blob: e16c2c61a44f3fa5eb856535674b25981476c9d0 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/locks.c
3 *
4 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
5 * Doug Evans (dje@spiff.uucp), August 07, 1992
6 *
7 * Deadlock detection added.
8 * FIXME: one thing isn't handled yet:
9 * - mandatory locks (requires lots of changes elsewhere)
10 * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994.
11 *
12 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
13 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
14 *
15 * Converted file_lock_table to a linked list from an array, which eliminates
16 * the limits on how many active file locks are open.
17 * Chad Page (pageone@netcom.com), November 27, 1994
18 *
19 * Removed dependency on file descriptors. dup()'ed file descriptors now
20 * get the same locks as the original file descriptors, and a close() on
21 * any file descriptor removes ALL the locks on the file for the current
22 * process. Since locks still depend on the process id, locks are inherited
23 * after an exec() but not after a fork(). This agrees with POSIX, and both
24 * BSD and SVR4 practice.
25 * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995
26 *
27 * Scrapped free list which is redundant now that we allocate locks
28 * dynamically with kmalloc()/kfree().
29 * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995
30 *
31 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
32 *
33 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
34 * fcntl() system call. They have the semantics described above.
35 *
36 * FL_FLOCK locks are created with calls to flock(), through the flock()
37 * system call, which is new. Old C libraries implement flock() via fcntl()
38 * and will continue to use the old, broken implementation.
39 *
40 * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated
41 * with a file pointer (filp). As a result they can be shared by a parent
42 * process and its children after a fork(). They are removed when the last
43 * file descriptor referring to the file pointer is closed (unless explicitly
44 * unlocked).
45 *
46 * FL_FLOCK locks never deadlock, an existing lock is always removed before
47 * upgrading from shared to exclusive (or vice versa). When this happens
48 * any processes blocked by the current lock are woken up and allowed to
49 * run before the new lock is applied.
50 * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995
51 *
52 * Removed some race conditions in flock_lock_file(), marked other possible
53 * races. Just grep for FIXME to see them.
54 * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996.
55 *
56 * Addressed Dmitry's concerns. Deadlock checking no longer recursive.
57 * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep
58 * once we've checked for blocking and deadlocking.
59 * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996.
60 *
61 * Initial implementation of mandatory locks. SunOS turned out to be
62 * a rotten model, so I implemented the "obvious" semantics.
Paul Bolle395cf962011-08-15 02:02:26 +020063 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
65 *
66 * Don't allow mandatory locks on mmap()'ed files. Added simple functions to
67 * check if a file has mandatory locks, used by mmap(), open() and creat() to
68 * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference
69 * Manual, Section 2.
70 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
71 *
72 * Tidied up block list handling. Added '/proc/locks' interface.
73 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
74 *
75 * Fixed deadlock condition for pathological code that mixes calls to
76 * flock() and fcntl().
77 * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996.
78 *
79 * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use
80 * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to
81 * guarantee sensible behaviour in the case where file system modules might
82 * be compiled with different options than the kernel itself.
83 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
84 *
85 * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel
86 * (Thomas.Meckel@mni.fh-giessen.de) for spotting this.
87 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
88 *
89 * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK
90 * locks. Changed process synchronisation to avoid dereferencing locks that
91 * have already been freed.
92 * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996.
93 *
94 * Made the block list a circular list to minimise searching in the list.
95 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
96 *
97 * Made mandatory locking a mount option. Default is not to allow mandatory
98 * locking.
99 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
100 *
101 * Some adaptations for NFS support.
102 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
103 *
104 * Fixed /proc/locks interface so that we can't overrun the buffer we are handed.
105 * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997.
106 *
107 * Use slab allocator instead of kmalloc/kfree.
108 * Use generic list implementation from <linux/list.h>.
109 * Sped up posix_locks_deadlock by only considering blocked locks.
110 * Matthew Wilcox <willy@debian.org>, March, 2000.
111 *
112 * Leases and LOCK_MAND
113 * Matthew Wilcox <willy@debian.org>, June, 2000.
114 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
115 */
116
117#include <linux/capability.h>
118#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -0400119#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120#include <linux/fs.h>
121#include <linux/init.h>
122#include <linux/module.h>
123#include <linux/security.h>
124#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125#include <linux/syscalls.h>
126#include <linux/time.h>
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700127#include <linux/rcupdate.h>
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000128#include <linux/pid_namespace.h>
Jeff Layton48f74182013-06-21 08:58:18 -0400129#include <linux/hashtable.h>
Jeff Layton7012b022013-06-21 08:58:22 -0400130#include <linux/percpu.h>
131#include <linux/lglock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Jeff Layton62af4f12014-05-09 14:13:05 -0400133#define CREATE_TRACE_POINTS
134#include <trace/events/filelock.h>
135
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136#include <asm/uaccess.h>
137
138#define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
139#define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
J. Bruce Fields617588d2011-07-01 15:18:34 -0400140#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG))
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400141#define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -0400143static bool lease_breaking(struct file_lock *fl)
144{
J. Bruce Fields778fc542011-07-26 18:25:49 -0400145 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
146}
147
148static int target_leasetype(struct file_lock *fl)
149{
150 if (fl->fl_flags & FL_UNLOCK_PENDING)
151 return F_UNLCK;
152 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
153 return F_RDLCK;
154 return fl->fl_type;
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -0400155}
156
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157int leases_enable = 1;
158int lease_break_time = 45;
159
160#define for_each_lock(inode, lockp) \
161 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
162
Jeff Layton1c8c6012013-06-21 08:58:15 -0400163/*
Jeff Layton7012b022013-06-21 08:58:22 -0400164 * The global file_lock_list is only used for displaying /proc/locks, so we
165 * keep a list on each CPU, with each list protected by its own spinlock via
166 * the file_lock_lglock. Note that alterations to the list also require that
167 * the relevant i_lock is held.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400168 */
Jeff Layton7012b022013-06-21 08:58:22 -0400169DEFINE_STATIC_LGLOCK(file_lock_lglock);
170static DEFINE_PER_CPU(struct hlist_head, file_lock_list);
Jeff Layton88974692013-06-21 08:58:14 -0400171
Jeff Layton1c8c6012013-06-21 08:58:15 -0400172/*
Jeff Layton48f74182013-06-21 08:58:18 -0400173 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
Jeff Layton7b2296a2013-06-21 08:58:20 -0400174 * It is protected by blocked_lock_lock.
Jeff Layton48f74182013-06-21 08:58:18 -0400175 *
176 * We hash locks by lockowner in order to optimize searching for the lock a
177 * particular lockowner is waiting on.
178 *
179 * FIXME: make this value scale via some heuristic? We generally will want more
180 * buckets when we have more lockowners holding locks, but that's a little
181 * difficult to determine without knowing what the workload will look like.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400182 */
Jeff Layton48f74182013-06-21 08:58:18 -0400183#define BLOCKED_HASH_BITS 7
184static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
Jeff Layton88974692013-06-21 08:58:14 -0400185
Jeff Layton1c8c6012013-06-21 08:58:15 -0400186/*
Jeff Layton7b2296a2013-06-21 08:58:20 -0400187 * This lock protects the blocked_hash. Generally, if you're accessing it, you
188 * want to be holding this lock.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400189 *
190 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
191 * pointer for file_lock structures that are acting as lock requests (in
192 * contrast to those that are acting as records of acquired locks).
193 *
194 * Note that when we acquire this lock in order to change the above fields,
195 * we often hold the i_lock as well. In certain cases, when reading the fields
196 * protected by this lock, we can skip acquiring it iff we already hold the
197 * i_lock.
198 *
199 * In particular, adding an entry to the fl_block list requires that you hold
200 * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting
201 * an entry from the list however only requires the file_lock_lock.
202 */
Jeff Layton7b2296a2013-06-21 08:58:20 -0400203static DEFINE_SPINLOCK(blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
Christoph Lametere18b8902006-12-06 20:33:20 -0800205static struct kmem_cache *filelock_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
Miklos Szerediee19cc42011-07-07 13:06:09 +0200207static void locks_init_lock_heads(struct file_lock *fl)
Miklos Szeredia51cb912011-07-06 12:33:55 +0200208{
Jeff Layton139ca042013-06-21 08:58:17 -0400209 INIT_HLIST_NODE(&fl->fl_link);
Miklos Szerediee19cc42011-07-07 13:06:09 +0200210 INIT_LIST_HEAD(&fl->fl_block);
211 init_waitqueue_head(&fl->fl_wait);
Miklos Szeredia51cb912011-07-06 12:33:55 +0200212}
213
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214/* Allocate an empty lock structure. */
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +0200215struct file_lock *locks_alloc_lock(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216{
Miklos Szerediee19cc42011-07-07 13:06:09 +0200217 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
Miklos Szeredia51cb912011-07-06 12:33:55 +0200218
219 if (fl)
Miklos Szerediee19cc42011-07-07 13:06:09 +0200220 locks_init_lock_heads(fl);
Miklos Szeredia51cb912011-07-06 12:33:55 +0200221
222 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +0200224EXPORT_SYMBOL_GPL(locks_alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
Felix Blyakhera9e61e22009-03-31 15:12:56 -0500226void locks_release_private(struct file_lock *fl)
Trond Myklebust47831f32006-03-20 13:44:05 -0500227{
228 if (fl->fl_ops) {
229 if (fl->fl_ops->fl_release_private)
230 fl->fl_ops->fl_release_private(fl);
231 fl->fl_ops = NULL;
232 }
Trond Myklebust47831f32006-03-20 13:44:05 -0500233
Kinglong Mee5c97d7b2014-08-22 10:18:43 -0400234 if (fl->fl_lmops) {
235 if (fl->fl_lmops->lm_put_owner)
236 fl->fl_lmops->lm_put_owner(fl);
237 fl->fl_lmops = NULL;
238 }
Trond Myklebust47831f32006-03-20 13:44:05 -0500239}
Felix Blyakhera9e61e22009-03-31 15:12:56 -0500240EXPORT_SYMBOL_GPL(locks_release_private);
Trond Myklebust47831f32006-03-20 13:44:05 -0500241
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242/* Free a lock which is not in use. */
J. Bruce Fields05fa3132010-10-30 17:31:15 -0400243void locks_free_lock(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244{
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800245 BUG_ON(waitqueue_active(&fl->fl_wait));
246 BUG_ON(!list_empty(&fl->fl_block));
Jeff Layton139ca042013-06-21 08:58:17 -0400247 BUG_ON(!hlist_unhashed(&fl->fl_link));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
Trond Myklebust47831f32006-03-20 13:44:05 -0500249 locks_release_private(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 kmem_cache_free(filelock_cache, fl);
251}
J. Bruce Fields05fa3132010-10-30 17:31:15 -0400252EXPORT_SYMBOL(locks_free_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
Jeff Laytoned9814d2014-08-11 14:20:31 -0400254static void
255locks_dispose_list(struct list_head *dispose)
256{
257 struct file_lock *fl;
258
259 while (!list_empty(dispose)) {
260 fl = list_first_entry(dispose, struct file_lock, fl_block);
261 list_del_init(&fl->fl_block);
262 locks_free_lock(fl);
263 }
264}
265
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266void locks_init_lock(struct file_lock *fl)
267{
Miklos Szerediee19cc42011-07-07 13:06:09 +0200268 memset(fl, 0, sizeof(struct file_lock));
269 locks_init_lock_heads(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270}
271
272EXPORT_SYMBOL(locks_init_lock);
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274/*
275 * Initialize a new lock from an existing file_lock structure.
