct file_lock *fl)
a51cb91d 335{
139ca04e 336 INIT_HLIST_NODE(&fl->fl_link);
6dee60f6 337 INIT_LIST_HEAD(&fl->fl_list);
ada5c1da
N
338 INIT_LIST_HEAD(&fl->fl_blocked_requests);
339 INIT_LIST_HEAD(&fl->fl_blocked_member);
ee19cc40 340 init_waitqueue_head(&fl->fl_wait);
a51cb91d
MS
341}
342
1da177e4 343/* Allocate an empty lock structure. */
c5b1f0d9 344struct file_lock *locks_alloc_lock(void)
1da177e4 345{
ee19cc40 346 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
a51cb91d
MS
347
348 if (fl)
ee19cc40 349 locks_init_lock_heads(fl);
a51cb91d
MS
350
351 return fl;
1da177e4 352}
c5b1f0d9 353EXPORT_SYMBOL_GPL(locks_alloc_lock);
1da177e4 354
a9e61e25 355void locks_release_private(struct file_lock *fl)
47831f35 356{
5926459e
N
357 BUG_ON(waitqueue_active(&fl->fl_wait));
358 BUG_ON(!list_empty(&fl->fl_list));
359 BUG_ON(!list_empty(&fl->fl_blocked_requests));
360 BUG_ON(!list_empty(&fl->fl_blocked_member));
361 BUG_ON(!hlist_unhashed(&fl->fl_link));
362
47831f35
TM
363 if (fl->fl_ops) {
364 if (fl->fl_ops->fl_release_private)
365 fl->fl_ops->fl_release_private(fl);
366 fl->fl_ops = NULL;
367 }
47831f35 368
5c97d7b1 369 if (fl->fl_lmops) {
cae80b30
JL
370 if (fl->fl_lmops->lm_put_owner) {
371 fl->fl_lmops->lm_put_owner(fl->fl_owner);
372 fl->fl_owner = NULL;
373 }
5c97d7b1
KM
374 fl->fl_lmops = NULL;
375 }
47831f35 376}
a9e61e25 377EXPORT_SYMBOL_GPL(locks_release_private);
47831f35 378
1da177e4 379/* Free a lock which is not in use. */
05fa3135 380void locks_free_lock(struct file_lock *fl)
1da177e4 381{
47831f35 382 locks_release_private(fl);
1da177e4
LT
383 kmem_cache_free(filelock_cache, fl);
384}
05fa3135 385EXPORT_SYMBOL(locks_free_lock);
1da177e4 386
ed9814d8
JL
387static void
388locks_dispose_list(struct list_head *dispose)
389{
390 struct file_lock *fl;
391
392 while (!list_empty(dispose)) {
6dee60f6
JL
393 fl = list_first_entry(dispose, struct file_lock, fl_list);
394 list_del_init(&fl->fl_list);
ed9814d8
JL
395 locks_free_lock(fl);
396 }
397}
398
1da177e4
LT
399void locks_init_lock(struct file_lock *fl)
400{
ee19cc40
MS
401 memset(fl, 0, sizeof(struct file_lock));
402 locks_init_lock_heads(fl);
1da177e4 403}
1da177e4
LT
404EXPORT_SYMBOL(locks_init_lock);
405
1da177e4
LT
406/*
407 * Initialize a new lock from an existing file_lock structure.
408 */
3fe0fff1 409void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
1da177e4
LT
410{
411 new->fl_owner = fl->fl_owner;
412 new->fl_pid = fl->fl_pid;
0996905f 413 new->fl_file = NULL;
1da177e4
LT
414 new->fl_flags = fl->fl_flags;
415 new->fl_type = fl->fl_type;
416 new->fl_start = fl->fl_start;
417 new->fl_end = fl->fl_end;
f328296e 418 new->fl_lmops = fl->fl_lmops;
0996905f 419 new->fl_ops = NULL;
f328296e
KM
420
421 if (fl->fl_lmops) {
422 if (fl->fl_lmops->lm_get_owner)
cae80b30 423 fl->fl_lmops->lm_get_owner(fl->fl_owner);
f328296e 424 }
0996905f 425}
3fe0fff1 426EXPORT_SYMBOL(locks_copy_conflock);
0996905f
TM
427
428void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
429{
566709bd
JL
430 /* "new" must be a freshly-initialized lock */
431 WARN_ON_ONCE(new->fl_ops);
0996905f 432
3fe0fff1 433 locks_copy_conflock(new, fl);
f328296e 434
0996905f 435 new->fl_file = fl->fl_file;
1da177e4 436 new->fl_ops = fl->fl_ops;
47831f35 437
f328296e
KM
438 if (fl->fl_ops) {
439 if (fl->fl_ops->fl_copy_lock)
440 fl->fl_ops->fl_copy_lock(new, fl);
441 }
1da177e4 442}
1da177e4
LT
443EXPORT_SYMBOL(locks_copy_lock);
444
5946c431
N
445static void locks_move_blocks(struct file_lock *new, struct file_lock *fl)
446{
447 struct file_lock *f;
448
449 /*
450 * As ctx->flc_lock is held, new requests cannot be added to
451 * ->fl_blocked_requests, so we don't need a lock to check if it
452 * is empty.
453 */
454 if (list_empty(&fl->fl_blocked_requests))
455 return;
456 spin_lock(&blocked_lock_lock);
457 list_splice_init(&fl->fl_blocked_requests, &new->fl_blocked_requests);
bf77ae4c 458 list_for_each_entry(f, &new->fl_blocked_requests, fl_blocked_member)
5946c431
N
459 f->fl_blocker = new;
460 spin_unlock(&blocked_lock_lock);
461}
462
1da177e4
LT
463static inline int flock_translate_cmd(int cmd) {
464 if (cmd & LOCK_MAND)
465 return cmd & (LOCK_MAND | LOCK_RW);
466 switch (cmd) {
467 case LOCK_SH:
468 return F_RDLCK;
469 case LOCK_EX:
470 return F_WRLCK;
471 case LOCK_UN:
472 return F_UNLCK;
473 }
474 return -EINVAL;
475}
476
477/* Fill in a file_lock structure with an appropriate FLOCK lock. */
6e129d00 478static struct file_lock *
d6367d62 479flock_make_lock(struct file *filp, unsigned int cmd, struct file_lock *fl)
1da177e4 480{
1da177e4 481 int type = flock_translate_cmd(cmd);
6e129d00 482
1da177e4 483 if (type < 0)
6e129d00 484 return ERR_PTR(type);
7bbd1fc0 485
d6367d62
N
486 if (fl == NULL) {
487 fl = locks_alloc_lock();
488 if (fl == NULL)
489 return ERR_PTR(-ENOMEM);
490 } else {
491 locks_init_lock(fl);
492 }
1da177e4
LT
493
494 fl->fl_file = filp;
73a8f5f7 495 fl->fl_owner = filp;
1da177e4
LT
496 fl->fl_pid = current->tgid;
497 fl->fl_flags = FL_FLOCK;
498 fl->fl_type = type;
499 fl->fl_end = OFFSET_MAX;
7bbd1fc0 500
6e129d00 501 return fl;
1da177e4
LT
502}
503
0ec4f431 504static int assign_type(struct file_lock *fl, long type)
1da177e4
LT
505{
506 switch (type) {
507 case F_RDLCK:
508 case F_WRLCK:
509 case F_UNLCK:
510 fl->fl_type = type;
511 break;
512 default:
513 return -EINVAL;
514 }
515 return 0;
516}
517
ef12e72a
BF
518static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
519 struct flock64 *l)
1da177e4 520{
1da177e4 521 switch (l->l_whence) {
f5579f8c 522 case SEEK_SET:
ef12e72a 523 fl->fl_start = 0;
1da177e4 524 break;
f5579f8c 525 case SEEK_CUR:
ef12e72a 526 fl->fl_start = filp->f_pos;
1da177e4 527 break;
f5579f8c 528 case SEEK_END:
ef12e72a 529 fl->fl_start = i_size_read(file_inode(filp));
1da177e4
LT
530 break;
531 default:
532 return -EINVAL;
533 }
ef12e72a
BF
534 if (l->l_start > OFFSET_MAX - fl->fl_start)
535 return -EOVERFLOW;
536 fl->fl_start += l->l_start;
537 if (fl->fl_start < 0)
538 return -EINVAL;
1da177e4
LT
539
540 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
541 POSIX-2001 defines it. */
4c780a46 542 if (l->l_len > 0) {
ef12e72a
BF
543 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
544 return -EOVERFLOW;
16238415 545 fl->fl_end = fl->fl_start + (l->l_len - 1);
ef12e72a 546
4c780a46 547 } else if (l->l_len < 0) {
ef12e72a 548 if (fl->fl_start + l->l_len < 0)
4c780a46 549 return -EINVAL;
ef12e72a
BF
550 fl->fl_end = fl->fl_start - 1;
551 fl->fl_start += l->l_len;
552 } else
553 fl->fl_end = OFFSET_MAX;
554
1da177e4
LT
555 fl->fl_owner = current->files;
556 fl->fl_pid = current->tgid;
557 fl->fl_file = filp;
558 fl->fl_flags = FL_POSIX;
559 fl->fl_ops = NULL;
560 fl->fl_lmops = NULL;
561
562 return assign_type(fl, l->l_type);
563}
564
ef12e72a
BF
565/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
566 * style lock.
567 */
568static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
569 struct flock *l)
1da177e4 570{
ef12e72a
BF
571 struct flock64 ll = {
572 .l_type = l->l_type,
573 .l_whence = l->l_whence,
574 .l_start = l->l_start,
575 .l_len = l->l_len,
576 };
577
578 return flock64_to_posix_lock(filp, fl, &ll);
1da177e4 579}
1da177e4
LT
580
581/* default lease lock manager operations */
4d01b7f5
JL
582static bool
583lease_break_callback(struct file_lock *fl)
1da177e4
LT
584{
585 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
4d01b7f5 586 return false;
1da177e4
LT
587}
588
1c7dd2ff
JL
589static void
590lease_setup(struct file_lock *fl, void **priv)
591{
592 struct file *filp = fl->fl_file;
593 struct fasync_struct *fa = *priv;
594
595 /*
596 * fasync_insert_entry() returns the old entry if any. If there was no
597 * old entry, then it used "priv" and inserted it into the fasync list.
598 * Clear the pointer to indicate that it shouldn't be freed.
599 */
600 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
601 *priv = NULL;
602
01919134 603 __f_setown(filp, task_pid(current), PIDTYPE_TGID, 0);
1c7dd2ff
JL
604}
605
7b021967 606static const struct lock_manager_operations lease_manager_ops = {
8fb47a4f 607 .lm_break = lease_break_callback,
8fb47a4f 608 .lm_change = lease_modify,
1c7dd2ff 609 .lm_setup = lease_setup,
1da177e4
LT
610};
611
612/*
613 * Initialize a lease, use the default lock manager operations
614 */
0ec4f431 615static int lease_init(struct file *filp, long type, struct file_lock *fl)
447a5647 616{
75dff55a
TM
617 if (assign_type(fl, type) != 0)
618 return -EINVAL;
619
7ca76311 620 fl->fl_owner = filp;
1da177e4
LT
621 fl->fl_pid = current->tgid;
622
623 fl->fl_file = filp;
624 fl->fl_flags = FL_LEASE;
1da177e4
LT
625 fl->fl_start = 0;
626 fl->fl_end = OFFSET_MAX;
627 fl->fl_ops = NULL;
628 fl->fl_lmops = &lease_manager_ops;
629 return 0;
630}
631
632/* Allocate a file_lock initialised to this type of lease */
0ec4f431 633static struct file_lock *lease_alloc(struct file *filp, long type)
1da177e4
LT
634{
635 struct file_lock *fl = locks_alloc_lock();
75dff55a 636 int error = -ENOMEM;
1da177e4
LT
637
638 if (fl == NULL)
e32b8ee2 639 return ERR_PTR(error);
1da177e4
LT
640
641 error = lease_init(filp, type, fl);
75dff55a
TM
642 if (error) {
643 locks_free_lock(fl);
e32b8ee2 644 return ERR_PTR(error);
75dff55a 645 }
e32b8ee2 646 return fl;
1da177e4
LT
647}
648
649/* Check if two locks overlap each other.
650 */
651static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
652{
653 return ((fl1->fl_end >= fl2->fl_start) &&
654 (fl2->fl_end >= fl1->fl_start));
655}
656
657/*
658 * Check whether two locks have the same owner.
659 */
33443c42 660static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
1da177e4 661{
1da177e4
LT
662 return fl1->fl_owner == fl2->fl_owner;
663}
664
6109c850 665/* Must be called with the flc_lock held! */
6ca10ed8 666static void locks_insert_global_locks(struct file_lock *fl)
88974691 667{
7c3f654d
PZ
668 struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list);
669
aba37660
PZ
670 percpu_rwsem_assert_held(&file_rwsem);
671
7c3f654d 672 spin_lock(&fll->lock);
7012b02a 673 fl->fl_link_cpu = smp_processor_id();
7c3f654d
PZ
674 hlist_add_head(&fl->fl_link, &fll->hlist);
675 spin_unlock(&fll->lock);
88974691
JL
676}
677
6109c850 678/* Must be called with the flc_lock held! */
6ca10ed8 679static void locks_delete_global_locks(struct file_lock *fl)
88974691 680{
7c3f654d
PZ
681 struct file_lock_list_struct *fll;
682
aba37660
PZ
683 percpu_rwsem_assert_held(&file_rwsem);
684
7012b02a
JL
685 /*
686 * Avoid taking lock if already unhashed. This is safe since this check
6109c850 687 * is done while holding the flc_lock, and new insertions into the list
7012b02a
JL
688 * also require that it be held.
