blob: 314135ad820b7275ec8a8d8dd398f016a113e3d2 [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 */
Jeff Layton4d01b7f2014-09-01 15:06:54 -0400430static bool
431lease_break_callback(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432{
433 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
Jeff Layton4d01b7f2014-09-01 15:06:54 -0400434 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435}
436
Jeff Layton1c7dd2f2014-08-22 10:55:47 -0400437static void
438lease_setup(struct file_lock *fl, void **priv)
439{
440 struct file *filp = fl->fl_file;
441 struct fasync_struct *fa = *priv;
442
443 /*
444 * fasync_insert_entry() returns the old entry if any. If there was no
445 * old entry, then it used "priv" and inserted it into the fasync list.
446 * Clear the pointer to indicate that it shouldn't be freed.
447 */
448 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
449 *priv = NULL;
450
451 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
452}
453
Alexey Dobriyan7b021962009-09-21 17:01:12 -0700454static const struct lock_manager_operations lease_manager_ops = {
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400455 .lm_break = lease_break_callback,
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400456 .lm_change = lease_modify,
Jeff Layton1c7dd2f2014-08-22 10:55:47 -0400457 .lm_setup = lease_setup,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458};
459
460/*
461 * Initialize a lease, use the default lock manager operations
462 */
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400463static int lease_init(struct file *filp, long type, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400465 if (assign_type(fl, type) != 0)
466 return -EINVAL;
467
Jeff Layton7ca76312014-09-01 19:04:48 -0400468 fl->fl_owner = filp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 fl->fl_pid = current->tgid;
470
471 fl->fl_file = filp;
472 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 fl->fl_start = 0;
474 fl->fl_end = OFFSET_MAX;
475 fl->fl_ops = NULL;
476 fl->fl_lmops = &lease_manager_ops;
477 return 0;
478}
479
480/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400481static struct file_lock *lease_alloc(struct file *filp, long type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482{
483 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400484 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
486 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500487 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
489 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400490 if (error) {
491 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500492 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400493 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500494 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495}
496
497/* Check if two locks overlap each other.
498 */
499static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
500{
501 return ((fl1->fl_end >= fl2->fl_start) &&
502 (fl2->fl_end >= fl1->fl_start));
503}
504
505/*
506 * Check whether two locks have the same owner.
507 */
Matt Mackall33443c42006-01-08 01:05:22 -0800508static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509{
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400510 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 return fl2->fl_lmops == fl1->fl_lmops &&
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400512 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 return fl1->fl_owner == fl2->fl_owner;
514}
515
Jeff Layton7012b022013-06-21 08:58:22 -0400516/* Must be called with the i_lock held! */
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500517static void locks_insert_global_locks(struct file_lock *fl)
Jeff Layton88974692013-06-21 08:58:14 -0400518{
Jeff Layton7012b022013-06-21 08:58:22 -0400519 lg_local_lock(&file_lock_lglock);
520 fl->fl_link_cpu = smp_processor_id();
521 hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list));
522 lg_local_unlock(&file_lock_lglock);
Jeff Layton88974692013-06-21 08:58:14 -0400523}
524
Jeff Layton7012b022013-06-21 08:58:22 -0400525/* Must be called with the i_lock held! */
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500526static void locks_delete_global_locks(struct file_lock *fl)
Jeff Layton88974692013-06-21 08:58:14 -0400527{
Jeff Layton7012b022013-06-21 08:58:22 -0400528 /*
529 * Avoid taking lock if already unhashed. This is safe since this check
530 * is done while holding the i_lock, and new insertions into the list
531 * also require that it be held.