276 */
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400277void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
279 new->fl_owner = fl->fl_owner;
280 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500281 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 new->fl_flags = fl->fl_flags;
283 new->fl_type = fl->fl_type;
284 new->fl_start = fl->fl_start;
285 new->fl_end = fl->fl_end;
Kinglong Meef3282962014-08-22 10:18:43 -0400286 new->fl_lmops = fl->fl_lmops;
Trond Myklebust09969052006-03-20 13:44:38 -0500287 new->fl_ops = NULL;
Kinglong Meef3282962014-08-22 10:18:43 -0400288
289 if (fl->fl_lmops) {
290 if (fl->fl_lmops->lm_get_owner)
291 fl->fl_lmops->lm_get_owner(new, fl);
292 }
Trond Myklebust09969052006-03-20 13:44:38 -0500293}
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400294EXPORT_SYMBOL(locks_copy_conflock);
Trond Myklebust09969052006-03-20 13:44:38 -0500295
296void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
297{
Jeff Layton566709b2014-08-11 14:09:35 -0400298 /* "new" must be a freshly-initialized lock */
299 WARN_ON_ONCE(new->fl_ops);
Trond Myklebust09969052006-03-20 13:44:38 -0500300
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400301 locks_copy_conflock(new, fl);
Kinglong Meef3282962014-08-22 10:18:43 -0400302
Trond Myklebust09969052006-03-20 13:44:38 -0500303 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 new->fl_ops = fl->fl_ops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500305
Kinglong Meef3282962014-08-22 10:18:43 -0400306 if (fl->fl_ops) {
307 if (fl->fl_ops->fl_copy_lock)
308 fl->fl_ops->fl_copy_lock(new, fl);
309 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310}
311
312EXPORT_SYMBOL(locks_copy_lock);
313
314static inline int flock_translate_cmd(int cmd) {
315 if (cmd & LOCK_MAND)
316 return cmd & (LOCK_MAND | LOCK_RW);
317 switch (cmd) {
318 case LOCK_SH:
319 return F_RDLCK;
320 case LOCK_EX:
321 return F_WRLCK;
322 case LOCK_UN:
323 return F_UNLCK;
324 }
325 return -EINVAL;
326}
327
328/* Fill in a file_lock structure with an appropriate FLOCK lock. */
329static int flock_make_lock(struct file *filp, struct file_lock **lock,
330 unsigned int cmd)
331{
332 struct file_lock *fl;
333 int type = flock_translate_cmd(cmd);
334 if (type < 0)
335 return type;
336
337 fl = locks_alloc_lock();
338 if (fl == NULL)
339 return -ENOMEM;
340
341 fl->fl_file = filp;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +0200342 fl->fl_owner = filp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 fl->fl_pid = current->tgid;
344 fl->fl_flags = FL_FLOCK;
345 fl->fl_type = type;
346 fl->fl_end = OFFSET_MAX;
347
348 *lock = fl;
349 return 0;
350}
351
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400352static int assign_type(struct file_lock *fl, long type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353{
354 switch (type) {
355 case F_RDLCK:
356 case F_WRLCK:
357 case F_UNLCK:
358 fl->fl_type = type;
359 break;
360 default:
361 return -EINVAL;
362 }
363 return 0;
364}
365
J. Bruce Fieldsef12e722014-02-03 12:13:08 -0500366static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
367 struct flock64 *l)
368{
369 switch (l->l_whence) {
370 case SEEK_SET:
371 fl->fl_start = 0;
372 break;
373 case SEEK_CUR:
374 fl->fl_start = filp->f_pos;
375 break;
376 case SEEK_END:
377 fl->fl_start = i_size_read(file_inode(filp));
378 break;
379 default:
380 return -EINVAL;
381 }
382 if (l->l_start > OFFSET_MAX - fl->fl_start)
383 return -EOVERFLOW;
384 fl->fl_start += l->l_start;
385 if (fl->fl_start < 0)
386 return -EINVAL;
387
388 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
389 POSIX-2001 defines it. */
390 if (l->l_len > 0) {
391 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
392 return -EOVERFLOW;
393 fl->fl_end = fl->fl_start + l->l_len - 1;
394
395 } else if (l->l_len < 0) {
396 if (fl->fl_start + l->l_len < 0)
397 return -EINVAL;
398 fl->fl_end = fl->fl_start - 1;
399 fl->fl_start += l->l_len;
400 } else
401 fl->fl_end = OFFSET_MAX;
402
403 fl->fl_owner = current->files;
404 fl->fl_pid = current->tgid;
405 fl->fl_file = filp;
406 fl->fl_flags = FL_POSIX;
407 fl->fl_ops = NULL;
408 fl->fl_lmops = NULL;
409
410 return assign_type(fl, l->l_type);
411}
412
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
414 * style lock.
415 */
416static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
417 struct flock *l)
418{
J. Bruce Fieldsef12e722014-02-03 12:13:08 -0500419 struct flock64 ll = {
420 .l_type = l->l_type,
421 .l_whence = l->l_whence,
422 .l_start = l->l_start,
423 .l_len = l->l_len,
424 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
J. Bruce Fieldsef12e722014-02-03 12:13:08 -0500426 return flock64_to_posix_lock(filp, fl, &ll);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427}
428
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429/* default lease lock manager operations */
430static void lease_break_callback(struct file_lock *fl)
431{
432 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
433}
434
Alexey Dobriyan7b021962009-09-21 17:01:12 -0700435static const struct lock_manager_operations lease_manager_ops = {
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400436 .lm_break = lease_break_callback,
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400437 .lm_change = lease_modify,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438};
439
440/*
441 * Initialize a lease, use the default lock manager operations
442 */
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400443static int lease_init(struct file *filp, long type, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400445 if (assign_type(fl, type) != 0)
446 return -EINVAL;
447
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +0200448 fl->fl_owner = current->files;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 fl->fl_pid = current->tgid;
450
451 fl->fl_file = filp;
452 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 fl->fl_start = 0;
454 fl->fl_end = OFFSET_MAX;
455 fl->fl_ops = NULL;
456 fl->fl_lmops = &lease_manager_ops;
457 return 0;
458}
459
460/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400461static struct file_lock *lease_alloc(struct file *filp, long type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462{
463 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400464 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500467 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
469 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400470 if (error) {
471 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500472 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400473 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500474 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475}
476
477/* Check if two locks overlap each other.
478 */
479static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
480{
481 return ((fl1->fl_end >= fl2->fl_start) &&
482 (fl2->fl_end >= fl1->fl_start));
483}
484
485/*
486 * Check whether two locks have the same owner.
487 */
Matt Mackall33443c42006-01-08 01:05:22 -0800488static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400490 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 return fl2->fl_lmops == fl1->fl_lmops &&
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400492 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 return fl1->fl_owner == fl2->fl_owner;
494}
495
Jeff Layton7012b022013-06-21 08:58:22 -0400496/* Must be called with the i_lock held! */
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500497static void locks_insert_global_locks(struct file_lock *fl)
Jeff Layton88974692013-06-21 08:58:14 -0400498{
Jeff Layton7012b022013-06-21 08:58:22 -0400499 lg_local_lock(&file_lock_lglock);
500 fl->fl_link_cpu = smp_processor_id();
501 hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list));
502 lg_local_unlock(&file_lock_lglock);
Jeff Layton88974692013-06-21 08:58:14 -0400503}
504
Jeff Layton7012b022013-06-21 08:58:22 -0400505/* Must be called with the i_lock held! */
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500506static void locks_delete_global_locks(struct file_lock *fl)
Jeff Layton88974692013-06-21 08:58:14 -0400507{
Jeff Layton7012b022013-06-21 08:58:22 -0400508 /*
509 * Avoid taking lock if already unhashed. This is safe since this check
510 * is done while holding the i_lock, and new insertions into the list
511 * also require that it be held.
512 */
513 if (hlist_unhashed(&fl->fl_link))
514 return;
515 lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu);
Jeff Layton139ca042013-06-21 08:58:17 -0400516 hlist_del_init(&fl->fl_link);
Jeff Layton7012b022013-06-21 08:58:22 -0400517 lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu);
Jeff Layton88974692013-06-21 08:58:14 -0400518}
519
Jeff Layton3999e492013-06-21 08:58:19 -0400520static unsigned long
521posix_owner_key(struct file_lock *fl)
522{
523 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
524 return fl->fl_lmops->lm_owner_key(fl);
525 return (unsigned long)fl->fl_owner;
526}
527
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500528static void locks_insert_global_blocked(struct file_lock *waiter)
Jeff Layton88974692013-06-21 08:58:14 -0400529{
Jeff Layton3999e492013-06-21 08:58:19 -0400530 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
Jeff Layton88974692013-06-21 08:58:14 -0400531}
532
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500533static void locks_delete_global_blocked(struct file_lock *waiter)
Jeff Layton88974692013-06-21 08:58:14 -0400534{
Jeff Layton48f74182013-06-21 08:58:18 -0400535 hash_del(&waiter->fl_link);
Jeff Layton88974692013-06-21 08:58:14 -0400536}
537
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538/* Remove waiter from blocker's block list.
539 * When blocker ends up pointing to itself then the list is empty.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400540 *
Jeff Layton7b2296a2013-06-21 08:58:20 -0400541 * Must be called with blocked_lock_lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 */
Matt Mackall33443c42006-01-08 01:05:22 -0800543static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544{
Jeff Layton88974692013-06-21 08:58:14 -0400545 locks_delete_global_blocked(waiter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 list_del_init(&waiter->fl_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 waiter->fl_next = NULL;
548}
549
Jeff Layton1a9e64a2013-06-21 08:58:10 -0400550static void locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551{
Jeff Layton7b2296a2013-06-21 08:58:20 -0400552 spin_lock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 __locks_delete_block(waiter);
Jeff Layton7b2296a2013-06-21 08:58:20 -0400554 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555}
556
557/* Insert waiter into blocker's block list.
558 * We use a circular list so that processes can be easily woken up in
559 * the order they blocked. The documentation doesn't require this but
560 * it seems like the reasonable thing to do.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400561 *
Jeff Layton7b2296a2013-06-21 08:58:20 -0400562 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
Jeff Layton46dad762014-02-03 12:13:06 -0500563 * list itself is protected by the blocked_lock_lock, but by ensuring that the
Jeff Layton7b2296a2013-06-21 08:58:20 -0400564 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
Jeff Layton4e8c7652013-06-21 08:58:16 -0400565 * in some cases when we see that the fl_block list is empty.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 */
Jeff Layton1c8c6012013-06-21 08:58:15 -0400567static void __locks_insert_block(struct file_lock *blocker,
568 struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800570 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 waiter->fl_next = blocker;
Jeff Layton88974692013-06-21 08:58:14 -0400572 list_add_tail(&waiter->fl_block, &blocker->fl_block);
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400573 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
Jeff Layton1c8c6012013-06-21 08:58:15 -0400574 locks_insert_global_blocked(waiter);
575}
576
577/* Must be called with i_lock held. */
578static void locks_insert_block(struct file_lock *blocker,
579 struct file_lock *waiter)
580{
Jeff Layton7b2296a2013-06-21 08:58:20 -0400581 spin_lock(&blocked_lock_lock);
Jeff Layton1c8c6012013-06-21 08:58:15 -0400582 __locks_insert_block(blocker, waiter);
Jeff Layton7b2296a2013-06-21 08:58:20 -0400583 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584}
585
Jeff Layton1cb36012013-06-21 08:58:12 -0400586/*
587 * Wake up processes blocked waiting for blocker.