689 */
690 if (hlist_unhashed(&fl->fl_link))
691 return;
7c3f654d
PZ
692
693 fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu);
694 spin_lock(&fll->lock);
139ca04e 695 hlist_del_init(&fl->fl_link);
7c3f654d 696 spin_unlock(&fll->lock);
88974691
JL
697}
698
3999e493
JL
699static unsigned long
700posix_owner_key(struct file_lock *fl)
701{
3999e493
JL
702 return (unsigned long)fl->fl_owner;
703}
704
6ca10ed8 705static void locks_insert_global_blocked(struct file_lock *waiter)
88974691 706{
663d5af7
DW
707 lockdep_assert_held(&blocked_lock_lock);
708
3999e493 709 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
88974691
JL
710}
711
6ca10ed8 712static void locks_delete_global_blocked(struct file_lock *waiter)
88974691 713{
663d5af7
DW
714 lockdep_assert_held(&blocked_lock_lock);
715
48f74186 716 hash_del(&waiter->fl_link);
88974691
JL
717}
718
1da177e4
LT
719/* Remove waiter from blocker's block list.
720 * When blocker ends up pointing to itself then the list is empty.
1c8c601a 721 *
7b2296af 722 * Must be called with blocked_lock_lock held.
1da177e4 723 */
33443c42 724static void __locks_delete_block(struct file_lock *waiter)
1da177e4 725{
88974691 726 locks_delete_global_blocked(waiter);
ada5c1da 727 list_del_init(&waiter->fl_blocked_member);
1da177e4
LT
728}
729
ad6bbd8b
N
730static void __locks_wake_up_blocks(struct file_lock *blocker)
731{
732 while (!list_empty(&blocker->fl_blocked_requests)) {
733 struct file_lock *waiter;
734
735 waiter = list_first_entry(&blocker->fl_blocked_requests,
736 struct file_lock, fl_blocked_member);
737 __locks_delete_block(waiter);
738 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
739 waiter->fl_lmops->lm_notify(waiter);
740 else
741 wake_up(&waiter->fl_wait);
dcf23ac3
LT
742
743 /*
744 * The setting of fl_blocker to NULL marks the "done"
745 * point in deleting a block. Paired with acquire at the top
746 * of locks_delete_block().
747 */
748 smp_store_release(&waiter->fl_blocker, NULL);
ad6bbd8b
N
749 }
750}
751
cb03f94f 752/**
529adfe8 753 * locks_delete_block - stop waiting for a file lock
cb03f94f
N
754 * @waiter: the lock which was waiting
755 *
756 * lockd/nfsd need to disconnect the lock while working on it.
757 */
758int locks_delete_block(struct file_lock *waiter)
1da177e4 759{
cb03f94f
N
760 int status = -ENOENT;
761
dcf23ac3
LT
762 /*
763 * If fl_blocker is NULL, it won't be set again as this thread "owns"
764 * the lock and is the only one that might try to claim the lock.
765 *
766 * We use acquire/release to manage fl_blocker so that we can
767 * optimize away taking the blocked_lock_lock in many cases.
768 *
769 * The smp_load_acquire guarantees two things:
770 *
771 * 1/ that fl_blocked_requests can be tested locklessly. If something
772 * was recently added to that list it must have been in a locked region
773 * *before* the locked region when fl_blocker was set to NULL.
774 *
775 * 2/ that no other thread is accessing 'waiter', so it is safe to free
776 * it. __locks_wake_up_blocks is careful not to touch waiter after
777 * fl_blocker is released.
778 *
779 * If a lockless check of fl_blocker shows it to be NULL, we know that
780 * no new locks can be inserted into its fl_blocked_requests list, and
781 * can avoid doing anything further if the list is empty.
782 */
783 if (!smp_load_acquire(&waiter->fl_blocker) &&
784 list_empty(&waiter->fl_blocked_requests))
785 return status;
786
7b2296af 787 spin_lock(&blocked_lock_lock);
cb03f94f
N
788 if (waiter->fl_blocker)
789 status = 0;
5946c431 790 __locks_wake_up_blocks(waiter);
1da177e4 791 __locks_delete_block(waiter);
dcf23ac3
LT
792
793 /*
794 * The setting of fl_blocker to NULL marks the "done" point in deleting
795 * a block. Paired with acquire at the top of this function.
796 */
797 smp_store_release(&waiter->fl_blocker, NULL);
7b2296af 798 spin_unlock(&blocked_lock_lock);
cb03f94f 799 return status;
1da177e4 800}
cb03f94f 801EXPORT_SYMBOL(locks_delete_block);
1da177e4
LT
802
803/* Insert waiter into blocker's block list.
804 * We use a circular list so that processes can be easily woken up in
805 * the order they blocked. The documentation doesn't require this but
806 * it seems like the reasonable thing to do.
1c8c601a 807 *
6109c850 808 * Must be called with both the flc_lock and blocked_lock_lock held. The
ada5c1da
N
809 * fl_blocked_requests list itself is protected by the blocked_lock_lock,
810 * but by ensuring that the flc_lock is also held on insertions we can avoid
811 * taking the blocked_lock_lock in some cases when we see that the
812 * fl_blocked_requests list is empty.
fd7732e0
N
813 *
814 * Rather than just adding to the list, we check for conflicts with any existing
815 * waiters, and add beneath any waiter that blocks the new waiter.
816 * Thus wakeups don't happen until needed.
1da177e4 817 */
1c8c601a 818static void __locks_insert_block(struct file_lock *blocker,
fd7732e0
N
819 struct file_lock *waiter,
820 bool conflict(struct file_lock *,
821 struct file_lock *))
1da177e4 822{
fd7732e0 823 struct file_lock *fl;
ada5c1da 824 BUG_ON(!list_empty(&waiter->fl_blocked_member));
fd7732e0
N
825
826new_blocker:
827 list_for_each_entry(fl, &blocker->fl_blocked_requests, fl_blocked_member)
828 if (conflict(fl, waiter)) {
829 blocker = fl;
830 goto new_blocker;
831 }
ada5c1da
N
832 waiter->fl_blocker = blocker;
833 list_add_tail(&waiter->fl_blocked_member, &blocker->fl_blocked_requests);
cff2fce5 834 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
1c8c601a 835 locks_insert_global_blocked(waiter);
5946c431
N
836
837 /* The requests in waiter->fl_blocked are known to conflict with
838 * waiter, but might not conflict with blocker, or the requests
839 * and lock which block it. So they all need to be woken.
840 */
841 __locks_wake_up_blocks(waiter);
1c8c601a
JL
842}
843
6109c850 844/* Must be called with flc_lock held. */
1c8c601a 845static void locks_insert_block(struct file_lock *blocker,
fd7732e0
N
846 struct file_lock *waiter,
847 bool conflict(struct file_lock *,
848 struct file_lock *))
1c8c601a 849{
7b2296af 850 spin_lock(&blocked_lock_lock);
fd7732e0 851 __locks_insert_block(blocker, waiter, conflict);
7b2296af 852 spin_unlock(&blocked_lock_lock);
1da177e4
LT
853}
854
1cb36012
JL
855/*
856 * Wake up processes blocked waiting for blocker.
857 *
6109c850 858 * Must be called with the inode->flc_lock held!
1da177e4
LT
859 */
860static void locks_wake_up_blocks(struct file_lock *blocker)
861{
4e8c765d
JL
862 /*
863 * Avoid taking global lock if list is empty. This is safe since new
6109c850 864 * blocked requests are only added to the list under the flc_lock, and
ada5c1da
N
865 * the flc_lock is always held here. Note that removal from the
866 * fl_blocked_requests list does not require the flc_lock, so we must
867 * recheck list_empty() after acquiring the blocked_lock_lock.
4e8c765d 868 */
ada5c1da 869 if (list_empty(&blocker->fl_blocked_requests))
4e8c765d
JL
870 return;
871
7b2296af 872 spin_lock(&blocked_lock_lock);
ad6bbd8b 873 __locks_wake_up_blocks(blocker);
7b2296af 874 spin_unlock(&blocked_lock_lock);
1da177e4
LT
875}
876
5263e31e 877static void
e084c1bd 878locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
5263e31e 879{
5263e31e
JL
880 list_add_tail(&fl->fl_list, before);
881 locks_insert_global_locks(fl);
882}
883
8634b51f 884static void
e084c1bd 885locks_unlink_lock_ctx(struct file_lock *fl)
1da177e4 886{
88974691 887 locks_delete_global_locks(fl);
8634b51f 888 list_del_init(&fl->fl_list);
1da177e4 889 locks_wake_up_blocks(fl);
24cbe784
JL
890}
891
8634b51f 892static void
e084c1bd 893locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
24cbe784 894{
e084c1bd 895 locks_unlink_lock_ctx(fl);
ed9814d8 896 if (dispose)
6dee60f6 897 list_add(&fl->fl_list, dispose);
ed9814d8
JL
898 else
899 locks_free_lock(fl);
1da177e4
LT
900}
901
902/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
903 * checks for shared/exclusive status of overlapping locks.
904 */
c0e15908
N
905static bool locks_conflict(struct file_lock *caller_fl,
906 struct file_lock *sys_fl)
1da177e4
LT
907{
908 if (sys_fl->fl_type == F_WRLCK)
c0e15908 909 return true;
1da177e4 910 if (caller_fl->fl_type == F_WRLCK)
c0e15908
N
911 return true;
912 return false;
1da177e4
LT
913}
914
915/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
916 * checking before calling the locks_conflict().
917 */
c0e15908
N
918static bool posix_locks_conflict(struct file_lock *caller_fl,
919 struct file_lock *sys_fl)
1da177e4
LT
920{
921 /* POSIX locks owned by the same process do not conflict with
922 * each other.
923 */
9b8c8695 924 if (posix_same_owner(caller_fl, sys_fl))
c0e15908 925 return false;
1da177e4
LT
926
927 /* Check whether they overlap */
928 if (!locks_overlap(caller_fl, sys_fl))
c0e15908 929 return false;
1da177e4 930
c0e15908 931 return locks_conflict(caller_fl, sys_fl);
1da177e4
LT
932}
933
934/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
935 * checking before calling the locks_conflict().
936 */
c0e15908
N
937static bool flock_locks_conflict(struct file_lock *caller_fl,
938 struct file_lock *sys_fl)
1da177e4
LT
939{
940 /* FLOCK locks referring to the same filp do not conflict with
941 * each other.
942 */
9b8c8695 943 if (caller_fl->fl_file == sys_fl->fl_file)
c0e15908 944 return false;
1da177e4 945 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
c0e15908 946 return false;
1da177e4 947
c0e15908 948 return locks_conflict(caller_fl, sys_fl);
1da177e4
LT
949}
950
6d34ac19 951void
9d6a8c5c 952posix_test_lock(struct file *filp, struct file_lock *fl)
1da177e4
LT
953{
954 struct file_lock *cfl;
bd61e0a9 955 struct file_lock_context *ctx;
c568d683 956 struct inode *inode = locks_inode(filp);
1da177e4 957
128a3785 958 ctx = smp_load_acquire(&inode->i_flctx);
bd61e0a9
JL
959 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
960 fl->fl_type = F_UNLCK;
961 return;
962 }
963
6109c850 964 spin_lock(&ctx->flc_lock);
bd61e0a9
JL
965 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
966 if (posix_locks_conflict(fl, cfl)) {
967 locks_copy_conflock(fl, cfl);
bd61e0a9
JL
968 goto out;
969 }
1da177e4 970 }
bd61e0a9
JL
971 fl->fl_type = F_UNLCK;
972out:
6109c850 973 spin_unlock(&ctx->flc_lock);
6d34ac19 974 return;
1da177e4 975}
1da177e4
LT
976EXPORT_SYMBOL(posix_test_lock);
977
b533184f
BF
978/*
979 * Deadlock detection:
980 *
981 * We attempt to detect deadlocks that are due purely to posix file
982 * locks.
1da177e4 983 *
b533184f
BF
984 * We assume that a task can be waiting for at most one lock at a time.
985 * So for any acquired lock, the process holding that lock may be
986 * waiting on at most one other lock. That lock in turns may be held by
987 * someone waiting for at most one other lock. Given a requested lock
988 * caller_fl which is about to wait for a conflicting lock block_fl, we
989 * follow this chain of waiters to ensure we are not about to create a
990 * cycle.
1da177e4 991 *
b533184f
BF
992 * Since we do this before we ever put a process to sleep on a lock, we
993 * are ensured that there is never a cycle; that is what guarantees that
994 * the while() loop in posix_locks_deadlock() eventually completes.
97855b49 995 *
b533184f
BF
996 * Note: the above assumption may not be true when handling lock
997 * requests from a broken NFS client. It may also fail in the presence
998 * of tasks (such as posix threads) sharing the same open file table.
b533184f 999 * To handle those cases, we just bail out after a few iterations.