532 */
533 if (hlist_unhashed(&fl->fl_link))
534 return;
535 lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu);
Jeff Layton139ca042013-06-21 08:58:17 -0400536 hlist_del_init(&fl->fl_link);
Jeff Layton7012b022013-06-21 08:58:22 -0400537 lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu);
Jeff Layton88974692013-06-21 08:58:14 -0400538}
539
Jeff Layton3999e492013-06-21 08:58:19 -0400540static unsigned long
541posix_owner_key(struct file_lock *fl)
542{
543 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
544 return fl->fl_lmops->lm_owner_key(fl);
545 return (unsigned long)fl->fl_owner;
546}
547
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500548static void locks_insert_global_blocked(struct file_lock *waiter)
Jeff Layton88974692013-06-21 08:58:14 -0400549{
Jeff Layton3999e492013-06-21 08:58:19 -0400550 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
Jeff Layton88974692013-06-21 08:58:14 -0400551}
552
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500553static void locks_delete_global_blocked(struct file_lock *waiter)
Jeff Layton88974692013-06-21 08:58:14 -0400554{
Jeff Layton48f74182013-06-21 08:58:18 -0400555 hash_del(&waiter->fl_link);
Jeff Layton88974692013-06-21 08:58:14 -0400556}
557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558/* Remove waiter from blocker's block list.
559 * When blocker ends up pointing to itself then the list is empty.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400560 *
Jeff Layton7b2296a2013-06-21 08:58:20 -0400561 * Must be called with blocked_lock_lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 */
Matt Mackall33443c42006-01-08 01:05:22 -0800563static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564{
Jeff Layton88974692013-06-21 08:58:14 -0400565 locks_delete_global_blocked(waiter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 list_del_init(&waiter->fl_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 waiter->fl_next = NULL;
568}
569
Jeff Layton1a9e64a2013-06-21 08:58:10 -0400570static void locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571{
Jeff Layton7b2296a2013-06-21 08:58:20 -0400572 spin_lock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 __locks_delete_block(waiter);
Jeff Layton7b2296a2013-06-21 08:58:20 -0400574 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575}
576
577/* Insert waiter into blocker's block list.
578 * We use a circular list so that processes can be easily woken up in
579 * the order they blocked. The documentation doesn't require this but
580 * it seems like the reasonable thing to do.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400581 *
Jeff Layton7b2296a2013-06-21 08:58:20 -0400582 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
Jeff Layton46dad762014-02-03 12:13:06 -0500583 * list itself is protected by the blocked_lock_lock, but by ensuring that the
Jeff Layton7b2296a2013-06-21 08:58:20 -0400584 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
Jeff Layton4e8c7652013-06-21 08:58:16 -0400585 * in some cases when we see that the fl_block list is empty.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 */
Jeff Layton1c8c6012013-06-21 08:58:15 -0400587static void __locks_insert_block(struct file_lock *blocker,
588 struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800590 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 waiter->fl_next = blocker;
Jeff Layton88974692013-06-21 08:58:14 -0400592 list_add_tail(&waiter->fl_block, &blocker->fl_block);
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400593 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
Jeff Layton1c8c6012013-06-21 08:58:15 -0400594 locks_insert_global_blocked(waiter);
595}
596
597/* Must be called with i_lock held. */
598static void locks_insert_block(struct file_lock *blocker,
599 struct file_lock *waiter)
600{
Jeff Layton7b2296a2013-06-21 08:58:20 -0400601 spin_lock(&blocked_lock_lock);
Jeff Layton1c8c6012013-06-21 08:58:15 -0400602 __locks_insert_block(blocker, waiter);
Jeff Layton7b2296a2013-06-21 08:58:20 -0400603 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604}
605
Jeff Layton1cb36012013-06-21 08:58:12 -0400606/*
607 * Wake up processes blocked waiting for blocker.
608 *
Jeff Layton1c8c6012013-06-21 08:58:15 -0400609 * Must be called with the inode->i_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 */
611static void locks_wake_up_blocks(struct file_lock *blocker)
612{
Jeff Layton4e8c7652013-06-21 08:58:16 -0400613 /*
614 * Avoid taking global lock if list is empty. This is safe since new
615 * blocked requests are only added to the list under the i_lock, and
616 * the i_lock is always held here. Note that removal from the fl_block
617 * list does not require the i_lock, so we must recheck list_empty()
Jeff Layton7b2296a2013-06-21 08:58:20 -0400618 * after acquiring the blocked_lock_lock.