588 *
Jeff Layton1c8c6012013-06-21 08:58:15 -0400589 * Must be called with the inode->i_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 */
591static void locks_wake_up_blocks(struct file_lock *blocker)
592{
Jeff Layton4e8c7652013-06-21 08:58:16 -0400593 /*
594 * Avoid taking global lock if list is empty. This is safe since new
595 * blocked requests are only added to the list under the i_lock, and
596 * the i_lock is always held here. Note that removal from the fl_block
597 * list does not require the i_lock, so we must recheck list_empty()
Jeff Layton7b2296a2013-06-21 08:58:20 -0400598 * after acquiring the blocked_lock_lock.
Jeff Layton4e8c7652013-06-21 08:58:16 -0400599 */
600 if (list_empty(&blocker->fl_block))
601 return;
602
Jeff Layton7b2296a2013-06-21 08:58:20 -0400603 spin_lock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400605 struct file_lock *waiter;
606
607 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 struct file_lock, fl_block);
609 __locks_delete_block(waiter);
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400610 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
611 waiter->fl_lmops->lm_notify(waiter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 else
613 wake_up(&waiter->fl_wait);
614 }
Jeff Layton7b2296a2013-06-21 08:58:20 -0400615 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616}
617
618/* Insert file lock fl into an inode's lock list at the position indicated
619 * by pos. At the same time add the lock to the global file lock list.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400620 *
621 * Must be called with the i_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 */
623static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
624{
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000625 fl->fl_nspid = get_pid(task_tgid(current));
626
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 /* insert into file's list */
628 fl->fl_next = *pos;
629 *pos = fl;
Jeff Layton88974692013-06-21 08:58:14 -0400630
631 locks_insert_global_locks(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632}
633
Jeff Layton24cbe782014-02-03 12:13:06 -0500634/**
635 * locks_delete_lock - Delete a lock and then free it.
636 * @thisfl_p: pointer that points to the fl_next field of the previous
637 * inode->i_flock list entry
638 *
639 * Unlink a lock from all lists and free the namespace reference, but don't
640 * free it yet. Wake up processes that are blocked waiting for this lock and
641 * notify the FS that the lock has been cleared.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400642 *
643 * Must be called with the i_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 */
Jeff Layton24cbe782014-02-03 12:13:06 -0500645static void locks_unlink_lock(struct file_lock **thisfl_p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646{
647 struct file_lock *fl = *thisfl_p;
648
Jeff Layton88974692013-06-21 08:58:14 -0400649 locks_delete_global_locks(fl);
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 *thisfl_p = fl->fl_next;
652 fl->fl_next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000654 if (fl->fl_nspid) {
655 put_pid(fl->fl_nspid);
656 fl->fl_nspid = NULL;
657 }
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 locks_wake_up_blocks(fl);
Jeff Layton24cbe782014-02-03 12:13:06 -0500660}
661
662/*
663 * Unlink a lock from all lists and free it.
664 *
665 * Must be called with i_lock held!
666 */
Jeff Laytoned9814d2014-08-11 14:20:31 -0400667static void locks_delete_lock(struct file_lock **thisfl_p,
668 struct list_head *dispose)
Jeff Layton24cbe782014-02-03 12:13:06 -0500669{
670 struct file_lock *fl = *thisfl_p;
671
672 locks_unlink_lock(thisfl_p);
Jeff Laytoned9814d2014-08-11 14:20:31 -0400673 if (dispose)
674 list_add(&fl->fl_block, dispose);
675 else
676 locks_free_lock(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677}
678
679/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
680 * checks for shared/exclusive status of overlapping locks.
681 */
682static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
683{
684 if (sys_fl->fl_type == F_WRLCK)
685 return 1;
686 if (caller_fl->fl_type == F_WRLCK)
687 return 1;
688 return 0;
689}
690
691/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
692 * checking before calling the locks_conflict().
693 */
694static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
695{
696 /* POSIX locks owned by the same process do not conflict with
697 * each other.
698 */
699 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
700 return (0);
701
702 /* Check whether they overlap */
703 if (!locks_overlap(caller_fl, sys_fl))
704 return 0;
705
706 return (locks_conflict(caller_fl, sys_fl));
707}
708
709/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
710 * checking before calling the locks_conflict().
711 */
712static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
713{
714 /* FLOCK locks referring to the same filp do not conflict with
715 * each other.
716 */
717 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
718 return (0);
719 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
720 return 0;
721
722 return (locks_conflict(caller_fl, sys_fl));
723}
724
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400725void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500726posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727{
728 struct file_lock *cfl;
Jeff Layton1c8c6012013-06-21 08:58:15 -0400729 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730
Jeff Layton1c8c6012013-06-21 08:58:15 -0400731 spin_lock(&inode->i_lock);
Al Viro496ad9a2013-01-23 17:07:38 -0500732 for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 if (!IS_POSIX(cfl))
734 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400735 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 break;
737 }
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000738 if (cfl) {
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400739 locks_copy_conflock(fl, cfl);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000740 if (cfl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800741 fl->fl_pid = pid_vnr(cfl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000742 } else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400743 fl->fl_type = F_UNLCK;
Jeff Layton1c8c6012013-06-21 08:58:15 -0400744 spin_unlock(&inode->i_lock);
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400745 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747EXPORT_SYMBOL(posix_test_lock);
748
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400749/*
750 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400752 * We attempt to detect deadlocks that are due purely to posix file
753 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400755 * We assume that a task can be waiting for at most one lock at a time.
756 * So for any acquired lock, the process holding that lock may be
757 * waiting on at most one other lock. That lock in turns may be held by
758 * someone waiting for at most one other lock. Given a requested lock
759 * caller_fl which is about to wait for a conflicting lock block_fl, we
760 * follow this chain of waiters to ensure we are not about to create a
761 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400762 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400763 * Since we do this before we ever put a process to sleep on a lock, we
764 * are ensured that there is never a cycle; that is what guarantees that
765 * the while() loop in posix_locks_deadlock() eventually completes.
766 *
767 * Note: the above assumption may not be true when handling lock
768 * requests from a broken NFS client. It may also fail in the presence
769 * of tasks (such as posix threads) sharing the same open file table.
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400770 * To handle those cases, we just bail out after a few iterations.
Jeff Layton57b65322014-02-03 12:13:09 -0500771 *
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400772 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
Jeff Layton57b65322014-02-03 12:13:09 -0500773 * Because the owner is not even nominally tied to a thread of
774 * execution, the deadlock detection below can't reasonably work well. Just
775 * skip it for those.
776 *
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400777 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
Jeff Layton57b65322014-02-03 12:13:09 -0500778 * locks that just checks for the case where two tasks are attempting to
779 * upgrade from read to write locks on the same inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400781
782#define MAX_DEADLK_ITERATIONS 10
783
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400784/* Find a lock that the owner of the given block_fl is blocking on. */
785static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
786{
787 struct file_lock *fl;
788
Jeff Layton3999e492013-06-21 08:58:19 -0400789 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400790 if (posix_same_owner(fl, block_fl))
791 return fl->fl_next;
792 }
793 return NULL;
794}
795
Jeff Layton7b2296a2013-06-21 08:58:20 -0400796/* Must be called with the blocked_lock_lock held! */
Adrian Bunkb0904e12006-06-23 02:05:13 -0700797static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 struct file_lock *block_fl)
799{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400800 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
Jeff Layton57b65322014-02-03 12:13:09 -0500802 /*
803 * This deadlock detector can't reasonably detect deadlocks with
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400804 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
Jeff Layton57b65322014-02-03 12:13:09 -0500805 */
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400806 if (IS_OFDLCK(caller_fl))
Jeff Layton57b65322014-02-03 12:13:09 -0500807 return 0;
808
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400809 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
810 if (i++ > MAX_DEADLK_ITERATIONS)
811 return 0;
812 if (posix_same_owner(caller_fl, block_fl))
813 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 }
815 return 0;
816}
817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400819 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400820 *
821 * Note that if called with an FL_EXISTS argument, the caller may determine
822 * whether or not a lock was successfully freed by testing the return
823 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800825static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800827 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 struct file_lock **before;
Al Viro496ad9a2013-01-23 17:07:38 -0500829 struct inode * inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 int error = 0;
831 int found = 0;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400832 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200834 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
835 new_fl = locks_alloc_lock();
836 if (!new_fl)
837 return -ENOMEM;
838 }
839
Jeff Layton1c8c6012013-06-21 08:58:15 -0400840 spin_lock(&inode->i_lock);
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400841 if (request->fl_flags & FL_ACCESS)
842 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400843
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 for_each_lock(inode, before) {
845 struct file_lock *fl = *before;
846 if (IS_POSIX(fl))
847 break;
848 if (IS_LEASE(fl))
849 continue;
850 if (filp != fl->fl_file)
851 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800852 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 goto out;
854 found = 1;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400855 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 break;
857 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Trond Myklebustf475ae92006-06-29 16:38:32 -0400859 if (request->fl_type == F_UNLCK) {
860 if ((request->fl_flags & FL_EXISTS) && !found)
861 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800862 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400863 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865 /*
866 * If a higher-priority process was blocked on the old file lock,
867 * give it the opportunity to lock the file.
868 */
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200869 if (found) {
Jeff Layton1c8c6012013-06-21 08:58:15 -0400870 spin_unlock(&inode->i_lock);
Frederic Weisbeckerdef01bc2009-07-16 15:44:29 +0200871 cond_resched();
Jeff Layton1c8c6012013-06-21 08:58:15 -0400872 spin_lock(&inode->i_lock);
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200873 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400875find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 for_each_lock(inode, before) {
877 struct file_lock *fl = *before;
878 if (IS_POSIX(fl))
879 break;
880 if (IS_LEASE(fl))
881 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800882 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 continue;
884 error = -EAGAIN;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700885 if (!(request->fl_flags & FL_SLEEP))
886 goto out;
887 error = FILE_LOCK_DEFERRED;
888 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 goto out;
890 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400891 if (request->fl_flags & FL_ACCESS)
892 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800893 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700894 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800895 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400896 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897
898out:
Jeff Layton1c8c6012013-06-21 08:58:15 -0400899 spin_unlock(&inode->i_lock);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800900 if (new_fl)
901 locks_free_lock(new_fl);
Jeff Laytoned9814d2014-08-11 14:20:31 -0400902 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 return error;
904}
905
Marc Eshel150b3932007-01-18 16:15:35 -0500906static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907{
908 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700909 struct file_lock *new_fl = NULL;
910 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 struct file_lock *left = NULL;
912 struct file_lock *right = NULL;
913 struct file_lock **before;
Jeff Laytonb9746ef2013-06-21 08:58:13 -0400914 int error;
915 bool added = false;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400916 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
918 /*
919 * We may need two file_lock structures for this operation,
920 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700921 *
922 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700924 if (!(request->fl_flags & FL_ACCESS) &&
925 (request->fl_type != F_UNLCK ||
926 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
927 new_fl = locks_alloc_lock();
928 new_fl2 = locks_alloc_lock();
929 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
Jeff Layton1c8c6012013-06-21 08:58:15 -0400931 spin_lock(&inode->i_lock);
Jeff Layton1cb36012013-06-21 08:58:12 -0400932 /*
933 * New lock request. Walk all POSIX locks and look for conflicts. If
934 * there are any, either return error or put the request on the
Jeff Layton48f74182013-06-21 08:58:18 -0400935 * blocker's list of waiters and the global blocked_hash.