57b65325 1000 *
cff2fce5 1001 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
57b65325
JL
1002 * Because the owner is not even nominally tied to a thread of
1003 * execution, the deadlock detection below can't reasonably work well. Just
1004 * skip it for those.
1005 *
cff2fce5 1006 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
57b65325
JL
1007 * locks that just checks for the case where two tasks are attempting to
1008 * upgrade from read to write locks on the same inode.
1da177e4 1009 */
97855b49
BF
1010
1011#define MAX_DEADLK_ITERATIONS 10
1012
b533184f
BF
1013/* Find a lock that the owner of the given block_fl is blocking on. */
1014static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
1015{
1016 struct file_lock *fl;
1017
3999e493 1018 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
5946c431
N
1019 if (posix_same_owner(fl, block_fl)) {
1020 while (fl->fl_blocker)
1021 fl = fl->fl_blocker;
1022 return fl;
1023 }
b533184f
BF
1024 }
1025 return NULL;
1026}
1027
7b2296af 1028/* Must be called with the blocked_lock_lock held! */
b0904e14 1029static int posix_locks_deadlock(struct file_lock *caller_fl,
1da177e4
LT
1030 struct file_lock *block_fl)
1031{
97855b49 1032 int i = 0;
1da177e4 1033
663d5af7
DW
1034 lockdep_assert_held(&blocked_lock_lock);
1035
57b65325
JL
1036 /*
1037 * This deadlock detector can't reasonably detect deadlocks with
cff2fce5 1038 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
57b65325 1039 */
cff2fce5 1040 if (IS_OFDLCK(caller_fl))
57b65325
JL
1041 return 0;
1042
b533184f
BF
1043 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
1044 if (i++ > MAX_DEADLK_ITERATIONS)
1045 return 0;
1046 if (posix_same_owner(caller_fl, block_fl))
1047 return 1;
1da177e4
LT
1048 }
1049 return 0;
1050}
1051
1da177e4 1052/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
02888f41 1053 * after any leases, but before any posix locks.
f475ae95
TM
1054 *
1055 * Note that if called with an FL_EXISTS argument, the caller may determine
1056 * whether or not a lock was successfully freed by testing the return
1057 * value for -ENOENT.
1da177e4 1058 */
bcd7f78d 1059static int flock_lock_inode(struct inode *inode, struct file_lock *request)
1da177e4 1060{
993dfa87 1061 struct file_lock *new_fl = NULL;
5263e31e
JL
1062 struct file_lock *fl;
1063 struct file_lock_context *ctx;
1da177e4 1064 int error = 0;
5263e31e 1065 bool found = false;
ed9814d8 1066 LIST_HEAD(dispose);
1da177e4 1067
5c1c669a
JL
1068 ctx = locks_get_lock_context(inode, request->fl_type);
1069 if (!ctx) {
1070 if (request->fl_type != F_UNLCK)
1071 return -ENOMEM;
1072 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
1073 }
5263e31e 1074
b89f4321 1075 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
84d535ad 1076 new_fl = locks_alloc_lock();
b89f4321
AB
1077 if (!new_fl)
1078 return -ENOMEM;
84d535ad
PE
1079 }
1080
02e525b2 1081 percpu_down_read(&file_rwsem);
6109c850 1082 spin_lock(&ctx->flc_lock);
b89f4321
AB
1083 if (request->fl_flags & FL_ACCESS)
1084 goto find_conflict;
1085
5263e31e 1086 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
bcd7f78d 1087 if (request->fl_file != fl->fl_file)
1da177e4 1088 continue;
993dfa87 1089 if (request->fl_type == fl->fl_type)
1da177e4 1090 goto out;
5263e31e 1091 found = true;
e084c1bd 1092 locks_delete_lock_ctx(fl, &dispose);
1da177e4
LT
1093 break;
1094 }
1da177e4 1095
f475ae95
TM
1096 if (request->fl_type == F_UNLCK) {
1097 if ((request->fl_flags & FL_EXISTS) && !found)
1098 error = -ENOENT;
993dfa87 1099 goto out;
f475ae95 1100 }
1da177e4 1101
f07f18dd 1102find_conflict:
5263e31e 1103 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
993dfa87 1104 if (!flock_locks_conflict(request, fl))
1da177e4
LT
1105 continue;
1106 error = -EAGAIN;
bde74e4b
MS
1107 if (!(request->fl_flags & FL_SLEEP))
1108 goto out;
1109 error = FILE_LOCK_DEFERRED;
fd7732e0 1110 locks_insert_block(fl, request, flock_locks_conflict);
1da177e4
LT
1111 goto out;
1112 }
f07f18dd
TM
1113 if (request->fl_flags & FL_ACCESS)
1114 goto out;
993dfa87 1115 locks_copy_lock(new_fl, request);
5946c431 1116 locks_move_blocks(new_fl, request);
e084c1bd 1117 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
993dfa87 1118 new_fl = NULL;
9cedc194 1119 error = 0;
1da177e4
LT
1120
1121out:
6109c850 1122 spin_unlock(&ctx->flc_lock);
02e525b2 1123 percpu_up_read(&file_rwsem);
993dfa87
TM
1124 if (new_fl)
1125 locks_free_lock(new_fl);
ed9814d8 1126 locks_dispose_list(&dispose);
c883da31 1127 trace_flock_lock_inode(inode, request, error);
1da177e4
LT
1128 return error;
1129}
1130
b4d629a3
JL
1131static int posix_lock_inode(struct inode *inode, struct file_lock *request,
1132 struct file_lock *conflock)
1da177e4 1133{
bd61e0a9 1134 struct file_lock *fl, *tmp;
39005d02
MS
1135 struct file_lock *new_fl = NULL;
1136 struct file_lock *new_fl2 = NULL;
1da177e4
LT
1137 struct file_lock *left = NULL;
1138 struct file_lock *right = NULL;
bd61e0a9 1139 struct file_lock_context *ctx;
b9746ef8
JL
1140 int error;
1141 bool added = false;
ed9814d8 1142 LIST_HEAD(dispose);
1da177e4 1143
5c1c669a 1144 ctx = locks_get_lock_context(inode, request->fl_type);
bd61e0a9 1145 if (!ctx)
5c1c669a 1146 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
bd61e0a9 1147
1da177e4
LT
1148 /*
1149 * We may need two file_lock structures for this operation,
1150 * so we get them in advance to avoid races.
39005d02
MS
1151 *
1152 * In some cases we can be sure, that no new locks will be needed
1da177e4 1153 */
39005d02
MS
1154 if (!(request->fl_flags & FL_ACCESS) &&
1155 (request->fl_type != F_UNLCK ||
1156 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
1157 new_fl = locks_alloc_lock();
1158 new_fl2 = locks_alloc_lock();
1159 }
1da177e4 1160
02e525b2 1161 percpu_down_read(&file_rwsem);
6109c850 1162 spin_lock(&ctx->flc_lock);
1cb36012
JL
1163 /*
1164 * New lock request. Walk all POSIX locks and look for conflicts. If
1165 * there are any, either return error or put the request on the
48f74186 1166 * blocker's list of waiters and the global blocked_hash.
1cb36012 1167 */
1da177e4 1168 if (request->fl_type != F_UNLCK) {
bd61e0a9 1169 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1da177e4
LT
1170 if (!posix_locks_conflict(request, fl))
1171 continue;
5842add2 1172 if (conflock)
3fe0fff1 1173 locks_copy_conflock(conflock, fl);
1da177e4
LT
1174 error = -EAGAIN;
1175 if (!(request->fl_flags & FL_SLEEP))
1176 goto out;
1c8c601a
JL
1177 /*
1178 * Deadlock detection and insertion into the blocked
1179 * locks list must be done while holding the same lock!
1180 */
1da177e4 1181 error = -EDEADLK;
7b2296af 1182 spin_lock(&blocked_lock_lock);
945ab8f6
JL
1183 /*
1184 * Ensure that we don't find any locks blocked on this
1185 * request during deadlock detection.
1186 */
1187 __locks_wake_up_blocks(request);
1c8c601a
JL
1188 if (likely(!posix_locks_deadlock(request, fl))) {
1189 error = FILE_LOCK_DEFERRED;
fd7732e0
N
1190 __locks_insert_block(fl, request,
1191 posix_locks_conflict);
1c8c601a 1192 }
7b2296af 1193 spin_unlock(&blocked_lock_lock);
1da177e4 1194 goto out;
7bbd1fc0
N
1195 }
1196 }
1da177e4
LT
1197
1198 /* If we're just looking for a conflict, we're done. */
1199 error = 0;
1200 if (request->fl_flags & FL_ACCESS)
1201 goto out;
1202
bd61e0a9
JL
1203 /* Find the first old lock with the same owner as the new lock */
1204 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1205 if (posix_same_owner(request, fl))
1206 break;
1da177e4
LT
1207 }
1208
1cb36012 1209 /* Process locks with this owner. */
bd61e0a9
JL
1210 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1211 if (!posix_same_owner(request, fl))
1212 break;
1213
1214 /* Detect adjacent or overlapping regions (if same lock type) */
1da177e4 1215 if (request->fl_type == fl->fl_type) {
449231d6
OK
1216 /* In all comparisons of start vs end, use
1217 * "start - 1" rather than "end + 1". If end
1218 * is OFFSET_MAX, end + 1 will become negative.
1219 */
1da177e4 1220 if (fl->fl_end < request->fl_start - 1)
bd61e0a9 1221 continue;
1da177e4
LT
1222 /* If the next lock in the list has entirely bigger
1223 * addresses than the new one, insert the lock here.
1224 */
449231d6 1225 if (fl->fl_start - 1 > request->fl_end)
1da177e4
LT
1226 break;
1227
1228 /* If we come here, the new and old lock are of the
1229 * same type and adjacent or overlapping. Make one
1230 * lock yielding from the lower start address of both
1231 * locks to the higher end address.
1232 */
1233 if (fl->fl_start > request->fl_start)
1234 fl->fl_start = request->fl_start;
1235 else
1236 request->fl_start = fl->fl_start;
1237 if (fl->fl_end < request->fl_end)
1238 fl->fl_end = request->fl_end;
1239 else
1240 request->fl_end = fl->fl_end;
1241 if (added) {
e084c1bd 1242 locks_delete_lock_ctx(fl, &dispose);
1da177e4
LT
1243 continue;
1244 }
1245 request = fl;
b9746ef8 1246 added = true;
bd61e0a9 1247 } else {
1da177e4
LT
1248 /* Processing for different lock types is a bit
1249 * more complex.
1250 */
1251 if (fl->fl_end < request->fl_start)
bd61e0a9 1252 continue;
1da177e4
LT
1253 if (fl->fl_start > request->fl_end)
1254 break;
1255 if (request->fl_type == F_UNLCK)
b9746ef8 1256 added = true;
1da177e4
LT
1257 if (fl->fl_start < request->fl_start)
1258 left = fl;
1259 /* If the next lock in the list has a higher end
1260 * address than the new one, insert the new one here.
1261 */
1262 if (fl->fl_end > request->fl_end) {
1263 right = fl;
1264 break;
1265 }
1266 if (fl->fl_start >= request->fl_start) {
1267 /* The new lock completely replaces an old
1268 * one (This may happen several times).
1269 */
1270 if (added) {
e084c1bd 1271 locks_delete_lock_ctx(fl, &dispose);
1da177e4
LT
1272 continue;
1273 }
b84d49f9
JL
1274 /*
1275 * Replace the old lock with new_fl, and
1276 * remove the old one. It's safe to do the
1277 * insert here since we know that we won't be
1278 * using new_fl later, and that the lock is
1279 * just replacing an existing lock.
1da177e4 1280 */
b84d49f9
JL
1281 error = -ENOLCK;
1282 if (!new_fl)
1283 goto out;
1284 locks_copy_lock(new_fl, request);
5ef15968 1285 locks_move_blocks(new_fl, request);
b84d49f9
JL
1286 request = new_fl;
1287 new_fl = NULL;
e084c1bd
JL
1288 locks_insert_lock_ctx(request, &fl->fl_list);
1289 locks_delete_lock_ctx(fl, &dispose);
b9746ef8 1290 added = true;
1da177e4
LT
1291 }
1292 }
1da177e4
LT
1293 }
1294
0d9a490a 1295 /*
1cb36012
JL
1296 * The above code only modifies existing locks in case of merging or
1297 * replacing. If new lock(s) need to be inserted all modifications are
1298 * done below this, so it's safe yet to bail out.
0d9a490a
MS
1299 */
1300 error = -ENOLCK; /* "no luck" */
1301 if (right && left == right && !new_fl2)
1302 goto out;
1303
1da177e4
LT
1304 error = 0;
1305 if (!added) {
f475ae95
TM
1306 if (request->fl_type == F_UNLCK) {
1307 if (request->fl_flags & FL_EXISTS)
1308 error = -ENOENT;
1da177e4 1309 goto out;
f475ae95 1310 }
0d9a490a
MS
1311
1312 if (!new_fl) {
1313 error = -ENOLCK;
1314 goto out;
1315 }
1da177e4 1316 locks_copy_lock(new_fl, request);
5946c431 1317 locks_move_blocks(new_fl, request);
e084c1bd 1318 locks_insert_lock_ctx(new_fl, &fl->fl_list);
2e2f756f 1319 fl = new_fl;
1da177e4
LT
1320 new_fl = NULL;
1321 }
1322 if (right) {
1323 if (left == right) {
1324 /* The new lock breaks the old one in two pieces,
1325 * so we have to use the second new lock.