Jeff Layton4e8c7652013-06-21 08:58:16 -0400619 */
620 if (list_empty(&blocker->fl_block))
621 return;
622
Jeff Layton7b2296a2013-06-21 08:58:20 -0400623 spin_lock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400625 struct file_lock *waiter;
626
627 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 struct file_lock, fl_block);
629 __locks_delete_block(waiter);
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400630 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
631 waiter->fl_lmops->lm_notify(waiter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 else
633 wake_up(&waiter->fl_wait);
634 }
Jeff Layton7b2296a2013-06-21 08:58:20 -0400635 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
637
638/* Insert file lock fl into an inode's lock list at the position indicated
639 * by pos. At the same time add the lock to the global file lock list.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400640 *
641 * Must be called with the i_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 */
643static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
644{
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000645 fl->fl_nspid = get_pid(task_tgid(current));
646
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 /* insert into file's list */
648 fl->fl_next = *pos;
649 *pos = fl;
Jeff Layton88974692013-06-21 08:58:14 -0400650
651 locks_insert_global_locks(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652}
653
Jeff Layton24cbe782014-02-03 12:13:06 -0500654/**
655 * locks_delete_lock - Delete a lock and then free it.
656 * @thisfl_p: pointer that points to the fl_next field of the previous
657 * inode->i_flock list entry
658 *
659 * Unlink a lock from all lists and free the namespace reference, but don't
660 * free it yet. Wake up processes that are blocked waiting for this lock and
661 * notify the FS that the lock has been cleared.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400662 *
663 * Must be called with the i_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 */
Jeff Layton24cbe782014-02-03 12:13:06 -0500665static void locks_unlink_lock(struct file_lock **thisfl_p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666{
667 struct file_lock *fl = *thisfl_p;
668
Jeff Layton88974692013-06-21 08:58:14 -0400669 locks_delete_global_locks(fl);
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 *thisfl_p = fl->fl_next;
672 fl->fl_next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000674 if (fl->fl_nspid) {
675 put_pid(fl->fl_nspid);
676 fl->fl_nspid = NULL;
677 }
678
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 locks_wake_up_blocks(fl);
Jeff Layton24cbe782014-02-03 12:13:06 -0500680}
681
682/*
683 * Unlink a lock from all lists and free it.
684 *
685 * Must be called with i_lock held!
686 */
Jeff Laytoned9814d2014-08-11 14:20:31 -0400687static void locks_delete_lock(struct file_lock **thisfl_p,
688 struct list_head *dispose)
Jeff Layton24cbe782014-02-03 12:13:06 -0500689{
690 struct file_lock *fl = *thisfl_p;
691
692 locks_unlink_lock(thisfl_p);
Jeff Laytoned9814d2014-08-11 14:20:31 -0400693 if (dispose)
694 list_add(&fl->fl_block, dispose);
695 else
696 locks_free_lock(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697}
698
699/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
700 * checks for shared/exclusive status of overlapping locks.
701 */
702static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
703{
704 if (sys_fl->fl_type == F_WRLCK)
705 return 1;
706 if (caller_fl->fl_type == F_WRLCK)
707 return 1;
708 return 0;
709}
710
711/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
712 * checking before calling the locks_conflict().
713 */
714static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
715{
716 /* POSIX locks owned by the same process do not conflict with
717 * each other.
718 */
719 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
720 return (0);
721
722 /* Check whether they overlap */
723 if (!locks_overlap(caller_fl, sys_fl))
724 return 0;
725
726 return (locks_conflict(caller_fl, sys_fl));
727}
728
729/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
730 * checking before calling the locks_conflict().
731 */
732static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
733{
734 /* FLOCK locks referring to the same filp do not conflict with
735 * each other.