Jeff Layton1cb36012013-06-21 08:58:12 -0400936 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 if (request->fl_type != F_UNLCK) {
938 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500939 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 if (!IS_POSIX(fl))
941 continue;
942 if (!posix_locks_conflict(request, fl))
943 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800944 if (conflock)
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400945 locks_copy_conflock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 error = -EAGAIN;
947 if (!(request->fl_flags & FL_SLEEP))
948 goto out;
Jeff Layton1c8c6012013-06-21 08:58:15 -0400949 /*
950 * Deadlock detection and insertion into the blocked
951 * locks list must be done while holding the same lock!
952 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 error = -EDEADLK;
Jeff Layton7b2296a2013-06-21 08:58:20 -0400954 spin_lock(&blocked_lock_lock);
Jeff Layton1c8c6012013-06-21 08:58:15 -0400955 if (likely(!posix_locks_deadlock(request, fl))) {
956 error = FILE_LOCK_DEFERRED;
957 __locks_insert_block(fl, request);
958 }
Jeff Layton7b2296a2013-06-21 08:58:20 -0400959 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 goto out;
961 }
962 }
963
964 /* If we're just looking for a conflict, we're done. */
965 error = 0;
966 if (request->fl_flags & FL_ACCESS)
967 goto out;
968
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 * Find the first old lock with the same owner as the new lock.
971 */
972
973 before = &inode->i_flock;
974
975 /* First skip locks owned by other processes. */
976 while ((fl = *before) && (!IS_POSIX(fl) ||
977 !posix_same_owner(request, fl))) {
978 before = &fl->fl_next;
979 }
980
Jeff Layton1cb36012013-06-21 08:58:12 -0400981 /* Process locks with this owner. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 while ((fl = *before) && posix_same_owner(request, fl)) {
983 /* Detect adjacent or overlapping regions (if same lock type)
984 */
985 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700986 /* In all comparisons of start vs end, use
987 * "start - 1" rather than "end + 1". If end
988 * is OFFSET_MAX, end + 1 will become negative.
989 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 if (fl->fl_end < request->fl_start - 1)
991 goto next_lock;
992 /* If the next lock in the list has entirely bigger
993 * addresses than the new one, insert the lock here.
994 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700995 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 break;
997
998 /* If we come here, the new and old lock are of the
999 * same type and adjacent or overlapping. Make one
1000 * lock yielding from the lower start address of both
1001 * locks to the higher end address.
1002 */
1003 if (fl->fl_start > request->fl_start)
1004 fl->fl_start = request->fl_start;
1005 else
1006 request->fl_start = fl->fl_start;
1007 if (fl->fl_end < request->fl_end)
1008 fl->fl_end = request->fl_end;
1009 else
1010 request->fl_end = fl->fl_end;
1011 if (added) {
Jeff Laytoned9814d2014-08-11 14:20:31 -04001012 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 continue;
1014 }
1015 request = fl;
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001016 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 }
1018 else {
1019 /* Processing for different lock types is a bit
1020 * more complex.
1021 */
1022 if (fl->fl_end < request->fl_start)
1023 goto next_lock;
1024 if (fl->fl_start > request->fl_end)
1025 break;
1026 if (request->fl_type == F_UNLCK)
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001027 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 if (fl->fl_start < request->fl_start)
1029 left = fl;
1030 /* If the next lock in the list has a higher end
1031 * address than the new one, insert the new one here.
1032 */
1033 if (fl->fl_end > request->fl_end) {
1034 right = fl;
1035 break;
1036 }
1037 if (fl->fl_start >= request->fl_start) {
1038 /* The new lock completely replaces an old
1039 * one (This may happen several times).
1040 */
1041 if (added) {
Jeff Laytoned9814d2014-08-11 14:20:31 -04001042 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 continue;
1044 }
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001045 /*
1046 * Replace the old lock with new_fl, and
1047 * remove the old one. It's safe to do the
1048 * insert here since we know that we won't be
1049 * using new_fl later, and that the lock is
1050 * just replacing an existing lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 */
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001052 error = -ENOLCK;
1053 if (!new_fl)
1054 goto out;
1055 locks_copy_lock(new_fl, request);
1056 request = new_fl;
1057 new_fl = NULL;
Jeff Laytoned9814d2014-08-11 14:20:31 -04001058 locks_delete_lock(before, &dispose);
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001059 locks_insert_lock(before, request);
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001060 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 }
1062 }
1063 /* Go on to next lock.
1064 */
1065 next_lock:
1066 before = &fl->fl_next;
1067 }
1068
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001069 /*
Jeff Layton1cb36012013-06-21 08:58:12 -04001070 * The above code only modifies existing locks in case of merging or
1071 * replacing. If new lock(s) need to be inserted all modifications are
1072 * done below this, so it's safe yet to bail out.
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001073 */
1074 error = -ENOLCK; /* "no luck" */
1075 if (right && left == right && !new_fl2)
1076 goto out;
1077
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 error = 0;
1079 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -04001080 if (request->fl_type == F_UNLCK) {
1081 if (request->fl_flags & FL_EXISTS)
1082 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -04001084 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001085
1086 if (!new_fl) {
1087 error = -ENOLCK;
1088 goto out;
1089 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 locks_copy_lock(new_fl, request);
1091 locks_insert_lock(before, new_fl);
1092 new_fl = NULL;
1093 }
1094 if (right) {
1095 if (left == right) {
1096 /* The new lock breaks the old one in two pieces,
1097 * so we have to use the second new lock.
1098 */
1099 left = new_fl2;
1100 new_fl2 = NULL;
1101 locks_copy_lock(left, right);
1102 locks_insert_lock(before, left);
1103 }
1104 right->fl_start = request->fl_end + 1;
1105 locks_wake_up_blocks(right);
1106 }
1107 if (left) {
1108 left->fl_end = request->fl_start - 1;
1109 locks_wake_up_blocks(left);
1110 }
1111 out:
Jeff Layton1c8c6012013-06-21 08:58:15 -04001112 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 /*
1114 * Free any unused locks.
1115 */
1116 if (new_fl)
1117 locks_free_lock(new_fl);
1118 if (new_fl2)
1119 locks_free_lock(new_fl2);
Jeff Laytoned9814d2014-08-11 14:20:31 -04001120 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 return error;
1122}
1123
1124/**
1125 * posix_lock_file - Apply a POSIX-style lock to a file
1126 * @filp: The file to apply the lock to
1127 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001128 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 *
1130 * Add a POSIX style lock to a file.
1131 * We merge adjacent & overlapping locks whenever possible.
1132 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001133 *
1134 * Note that if called with an FL_EXISTS argument, the caller may determine
1135 * whether or not a lock was successfully freed by testing the return
1136 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 */
Marc Eshel150b3932007-01-18 16:15:35 -05001138int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001139 struct file_lock *conflock)
1140{
Al Viro496ad9a2013-01-23 17:07:38 -05001141 return __posix_lock_file(file_inode(filp), fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001142}
Marc Eshel150b3932007-01-18 16:15:35 -05001143EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
1145/**
1146 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1147 * @filp: The file to apply the lock to
1148 * @fl: The lock to be applied
1149 *
1150 * Add a POSIX style lock to a file.
1151 * We merge adjacent & overlapping locks whenever possible.
1152 * POSIX locks are sorted by owner task, then by starting address
1153 */
1154int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1155{
1156 int error;
1157 might_sleep ();
1158 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001159 error = posix_lock_file(filp, fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001160 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 break;
1162 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1163 if (!error)
1164 continue;
1165
1166 locks_delete_block(fl);
1167 break;
1168 }
1169 return error;
1170}
1171EXPORT_SYMBOL(posix_lock_file_wait);
1172
1173/**
1174 * locks_mandatory_locked - Check for an active lock
Jeff Laytond7a06982014-03-10 09:54:15 -04001175 * @file: the file to check
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 *
1177 * Searches the inode's list of locks to find any POSIX locks which conflict.
1178 * This function is called from locks_verify_locked() only.
1179 */
Jeff Laytond7a06982014-03-10 09:54:15 -04001180int locks_mandatory_locked(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181{
Jeff Laytond7a06982014-03-10 09:54:15 -04001182 struct inode *inode = file_inode(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 struct file_lock *fl;
1184
1185 /*
1186 * Search the lock list for this inode for any POSIX locks.
1187 */
Jeff Layton1c8c6012013-06-21 08:58:15 -04001188 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1190 if (!IS_POSIX(fl))
1191 continue;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001192 if (fl->fl_owner != current->files &&
1193 fl->fl_owner != file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 break;
1195 }
Jeff Layton1c8c6012013-06-21 08:58:15 -04001196 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 return fl ? -EAGAIN : 0;
1198}
1199
1200/**
1201 * locks_mandatory_area - Check for a conflicting lock
1202 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1203 * for shared
1204 * @inode: the file to check
1205 * @filp: how the file was opened (if it was)
1206 * @offset: start of area to check
1207 * @count: length of area to check
1208 *
1209 * Searches the inode's list of locks to find any POSIX locks which conflict.
1210 * This function is called from rw_verify_area() and
1211 * locks_verify_truncate().
1212 */
1213int locks_mandatory_area(int read_write, struct inode *inode,
1214 struct file *filp, loff_t offset,
1215 size_t count)
1216{
1217 struct file_lock fl;
1218 int error;
Jeff Layton29723ad2014-03-10 09:54:19 -04001219 bool sleep = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220
1221 locks_init_lock(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 fl.fl_pid = current->tgid;
1223 fl.fl_file = filp;
1224 fl.fl_flags = FL_POSIX | FL_ACCESS;
1225 if (filp && !(filp->f_flags & O_NONBLOCK))
Jeff Layton29723ad2014-03-10 09:54:19 -04001226 sleep = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1228 fl.fl_start = offset;
1229 fl.fl_end = offset + count - 1;
1230
1231 for (;;) {
Jeff Layton29723ad2014-03-10 09:54:19 -04001232 if (filp) {
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001233 fl.fl_owner = filp;
Jeff Layton29723ad2014-03-10 09:54:19 -04001234 fl.fl_flags &= ~FL_SLEEP;
1235 error = __posix_lock_file(inode, &fl, NULL);
1236 if (!error)
1237 break;
1238 }
1239
1240 if (sleep)
1241 fl.fl_flags |= FL_SLEEP;
1242 fl.fl_owner = current->files;
Marc Eshel150b3932007-01-18 16:15:35 -05001243 error = __posix_lock_file(inode, &fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001244 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 break;
1246 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1247 if (!error) {
1248 /*
1249 * If we've been sleeping someone might have
1250 * changed the permissions behind our back.