1326 */
1327 left = new_fl2;
1328 new_fl2 = NULL;
1329 locks_copy_lock(left, right);
e084c1bd 1330 locks_insert_lock_ctx(left, &fl->fl_list);
1da177e4
LT
1331 }
1332 right->fl_start = request->fl_end + 1;
1333 locks_wake_up_blocks(right);
1334 }
1335 if (left) {
1336 left->fl_end = request->fl_start - 1;
1337 locks_wake_up_blocks(left);
1338 }
1339 out:
6109c850 1340 spin_unlock(&ctx->flc_lock);
02e525b2 1341 percpu_up_read(&file_rwsem);
1da177e4
LT
1342 /*
1343 * Free any unused locks.
1344 */
1345 if (new_fl)
1346 locks_free_lock(new_fl);
1347 if (new_fl2)
1348 locks_free_lock(new_fl2);
ed9814d8 1349 locks_dispose_list(&dispose);
1890910f
JL
1350 trace_posix_lock_inode(inode, request, error);
1351
1da177e4
LT
1352 return error;
1353}
1354
1355/**
1356 * posix_lock_file - Apply a POSIX-style lock to a file
1357 * @filp: The file to apply the lock to
1358 * @fl: The lock to be applied
150b3934 1359 * @conflock: Place to return a copy of the conflicting lock, if found.
1da177e4
LT
1360 *
1361 * Add a POSIX style lock to a file.
1362 * We merge adjacent & overlapping locks whenever possible.
1363 * POSIX locks are sorted by owner task, then by starting address
f475ae95
TM
1364 *
1365 * Note that if called with an FL_EXISTS argument, the caller may determine
1366 * whether or not a lock was successfully freed by testing the return
1367 * value for -ENOENT.
1da177e4 1368 */
150b3934 1369int posix_lock_file(struct file *filp, struct file_lock *fl,
5842add2
AA
1370 struct file_lock *conflock)
1371{
c568d683 1372 return posix_lock_inode(locks_inode(filp), fl, conflock);
1da177e4 1373}
150b3934 1374EXPORT_SYMBOL(posix_lock_file);
1da177e4
LT
1375
1376/**
29d01b22
JL
1377 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1378 * @inode: inode of file to which lock request should be applied
1da177e4
LT
1379 * @fl: The lock to be applied
1380 *
616fb38f 1381 * Apply a POSIX style lock request to an inode.
1da177e4 1382 */
616fb38f 1383static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1da177e4
LT
1384{
1385 int error;
1386 might_sleep ();
1387 for (;;) {
b4d629a3 1388 error = posix_lock_inode(inode, fl, NULL);
bde74e4b 1389 if (error != FILE_LOCK_DEFERRED)
1da177e4 1390 break;
dcf23ac3
LT
1391 error = wait_event_interruptible(fl->fl_wait,
1392 list_empty(&fl->fl_blocked_member));
16306a61
N
1393 if (error)
1394 break;
1da177e4 1395 }
16306a61 1396 locks_delete_block(fl);
1da177e4
LT
1397 return error;
1398}
29d01b22 1399
778fc546
BF
1400static void lease_clear_pending(struct file_lock *fl, int arg)
1401{
1402 switch (arg) {
1403 case F_UNLCK:
1404 fl->fl_flags &= ~FL_UNLOCK_PENDING;
df561f66 1405 fallthrough;
778fc546
BF
1406 case F_RDLCK:
1407 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1408 }
1409}
1410
1da177e4 1411/* We already had a lease on this file; just change its type */
7448cc37 1412int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1da177e4 1413{
1da177e4
LT
1414 int error = assign_type(fl, arg);
1415
1416 if (error)
1417 return error;
778fc546 1418 lease_clear_pending(fl, arg);
1da177e4 1419 locks_wake_up_blocks(fl);
3b6e2723
FB
1420 if (arg == F_UNLCK) {
1421 struct file *filp = fl->fl_file;
1422
1423 f_delown(filp);
1424 filp->f_owner.signum = 0;
96d6d59c
BF
1425 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1426 if (fl->fl_fasync != NULL) {
1427 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1428 fl->fl_fasync = NULL;
1429 }
e084c1bd 1430 locks_delete_lock_ctx(fl, dispose);
3b6e2723 1431 }
1da177e4
LT
1432 return 0;
1433}
1da177e4
LT
1434EXPORT_SYMBOL(lease_modify);
1435
778fc546
BF
1436static bool past_time(unsigned long then)
1437{
1438 if (!then)
1439 /* 0 is a special value meaning "this never expires": */
1440 return false;
1441 return time_after(jiffies, then);
1442}
1443
c45198ed 1444static void time_out_leases(struct inode *inode, struct list_head *dispose)
1da177e4 1445{
8634b51f
JL
1446 struct file_lock_context *ctx = inode->i_flctx;
1447 struct file_lock *fl, *tmp;
1da177e4 1448
6109c850 1449 lockdep_assert_held(&ctx->flc_lock);
f82b4b67 1450
8634b51f 1451 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
62af4f1f 1452 trace_time_out_leases(inode, fl);
778fc546 1453 if (past_time(fl->fl_downgrade_time))
7448cc37 1454 lease_modify(fl, F_RDLCK, dispose);
778fc546 1455 if (past_time(fl->fl_break_time))
7448cc37 1456 lease_modify(fl, F_UNLCK, dispose);
1da177e4
LT
1457 }
1458}
1459
df4e8d2c
BF
1460static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1461{
d51f527f
IW
1462 bool rc;
1463
28df3d15
BF
1464 if (lease->fl_lmops->lm_breaker_owns_lease
1465 && lease->fl_lmops->lm_breaker_owns_lease(lease))
1466 return false;
d51f527f
IW
1467 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT)) {
1468 rc = false;
1469 goto trace;
1470 }
1471 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) {
1472 rc = false;
1473 goto trace;
1474 }
1475
1476 rc = locks_conflict(breaker, lease);
1477trace:
1478 trace_leases_conflict(rc, lease, breaker);
1479 return rc;
df4e8d2c
BF
1480}
1481
03d12ddf
JL
1482static bool
1483any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1484{
8634b51f 1485 struct file_lock_context *ctx = inode->i_flctx;
03d12ddf
JL
1486 struct file_lock *fl;
1487
6109c850 1488 lockdep_assert_held(&ctx->flc_lock);
03d12ddf 1489
8634b51f 1490 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
03d12ddf
JL
1491 if (leases_conflict(fl, breaker))
1492 return true;
1493 }
1494 return false;
1495}
1496
1da177e4
LT
1497/**
1498 * __break_lease - revoke all outstanding leases on file
1499 * @inode: the inode of the file to return
df4e8d2c
BF
1500 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1501 * break all leases
1502 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1503 * only delegations
1da177e4 1504 *
87250dd2 1505 * break_lease (inlined for speed) has checked there already is at least
1506 * some kind of lock (maybe a lease) on this file. Leases are broken on
1507 * a call to open() or truncate(). This function can sleep unless you
1da177e4
LT
1508 * specified %O_NONBLOCK to your open().
1509 */
df4e8d2c 1510int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1da177e4 1511{
778fc546 1512 int error = 0;
128a3785 1513 struct file_lock_context *ctx;
a901125c 1514 struct file_lock *new_fl, *fl, *tmp;
1da177e4 1515 unsigned long break_time;
8737c930 1516 int want_write = (mode & O_ACCMODE) != O_RDONLY;
c45198ed 1517 LIST_HEAD(dispose);
1da177e4 1518
8737c930 1519 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
6d4b9e38
LT
1520 if (IS_ERR(new_fl))
1521 return PTR_ERR(new_fl);
df4e8d2c 1522 new_fl->fl_flags = type;
1da177e4 1523
8634b51f 1524 /* typically we will check that ctx is non-NULL before calling */
128a3785 1525 ctx = smp_load_acquire(&inode->i_flctx);
8634b51f
JL
1526 if (!ctx) {
1527 WARN_ON_ONCE(1);
cfddf9f4 1528 goto free_lock;
8634b51f
JL
1529 }
1530
02e525b2 1531 percpu_down_read(&file_rwsem);
6109c850 1532 spin_lock(&ctx->flc_lock);
1da177e4 1533
c45198ed 1534 time_out_leases(inode, &dispose);
1da177e4 1535
03d12ddf 1536 if (!any_leases_conflict(inode, new_fl))
778fc546
BF
1537 goto out;
1538
1da177e4
LT
1539 break_time = 0;
1540 if (lease_break_time > 0) {
1541 break_time = jiffies + lease_break_time * HZ;
1542 if (break_time == 0)
1543 break_time++; /* so that 0 means no break time */
1544 }
1545
a901125c 1546 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
df4e8d2c
BF
1547 if (!leases_conflict(fl, new_fl))
1548 continue;
778fc546
BF
1549 if (want_write) {
1550 if (fl->fl_flags & FL_UNLOCK_PENDING)
1551 continue;
1552 fl->fl_flags |= FL_UNLOCK_PENDING;
1da177e4 1553 fl->fl_break_time = break_time;
778fc546 1554 } else {
8634b51f 1555 if (lease_breaking(fl))
778fc546
BF
1556 continue;
1557 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1558 fl->fl_downgrade_time = break_time;
1da177e4 1559 }
4d01b7f5 1560 if (fl->fl_lmops->lm_break(fl))
e084c1bd 1561 locks_delete_lock_ctx(fl, &dispose);
1da177e4
LT
1562 }
1563
8634b51f 1564 if (list_empty(&ctx->flc_lease))
4d01b7f5
JL
1565 goto out;
1566
843c6b2f 1567 if (mode & O_NONBLOCK) {
62af4f1f 1568 trace_break_lease_noblock(inode, new_fl);
1da177e4
LT
1569 error = -EWOULDBLOCK;
1570 goto out;
1571 }
1572
1573restart:
8634b51f
JL
1574 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1575 break_time = fl->fl_break_time;
f1c6bb2c 1576 if (break_time != 0)
1da177e4 1577 break_time -= jiffies;
f1c6bb2c
JL
1578 if (break_time == 0)
1579 break_time++;
fd7732e0 1580 locks_insert_block(fl, new_fl, leases_conflict);
62af4f1f 1581 trace_break_lease_block(inode, new_fl);
6109c850 1582 spin_unlock(&ctx->flc_lock);
02e525b2 1583 percpu_up_read(&file_rwsem);
aba37660 1584
c45198ed 1585 locks_dispose_list(&dispose);
4321e01e 1586 error = wait_event_interruptible_timeout(new_fl->fl_wait,
dcf23ac3
LT
1587 list_empty(&new_fl->fl_blocked_member),
1588 break_time);
aba37660 1589
02e525b2 1590 percpu_down_read(&file_rwsem);
6109c850 1591 spin_lock(&ctx->flc_lock);
62af4f1f 1592 trace_break_lease_unblock(inode, new_fl);
1c8c601a 1593 locks_delete_block(new_fl);
1da177e4 1594 if (error >= 0) {
778fc546
BF
1595 /*
1596 * Wait for the next conflicting lease that has not been
1597 * broken yet
1598 */
03d12ddf
JL
1599 if (error == 0)
1600 time_out_leases(inode, &dispose);
1601 if (any_leases_conflict(inode, new_fl))
1602 goto restart;
1da177e4
LT
1603 error = 0;
1604 }
1da177e4 1605out:
6109c850 1606 spin_unlock(&ctx->flc_lock);
02e525b2 1607 percpu_up_read(&file_rwsem);
c45198ed 1608 locks_dispose_list(&dispose);
cfddf9f4 1609free_lock:
6d4b9e38 1610 locks_free_lock(new_fl);
1da177e4
LT
1611 return error;
1612}
1da177e4
LT
1613EXPORT_SYMBOL(__break_lease);
1614
1615/**
76c47948 1616 * lease_get_mtime - update modified time of an inode with exclusive lease
1da177e4 1617 * @inode: the inode
76c47948 1618 * @time: pointer to a timespec which contains the last modified time
1da177e4
LT
1619 *
1620 * This is to force NFS clients to flush their caches for files with
1621 * exclusive leases. The justification is that if someone has an
a6b91919 1622 * exclusive lease, then they could be modifying it.
1da177e4 1623 */
95582b00 1624void lease_get_mtime(struct inode *inode, struct timespec64 *time)
1da177e4 1625{
bfe86024 1626 bool has_lease = false;
128a3785 1627 struct file_lock_context *ctx;
8634b51f 1628 struct file_lock *fl;
bfe86024 1629
128a3785 1630 ctx = smp_load_acquire(&inode->i_flctx);
8634b51f 1631 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
6109c850 1632 spin_lock(&ctx->flc_lock);
8ace5dfb
GT
1633 fl = list_first_entry_or_null(&ctx->flc_lease,
1634 struct file_lock, fl_list);
1635 if (fl && (fl->fl_type == F_WRLCK))
1636 has_lease = true;
6109c850 1637 spin_unlock(&ctx->flc_lock);
bfe86024
JL
1638 }
1639
1640 if (has_lease)
c2050a45 1641 *time = current_time(inode);
1da177e4 1642}
1da177e4
LT
1643EXPORT_SYMBOL(lease_get_mtime);
1644
1645/**
1646 * fcntl_getlease - Enquire what lease is currently active
1647 * @filp: the file
1648 *
1649 * The value returned by this function will be one of
1650 * (if no lease break is pending):
1651 *
1652 * %F_RDLCK to indicate a shared lease is held.