736 */
737 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
738 return (0);
739 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
740 return 0;
741
742 return (locks_conflict(caller_fl, sys_fl));
743}
744
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400745void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500746posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747{
748 struct file_lock *cfl;
Jeff Layton1c8c6012013-06-21 08:58:15 -0400749 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
Jeff Layton1c8c6012013-06-21 08:58:15 -0400751 spin_lock(&inode->i_lock);
Al Viro496ad9a2013-01-23 17:07:38 -0500752 for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 if (!IS_POSIX(cfl))
754 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400755 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 break;
757 }
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000758 if (cfl) {
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400759 locks_copy_conflock(fl, cfl);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000760 if (cfl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800761 fl->fl_pid = pid_vnr(cfl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000762 } else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400763 fl->fl_type = F_UNLCK;
Jeff Layton1c8c6012013-06-21 08:58:15 -0400764 spin_unlock(&inode->i_lock);
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400765 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767EXPORT_SYMBOL(posix_test_lock);
768
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400769/*
770 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400772 * We attempt to detect deadlocks that are due purely to posix file
773 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400775 * We assume that a task can be waiting for at most one lock at a time.
776 * So for any acquired lock, the process holding that lock may be
777 * waiting on at most one other lock. That lock in turns may be held by
778 * someone waiting for at most one other lock. Given a requested lock
779 * caller_fl which is about to wait for a conflicting lock block_fl, we
780 * follow this chain of waiters to ensure we are not about to create a
781 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400782 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400783 * Since we do this before we ever put a process to sleep on a lock, we
784 * are ensured that there is never a cycle; that is what guarantees that
785 * the while() loop in posix_locks_deadlock() eventually completes.
786 *
787 * Note: the above assumption may not be true when handling lock
788 * requests from a broken NFS client. It may also fail in the presence
789 * of tasks (such as posix threads) sharing the same open file table.
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400790 * To handle those cases, we just bail out after a few iterations.
Jeff Layton57b65322014-02-03 12:13:09 -0500791 *
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400792 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
Jeff Layton57b65322014-02-03 12:13:09 -0500793 * Because the owner is not even nominally tied to a thread of
794 * execution, the deadlock detection below can't reasonably work well. Just
795 * skip it for those.
796 *
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400797 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
Jeff Layton57b65322014-02-03 12:13:09 -0500798 * locks that just checks for the case where two tasks are attempting to
799 * upgrade from read to write locks on the same inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400801
802#define MAX_DEADLK_ITERATIONS 10
803
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400804/* Find a lock that the owner of the given block_fl is blocking on. */
805static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
806{
807 struct file_lock *fl;
808
Jeff Layton3999e492013-06-21 08:58:19 -0400809 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400810 if (posix_same_owner(fl, block_fl))
811 return fl->fl_next;
812 }
813 return NULL;
814}
815
Jeff Layton7b2296a2013-06-21 08:58:20 -0400816/* Must be called with the blocked_lock_lock held! */
Adrian Bunkb0904e12006-06-23 02:05:13 -0700817static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 struct file_lock *block_fl)
819{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400820 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821
Jeff Layton57b65322014-02-03 12:13:09 -0500822 /*
823 * This deadlock detector can't reasonably detect deadlocks with
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400824 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
Jeff Layton57b65322014-02-03 12:13:09 -0500825 */
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400826 if (IS_OFDLCK(caller_fl))
Jeff Layton57b65322014-02-03 12:13:09 -0500827 return 0;
828
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400829 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
830 if (i++ > MAX_DEADLK_ITERATIONS)
831 return 0;
832 if (posix_same_owner(caller_fl, block_fl))
833 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 }
835 return 0;
836}
837
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400839 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400840 *
841 * Note that if called with an FL_EXISTS argument, the caller may determine
842 * whether or not a lock was successfully freed by testing the return
843 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800845static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800847 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 struct file_lock **before;
Al Viro496ad9a2013-01-23 17:07:38 -0500849 struct inode * inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 int error = 0;
851 int found = 0;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400852 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200854 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
855 new_fl = locks_alloc_lock();
856 if (!new_fl)
857 return -ENOMEM;
858 }
859
Jeff Layton1c8c6012013-06-21 08:58:15 -0400860 spin_lock(&inode->i_lock);
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400861 if (request->fl_flags & FL_ACCESS)
862 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400863
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 for_each_lock(inode, before) {
865 struct file_lock *fl = *before;
866 if (IS_POSIX(fl))
867 break;
868 if (IS_LEASE(fl))
869 continue;
870 if (filp != fl->fl_file)
871 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800872 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 goto out;
874 found = 1;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400875 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 break;
877 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Trond Myklebustf475ae92006-06-29 16:38:32 -0400879 if (request->fl_type == F_UNLCK) {
880 if ((request->fl_flags & FL_EXISTS) && !found)
881 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800882 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400883 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884
885 /*
886 * If a higher-priority process was blocked on the old file lock,
887 * give it the opportunity to lock the file.