1251 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001252 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 continue;
1254 }
1255
1256 locks_delete_block(&fl);
1257 break;
1258 }
1259
1260 return error;
1261}
1262
1263EXPORT_SYMBOL(locks_mandatory_area);
1264
J. Bruce Fields778fc542011-07-26 18:25:49 -04001265static void lease_clear_pending(struct file_lock *fl, int arg)
1266{
1267 switch (arg) {
1268 case F_UNLCK:
1269 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1270 /* fall through: */
1271 case F_RDLCK:
1272 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1273 }
1274}
1275
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276/* We already had a lease on this file; just change its type */
1277int lease_modify(struct file_lock **before, int arg)
1278{
1279 struct file_lock *fl = *before;
1280 int error = assign_type(fl, arg);
1281
1282 if (error)
1283 return error;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001284 lease_clear_pending(fl, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 locks_wake_up_blocks(fl);
Filipe Brandenburger3b6e2722012-07-27 00:42:52 -04001286 if (arg == F_UNLCK) {
1287 struct file *filp = fl->fl_file;
1288
1289 f_delown(filp);
1290 filp->f_owner.signum = 0;
J. Bruce Fields96d6d592012-07-27 16:18:00 -04001291 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1292 if (fl->fl_fasync != NULL) {
1293 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1294 fl->fl_fasync = NULL;
1295 }
Jeff Laytoned9814d2014-08-11 14:20:31 -04001296 locks_delete_lock(before, NULL);
Filipe Brandenburger3b6e2722012-07-27 00:42:52 -04001297 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 return 0;
1299}
1300
1301EXPORT_SYMBOL(lease_modify);
1302
J. Bruce Fields778fc542011-07-26 18:25:49 -04001303static bool past_time(unsigned long then)
1304{
1305 if (!then)
1306 /* 0 is a special value meaning "this never expires": */
1307 return false;
1308 return time_after(jiffies, then);
1309}
1310
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311static void time_out_leases(struct inode *inode)
1312{
1313 struct file_lock **before;
1314 struct file_lock *fl;
1315
1316 before = &inode->i_flock;
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04001317 while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) {
Jeff Layton62af4f12014-05-09 14:13:05 -04001318 trace_time_out_leases(inode, fl);
J. Bruce Fields778fc542011-07-26 18:25:49 -04001319 if (past_time(fl->fl_downgrade_time))
1320 lease_modify(before, F_RDLCK);
1321 if (past_time(fl->fl_break_time))
1322 lease_modify(before, F_UNLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 if (fl == *before) /* lease_modify may have freed fl */
1324 before = &fl->fl_next;
1325 }
1326}
1327
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001328static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1329{
1330 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1331 return false;
1332 return locks_conflict(breaker, lease);
1333}
1334
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335/**
1336 * __break_lease - revoke all outstanding leases on file
1337 * @inode: the inode of the file to return
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001338 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1339 * break all leases
1340 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1341 * only delegations
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 *
david m. richter87250dd2007-05-09 16:10:27 -04001343 * break_lease (inlined for speed) has checked there already is at least
1344 * some kind of lock (maybe a lease) on this file. Leases are broken on
1345 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 * specified %O_NONBLOCK to your open().
1347 */
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001348int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349{
J. Bruce Fields778fc542011-07-26 18:25:49 -04001350 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 struct file_lock *new_fl, *flock;
1352 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 unsigned long break_time;
1354 int i_have_this_lease = 0;
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001355 bool lease_conflict = false;
Al Viro8737c932009-12-24 06:47:55 -05001356 int want_write = (mode & O_ACCMODE) != O_RDONLY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357
Al Viro8737c932009-12-24 06:47:55 -05001358 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
Linus Torvalds6d4b9e32011-12-26 10:25:26 -08001359 if (IS_ERR(new_fl))
1360 return PTR_ERR(new_fl);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001361 new_fl->fl_flags = type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
Jeff Layton1c8c6012013-06-21 08:58:15 -04001363 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
1365 time_out_leases(inode);
1366
1367 flock = inode->i_flock;
1368 if ((flock == NULL) || !IS_LEASE(flock))
1369 goto out;
1370
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001371 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1372 if (leases_conflict(fl, new_fl)) {
1373 lease_conflict = true;
1374 if (fl->fl_owner == current->files)
1375 i_have_this_lease = 1;
1376 }
1377 }
1378 if (!lease_conflict)
J. Bruce Fields778fc542011-07-26 18:25:49 -04001379 goto out;
1380
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 break_time = 0;
1382 if (lease_break_time > 0) {
1383 break_time = jiffies + lease_break_time * HZ;
1384 if (break_time == 0)
1385 break_time++; /* so that 0 means no break time */
1386 }
1387
1388 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001389 if (!leases_conflict(fl, new_fl))
1390 continue;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001391 if (want_write) {
1392 if (fl->fl_flags & FL_UNLOCK_PENDING)
1393 continue;
1394 fl->fl_flags |= FL_UNLOCK_PENDING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 fl->fl_break_time = break_time;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001396 } else {
1397 if (lease_breaking(flock))
1398 continue;
1399 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1400 fl->fl_downgrade_time = break_time;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 }
J. Bruce Fields778fc542011-07-26 18:25:49 -04001402 fl->fl_lmops->lm_break(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 }
1404
1405 if (i_have_this_lease || (mode & O_NONBLOCK)) {
Jeff Layton62af4f12014-05-09 14:13:05 -04001406 trace_break_lease_noblock(inode, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 error = -EWOULDBLOCK;
1408 goto out;
1409 }
1410
1411restart:
1412 break_time = flock->fl_break_time;
Jeff Laytonf1c6bb22014-04-15 06:17:49 -04001413 if (break_time != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 break_time -= jiffies;
Jeff Laytonf1c6bb22014-04-15 06:17:49 -04001415 if (break_time == 0)
1416 break_time++;
Matthew Wilcox4321e012008-01-14 21:28:30 -07001417 locks_insert_block(flock, new_fl);
Jeff Layton62af4f12014-05-09 14:13:05 -04001418 trace_break_lease_block(inode, new_fl);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001419 spin_unlock(&inode->i_lock);
Matthew Wilcox4321e012008-01-14 21:28:30 -07001420 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1421 !new_fl->fl_next, break_time);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001422 spin_lock(&inode->i_lock);
Jeff Layton62af4f12014-05-09 14:13:05 -04001423 trace_break_lease_unblock(inode, new_fl);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001424 locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 if (error >= 0) {
1426 if (error == 0)
1427 time_out_leases(inode);
J. Bruce Fields778fc542011-07-26 18:25:49 -04001428 /*
1429 * Wait for the next conflicting lease that has not been
1430 * broken yet
1431 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1433 flock = flock->fl_next) {
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001434 if (leases_conflict(new_fl, flock))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 goto restart;
1436 }
1437 error = 0;
1438 }
1439
1440out:
Jeff Layton1c8c6012013-06-21 08:58:15 -04001441 spin_unlock(&inode->i_lock);
Linus Torvalds6d4b9e32011-12-26 10:25:26 -08001442 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 return error;
1444}
1445
1446EXPORT_SYMBOL(__break_lease);
1447
1448/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001449 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 * @inode: the inode
1451 * @time: pointer to a timespec which will contain the last modified time
1452 *
1453 * This is to force NFS clients to flush their caches for files with
1454 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001455 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 */
1457void lease_get_mtime(struct inode *inode, struct timespec *time)
1458{
Jeff Laytonbfe86022014-08-22 10:18:44 -04001459 bool has_lease = false;
1460 struct file_lock *flock;
1461
1462 if (inode->i_flock) {
1463 spin_lock(&inode->i_lock);
1464 flock = inode->i_flock;
1465 if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK))
1466 has_lease = true;
1467 spin_unlock(&inode->i_lock);
1468 }
1469
1470 if (has_lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 *time = current_fs_time(inode->i_sb);
1472 else
1473 *time = inode->i_mtime;
1474}
1475
1476EXPORT_SYMBOL(lease_get_mtime);
1477
1478/**
1479 * fcntl_getlease - Enquire what lease is currently active
1480 * @filp: the file
1481 *
1482 * The value returned by this function will be one of
1483 * (if no lease break is pending):
1484 *
1485 * %F_RDLCK to indicate a shared lease is held.
1486 *
1487 * %F_WRLCK to indicate an exclusive lease is held.
1488 *
1489 * %F_UNLCK to indicate no lease is held.
1490 *
1491 * (if a lease break is pending):
1492 *
1493 * %F_RDLCK to indicate an exclusive lease needs to be
1494 * changed to a shared lease (or removed).
1495 *
1496 * %F_UNLCK to indicate the lease needs to be removed.
1497 *
1498 * XXX: sfr & willy disagree over whether F_INPROGRESS
1499 * should be returned to userspace.
1500 */
1501int fcntl_getlease(struct file *filp)
1502{
1503 struct file_lock *fl;
Jeff Layton1c8c6012013-06-21 08:58:15 -04001504 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 int type = F_UNLCK;
1506
Jeff Layton1c8c6012013-06-21 08:58:15 -04001507 spin_lock(&inode->i_lock);
Al Viro496ad9a2013-01-23 17:07:38 -05001508 time_out_leases(file_inode(filp));
1509 for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 fl = fl->fl_next) {
1511 if (fl->fl_file == filp) {
J. Bruce Fields778fc542011-07-26 18:25:49 -04001512 type = target_leasetype(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 break;
1514 }
1515 }
Jeff Layton1c8c6012013-06-21 08:58:15 -04001516 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 return type;
1518}
1519
Jeff Layton24cbe782014-02-03 12:13:06 -05001520/**
1521 * check_conflicting_open - see if the given dentry points to a file that has
1522 * an existing open that would conflict with the
1523 * desired lease.
1524 * @dentry: dentry to check
1525 * @arg: type of lease that we're trying to acquire
1526 *
1527 * Check to see if there's an existing open fd on this file that would
1528 * conflict with the lease we're trying to set.
1529 */
1530static int
1531check_conflicting_open(const struct dentry *dentry, const long arg)
1532{
1533 int ret = 0;
1534 struct inode *inode = dentry->d_inode;
1535
1536 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1537 return -EAGAIN;
1538
1539 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1540 (atomic_read(&inode->i_count) > 1)))
1541 ret = -EAGAIN;
1542
1543 return ret;
1544}
1545
Jeff Laytond4f22d12013-06-21 08:58:11 -04001546static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001548 struct file_lock *fl, **before, **my_before = NULL, *lease;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001549 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 struct inode *inode = dentry->d_inode;
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001551 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001552 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553
J. Bruce Fields096657b2010-10-30 17:31:14 -04001554 lease = *flp;
Jeff Layton62af4f12014-05-09 14:13:05 -04001555 trace_generic_add_lease(inode, lease);
1556
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001557 /*
1558 * In the delegation case we need mutual exclusion with
1559 * a number of operations that take the i_mutex. We trylock
1560 * because delegations are an optional optimization, and if
1561 * there's some chance of a conflict--we'd rather not
1562 * bother, maybe that's a sign this just isn't a good file to
1563 * hand out a delegation on.
1564 */
1565 if (is_deleg && !mutex_trylock(&inode->i_mutex))
1566 return -EAGAIN;
1567
1568 if (is_deleg && arg == F_WRLCK) {
1569 /* Write delegations are not currently supported: */
Dan Carpenter4fdb7932013-11-13 10:56:27 +03001570 mutex_unlock(&inode->i_mutex);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001571 WARN_ON_ONCE(1);
1572 return -EINVAL;
1573 }
J. Bruce Fields096657b2010-10-30 17:31:14 -04001574
Jeff Layton24cbe782014-02-03 12:13:06 -05001575 error = check_conflicting_open(dentry, arg);
1576 if (error)
J. Bruce Fields096657b2010-10-30 17:31:14 -04001577 goto out;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001578
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 /*
1580 * At this point, we know that if there is an exclusive
1581 * lease on this file, then we hold it on this filp
1582 * (otherwise our open of this file would have blocked).
1583 * And if we are trying to acquire an exclusive lease,
1584 * then the file is not open by anyone (including us)
1585 * except for this filp.
1586 */
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001587 error = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 for (before = &inode->i_flock;
1589 ((fl = *before) != NULL) && IS_LEASE(fl);
1590 before = &fl->fl_next) {
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001591 if (fl->fl_file == filp) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 my_before = before;
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001593 continue;
1594 }
1595 /*
1596 * No exclusive leases if someone else has a lease on
1597 * this file:
1598 */
1599 if (arg == F_WRLCK)
1600 goto out;
1601 /*
1602 * Modifying our existing lease is OK, but no getting a
1603 * new lease if someone else is opening for write:
1604 */
1605 if (fl->fl_flags & FL_UNLOCK_PENDING)
1606 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 }
1608
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 if (my_before != NULL) {
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04001610 error = lease->fl_lmops->lm_change(my_before, arg);
Christoph Hellwig51ee4b82010-10-31 08:35:10 -04001611 if (!error)
1612 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 goto out;
1614 }
1615
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 error = -EINVAL;
1617 if (!leases_enable)
1618 goto out;
1619
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +02001620 locks_insert_lock(before, lease);
Jeff Layton24cbe782014-02-03 12:13:06 -05001621 /*
1622 * The check in break_lease() is lockless. It's possible for another
1623 * open to race in after we did the earlier check for a conflicting
1624 * open but before the lease was inserted. Check again for a
1625 * conflicting open and cancel the lease if there is one.