1653 *
1654 * %F_WRLCK to indicate an exclusive lease is held.
1655 *
1656 * %F_UNLCK to indicate no lease is held.
1657 *
1658 * (if a lease break is pending):
1659 *
1660 * %F_RDLCK to indicate an exclusive lease needs to be
1661 * changed to a shared lease (or removed).
1662 *
1663 * %F_UNLCK to indicate the lease needs to be removed.
1664 *
1665 * XXX: sfr & willy disagree over whether F_INPROGRESS
1666 * should be returned to userspace.
1667 */
1668int fcntl_getlease(struct file *filp)
1669{
1670 struct file_lock *fl;
c568d683 1671 struct inode *inode = locks_inode(filp);
128a3785 1672 struct file_lock_context *ctx;
1da177e4 1673 int type = F_UNLCK;
c45198ed 1674 LIST_HEAD(dispose);
1da177e4 1675
128a3785 1676 ctx = smp_load_acquire(&inode->i_flctx);
8634b51f 1677 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
02e525b2 1678 percpu_down_read(&file_rwsem);
6109c850 1679 spin_lock(&ctx->flc_lock);
c568d683 1680 time_out_leases(inode, &dispose);
8634b51f
JL
1681 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1682 if (fl->fl_file != filp)
1683 continue;
778fc546 1684 type = target_leasetype(fl);
1da177e4
LT
1685 break;
1686 }
6109c850 1687 spin_unlock(&ctx->flc_lock);
02e525b2 1688 percpu_up_read(&file_rwsem);
5f43086b 1689
8634b51f 1690 locks_dispose_list(&dispose);
1da177e4 1691 }
1da177e4
LT
1692 return type;
1693}
1694
24cbe784 1695/**
387e3746 1696 * check_conflicting_open - see if the given file points to an inode that has
7bbd1fc0
N
1697 * an existing open that would conflict with the
1698 * desired lease.
387e3746 1699 * @filp: file to check
24cbe784 1700 * @arg: type of lease that we're trying to acquire
7fadc59c 1701 * @flags: current lock flags
24cbe784
JL
1702 *
1703 * Check to see if there's an existing open fd on this file that would
1704 * conflict with the lease we're trying to set.
1705 */
1706static int
387e3746 1707check_conflicting_open(struct file *filp, const long arg, int flags)
24cbe784 1708{
387e3746
AG
1709 struct inode *inode = locks_inode(filp);
1710 int self_wcount = 0, self_rcount = 0;
24cbe784 1711
11afe9f7
CH
1712 if (flags & FL_LAYOUT)
1713 return 0;
aba2072f
BF
1714 if (flags & FL_DELEG)
1715 /* We leave these checks to the caller */
1716 return 0;
11afe9f7 1717
387e3746
AG
1718 if (arg == F_RDLCK)
1719 return inode_is_open_for_write(inode) ? -EAGAIN : 0;
1720 else if (arg != F_WRLCK)
1721 return 0;
1722
1723 /*
1724 * Make sure that only read/write count is from lease requestor.
1725 * Note that this will result in denying write leases when i_writecount
1726 * is negative, which is what we want. (We shouldn't grant write leases
1727 * on files open for execution.)
1728 */
1729 if (filp->f_mode & FMODE_WRITE)
1730 self_wcount = 1;
1731 else if (filp->f_mode & FMODE_READ)
1732 self_rcount = 1;
24cbe784 1733
387e3746
AG
1734 if (atomic_read(&inode->i_writecount) != self_wcount ||
1735 atomic_read(&inode->i_readcount) != self_rcount)
1736 return -EAGAIN;
24cbe784 1737
387e3746 1738 return 0;
24cbe784
JL
1739}
1740
e6f5c789
JL
1741static int
1742generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1da177e4 1743{
8634b51f 1744 struct file_lock *fl, *my_fl = NULL, *lease;
387e3746 1745 struct inode *inode = locks_inode(filp);
8634b51f 1746 struct file_lock_context *ctx;
df4e8d2c 1747 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
c1f24ef4 1748 int error;
c45198ed 1749 LIST_HEAD(dispose);
1da177e4 1750
096657b6 1751 lease = *flp;
62af4f1f
JL
1752 trace_generic_add_lease(inode, lease);
1753
5c1c669a
JL
1754 /* Note that arg is never F_UNLCK here */
1755 ctx = locks_get_lock_context(inode, arg);
8634b51f
JL
1756 if (!ctx)
1757 return -ENOMEM;
1758
df4e8d2c
BF
1759 /*
1760 * In the delegation case we need mutual exclusion with
1761 * a number of operations that take the i_mutex. We trylock
1762 * because delegations are an optional optimization, and if
1763 * there's some chance of a conflict--we'd rather not
1764 * bother, maybe that's a sign this just isn't a good file to
1765 * hand out a delegation on.
1766 */
5955102c 1767 if (is_deleg && !inode_trylock(inode))
df4e8d2c
BF
1768 return -EAGAIN;
1769
1770 if (is_deleg && arg == F_WRLCK) {
1771 /* Write delegations are not currently supported: */
5955102c 1772 inode_unlock(inode);
df4e8d2c
BF
1773 WARN_ON_ONCE(1);
1774 return -EINVAL;
1775 }
096657b6 1776
02e525b2 1777 percpu_down_read(&file_rwsem);
6109c850 1778 spin_lock(&ctx->flc_lock);
c45198ed 1779 time_out_leases(inode, &dispose);
387e3746 1780 error = check_conflicting_open(filp, arg, lease->fl_flags);
24cbe784 1781 if (error)
096657b6 1782 goto out;
6d5e8b05 1783
1da177e4
LT
1784 /*
1785 * At this point, we know that if there is an exclusive
1786 * lease on this file, then we hold it on this filp
1787 * (otherwise our open of this file would have blocked).
1788 * And if we are trying to acquire an exclusive lease,
1789 * then the file is not open by anyone (including us)
1790 * except for this filp.
1791 */
c1f24ef4 1792 error = -EAGAIN;
8634b51f 1793 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
2ab99ee1
CH
1794 if (fl->fl_file == filp &&
1795 fl->fl_owner == lease->fl_owner) {
8634b51f 1796 my_fl = fl;
c1f24ef4
BF
1797 continue;
1798 }
8634b51f 1799
c1f24ef4
BF
1800 /*
1801 * No exclusive leases if someone else has a lease on
1802 * this file:
1803 */
1804 if (arg == F_WRLCK)
1805 goto out;
1806 /*
1807 * Modifying our existing lease is OK, but no getting a
1808 * new lease if someone else is opening for write:
1809 */
1810 if (fl->fl_flags & FL_UNLOCK_PENDING)
1811 goto out;
1da177e4
LT
1812 }
1813
8634b51f 1814 if (my_fl != NULL) {
0164bf02
JL
1815 lease = my_fl;
1816 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1c7dd2ff
JL
1817 if (error)
1818 goto out;
1819 goto out_setup;
1da177e4
LT
1820 }
1821
1da177e4
LT
1822 error = -EINVAL;
1823 if (!leases_enable)
1824 goto out;
1825
e084c1bd 1826 locks_insert_lock_ctx(lease, &ctx->flc_lease);
24cbe784
JL
1827 /*
1828 * The check in break_lease() is lockless. It's possible for another
1829 * open to race in after we did the earlier check for a conflicting
1830 * open but before the lease was inserted. Check again for a
1831 * conflicting open and cancel the lease if there is one.
1832 *
1833 * We also add a barrier here to ensure that the insertion of the lock
1834 * precedes these checks.
1835 */
1836 smp_mb();
387e3746 1837 error = check_conflicting_open(filp, arg, lease->fl_flags);
8634b51f 1838 if (error) {
e084c1bd 1839 locks_unlink_lock_ctx(lease);
8634b51f
JL
1840 goto out;
1841 }
1c7dd2ff
JL
1842
1843out_setup:
1844 if (lease->fl_lmops->lm_setup)
1845 lease->fl_lmops->lm_setup(lease, priv);
1da177e4 1846out:
6109c850 1847 spin_unlock(&ctx->flc_lock);
02e525b2 1848 percpu_up_read(&file_rwsem);
c45198ed 1849 locks_dispose_list(&dispose);
df4e8d2c 1850 if (is_deleg)
5955102c 1851 inode_unlock(inode);
8634b51f 1852 if (!error && !my_fl)
1c7dd2ff 1853 *flp = NULL;
1da177e4
LT
1854 return error;
1855}
8335ebd9 1856
2ab99ee1 1857static int generic_delete_lease(struct file *filp, void *owner)
8335ebd9 1858{
0efaa7e8 1859 int error = -EAGAIN;
8634b51f 1860 struct file_lock *fl, *victim = NULL;
c568d683 1861 struct inode *inode = locks_inode(filp);
128a3785 1862 struct file_lock_context *ctx;
c45198ed 1863 LIST_HEAD(dispose);
8335ebd9 1864
128a3785 1865 ctx = smp_load_acquire(&inode->i_flctx);
8634b51f
JL
1866 if (!ctx) {
1867 trace_generic_delete_lease(inode, NULL);
1868 return error;
1869 }
1870
02e525b2 1871 percpu_down_read(&file_rwsem);
6109c850 1872 spin_lock(&ctx->flc_lock);
8634b51f 1873 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
2ab99ee1
CH
1874 if (fl->fl_file == filp &&
1875 fl->fl_owner == owner) {
8634b51f 1876 victim = fl;
0efaa7e8 1877 break;
8634b51f 1878 }
8335ebd9 1879 }
a9b1b455 1880 trace_generic_delete_lease(inode, victim);
8634b51f 1881 if (victim)
7448cc37 1882 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
6109c850 1883 spin_unlock(&ctx->flc_lock);
02e525b2 1884 percpu_up_read(&file_rwsem);
c45198ed 1885 locks_dispose_list(&dispose);
0efaa7e8 1886 return error;
8335ebd9
BF
1887}
1888
1889/**
1890 * generic_setlease - sets a lease on an open file
1c7dd2ff
JL
1891 * @filp: file pointer
1892 * @arg: type of lease to obtain
1893 * @flp: input - file_lock to use, output - file_lock inserted
1894 * @priv: private data for lm_setup (may be NULL if lm_setup
1895 * doesn't require it)
8335ebd9
BF
1896 *
1897 * The (input) flp->fl_lmops->lm_break function is required
1898 * by break_lease().
8335ebd9 1899 */
e6f5c789
JL
1900int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1901 void **priv)
8335ebd9 1902{
c568d683 1903 struct inode *inode = locks_inode(filp);
8335ebd9
BF
1904 int error;
1905
8e96e3b7 1906 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
8335ebd9
BF
1907 return -EACCES;
1908 if (!S_ISREG(inode->i_mode))
1909 return -EINVAL;
1910 error = security_file_lock(filp, arg);
1911 if (error)
1912 return error;
1913
8335ebd9
BF
1914 switch (arg) {
1915 case F_UNLCK:
2ab99ee1 1916 return generic_delete_lease(filp, *priv);
8335ebd9
BF
1917 case F_RDLCK:
1918 case F_WRLCK:
0efaa7e8
JL
1919 if (!(*flp)->fl_lmops->lm_break) {
1920 WARN_ON_ONCE(1);
1921 return -ENOLCK;
1922 }
11afe9f7 1923
e6f5c789 1924 return generic_add_lease(filp, arg, flp, priv);
8335ebd9 1925 default:
8d657eb3 1926 return -EINVAL;
8335ebd9
BF
1927 }
1928}
0af1a450 1929EXPORT_SYMBOL(generic_setlease);
1da177e4 1930
18f6622e
JL
1931#if IS_ENABLED(CONFIG_SRCU)
1932/*
1933 * Kernel subsystems can register to be notified on any attempt to set
1934 * a new lease with the lease_notifier_chain. This is used by (e.g.) nfsd
1935 * to close files that it may have cached when there is an attempt to set a
1936 * conflicting lease.
1937 */
1938static struct srcu_notifier_head lease_notifier_chain;
1939
1940static inline void
1941lease_notifier_chain_init(void)
1942{
1943 srcu_init_notifier_head(&lease_notifier_chain);
1944}
1945
1946static inline void
1947setlease_notifier(long arg, struct file_lock *lease)
1948{
1949 if (arg != F_UNLCK)
1950 srcu_notifier_call_chain(&lease_notifier_chain, arg, lease);
1951}
1952
1953int lease_register_notifier(struct notifier_block *nb)
1954{
1955 return srcu_notifier_chain_register(&lease_notifier_chain, nb);
1956}
1957EXPORT_SYMBOL_GPL(lease_register_notifier);
1958
1959void lease_unregister_notifier(struct notifier_block *nb)
1960{
1961 srcu_notifier_chain_unregister(&lease_notifier_chain, nb);
1962}
1963EXPORT_SYMBOL_GPL(lease_unregister_notifier);
1964
1965#else /* !IS_ENABLED(CONFIG_SRCU) */
1966static inline void
1967lease_notifier_chain_init(void)
1968{
1969}
1970
1971static inline void
1972setlease_notifier(long arg, struct file_lock *lease)
1973{
1974}
1975
1976int lease_register_notifier(struct notifier_block *nb)
1977{
1978 return 0;
1979}
1980EXPORT_SYMBOL_GPL(lease_register_notifier);
1981
1982void lease_unregister_notifier(struct notifier_block *nb)
1983{
1984}
1985EXPORT_SYMBOL_GPL(lease_unregister_notifier);
1986
1987#endif /* IS_ENABLED(CONFIG_SRCU) */
1988
b89f4321 1989/**
e51673aa 1990 * vfs_setlease - sets a lease on an open file
1c7dd2ff
JL
1991 * @filp: file pointer
1992 * @arg: type of lease to obtain
1993 * @lease: file_lock to use when adding a lease
1994 * @priv: private info for lm_setup when adding a lease (may be
7bbd1fc0 1995 * NULL if lm_setup doesn't require it)
e51673aa
JL
1996 *
1997 * Call this to establish a lease on the file. The "lease" argument is not
1998 * used for F_UNLCK requests and may be NULL. For commands that set or alter
80b79dd0
MCC
1999 * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be
2000 * set; if not, this function will return -ENOLCK (and generate a scary-looking
e51673aa 2001 * stack trace).