888 */
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200889 if (found) {
Jeff Layton1c8c6012013-06-21 08:58:15 -0400890 spin_unlock(&inode->i_lock);
Frederic Weisbeckerdef01bc2009-07-16 15:44:29 +0200891 cond_resched();
Jeff Layton1c8c6012013-06-21 08:58:15 -0400892 spin_lock(&inode->i_lock);
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200893 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400895find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 for_each_lock(inode, before) {
897 struct file_lock *fl = *before;
898 if (IS_POSIX(fl))
899 break;
900 if (IS_LEASE(fl))
901 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800902 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 continue;
904 error = -EAGAIN;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700905 if (!(request->fl_flags & FL_SLEEP))
906 goto out;
907 error = FILE_LOCK_DEFERRED;
908 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 goto out;
910 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400911 if (request->fl_flags & FL_ACCESS)
912 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800913 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700914 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800915 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400916 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
918out:
Jeff Layton1c8c6012013-06-21 08:58:15 -0400919 spin_unlock(&inode->i_lock);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800920 if (new_fl)
921 locks_free_lock(new_fl);
Jeff Laytoned9814d2014-08-11 14:20:31 -0400922 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 return error;
924}
925
Marc Eshel150b3932007-01-18 16:15:35 -0500926static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927{
928 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700929 struct file_lock *new_fl = NULL;
930 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 struct file_lock *left = NULL;
932 struct file_lock *right = NULL;
933 struct file_lock **before;
Jeff Laytonb9746ef2013-06-21 08:58:13 -0400934 int error;
935 bool added = false;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400936 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
938 /*
939 * We may need two file_lock structures for this operation,
940 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700941 *
942 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700944 if (!(request->fl_flags & FL_ACCESS) &&
945 (request->fl_type != F_UNLCK ||
946 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
947 new_fl = locks_alloc_lock();
948 new_fl2 = locks_alloc_lock();
949 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Jeff Layton1c8c6012013-06-21 08:58:15 -0400951 spin_lock(&inode->i_lock);
Jeff Layton1cb36012013-06-21 08:58:12 -0400952 /*
953 * New lock request. Walk all POSIX locks and look for conflicts. If
954 * there are any, either return error or put the request on the
Jeff Layton48f74182013-06-21 08:58:18 -0400955 * blocker's list of waiters and the global blocked_hash.
Jeff Layton1cb36012013-06-21 08:58:12 -0400956 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 if (request->fl_type != F_UNLCK) {
958 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500959 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 if (!IS_POSIX(fl))
961 continue;
962 if (!posix_locks_conflict(request, fl))
963 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800964 if (conflock)
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400965 locks_copy_conflock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 error = -EAGAIN;
967 if (!(request->fl_flags & FL_SLEEP))
968 goto out;
Jeff Layton1c8c6012013-06-21 08:58:15 -0400969 /*
970 * Deadlock detection and insertion into the blocked
971 * locks list must be done while holding the same lock!