1626 *
1627 * We also add a barrier here to ensure that the insertion of the lock
1628 * precedes these checks.
1629 */
1630 smp_mb();
1631 error = check_conflicting_open(dentry, arg);
1632 if (error)
Jeff Laytonf39b9132014-09-09 16:00:51 -04001633 locks_unlink_lock(before);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634out:
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001635 if (is_deleg)
1636 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 return error;
1638}
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001639
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001640static int generic_delete_lease(struct file *filp)
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001641{
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001642 int error = -EAGAIN;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001643 struct file_lock *fl, **before;
1644 struct dentry *dentry = filp->f_path.dentry;
1645 struct inode *inode = dentry->d_inode;
1646
1647 for (before = &inode->i_flock;
1648 ((fl = *before) != NULL) && IS_LEASE(fl);
1649 before = &fl->fl_next) {
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001650 if (fl->fl_file == filp)
1651 break;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001652 }
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001653 trace_generic_delete_lease(inode, fl);
1654 if (fl)
1655 error = fl->fl_lmops->lm_change(before, F_UNLCK);
1656 return error;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001657}
1658
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659/**
1660 * generic_setlease - sets a lease on an open file
1661 * @filp: file pointer
1662 * @arg: type of lease to obtain
1663 * @flp: input - file_lock to use, output - file_lock inserted
1664 *
1665 * The (input) flp->fl_lmops->lm_break function is required
1666 * by break_lease().
1667 *
Jeff Layton1c8c6012013-06-21 08:58:15 -04001668 * Called with inode->i_lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 */
1670int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
1671{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 struct dentry *dentry = filp->f_path.dentry;
1673 struct inode *inode = dentry->d_inode;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001674 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001676 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001677 return -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 if (!S_ISREG(inode->i_mode))
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001679 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 error = security_file_lock(filp, arg);
1681 if (error)
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001682 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683
1684 time_out_leases(inode);
1685
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001686 switch (arg) {
1687 case F_UNLCK:
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001688 return generic_delete_lease(filp);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001689 case F_RDLCK:
1690 case F_WRLCK:
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001691 if (!(*flp)->fl_lmops->lm_break) {
1692 WARN_ON_ONCE(1);
1693 return -ENOLCK;
1694 }
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001695 return generic_add_lease(filp, arg, flp);
1696 default:
Dave Jones8d657eb2012-07-13 13:35:36 -04001697 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001700EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
Arnd Bergmannb89f4322010-09-18 15:09:31 +02001702static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
1703{
Al Viro72c2d532013-09-22 16:27:52 -04001704 if (filp->f_op->setlease)
Arnd Bergmannb89f4322010-09-18 15:09:31 +02001705 return filp->f_op->setlease(filp, arg, lease);
1706 else
1707 return generic_setlease(filp, arg, lease);
1708}
1709
1710/**
Jeff Laytone51673a2014-08-22 18:13:28 -04001711 * vfs_setlease - sets a lease on an open file
1712 * @filp: file pointer
1713 * @arg: type of lease to obtain
1714 * @lease: file_lock to use when adding a lease
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 *
Jeff Laytone51673a2014-08-22 18:13:28 -04001716 * Call this to establish a lease on the file. The "lease" argument is not
1717 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1718 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1719 * if not, this function will return -ENOLCK (and generate a scary-looking
1720 * stack trace).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 */
1722
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001723int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724{
Jeff Layton1c8c6012013-06-21 08:58:15 -04001725 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 int error;
1727
Jeff Layton1c8c6012013-06-21 08:58:15 -04001728 spin_lock(&inode->i_lock);
Arnd Bergmannb89f4322010-09-18 15:09:31 +02001729 error = __vfs_setlease(filp, arg, lease);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001730 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732 return error;
1733}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001734EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001736static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737{
J. Bruce Fields3df057a2010-11-03 16:49:44 -04001738 struct file_lock *fl, *ret;
Jeff Layton1c8c6012013-06-21 08:58:15 -04001739 struct inode *inode = file_inode(filp);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001740 struct fasync_struct *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 int error;
1742
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +02001743 fl = lease_alloc(filp, arg);
1744 if (IS_ERR(fl))
1745 return PTR_ERR(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001747 new = fasync_alloc();
1748 if (!new) {
1749 locks_free_lock(fl);
1750 return -ENOMEM;
1751 }
J. Bruce Fields3df057a2010-11-03 16:49:44 -04001752 ret = fl;
Jeff Layton1c8c6012013-06-21 08:58:15 -04001753 spin_lock(&inode->i_lock);
J. Bruce Fields8896b932010-11-03 18:09:18 -04001754 error = __vfs_setlease(filp, arg, &ret);
Jeff Layton2dfb9282014-08-11 18:14:12 -04001755 if (error)
1756 goto out_unlock;
1757 if (ret == fl)
1758 fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001760 /*
1761 * fasync_insert_entry() returns the old entry if any.
1762 * If there was no old entry, then it used 'new' and
1763 * inserted it into the fasync list. Clear new so that
1764 * we don't release it here.
1765 */
J. Bruce Fields3df057a2010-11-03 16:49:44 -04001766 if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new))
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001767 new = NULL;
1768
Jeff Laytone0b93ed2014-08-22 11:27:32 -04001769 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Jeff Layton2dfb9282014-08-11 18:14:12 -04001770out_unlock:
Jeff Layton1c8c6012013-06-21 08:58:15 -04001771 spin_unlock(&inode->i_lock);
Jeff Layton2dfb9282014-08-11 18:14:12 -04001772 if (fl)
1773 locks_free_lock(fl);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001774 if (new)
1775 fasync_free(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 return error;
1777}
1778
1779/**
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001780 * fcntl_setlease - sets a lease on an open file
1781 * @fd: open file descriptor
1782 * @filp: file pointer
1783 * @arg: type of lease to obtain
1784 *
1785 * Call this fcntl to establish a lease on the file.
1786 * Note that you also need to call %F_SETSIG to
1787 * receive a signal when the lease is broken.
1788 */
1789int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1790{
1791 if (arg == F_UNLCK)
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001792 return vfs_setlease(filp, F_UNLCK, NULL);
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001793 return do_fcntl_add_lease(fd, filp, arg);
1794}
1795
1796/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1798 * @filp: The file to apply the lock to
1799 * @fl: The lock to be applied
1800 *
1801 * Add a FLOCK style lock to a file.
1802 */
1803int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1804{
1805 int error;
1806 might_sleep();
1807 for (;;) {
1808 error = flock_lock_file(filp, fl);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001809 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 break;
1811 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1812 if (!error)
1813 continue;
1814
1815 locks_delete_block(fl);
1816 break;
1817 }
1818 return error;
1819}
1820
1821EXPORT_SYMBOL(flock_lock_file_wait);
1822
1823/**
1824 * sys_flock: - flock() system call.
1825 * @fd: the file descriptor to lock.
1826 * @cmd: the type of lock to apply.
1827 *
1828 * Apply a %FL_FLOCK style lock to an open file descriptor.
1829 * The @cmd can be one of
1830 *
1831 * %LOCK_SH -- a shared lock.
1832 *
1833 * %LOCK_EX -- an exclusive lock.
1834 *
1835 * %LOCK_UN -- remove an existing lock.
1836 *
1837 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1838 *
1839 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1840 * processes read and write access respectively.
1841 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001842SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843{
Al Viro2903ff02012-08-28 12:52:22 -04001844 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 struct file_lock *lock;
1846 int can_sleep, unlock;
1847 int error;
1848
1849 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001850 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 goto out;
1852
1853 can_sleep = !(cmd & LOCK_NB);
1854 cmd &= ~LOCK_NB;
1855 unlock = (cmd == LOCK_UN);
1856
Al Viroaeb5d722008-09-02 15:28:45 -04001857 if (!unlock && !(cmd & LOCK_MAND) &&
Al Viro2903ff02012-08-28 12:52:22 -04001858 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 goto out_putf;
1860
Al Viro2903ff02012-08-28 12:52:22 -04001861 error = flock_make_lock(f.file, &lock, cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 if (error)
1863 goto out_putf;
1864 if (can_sleep)
1865 lock->fl_flags |= FL_SLEEP;
1866
Al Viro2903ff02012-08-28 12:52:22 -04001867 error = security_file_lock(f.file, lock->fl_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 if (error)
1869 goto out_free;
1870
Al Viro72c2d532013-09-22 16:27:52 -04001871 if (f.file->f_op->flock)
Al Viro2903ff02012-08-28 12:52:22 -04001872 error = f.file->f_op->flock(f.file,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 (can_sleep) ? F_SETLKW : F_SETLK,
1874 lock);
1875 else
Al Viro2903ff02012-08-28 12:52:22 -04001876 error = flock_lock_file_wait(f.file, lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877
1878 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001879 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
1881 out_putf:
Al Viro2903ff02012-08-28 12:52:22 -04001882 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 out:
1884 return error;
1885}
1886
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001887/**
1888 * vfs_test_lock - test file byte range lock
1889 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001890 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001891 *
1892 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1893 * setting conf->fl_type to something other than F_UNLCK.
1894 */
1895int vfs_test_lock(struct file *filp, struct file_lock *fl)
1896{
Al Viro72c2d532013-09-22 16:27:52 -04001897 if (filp->f_op->lock)
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001898 return filp->f_op->lock(filp, F_GETLK, fl);
1899 posix_test_lock(filp, fl);
1900 return 0;
1901}
1902EXPORT_SYMBOL_GPL(vfs_test_lock);
1903
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001904static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1905{
Jeff Laytoncff2fce2014-04-22 08:24:32 -04001906 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001907#if BITS_PER_LONG == 32
1908 /*
1909 * Make sure we can represent the posix lock via
1910 * legacy 32bit flock.
1911 */
1912 if (fl->fl_start > OFFT_OFFSET_MAX)
1913 return -EOVERFLOW;
1914 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1915 return -EOVERFLOW;
1916#endif
1917 flock->l_start = fl->fl_start;
1918 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1919 fl->fl_end - fl->fl_start + 1;
1920 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001921 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001922 return 0;
1923}
1924
1925#if BITS_PER_LONG == 32
1926static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1927{
Jeff Laytoncff2fce2014-04-22 08:24:32 -04001928 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001929 flock->l_start = fl->fl_start;
1930 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1931 fl->fl_end - fl->fl_start + 1;
1932 flock->l_whence = 0;
1933 flock->l_type = fl->fl_type;
1934}
1935#endif
1936
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937/* Report the first existing lock that would conflict with l.
1938 * This implements the F_GETLK command of fcntl().