1c7dd2ff
JL
2002 *
2003 * The "priv" pointer is passed directly to the lm_setup function as-is. It
2004 * may be NULL if the lm_setup operation doesn't require it.
1da177e4 2005 */
e6f5c789
JL
2006int
2007vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1da177e4 2008{
18f6622e
JL
2009 if (lease)
2010 setlease_notifier(arg, *lease);
de2a4a50 2011 if (filp->f_op->setlease)
f82b4b67 2012 return filp->f_op->setlease(filp, arg, lease, priv);
1c7dd2ff 2013 else
f82b4b67 2014 return generic_setlease(filp, arg, lease, priv);
1da177e4 2015}
a9933cea 2016EXPORT_SYMBOL_GPL(vfs_setlease);
1da177e4 2017
0ceaf6c7 2018static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1da177e4 2019{
1c7dd2ff 2020 struct file_lock *fl;
f7347ce4 2021 struct fasync_struct *new;
1da177e4
LT
2022 int error;
2023
c5b1f0d9
AB
2024 fl = lease_alloc(filp, arg);
2025 if (IS_ERR(fl))
2026 return PTR_ERR(fl);
1da177e4 2027
f7347ce4
LT
2028 new = fasync_alloc();
2029 if (!new) {
2030 locks_free_lock(fl);
2031 return -ENOMEM;
2032 }
1c7dd2ff 2033 new->fa_fd = fd;
f7347ce4 2034
1c7dd2ff 2035 error = vfs_setlease(filp, arg, &fl, (void **)&new);
2dfb928f
JL
2036 if (fl)
2037 locks_free_lock(fl);
f7347ce4
LT
2038 if (new)
2039 fasync_free(new);
1da177e4
LT
2040 return error;
2041}
2042
0ceaf6c7
BF
2043/**
2044 * fcntl_setlease - sets a lease on an open file
2045 * @fd: open file descriptor
2046 * @filp: file pointer
2047 * @arg: type of lease to obtain
2048 *
2049 * Call this fcntl to establish a lease on the file.
2050 * Note that you also need to call %F_SETSIG to
2051 * receive a signal when the lease is broken.
2052 */
2053int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
2054{
2055 if (arg == F_UNLCK)
2ab99ee1 2056 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
0ceaf6c7
BF
2057 return do_fcntl_add_lease(fd, filp, arg);
2058}
2059
1da177e4 2060/**
29d01b22
JL
2061 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
2062 * @inode: inode of the file to apply to
1da177e4
LT
2063 * @fl: The lock to be applied
2064 *
29d01b22 2065 * Apply a FLOCK style lock request to an inode.
1da177e4 2066 */
616fb38f 2067static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1da177e4
LT
2068{
2069 int error;
2070 might_sleep();
2071 for (;;) {
29d01b22 2072 error = flock_lock_inode(inode, fl);
bde74e4b 2073 if (error != FILE_LOCK_DEFERRED)
1da177e4 2074 break;
dcf23ac3
LT
2075 error = wait_event_interruptible(fl->fl_wait,
2076 list_empty(&fl->fl_blocked_member));
16306a61
N
2077 if (error)
2078 break;
1da177e4 2079 }
16306a61 2080 locks_delete_block(fl);
1da177e4
LT
2081 return error;
2082}
2083
e55c34a6
BC
2084/**
2085 * locks_lock_inode_wait - Apply a lock to an inode
2086 * @inode: inode of the file to apply to
2087 * @fl: The lock to be applied
2088 *
2089 * Apply a POSIX or FLOCK style lock request to an inode.
2090 */
2091int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
2092{
2093 int res = 0;
2094 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
2095 case FL_POSIX:
2096 res = posix_lock_inode_wait(inode, fl);
2097 break;
2098 case FL_FLOCK:
2099 res = flock_lock_inode_wait(inode, fl);
2100 break;
2101 default:
2102 BUG();
2103 }
2104 return res;
2105}
2106EXPORT_SYMBOL(locks_lock_inode_wait);
2107
1da177e4
LT
2108/**
2109 * sys_flock: - flock() system call.
2110 * @fd: the file descriptor to lock.
2111 * @cmd: the type of lock to apply.
2112 *
2113 * Apply a %FL_FLOCK style lock to an open file descriptor.
80b79dd0 2114 * The @cmd can be one of:
1da177e4 2115 *
80b79dd0
MCC
2116 * - %LOCK_SH -- a shared lock.
2117 * - %LOCK_EX -- an exclusive lock.
2118 * - %LOCK_UN -- remove an existing lock.
2119 * - %LOCK_MAND -- a 'mandatory' flock.
2120 * This exists to emulate Windows Share Modes.
1da177e4
LT
2121 *
2122 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
2123 * processes read and write access respectively.
2124 */
002c8976 2125SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1da177e4 2126{
2903ff01 2127 struct fd f = fdget(fd);
1da177e4
LT
2128 struct file_lock *lock;
2129 int can_sleep, unlock;
2130 int error;
2131
2132 error = -EBADF;
2903ff01 2133 if (!f.file)
1da177e4
LT
2134 goto out;
2135
2136 can_sleep = !(cmd & LOCK_NB);
2137 cmd &= ~LOCK_NB;
2138 unlock = (cmd == LOCK_UN);
2139
aeb5d727 2140 if (!unlock && !(cmd & LOCK_MAND) &&
2903ff01 2141 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
1da177e4
LT
2142 goto out_putf;
2143
d6367d62 2144 lock = flock_make_lock(f.file, cmd, NULL);
6e129d00
JL
2145 if (IS_ERR(lock)) {
2146 error = PTR_ERR(lock);
1da177e4 2147 goto out_putf;
6e129d00
JL
2148 }
2149
1da177e4
LT
2150 if (can_sleep)
2151 lock->fl_flags |= FL_SLEEP;
2152
2903ff01 2153 error = security_file_lock(f.file, lock->fl_type);
1da177e4
LT
2154 if (error)
2155 goto out_free;
2156
de2a4a50 2157 if (f.file->f_op->flock)
2903ff01 2158 error = f.file->f_op->flock(f.file,
1da177e4
LT
2159 (can_sleep) ? F_SETLKW : F_SETLK,
2160 lock);
2161 else
4f656367 2162 error = locks_lock_file_wait(f.file, lock);
1da177e4
LT
2163
2164 out_free:
993dfa87 2165 locks_free_lock(lock);
1da177e4
LT
2166
2167 out_putf:
2903ff01 2168 fdput(f);
1da177e4
LT
2169 out:
2170 return error;
2171}
2172
3ee17abd
BF
2173/**
2174 * vfs_test_lock - test file byte range lock
2175 * @filp: The file to test lock for
6924c554 2176 * @fl: The lock to test; also used to hold result
3ee17abd
BF
2177 *
2178 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2179 * setting conf->fl_type to something other than F_UNLCK.
2180 */
2181int vfs_test_lock(struct file *filp, struct file_lock *fl)
2182{
de2a4a50 2183 if (filp->f_op->lock)
3ee17abd
BF
2184 return filp->f_op->lock(filp, F_GETLK, fl);
2185 posix_test_lock(filp, fl);
2186 return 0;
2187}
2188EXPORT_SYMBOL_GPL(vfs_test_lock);
2189
9d5b86ac
BC
2190/**
2191 * locks_translate_pid - translate a file_lock's fl_pid number into a namespace
2192 * @fl: The file_lock who's fl_pid should be translated
2193 * @ns: The namespace into which the pid should be translated
2194 *
2195 * Used to tranlate a fl_pid into a namespace virtual pid number
2196 */
2197static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns)
2198{
2199 pid_t vnr;
2200 struct pid *pid;
2201
2202 if (IS_OFDLCK(fl))
2203 return -1;
2204 if (IS_REMOTELCK(fl))
2205 return fl->fl_pid;
826d7bc9
KK
2206 /*
2207 * If the flock owner process is dead and its pid has been already
2208 * freed, the translation below won't work, but we still want to show
2209 * flock owner pid number in init pidns.
2210 */
2211 if (ns == &init_pid_ns)
2212 return (pid_t)fl->fl_pid;
9d5b86ac
BC
2213
2214 rcu_read_lock();
2215 pid = find_pid_ns(fl->fl_pid, &init_pid_ns);
2216 vnr = pid_nr_ns(pid, ns);
2217 rcu_read_unlock();
2218 return vnr;
2219}
2220
c2fa1b8a
BF
2221static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
2222{
9d5b86ac 2223 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
c2fa1b8a
BF
2224#if BITS_PER_LONG == 32
2225 /*
2226 * Make sure we can represent the posix lock via
2227 * legacy 32bit flock.
2228 */
2229 if (fl->fl_start > OFFT_OFFSET_MAX)
2230 return -EOVERFLOW;
2231 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2232 return -EOVERFLOW;
2233#endif
2234 flock->l_start = fl->fl_start;
2235 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2236 fl->fl_end - fl->fl_start + 1;
2237 flock->l_whence = 0;
129a84de 2238 flock->l_type = fl->fl_type;
c2fa1b8a
BF
2239 return 0;
2240}
2241
2242#if BITS_PER_LONG == 32
2243static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2244{
9d5b86ac 2245 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
c2fa1b8a
BF
2246 flock->l_start = fl->fl_start;
2247 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2248 fl->fl_end - fl->fl_start + 1;
2249 flock->l_whence = 0;
2250 flock->l_type = fl->fl_type;
2251}
2252#endif
2253
1da177e4
LT
2254/* Report the first existing lock that would conflict with l.
2255 * This implements the F_GETLK command of fcntl().
2256 */
a75d30c7 2257int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock)
1da177e4 2258{
52306e88 2259 struct file_lock *fl;
1da177e4
LT
2260 int error;
2261
52306e88
BC
2262 fl = locks_alloc_lock();
2263 if (fl == NULL)
2264 return -ENOMEM;
1da177e4 2265 error = -EINVAL;
a75d30c7 2266 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
1da177e4
LT
2267 goto out;
2268
52306e88 2269 error = flock_to_posix_lock(filp, fl, flock);
1da177e4
LT
2270 if (error)
2271 goto out;
2272
0d3f7a2d 2273 if (cmd == F_OFD_GETLK) {
90478939 2274 error = -EINVAL;
a75d30c7 2275 if (flock->l_pid != 0)
90478939
JL
2276 goto out;
2277
52306e88
BC
2278 fl->fl_flags |= FL_OFDLCK;
2279 fl->fl_owner = filp;
5d50ffd7
JL
2280 }
2281
52306e88 2282 error = vfs_test_lock(filp, fl);
3ee17abd
BF
2283 if (error)
2284 goto out;
7bbd1fc0 2285
52306e88
BC
2286 flock->l_type = fl->fl_type;
2287 if (fl->fl_type != F_UNLCK) {
2288 error = posix_lock_to_flock(flock, fl);
c2fa1b8a 2289 if (error)
52306e88 2290 goto out;
1da177e4 2291 }
1da177e4 2292out:
52306e88 2293 locks_free_lock(fl);
1da177e4
LT
2294 return error;
2295}
2296
7723ec97
ME
2297/**
2298 * vfs_lock_file - file byte range lock
2299 * @filp: The file to apply the lock to
2300 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2301 * @fl: The lock to be applied
150b3934
ME
2302 * @conf: Place to return a copy of the conflicting lock, if found.
2303 *
2304 * A caller that doesn't care about the conflicting lock may pass NULL
2305 * as the final argument.
2306 *
2307 * If the filesystem defines a private ->lock() method, then @conf will
2308 * be left unchanged; so a caller that cares should initialize it to
2309 * some acceptable default.
2beb6614
ME
2310 *
2311 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2312 * locks, the ->lock() interface may return asynchronously, before the lock has
2313 * been granted or denied by the underlying filesystem, if (and only if)
8fb47a4f 2314 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2beb6614
ME
2315 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2316 * the request is for a blocking lock. When ->lock() does return asynchronously,
8fb47a4f 2317 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2beb6614
ME
2318 * request completes.
2319 * If the request is for non-blocking lock the file system should return
bde74e4b
MS
2320 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2321 * with the result. If the request timed out the callback routine will return a
2beb6614
ME
2322 * nonzero return code and the file system should release the lock. The file
2323 * system is also responsible to keep a corresponding posix lock when it
2324 * grants a lock so the VFS can find out which locks are locally held and do
2325 * the correct lock cleanup when required.