972 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 error = -EDEADLK;
Jeff Layton7b2296a2013-06-21 08:58:20 -0400974 spin_lock(&blocked_lock_lock);
Jeff Layton1c8c6012013-06-21 08:58:15 -0400975 if (likely(!posix_locks_deadlock(request, fl))) {
976 error = FILE_LOCK_DEFERRED;
977 __locks_insert_block(fl, request);
978 }
Jeff Layton7b2296a2013-06-21 08:58:20 -0400979 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 goto out;
981 }
982 }
983
984 /* If we're just looking for a conflict, we're done. */
985 error = 0;
986 if (request->fl_flags & FL_ACCESS)
987 goto out;
988
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 * Find the first old lock with the same owner as the new lock.
991 */
992
993 before = &inode->i_flock;
994
995 /* First skip locks owned by other processes. */
996 while ((fl = *before) && (!IS_POSIX(fl) ||
997 !posix_same_owner(request, fl))) {
998 before = &fl->fl_next;
999 }
1000
Jeff Layton1cb36012013-06-21 08:58:12 -04001001 /* Process locks with this owner. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 while ((fl = *before) && posix_same_owner(request, fl)) {
1003 /* Detect adjacent or overlapping regions (if same lock type)
1004 */
1005 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -07001006 /* In all comparisons of start vs end, use
1007 * "start - 1" rather than "end + 1". If end
1008 * is OFFSET_MAX, end + 1 will become negative.
1009 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 if (fl->fl_end < request->fl_start - 1)
1011 goto next_lock;
1012 /* If the next lock in the list has entirely bigger
1013 * addresses than the new one, insert the lock here.
1014 */
Olaf Kirch449231d2005-08-25 16:25:35 -07001015 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 break;
1017
1018 /* If we come here, the new and old lock are of the
1019 * same type and adjacent or overlapping. Make one
1020 * lock yielding from the lower start address of both
1021 * locks to the higher end address.
1022 */
1023 if (fl->fl_start > request->fl_start)
1024 fl->fl_start = request->fl_start;
1025 else
1026 request->fl_start = fl->fl_start;
1027 if (fl->fl_end < request->fl_end)
1028 fl->fl_end = request->fl_end;
1029 else
1030 request->fl_end = fl->fl_end;
1031 if (added) {
Jeff Laytoned9814d2014-08-11 14:20:31 -04001032 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 continue;
1034 }
1035 request = fl;
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001036 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 }
1038 else {
1039 /* Processing for different lock types is a bit
1040 * more complex.
1041 */
1042 if (fl->fl_end < request->fl_start)
1043 goto next_lock;
1044 if (fl->fl_start > request->fl_end)
1045 break;
1046 if (request->fl_type == F_UNLCK)
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001047 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 if (fl->fl_start < request->fl_start)
1049 left = fl;
1050 /* If the next lock in the list has a higher end
1051 * address than the new one, insert the new one here.
1052 */
1053 if (fl->fl_end > request->fl_end) {
1054 right = fl;
1055 break;
1056 }
1057 if (fl->fl_start >= request->fl_start) {
1058 /* The new lock completely replaces an old
1059 * one (This may happen several times).
1060 */
1061 if (added) {
Jeff Laytoned9814d2014-08-11 14:20:31 -04001062 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 continue;
1064 }
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001065 /*
1066 * Replace the old lock with new_fl, and
1067 * remove the old one. It's safe to do the
1068 * insert here since we know that we won't be
1069 * using new_fl later, and that the lock is
1070 * just replacing an existing lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 */
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001072 error = -ENOLCK;
1073 if (!new_fl)
1074 goto out;
1075 locks_copy_lock(new_fl, request);
1076 request = new_fl;
1077 new_fl = NULL;
Jeff Laytoned9814d2014-08-11 14:20:31 -04001078 locks_delete_lock(before, &dispose);
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001079 locks_insert_lock(before, request);
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001080 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 }
1082 }
1083 /* Go on to next lock.
1084 */
1085 next_lock:
1086 before = &fl->fl_next;
1087 }
1088
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001089 /*
Jeff Layton1cb36012013-06-21 08:58:12 -04001090 * The above code only modifies existing locks in case of merging or
1091 * replacing. If new lock(s) need to be inserted all modifications are
1092 * done below this, so it's safe yet to bail out.