1939 */
Jeff Laytonc1e62b82014-02-03 12:13:09 -05001940int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001942 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 struct flock flock;
1944 int error;
1945
1946 error = -EFAULT;
1947 if (copy_from_user(&flock, l, sizeof(flock)))
1948 goto out;
1949 error = -EINVAL;
1950 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1951 goto out;
1952
1953 error = flock_to_posix_lock(filp, &file_lock, &flock);
1954 if (error)
1955 goto out;
1956
Jeff Layton0d3f7a22014-04-22 08:23:58 -04001957 if (cmd == F_OFD_GETLK) {
Jeff Layton90478932014-03-04 10:30:23 -05001958 error = -EINVAL;
1959 if (flock.l_pid != 0)
1960 goto out;
1961
Jeff Layton5d50ffd2014-02-03 12:13:10 -05001962 cmd = F_GETLK;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04001963 file_lock.fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001964 file_lock.fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05001965 }
1966
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001967 error = vfs_test_lock(filp, &file_lock);
1968 if (error)
1969 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001971 flock.l_type = file_lock.fl_type;
1972 if (file_lock.fl_type != F_UNLCK) {
1973 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001974 if (error)
Kinglong Meef3282962014-08-22 10:18:43 -04001975 goto rel_priv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 }
1977 error = -EFAULT;
1978 if (!copy_to_user(l, &flock, sizeof(flock)))
1979 error = 0;
Kinglong Meef3282962014-08-22 10:18:43 -04001980rel_priv:
1981 locks_release_private(&file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982out:
1983 return error;
1984}
1985
Marc Eshel7723ec92007-01-18 15:08:55 -05001986/**
1987 * vfs_lock_file - file byte range lock
1988 * @filp: The file to apply the lock to
1989 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1990 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001991 * @conf: Place to return a copy of the conflicting lock, if found.
1992 *
1993 * A caller that doesn't care about the conflicting lock may pass NULL
1994 * as the final argument.
1995 *
1996 * If the filesystem defines a private ->lock() method, then @conf will
1997 * be left unchanged; so a caller that cares should initialize it to
1998 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001999 *
2000 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2001 * locks, the ->lock() interface may return asynchronously, before the lock has
2002 * been granted or denied by the underlying filesystem, if (and only if)
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002003 * lm_grant is set. Callers expecting ->lock() to return asynchronously
Marc Eshel2beb6612006-12-05 23:31:28 -05002004 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2005 * the request is for a blocking lock. When ->lock() does return asynchronously,
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002006 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
Marc Eshel2beb6612006-12-05 23:31:28 -05002007 * request completes.
2008 * If the request is for non-blocking lock the file system should return
Miklos Szeredibde74e42008-07-25 01:48:57 -07002009 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2010 * with the result. If the request timed out the callback routine will return a
Marc Eshel2beb6612006-12-05 23:31:28 -05002011 * nonzero return code and the file system should release the lock. The file
2012 * system is also responsible to keep a corresponding posix lock when it
2013 * grants a lock so the VFS can find out which locks are locally held and do
2014 * the correct lock cleanup when required.
2015 * The underlying filesystem must not drop the kernel lock or call
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002016 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
Marc Eshel2beb6612006-12-05 23:31:28 -05002017 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05002018 */
Marc Eshel150b3932007-01-18 16:15:35 -05002019int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05002020{
Al Viro72c2d532013-09-22 16:27:52 -04002021 if (filp->f_op->lock)
Marc Eshel7723ec92007-01-18 15:08:55 -05002022 return filp->f_op->lock(filp, cmd, fl);
2023 else
Marc Eshel150b3932007-01-18 16:15:35 -05002024 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05002025}
2026EXPORT_SYMBOL_GPL(vfs_lock_file);
2027
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002028static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2029 struct file_lock *fl)
2030{
2031 int error;
2032
2033 error = security_file_lock(filp, fl->fl_type);
2034 if (error)
2035 return error;
2036
Miklos Szeredi764c76b2008-07-25 01:48:58 -07002037 for (;;) {
2038 error = vfs_lock_file(filp, cmd, fl, NULL);
2039 if (error != FILE_LOCK_DEFERRED)
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002040 break;
Miklos Szeredi764c76b2008-07-25 01:48:58 -07002041 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2042 if (!error)
2043 continue;
2044
2045 locks_delete_block(fl);
2046 break;
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002047 }
2048
2049 return error;
2050}
2051
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002052/* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2053static int
2054check_fmode_for_setlk(struct file_lock *fl)
2055{
2056 switch (fl->fl_type) {
2057 case F_RDLCK:
2058 if (!(fl->fl_file->f_mode & FMODE_READ))
2059 return -EBADF;
2060 break;
2061 case F_WRLCK:
2062 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2063 return -EBADF;
2064 }
2065 return 0;
2066}
2067
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068/* Apply the lock described by l to an open file descriptor.
2069 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2070 */
Peter Staubachc2936212005-07-27 11:45:09 -07002071int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2072 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073{
2074 struct file_lock *file_lock = locks_alloc_lock();
2075 struct flock flock;
2076 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04002077 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 int error;
2079
2080 if (file_lock == NULL)
2081 return -ENOLCK;
2082
2083 /*
2084 * This might block, so we do it before checking the inode.
2085 */
2086 error = -EFAULT;
2087 if (copy_from_user(&flock, l, sizeof(flock)))
2088 goto out;
2089
Al Viro496ad9a2013-01-23 17:07:38 -05002090 inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091
2092 /* Don't allow mandatory locks on files that may be memory mapped
2093 * and shared.
2094 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002095 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 error = -EAGAIN;
2097 goto out;
2098 }
2099
Peter Staubachc2936212005-07-27 11:45:09 -07002100again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 error = flock_to_posix_lock(filp, file_lock, &flock);
2102 if (error)
2103 goto out;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002104
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002105 error = check_fmode_for_setlk(file_lock);
2106 if (error)
2107 goto out;
2108
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002109 /*
2110 * If the cmd is requesting file-private locks, then set the
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002111 * FL_OFDLCK flag and override the owner.
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002112 */
2113 switch (cmd) {
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002114 case F_OFD_SETLK:
Jeff Layton90478932014-03-04 10:30:23 -05002115 error = -EINVAL;
2116 if (flock.l_pid != 0)
2117 goto out;
2118
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002119 cmd = F_SETLK;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002120 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002121 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002122 break;
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002123 case F_OFD_SETLKW:
Jeff Layton90478932014-03-04 10:30:23 -05002124 error = -EINVAL;
2125 if (flock.l_pid != 0)
2126 goto out;
2127
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002128 cmd = F_SETLKW;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002129 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002130 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002131 /* Fallthrough */
2132 case F_SETLKW:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 file_lock->fl_flags |= FL_SLEEP;
2134 }
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002135
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002136 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137
Peter Staubachc2936212005-07-27 11:45:09 -07002138 /*
2139 * Attempt to detect a close/fcntl race and recover by
2140 * releasing the lock that was just acquired.
2141 */
Al Viro0b2bac22008-05-06 13:58:34 -04002142 /*
2143 * we need that spin_lock here - it prevents reordering between
2144 * update of inode->i_flock and check for it done in close().
2145 * rcu_read_lock() wouldn't do.
2146 */
2147 spin_lock(&current->files->file_lock);
2148 f = fcheck(fd);
2149 spin_unlock(&current->files->file_lock);
2150 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07002151 flock.l_type = F_UNLCK;
2152 goto again;
2153 }
2154
2155out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 locks_free_lock(file_lock);
2157 return error;
2158}
2159
2160#if BITS_PER_LONG == 32
2161/* Report the first existing lock that would conflict with l.
2162 * This implements the F_GETLK command of fcntl().
2163 */
Jeff Laytonc1e62b82014-02-03 12:13:09 -05002164int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002166 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167 struct flock64 flock;
2168 int error;
2169
2170 error = -EFAULT;
2171 if (copy_from_user(&flock, l, sizeof(flock)))
2172 goto out;
2173 error = -EINVAL;
2174 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2175 goto out;
2176
2177 error = flock64_to_posix_lock(filp, &file_lock, &flock);
2178 if (error)
2179 goto out;
2180
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002181 if (cmd == F_OFD_GETLK) {
Jeff Layton90478932014-03-04 10:30:23 -05002182 error = -EINVAL;
2183 if (flock.l_pid != 0)
2184 goto out;
2185
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002186 cmd = F_GETLK64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002187 file_lock.fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002188 file_lock.fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002189 }
2190
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002191 error = vfs_test_lock(filp, &file_lock);
2192 if (error)
2193 goto out;
2194
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002195 flock.l_type = file_lock.fl_type;
2196 if (file_lock.fl_type != F_UNLCK)
2197 posix_lock_to_flock64(&flock, &file_lock);
2198
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199 error = -EFAULT;
2200 if (!copy_to_user(l, &flock, sizeof(flock)))
2201 error = 0;
Kinglong Meef3282962014-08-22 10:18:43 -04002202
2203 locks_release_private(&file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204out:
2205 return error;
2206}
2207
2208/* Apply the lock described by l to an open file descriptor.
2209 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2210 */
Peter Staubachc2936212005-07-27 11:45:09 -07002211int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2212 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213{
2214 struct file_lock *file_lock = locks_alloc_lock();
2215 struct flock64 flock;
2216 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04002217 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 int error;
2219
2220 if (file_lock == NULL)
2221 return -ENOLCK;
2222
2223 /*
2224 * This might block, so we do it before checking the inode.
2225 */
2226 error = -EFAULT;
2227 if (copy_from_user(&flock, l, sizeof(flock)))
2228 goto out;
2229
Al Viro496ad9a2013-01-23 17:07:38 -05002230 inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231
2232 /* Don't allow mandatory locks on files that may be memory mapped
2233 * and shared.
2234 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002235 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236 error = -EAGAIN;
2237 goto out;
2238 }
2239
Peter Staubachc2936212005-07-27 11:45:09 -07002240again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 error = flock64_to_posix_lock(filp, file_lock, &flock);
2242 if (error)
2243 goto out;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002244
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002245 error = check_fmode_for_setlk(file_lock);
2246 if (error)
2247 goto out;
2248
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002249 /*
2250 * If the cmd is requesting file-private locks, then set the
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002251 * FL_OFDLCK flag and override the owner.
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002252 */
2253 switch (cmd) {
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002254 case F_OFD_SETLK:
Jeff Layton90478932014-03-04 10:30:23 -05002255 error = -EINVAL;
2256 if (flock.l_pid != 0)
2257 goto out;
2258
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002259 cmd = F_SETLK64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002260 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002261 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002262 break;
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002263 case F_OFD_SETLKW:
Jeff Layton90478932014-03-04 10:30:23 -05002264 error = -EINVAL;
2265 if (flock.l_pid != 0)
2266 goto out;
2267
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002268 cmd = F_SETLKW64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002269 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002270 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002271 /* Fallthrough */
2272 case F_SETLKW64:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002273 file_lock->fl_flags |= FL_SLEEP;
2274 }
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002275
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002276 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277
Peter Staubachc2936212005-07-27 11:45:09 -07002278 /*
2279 * Attempt to detect a close/fcntl race and recover by
2280 * releasing the lock that was just acquired.
2281 */
Al Viro0b2bac22008-05-06 13:58:34 -04002282 spin_lock(&current->files->file_lock);
2283 f = fcheck(fd);
2284 spin_unlock(&current->files->file_lock);
2285 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07002286 flock.l_type = F_UNLCK;
2287 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288 }
2289
2290out:
2291 locks_free_lock(file_lock);
2292 return error;
2293}
2294#endif /* BITS_PER_LONG == 32 */
2295
2296/*
2297 * This function is called when the file is being removed
2298 * from the task's fd array. POSIX locks belonging to this task
2299 * are deleted at this time.
2300 */
2301void locks_remove_posix(struct file *filp, fl_owner_t owner)
2302{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07002303 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304
2305 /*
2306 * If there are no locks held on this file, we don't need to call
2307 * posix_lock_file(). Another process could be setting a lock on this
2308 * file at the same time, but we wouldn't remove that lock anyway.