2326 * The underlying filesystem must not drop the kernel lock or call
8fb47a4f 2327 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2beb6614 2328 * return code.
7723ec97 2329 */
150b3934 2330int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
7723ec97 2331{
de2a4a50 2332 if (filp->f_op->lock)
7723ec97
ME
2333 return filp->f_op->lock(filp, cmd, fl);
2334 else
150b3934 2335 return posix_lock_file(filp, fl, conf);
7723ec97
ME
2336}
2337EXPORT_SYMBOL_GPL(vfs_lock_file);
2338
b648a6de
MS
2339static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2340 struct file_lock *fl)
2341{
2342 int error;
2343
2344 error = security_file_lock(filp, fl->fl_type);
2345 if (error)
2346 return error;
2347
764c76b3
MS
2348 for (;;) {
2349 error = vfs_lock_file(filp, cmd, fl, NULL);
2350 if (error != FILE_LOCK_DEFERRED)
b648a6de 2351 break;
dcf23ac3
LT
2352 error = wait_event_interruptible(fl->fl_wait,
2353 list_empty(&fl->fl_blocked_member));
16306a61
N
2354 if (error)
2355 break;
b648a6de 2356 }
16306a61 2357 locks_delete_block(fl);
b648a6de
MS
2358
2359 return error;
2360}
2361
6ca7d910 2362/* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
cf01f4ee
JL
2363static int
2364check_fmode_for_setlk(struct file_lock *fl)
2365{
2366 switch (fl->fl_type) {
2367 case F_RDLCK:
2368 if (!(fl->fl_file->f_mode & FMODE_READ))
2369 return -EBADF;
2370 break;
2371 case F_WRLCK:
2372 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2373 return -EBADF;
2374 }
2375 return 0;
2376}
2377
1da177e4
LT
2378/* Apply the lock described by l to an open file descriptor.
2379 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2380 */
c293621b 2381int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
a75d30c7 2382 struct flock *flock)
1da177e4
LT
2383{
2384 struct file_lock *file_lock = locks_alloc_lock();
a75d30c7 2385 struct inode *inode = locks_inode(filp);
0b2bac2f 2386 struct file *f;
1da177e4
LT
2387 int error;
2388
2389 if (file_lock == NULL)
2390 return -ENOLCK;
2391
a75d30c7 2392 error = flock_to_posix_lock(filp, file_lock, flock);
1da177e4
LT
2393 if (error)
2394 goto out;
5d50ffd7 2395
cf01f4ee
JL
2396 error = check_fmode_for_setlk(file_lock);
2397 if (error)
2398 goto out;
2399
5d50ffd7
JL
2400 /*
2401 * If the cmd is requesting file-private locks, then set the
cff2fce5 2402 * FL_OFDLCK flag and override the owner.
5d50ffd7
JL
2403 */
2404 switch (cmd) {
0d3f7a2d 2405 case F_OFD_SETLK:
90478939 2406 error = -EINVAL;
a75d30c7 2407 if (flock->l_pid != 0)
90478939
JL
2408 goto out;
2409
5d50ffd7 2410 cmd = F_SETLK;
cff2fce5 2411 file_lock->fl_flags |= FL_OFDLCK;
73a8f5f7 2412 file_lock->fl_owner = filp;
5d50ffd7 2413 break;
0d3f7a2d 2414 case F_OFD_SETLKW:
90478939 2415 error = -EINVAL;
a75d30c7 2416 if (flock->l_pid != 0)
90478939
JL
2417 goto out;
2418
5d50ffd7 2419 cmd = F_SETLKW;
cff2fce5 2420 file_lock->fl_flags |= FL_OFDLCK;
73a8f5f7 2421 file_lock->fl_owner = filp;
df561f66 2422 fallthrough;
5d50ffd7 2423 case F_SETLKW:
1da177e4
LT
2424 file_lock->fl_flags |= FL_SLEEP;
2425 }
5d50ffd7 2426
b648a6de 2427 error = do_lock_file_wait(filp, cmd, file_lock);
1da177e4 2428
c293621b 2429 /*
0752ba80
JL
2430 * Attempt to detect a close/fcntl race and recover by releasing the
2431 * lock that was just acquired. There is no need to do that when we're
2432 * unlocking though, or for OFD locks.
c293621b 2433 */
0752ba80
JL
2434 if (!error && file_lock->fl_type != F_UNLCK &&
2435 !(file_lock->fl_flags & FL_OFDLCK)) {
120ce2b0 2436 struct files_struct *files = current->files;
7f3697e2
JL
2437 /*
2438 * We need that spin_lock here - it prevents reordering between
2439 * update of i_flctx->flc_posix and check for it done in
2440 * close(). rcu_read_lock() wouldn't do.
2441 */
120ce2b0
EB
2442 spin_lock(&files->file_lock);
2443 f = files_lookup_fd_locked(files, fd);
2444 spin_unlock(&files->file_lock);
7f3697e2
JL
2445 if (f != filp) {
2446 file_lock->fl_type = F_UNLCK;
2447 error = do_lock_file_wait(filp, cmd, file_lock);
2448 WARN_ON_ONCE(error);
2449 error = -EBADF;
2450 }
1da177e4 2451 }
c293621b 2452out:
1890910f 2453 trace_fcntl_setlk(inode, file_lock, error);
1da177e4
LT
2454 locks_free_lock(file_lock);
2455 return error;
2456}
2457
2458#if BITS_PER_LONG == 32
2459/* Report the first existing lock that would conflict with l.
2460 * This implements the F_GETLK command of fcntl().
2461 */
a75d30c7 2462int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock)
1da177e4 2463{
52306e88 2464 struct file_lock *fl;
1da177e4
LT
2465 int error;
2466
52306e88
BC
2467 fl = locks_alloc_lock();
2468 if (fl == NULL)
2469 return -ENOMEM;
2470
1da177e4 2471 error = -EINVAL;
a75d30c7 2472 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
1da177e4
LT
2473 goto out;
2474
52306e88 2475 error = flock64_to_posix_lock(filp, fl, flock);
1da177e4
LT
2476 if (error)
2477 goto out;
2478
0d3f7a2d 2479 if (cmd == F_OFD_GETLK) {
90478939 2480 error = -EINVAL;
a75d30c7 2481 if (flock->l_pid != 0)
90478939
JL
2482 goto out;
2483
5d50ffd7 2484 cmd = F_GETLK64;
52306e88
BC
2485 fl->fl_flags |= FL_OFDLCK;
2486 fl->fl_owner = filp;
5d50ffd7
JL
2487 }
2488
52306e88 2489 error = vfs_test_lock(filp, fl);
3ee17abd
BF
2490 if (error)
2491 goto out;
2492
52306e88
BC
2493 flock->l_type = fl->fl_type;
2494 if (fl->fl_type != F_UNLCK)
2495 posix_lock_to_flock64(flock, fl);
f328296e 2496
1da177e4 2497out:
52306e88 2498 locks_free_lock(fl);
1da177e4
LT
2499 return error;
2500}
2501
2502/* Apply the lock described by l to an open file descriptor.
2503 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2504 */
c293621b 2505int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
a75d30c7 2506 struct flock64 *flock)
1da177e4
LT
2507{
2508 struct file_lock *file_lock = locks_alloc_lock();
0b2bac2f 2509 struct file *f;
1da177e4
LT
2510 int error;
2511
2512 if (file_lock == NULL)
2513 return -ENOLCK;
2514
a75d30c7 2515 error = flock64_to_posix_lock(filp, file_lock, flock);
1da177e4
LT
2516 if (error)
2517 goto out;
5d50ffd7 2518
cf01f4ee
JL
2519 error = check_fmode_for_setlk(file_lock);
2520 if (error)
2521 goto out;
2522
5d50ffd7
JL
2523 /*
2524 * If the cmd is requesting file-private locks, then set the
cff2fce5 2525 * FL_OFDLCK flag and override the owner.
5d50ffd7
JL
2526 */
2527 switch (cmd) {
0d3f7a2d 2528 case F_OFD_SETLK:
90478939 2529 error = -EINVAL;
a75d30c7 2530 if (flock->l_pid != 0)
90478939
JL
2531 goto out;
2532
5d50ffd7 2533 cmd = F_SETLK64;
cff2fce5 2534 file_lock->fl_flags |= FL_OFDLCK;
73a8f5f7 2535 file_lock->fl_owner = filp;
5d50ffd7 2536 break;
0d3f7a2d 2537 case F_OFD_SETLKW:
90478939 2538 error = -EINVAL;
a75d30c7 2539 if (flock->l_pid != 0)
90478939
JL
2540 goto out;
2541
5d50ffd7 2542 cmd = F_SETLKW64;
cff2fce5 2543 file_lock->fl_flags |= FL_OFDLCK;
73a8f5f7 2544 file_lock->fl_owner = filp;
df561f66 2545 fallthrough;
5d50ffd7 2546 case F_SETLKW64:
1da177e4
LT
2547 file_lock->fl_flags |= FL_SLEEP;
2548 }
5d50ffd7 2549
b648a6de 2550 error = do_lock_file_wait(filp, cmd, file_lock);
1da177e4 2551
c293621b 2552 /*
0752ba80
JL
2553 * Attempt to detect a close/fcntl race and recover by releasing the
2554 * lock that was just acquired. There is no need to do that when we're
2555 * unlocking though, or for OFD locks.
c293621b 2556 */
0752ba80
JL
2557 if (!error && file_lock->fl_type != F_UNLCK &&
2558 !(file_lock->fl_flags & FL_OFDLCK)) {
120ce2b0 2559 struct files_struct *files = current->files;
7f3697e2
JL
2560 /*
2561 * We need that spin_lock here - it prevents reordering between
2562 * update of i_flctx->flc_posix and check for it done in
2563 * close(). rcu_read_lock() wouldn't do.
2564 */
120ce2b0
EB
2565 spin_lock(&files->file_lock);
2566 f = files_lookup_fd_locked(files, fd);
2567 spin_unlock(&files->file_lock);
7f3697e2
JL
2568 if (f != filp) {
2569 file_lock->fl_type = F_UNLCK;
2570 error = do_lock_file_wait(filp, cmd, file_lock);
2571 WARN_ON_ONCE(error);
2572 error = -EBADF;
2573 }
1da177e4 2574 }
1da177e4
LT
2575out:
2576 locks_free_lock(file_lock);
2577 return error;
2578}
2579#endif /* BITS_PER_LONG == 32 */
2580
2581/*
2582 * This function is called when the file is being removed
2583 * from the task's fd array. POSIX locks belonging to this task
2584 * are deleted at this time.
2585 */
2586void locks_remove_posix(struct file *filp, fl_owner_t owner)
2587{
1890910f 2588 int error;
c568d683 2589 struct inode *inode = locks_inode(filp);
ff7b86b8 2590 struct file_lock lock;
128a3785 2591 struct file_lock_context *ctx;
1da177e4
LT
2592
2593 /*
2594 * If there are no locks held on this file, we don't need to call
2595 * posix_lock_file(). Another process could be setting a lock on this
2596 * file at the same time, but we wouldn't remove that lock anyway.
2597 */
c568d683 2598 ctx = smp_load_acquire(&inode->i_flctx);
bd61e0a9 2599 if (!ctx || list_empty(&ctx->flc_posix))
1da177e4
LT
2600 return;
2601
d6367d62 2602 locks_init_lock(&lock);
1da177e4 2603 lock.fl_type = F_UNLCK;
75e1fcc0 2604 lock.fl_flags = FL_POSIX | FL_CLOSE;
1da177e4
LT
2605 lock.fl_start = 0;
2606 lock.fl_end = OFFSET_MAX;
2607 lock.fl_owner = owner;
2608 lock.fl_pid = current->tgid;
2609 lock.fl_file = filp;
2610 lock.fl_ops = NULL;
2611 lock.fl_lmops = NULL;
2612
1890910f 2613 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
1da177e4 2614
1da177e4
LT
2615 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2616 lock.fl_ops->fl_release_private(&lock);
c568d683 2617 trace_locks_remove_posix(inode, &lock, error);
1da177e4 2618}
1da177e4
LT
2619EXPORT_SYMBOL(locks_remove_posix);
2620
3d8e560d 2621/* The i_flctx must be valid when calling into here */
dd459bb1 2622static void
128a3785 2623locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
dd459bb1 2624{
d6367d62 2625 struct file_lock fl;
c568d683 2626 struct inode *inode = locks_inode(filp);
dd459bb1 2627
3d8e560d 2628 if (list_empty(&flctx->flc_flock))
dd459bb1
JL
2629 return;
2630
d6367d62
N
2631 flock_make_lock(filp, LOCK_UN, &fl);
2632 fl.fl_flags |= FL_CLOSE;
2633
de2a4a50 2634 if (filp->f_op->flock)
dd459bb1
JL
2635 filp->f_op->flock(filp, F_SETLKW, &fl);
2636 else
bcd7f78d 2637 flock_lock_inode(inode, &fl);
dd459bb1
JL
2638
2639 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2640 fl.fl_ops->fl_release_private(&fl);
2641}
2642
3d8e560d 2643/* The i_flctx must be valid when calling into here */
8634b51f 2644static void
128a3785 2645locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
8634b51f 2646{
8634b51f
JL
2647 struct file_lock *fl, *tmp;
2648 LIST_HEAD(dispose);
2649
3d8e560d 2650 if (list_empty(&ctx->flc_lease))
8634b51f
JL
2651 return;
2652
02e525b2 2653 percpu_down_read(&file_rwsem);
6109c850 2654 spin_lock(&ctx->flc_lock);
8634b51f 2655 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
c4e136cd
JL
2656 if (filp == fl->fl_file)
2657 lease_modify(fl, F_UNLCK, &dispose);
6109c850 2658 spin_unlock(&ctx->flc_lock);
02e525b2 2659 percpu_up_read(&file_rwsem);
5f43086b 2660
8634b51f
JL
2661 locks_dispose_list(&dispose);
2662}
2663
1da177e4
LT
2664/*
2665 * This function is called on the last close of an open file.