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001093 */
1094 error = -ENOLCK; /* "no luck" */
1095 if (right && left == right && !new_fl2)
1096 goto out;
1097
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 error = 0;
1099 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -04001100 if (request->fl_type == F_UNLCK) {
1101 if (request->fl_flags & FL_EXISTS)
1102 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -04001104 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001105
1106 if (!new_fl) {
1107 error = -ENOLCK;
1108 goto out;
1109 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 locks_copy_lock(new_fl, request);
1111 locks_insert_lock(before, new_fl);
1112 new_fl = NULL;
1113 }
1114 if (right) {
1115 if (left == right) {
1116 /* The new lock breaks the old one in two pieces,
1117 * so we have to use the second new lock.
1118 */
1119 left = new_fl2;
1120 new_fl2 = NULL;
1121 locks_copy_lock(left, right);
1122 locks_insert_lock(before, left);
1123 }
1124 right->fl_start = request->fl_end + 1;
1125 locks_wake_up_blocks(right);
1126 }
1127 if (left) {
1128 left->fl_end = request->fl_start - 1;
1129 locks_wake_up_blocks(left);
1130 }
1131 out:
Jeff Layton1c8c6012013-06-21 08:58:15 -04001132 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 /*
1134 * Free any unused locks.
1135 */
1136 if (new_fl)
1137 locks_free_lock(new_fl);
1138 if (new_fl2)
1139 locks_free_lock(new_fl2);
Jeff Laytoned9814d2014-08-11 14:20:31 -04001140 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 return error;
1142}
1143
1144/**
1145 * posix_lock_file - Apply a POSIX-style lock to a file
1146 * @filp: The file to apply the lock to
1147 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001148 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 *
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
Trond Myklebustf475ae92006-06-29 16:38:32 -04001153 *
1154 * Note that if called with an FL_EXISTS argument, the caller may determine
1155 * whether or not a lock was successfully freed by testing the return
1156 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 */
Marc Eshel150b3932007-01-18 16:15:35 -05001158int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001159 struct file_lock *conflock)
1160{
Al Viro496ad9a2013-01-23 17:07:38 -05001161 return __posix_lock_file(file_inode(filp), fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001162}
Marc Eshel150b3932007-01-18 16:15:35 -05001163EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
1165/**
1166 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1167 * @filp: The file to apply the lock to
1168 * @fl: The lock to be applied
1169 *
1170 * Add a POSIX style lock to a file.
1171 * We merge adjacent & overlapping locks whenever possible.
1172 * POSIX locks are sorted by owner task, then by starting address
1173 */
1174int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1175{
1176 int error;
1177 might_sleep ();
1178 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001179 error = posix_lock_file(filp, fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001180 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 break;
1182 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1183 if (!error)
1184 continue;
1185
1186 locks_delete_block(fl);
1187 break;
1188 }
1189 return error;
1190}
1191EXPORT_SYMBOL(posix_lock_file_wait);
1192
1193/**
1194 * locks_mandatory_locked - Check for an active lock
Jeff Laytond7a06982014-03-10 09:54:15 -04001195 * @file: the file to check
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 *
1197 * Searches the inode's list of locks to find any POSIX locks which conflict.
1198 * This function is called from locks_verify_locked() only.
1199 */
Jeff Laytond7a06982014-03-10 09:54:15 -04001200int locks_mandatory_locked(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201{
Jeff Laytond7a06982014-03-10 09:54:15 -04001202 struct inode *inode = file_inode(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 struct file_lock *fl;
1204
1205 /*
1206 * Search the lock list for this inode for any POSIX locks.
1207 */
Jeff Layton1c8c6012013-06-21 08:58:15 -04001208 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1210 if (!IS_POSIX(fl))
1211 continue;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001212 if (fl->fl_owner != current->files &&
1213 fl->fl_owner != file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 break;
1215 }
Jeff Layton1c8c6012013-06-21 08:58:15 -04001216 spin_unlock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 return fl ? -EAGAIN : 0;
1218}
1219
1220/**
1221 * locks_mandatory_area - Check for a conflicting lock
1222 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1223 * for shared
1224 * @inode: the file to check
1225 * @filp: how the file was opened (if it was)
1226 * @offset: start of area to check
1227 * @count: length of area to check
1228 *
1229 * Searches the inode's list of locks to find any POSIX locks which conflict.