2309 */
Al Viro496ad9a2013-01-23 17:07:38 -05002310 if (!file_inode(filp)->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 return;
2312
2313 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07002314 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315 lock.fl_start = 0;
2316 lock.fl_end = OFFSET_MAX;
2317 lock.fl_owner = owner;
2318 lock.fl_pid = current->tgid;
2319 lock.fl_file = filp;
2320 lock.fl_ops = NULL;
2321 lock.fl_lmops = NULL;
2322
Marc Eshel150b3932007-01-18 16:15:35 -05002323 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2326 lock.fl_ops->fl_release_private(&lock);
2327}
2328
2329EXPORT_SYMBOL(locks_remove_posix);
2330
2331/*
2332 * This function is called on the last close of an open file.
2333 */
Jeff Layton78ed8a12014-02-03 12:13:08 -05002334void locks_remove_file(struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335{
Al Viro496ad9a2013-01-23 17:07:38 -05002336 struct inode * inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337 struct file_lock *fl;
2338 struct file_lock **before;
Jeff Laytoned9814d2014-08-11 14:20:31 -04002339 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340
2341 if (!inode->i_flock)
2342 return;
2343
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002344 locks_remove_posix(filp, filp);
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002345
Al Viro72c2d532013-09-22 16:27:52 -04002346 if (filp->f_op->flock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347 struct file_lock fl = {
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002348 .fl_owner = filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002349 .fl_pid = current->tgid,
2350 .fl_file = filp,
2351 .fl_flags = FL_FLOCK,
2352 .fl_type = F_UNLCK,
2353 .fl_end = OFFSET_MAX,
2354 };
2355 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002356 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2357 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 }
2359
Jeff Layton1c8c6012013-06-21 08:58:15 -04002360 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361 before = &inode->i_flock;
2362
2363 while ((fl = *before) != NULL) {
2364 if (fl->fl_file == filp) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365 if (IS_LEASE(fl)) {
2366 lease_modify(before, F_UNLCK);
2367 continue;
2368 }
Jeff Layton8c3cac52014-02-03 12:13:07 -05002369
2370 /*
2371 * There's a leftover lock on the list of a type that
2372 * we didn't expect to see. Most likely a classic
2373 * POSIX lock that ended up not getting released
2374 * properly, or that raced onto the list somehow. Log
2375 * some info about it and then just remove it from
2376 * the list.
2377 */
2378 WARN(!IS_FLOCK(fl),
2379 "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n",
2380 MAJOR(inode->i_sb->s_dev),
2381 MINOR(inode->i_sb->s_dev), inode->i_ino,
2382 fl->fl_type, fl->fl_flags,
2383 fl->fl_start, fl->fl_end);
2384
Jeff Laytoned9814d2014-08-11 14:20:31 -04002385 locks_delete_lock(before, &dispose);
Jeff Layton8c3cac52014-02-03 12:13:07 -05002386 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387 }
2388 before = &fl->fl_next;
2389 }
Jeff Layton1c8c6012013-06-21 08:58:15 -04002390 spin_unlock(&inode->i_lock);
Jeff Laytoned9814d2014-08-11 14:20:31 -04002391 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392}
2393
2394/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395 * posix_unblock_lock - stop waiting for a file lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396 * @waiter: the lock which was waiting
2397 *
2398 * lockd needs to block waiting for locks.
2399 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002400int
Jeff Laytonf891a292013-06-21 08:58:09 -04002401posix_unblock_lock(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002403 int status = 0;
2404
Jeff Layton7b2296a2013-06-21 08:58:20 -04002405 spin_lock(&blocked_lock_lock);
J. Bruce Fields5996a292006-01-03 09:55:44 +01002406 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002408 else
2409 status = -ENOENT;
Jeff Layton7b2296a2013-06-21 08:58:20 -04002410 spin_unlock(&blocked_lock_lock);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002411 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002413EXPORT_SYMBOL(posix_unblock_lock);
2414
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002415/**
2416 * vfs_cancel_lock - file byte range unblock lock
2417 * @filp: The file to apply the unblock to
2418 * @fl: The lock to be unblocked
2419 *
2420 * Used by lock managers to cancel blocked requests
2421 */
2422int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2423{
Al Viro72c2d532013-09-22 16:27:52 -04002424 if (filp->f_op->lock)
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002425 return filp->f_op->lock(filp, F_CANCELLK, fl);
2426 return 0;
2427}
2428
2429EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2430
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002431#ifdef CONFIG_PROC_FS
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002432#include <linux/proc_fs.h>
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002433#include <linux/seq_file.h>
2434
Jeff Layton7012b022013-06-21 08:58:22 -04002435struct locks_iterator {
2436 int li_cpu;
2437 loff_t li_pos;
2438};
2439
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002440static void lock_get_status(struct seq_file *f, struct file_lock *fl,
Jerome Marchand99dc8292010-10-26 14:22:33 -07002441 loff_t id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442{
2443 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002444 unsigned int fl_pid;
2445
2446 if (fl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08002447 fl_pid = pid_vnr(fl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002448 else
2449 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450
2451 if (fl->fl_file != NULL)
Al Viro496ad9a2013-01-23 17:07:38 -05002452 inode = file_inode(fl->fl_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453
Jerome Marchand99dc8292010-10-26 14:22:33 -07002454 seq_printf(f, "%lld:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 if (IS_POSIX(fl)) {
Jeff Laytonc918d422014-02-03 12:13:09 -05002456 if (fl->fl_flags & FL_ACCESS)
Fabian Frederick5315c262014-05-09 14:13:05 -04002457 seq_puts(f, "ACCESS");
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002458 else if (IS_OFDLCK(fl))
Fabian Frederick5315c262014-05-09 14:13:05 -04002459 seq_puts(f, "OFDLCK");
Jeff Laytonc918d422014-02-03 12:13:09 -05002460 else
Fabian Frederick5315c262014-05-09 14:13:05 -04002461 seq_puts(f, "POSIX ");
Jeff Laytonc918d422014-02-03 12:13:09 -05002462
2463 seq_printf(f, " %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002464 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002465 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466 } else if (IS_FLOCK(fl)) {
2467 if (fl->fl_type & LOCK_MAND) {
Fabian Frederick5315c262014-05-09 14:13:05 -04002468 seq_puts(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002470 seq_puts(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471 }
2472 } else if (IS_LEASE(fl)) {
Jeff Layton8144f1f2014-08-11 13:36:54 -04002473 if (fl->fl_flags & FL_DELEG)
2474 seq_puts(f, "DELEG ");
2475 else
2476 seq_puts(f, "LEASE ");
2477
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04002478 if (lease_breaking(fl))
Fabian Frederick5315c262014-05-09 14:13:05 -04002479 seq_puts(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480 else if (fl->fl_file)
Fabian Frederick5315c262014-05-09 14:13:05 -04002481 seq_puts(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 else
Fabian Frederick5315c262014-05-09 14:13:05 -04002483 seq_puts(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002485 seq_puts(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486 }
2487 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002488 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002489 (fl->fl_type & LOCK_READ)
2490 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2491 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2492 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002493 seq_printf(f, "%s ",
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04002494 (lease_breaking(fl))
Jeff Layton0ee5c6d2012-08-02 15:46:30 -04002495 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2496 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497 }
2498 if (inode) {
2499#ifdef WE_CAN_BREAK_LSLK_NOW
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002500 seq_printf(f, "%d %s:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002501 inode->i_sb->s_id, inode->i_ino);
2502#else
2503 /* userspace relies on this representation of dev_t ;-( */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002504 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002505 MAJOR(inode->i_sb->s_dev),
2506 MINOR(inode->i_sb->s_dev), inode->i_ino);
2507#endif
2508 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002509 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510 }
2511 if (IS_POSIX(fl)) {
2512 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002513 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002515 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002517 seq_puts(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518 }
2519}
2520
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002521static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522{
Jeff Layton7012b022013-06-21 08:58:22 -04002523 struct locks_iterator *iter = f->private;
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002524 struct file_lock *fl, *bfl;
2525
Jeff Layton139ca042013-06-21 08:58:17 -04002526 fl = hlist_entry(v, struct file_lock, fl_link);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002527
Jeff Layton7012b022013-06-21 08:58:22 -04002528 lock_get_status(f, fl, iter->li_pos, "");
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002529
2530 list_for_each_entry(bfl, &fl->fl_block, fl_block)
Jeff Layton7012b022013-06-21 08:58:22 -04002531 lock_get_status(f, bfl, iter->li_pos, " ->");
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002532
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002533 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534}
2535
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002536static void *locks_start(struct seq_file *f, loff_t *pos)
Jeff Laytonb03dfdec2014-02-03 12:13:07 -05002537 __acquires(&blocked_lock_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538{
Jeff Layton7012b022013-06-21 08:58:22 -04002539 struct locks_iterator *iter = f->private;
Jerome Marchand99dc8292010-10-26 14:22:33 -07002540
Jeff Layton7012b022013-06-21 08:58:22 -04002541 iter->li_pos = *pos + 1;
2542 lg_global_lock(&file_lock_lglock);
Jeff Layton7b2296a2013-06-21 08:58:20 -04002543 spin_lock(&blocked_lock_lock);
Jeff Layton7012b022013-06-21 08:58:22 -04002544 return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545}
2546
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002547static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2548{
Jeff Layton7012b022013-06-21 08:58:22 -04002549 struct locks_iterator *iter = f->private;
2550
2551 ++iter->li_pos;
2552 return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002553}
2554
2555static void locks_stop(struct seq_file *f, void *v)
Jeff Laytonb03dfdec2014-02-03 12:13:07 -05002556 __releases(&blocked_lock_lock)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002557{
Jeff Layton7b2296a2013-06-21 08:58:20 -04002558 spin_unlock(&blocked_lock_lock);
Jeff Layton7012b022013-06-21 08:58:22 -04002559 lg_global_unlock(&file_lock_lglock);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002560}
2561
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002562static const struct seq_operations locks_seq_operations = {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002563 .start = locks_start,
2564 .next = locks_next,
2565 .stop = locks_stop,
2566 .show = locks_show,
2567};
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002568
2569static int locks_open(struct inode *inode, struct file *filp)
2570{
Jeff Layton7012b022013-06-21 08:58:22 -04002571 return seq_open_private(filp, &locks_seq_operations,
2572 sizeof(struct locks_iterator));
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002573}
2574
2575static const struct file_operations proc_locks_operations = {
2576 .open = locks_open,
2577 .read = seq_read,
2578 .llseek = seq_lseek,
Jerome Marchand99dc8292010-10-26 14:22:33 -07002579 .release = seq_release_private,
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002580};
2581
2582static int __init proc_locks_init(void)
2583{
2584 proc_create("locks", 0, NULL, &proc_locks_operations);
2585 return 0;
2586}
2587module_init(proc_locks_init);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002588#endif
2589
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590static int __init filelock_init(void)
2591{
Jeff Layton7012b022013-06-21 08:58:22 -04002592 int i;
2593
Linus Torvalds1da177e2005-04-16 15:20:36 -07002594 filelock_cache = kmem_cache_create("file_lock_cache",
Miklos Szerediee19cc42011-07-07 13:06:09 +02002595 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2596
Jeff Layton7012b022013-06-21 08:58:22 -04002597 lg_lock_init(&file_lock_lglock, "file_lock_lglock");
2598
2599 for_each_possible_cpu(i)
2600 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i));
2601
Linus Torvalds1da177e2005-04-16 15:20:36 -07002602 return 0;
2603}
2604
2605core_initcall(filelock_init);