2666 */
78ed8a13 2667void locks_remove_file(struct file *filp)
1da177e4 2668{
128a3785
DV
2669 struct file_lock_context *ctx;
2670
c568d683 2671 ctx = smp_load_acquire(&locks_inode(filp)->i_flctx);
128a3785 2672 if (!ctx)
3d8e560d
JL
2673 return;
2674
dd459bb1 2675 /* remove any OFD locks */
73a8f5f7 2676 locks_remove_posix(filp, filp);
5d50ffd7 2677
dd459bb1 2678 /* remove flock locks */
128a3785 2679 locks_remove_flock(filp, ctx);
dd459bb1 2680
8634b51f 2681 /* remove any leases */
128a3785 2682 locks_remove_lease(filp, ctx);
3953704f
BC
2683
2684 spin_lock(&ctx->flc_lock);
2685 locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX");
2686 locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK");
2687 locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE");
2688 spin_unlock(&ctx->flc_lock);
1da177e4
LT
2689}
2690
9b9d2ab4
ME
2691/**
2692 * vfs_cancel_lock - file byte range unblock lock
2693 * @filp: The file to apply the unblock to
2694 * @fl: The lock to be unblocked
2695 *
2696 * Used by lock managers to cancel blocked requests
2697 */
2698int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2699{
de2a4a50 2700 if (filp->f_op->lock)
9b9d2ab4
ME
2701 return filp->f_op->lock(filp, F_CANCELLK, fl);
2702 return 0;
2703}
9b9d2ab4
ME
2704EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2705
7f8ada98 2706#ifdef CONFIG_PROC_FS
d8ba7a36 2707#include <linux/proc_fs.h>
7f8ada98
PE
2708#include <linux/seq_file.h>
2709
7012b02a
JL
2710struct locks_iterator {
2711 int li_cpu;
2712 loff_t li_pos;
2713};
2714
7f8ada98 2715static void lock_get_status(struct seq_file *f, struct file_lock *fl,
b8da9b10 2716 loff_t id, char *pfx, int repeat)
1da177e4
LT
2717{
2718 struct inode *inode = NULL;
ab1f1611 2719 unsigned int fl_pid;
9d78edea 2720 struct pid_namespace *proc_pidns = proc_pid_ns(file_inode(f->file)->i_sb);
ab1f1611 2721
9d5b86ac
BC
2722 fl_pid = locks_translate_pid(fl, proc_pidns);
2723 /*
1cf8e5de
KK
2724 * If lock owner is dead (and pid is freed) or not visible in current
2725 * pidns, zero is shown as a pid value. Check lock info from
2726 * init_pid_ns to get saved lock pid value.
9d5b86ac 2727 */
1da177e4
LT
2728
2729 if (fl->fl_file != NULL)
c568d683 2730 inode = locks_inode(fl->fl_file);
1da177e4 2731
b8da9b10
LL
2732 seq_printf(f, "%lld: ", id);
2733
2734 if (repeat)
2735 seq_printf(f, "%*s", repeat - 1 + (int)strlen(pfx), pfx);
2736
1da177e4 2737 if (IS_POSIX(fl)) {
c918d42a 2738 if (fl->fl_flags & FL_ACCESS)
5315c26a 2739 seq_puts(f, "ACCESS");
cff2fce5 2740 else if (IS_OFDLCK(fl))
5315c26a 2741 seq_puts(f, "OFDLCK");
c918d42a 2742 else
5315c26a 2743 seq_puts(f, "POSIX ");
c918d42a
JL
2744
2745 seq_printf(f, " %s ",
f7e33bdb 2746 (inode == NULL) ? "*NOINODE*" : "ADVISORY ");
1da177e4
LT
2747 } else if (IS_FLOCK(fl)) {
2748 if (fl->fl_type & LOCK_MAND) {
5315c26a 2749 seq_puts(f, "FLOCK MSNFS ");
1da177e4 2750 } else {
5315c26a 2751 seq_puts(f, "FLOCK ADVISORY ");
1da177e4
LT
2752 }
2753 } else if (IS_LEASE(fl)) {
8144f1f6
JL
2754 if (fl->fl_flags & FL_DELEG)
2755 seq_puts(f, "DELEG ");
2756 else
2757 seq_puts(f, "LEASE ");
2758
ab83fa4b 2759 if (lease_breaking(fl))
5315c26a 2760 seq_puts(f, "BREAKING ");
1da177e4 2761 else if (fl->fl_file)
5315c26a 2762 seq_puts(f, "ACTIVE ");
1da177e4 2763 else
5315c26a 2764 seq_puts(f, "BREAKER ");
1da177e4 2765 } else {
5315c26a 2766 seq_puts(f, "UNKNOWN UNKNOWN ");
1da177e4
LT
2767 }
2768 if (fl->fl_type & LOCK_MAND) {
7f8ada98 2769 seq_printf(f, "%s ",
1da177e4
LT
2770 (fl->fl_type & LOCK_READ)
2771 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2772 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2773 } else {
43e4cb94
PB
2774 int type = IS_LEASE(fl) ? target_leasetype(fl) : fl->fl_type;
2775
2776 seq_printf(f, "%s ", (type == F_WRLCK) ? "WRITE" :
2777 (type == F_RDLCK) ? "READ" : "UNLCK");
1da177e4
LT
2778 }
2779 if (inode) {
3648888e 2780 /* userspace relies on this representation of dev_t */
98ca480a 2781 seq_printf(f, "%d %02x:%02x:%lu ", fl_pid,
1da177e4
LT
2782 MAJOR(inode->i_sb->s_dev),
2783 MINOR(inode->i_sb->s_dev), inode->i_ino);
1da177e4 2784 } else {
ab1f1611 2785 seq_printf(f, "%d <none>:0 ", fl_pid);
1da177e4
LT
2786 }
2787 if (IS_POSIX(fl)) {
2788 if (fl->fl_end == OFFSET_MAX)
7f8ada98 2789 seq_printf(f, "%Ld EOF\n", fl->fl_start);
1da177e4 2790 else
7f8ada98 2791 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
1da177e4 2792 } else {
5315c26a 2793 seq_puts(f, "0 EOF\n");
1da177e4
LT
2794 }
2795}
2796
b8da9b10
LL
2797static struct file_lock *get_next_blocked_member(struct file_lock *node)
2798{
2799 struct file_lock *tmp;
2800
2801 /* NULL node or root node */
2802 if (node == NULL || node->fl_blocker == NULL)
2803 return NULL;
2804
2805 /* Next member in the linked list could be itself */
2806 tmp = list_next_entry(node, fl_blocked_member);
2807 if (list_entry_is_head(tmp, &node->fl_blocker->fl_blocked_requests, fl_blocked_member)
2808 || tmp == node) {
2809 return NULL;
2810 }
2811
2812 return tmp;
2813}
2814
7f8ada98 2815static int locks_show(struct seq_file *f, void *v)
1da177e4 2816{
7012b02a 2817 struct locks_iterator *iter = f->private;
b8da9b10 2818 struct file_lock *cur, *tmp;
9d78edea 2819 struct pid_namespace *proc_pidns = proc_pid_ns(file_inode(f->file)->i_sb);
b8da9b10 2820 int level = 0;
1da177e4 2821
b8da9b10 2822 cur = hlist_entry(v, struct file_lock, fl_link);
1da177e4 2823
b8da9b10 2824 if (locks_translate_pid(cur, proc_pidns) == 0)
d67fd44f
NB
2825 return 0;
2826
b8da9b10
LL
2827 /* View this crossed linked list as a binary tree, the first member of fl_blocked_requests
2828 * is the left child of current node, the next silibing in fl_blocked_member is the
2829 * right child, we can alse get the parent of current node from fl_blocker, so this
2830 * question becomes traversal of a binary tree
2831 */
2832 while (cur != NULL) {
2833 if (level)
2834 lock_get_status(f, cur, iter->li_pos, "-> ", level);
2835 else
2836 lock_get_status(f, cur, iter->li_pos, "", level);
1da177e4 2837
b8da9b10
LL
2838 if (!list_empty(&cur->fl_blocked_requests)) {
2839 /* Turn left */
2840 cur = list_first_entry_or_null(&cur->fl_blocked_requests,
2841 struct file_lock, fl_blocked_member);
2842 level++;
2843 } else {
2844 /* Turn right */
2845 tmp = get_next_blocked_member(cur);
2846 /* Fall back to parent node */
2847 while (tmp == NULL && cur->fl_blocker != NULL) {
2848 cur = cur->fl_blocker;
2849 level--;
2850 tmp = get_next_blocked_member(cur);
2851 }
2852 cur = tmp;
2853 }
2854 }
094f2825 2855
7f8ada98
PE
2856 return 0;
2857}
1da177e4 2858
6c8c9031
AV
2859static void __show_fd_locks(struct seq_file *f,
2860 struct list_head *head, int *id,
2861 struct file *filp, struct files_struct *files)
2862{
2863 struct file_lock *fl;
2864
2865 list_for_each_entry(fl, head, fl_list) {
2866
2867 if (filp != fl->fl_file)
2868 continue;
2869 if (fl->fl_owner != files &&
2870 fl->fl_owner != filp)
2871 continue;
2872
2873 (*id)++;
2874 seq_puts(f, "lock:\t");
b8da9b10 2875 lock_get_status(f, fl, *id, "", 0);
6c8c9031
AV
2876 }
2877}
2878
2879void show_fd_locks(struct seq_file *f,
2880 struct file *filp, struct files_struct *files)
2881{
c568d683 2882 struct inode *inode = locks_inode(filp);
6c8c9031
AV
2883 struct file_lock_context *ctx;
2884 int id = 0;
2885
128a3785 2886 ctx = smp_load_acquire(&inode->i_flctx);
6c8c9031
AV
2887 if (!ctx)
2888 return;
2889
2890 spin_lock(&ctx->flc_lock);
2891 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2892 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2893 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2894 spin_unlock(&ctx->flc_lock);
2895}
2896
7f8ada98 2897static void *locks_start(struct seq_file *f, loff_t *pos)
b03dfdec 2898 __acquires(&blocked_lock_lock)
7f8ada98 2899{
7012b02a 2900 struct locks_iterator *iter = f->private;
99dc8292 2901
7012b02a 2902 iter->li_pos = *pos + 1;
aba37660 2903 percpu_down_write(&file_rwsem);
7b2296af 2904 spin_lock(&blocked_lock_lock);
7c3f654d 2905 return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos);
7f8ada98 2906}
1da177e4 2907
7f8ada98
PE
2908static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2909{
7012b02a
JL
2910 struct locks_iterator *iter = f->private;
2911
2912 ++iter->li_pos;
7c3f654d 2913 return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos);
7f8ada98 2914}
1da177e4 2915
7f8ada98 2916static void locks_stop(struct seq_file *f, void *v)
b03dfdec 2917 __releases(&blocked_lock_lock)
7f8ada98 2918{
7b2296af 2919 spin_unlock(&blocked_lock_lock);
aba37660 2920 percpu_up_write(&file_rwsem);
1da177e4
LT
2921}
2922
d8ba7a36 2923static const struct seq_operations locks_seq_operations = {
7f8ada98
PE
2924 .start = locks_start,
2925 .next = locks_next,
2926 .stop = locks_stop,
2927 .show = locks_show,
2928};
d8ba7a36 2929
d8ba7a36
AD
2930static int __init proc_locks_init(void)
2931{
44414d82
CH
2932 proc_create_seq_private("locks", 0, NULL, &locks_seq_operations,
2933 sizeof(struct locks_iterator), NULL);
d8ba7a36
AD
2934 return 0;
2935}
91899226 2936fs_initcall(proc_locks_init);
7f8ada98
PE
2937#endif
2938
1da177e4
LT
2939static int __init filelock_init(void)
2940{
7012b02a
JL
2941 int i;
2942
4a075e39 2943 flctx_cache = kmem_cache_create("file_lock_ctx",
3754707b 2944 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
4a075e39 2945
1da177e4 2946 filelock_cache = kmem_cache_create("file_lock_cache",
3754707b 2947 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
ee19cc40 2948
7c3f654d
PZ
2949 for_each_possible_cpu(i) {
2950 struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i);
2951
2952 spin_lock_init(&fll->lock);
2953 INIT_HLIST_HEAD(&fll->hlist);
2954 }
7012b02a 2955
18f6622e 2956 lease_notifier_chain_init();
1da177e4
LT
2957 return 0;
2958}
1da177e4 2959core_initcall(filelock_init);