1230 * This function is called from rw_verify_area() and
1231 * locks_verify_truncate().
1232 */
1233int locks_mandatory_area(int read_write, struct inode *inode,
1234 struct file *filp, loff_t offset,
1235 size_t count)
1236{
1237 struct file_lock fl;
1238 int error;
Jeff Layton29723ad2014-03-10 09:54:19 -04001239 bool sleep = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240
1241 locks_init_lock(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 fl.fl_pid = current->tgid;
1243 fl.fl_file = filp;
1244 fl.fl_flags = FL_POSIX | FL_ACCESS;
1245 if (filp && !(filp->f_flags & O_NONBLOCK))
Jeff Layton29723ad2014-03-10 09:54:19 -04001246 sleep = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1248 fl.fl_start = offset;
1249 fl.fl_end = offset + count - 1;
1250
1251 for (;;) {
Jeff Layton29723ad2014-03-10 09:54:19 -04001252 if (filp) {
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001253 fl.fl_owner = filp;
Jeff Layton29723ad2014-03-10 09:54:19 -04001254 fl.fl_flags &= ~FL_SLEEP;
1255 error = __posix_lock_file(inode, &fl, NULL);
1256 if (!error)
1257 break;
1258 }
1259
1260 if (sleep)
1261 fl.fl_flags |= FL_SLEEP;
1262 fl.fl_owner = current->files;
Marc Eshel150b3932007-01-18 16:15:35 -05001263 error = __posix_lock_file(inode, &fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001264 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 break;
1266 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1267 if (!error) {
1268 /*
1269 * If we've been sleeping someone might have
1270 * changed the permissions behind our back.
1271 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001272 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 continue;
1274 }
1275
1276 locks_delete_block(&fl);
1277 break;
1278 }
1279
1280 return error;
1281}
1282
1283EXPORT_SYMBOL(locks_mandatory_area);
1284
J. Bruce Fields778fc542011-07-26 18:25:49 -04001285static void lease_clear_pending(struct file_lock *fl, int arg)
1286{
1287 switch (arg) {
1288 case F_UNLCK:
1289 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1290 /* fall through: */
1291 case F_RDLCK:
1292 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1293 }
1294}
1295
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296/* We already had a lease on this file; just change its type */
Jeff Laytonc45198e2014-09-01 07:12:07 -04001297int lease_modify(struct file_lock **before, int arg, struct list_head *dispose)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298{
1299 struct file_lock *fl = *before;
1300 int error = assign_type(fl, arg);
1301
1302 if (error)
1303 return error;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001304 lease_clear_pending(fl, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 locks_wake_up_blocks(fl);
Filipe Brandenburger3b6e2722012-07-27 00:42:52 -04001306 if (arg == F_UNLCK) {
1307 struct file *filp = fl->fl_file;
1308
1309 f_delown(filp);
1310 filp->f_owner.signum = 0;
J. Bruce Fields96d6d592012-07-27 16:18:00 -04001311 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1312 if (fl->fl_fasync != NULL) {
1313 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1314 fl->fl_fasync = NULL;
1315 }
Jeff Laytonc45198e2014-09-01 07:12:07 -04001316 locks_delete_lock(before, dispose);
Filipe Brandenburger3b6e2722012-07-27 00:42:52 -04001317 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 return 0;
1319}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320EXPORT_SYMBOL(lease_modify);
1321
J. Bruce Fields778fc542011-07-26 18:25:49 -04001322static bool past_time(unsigned long then)
1323{
1324 if (!then)
1325 /* 0 is a special value meaning "this never expires": */
1326 return false;
1327 return time_after(jiffies, then);
1328}
1329
Jeff Laytonc45198e2014-09-01 07:12:07 -04001330static void time_out_leases(struct inode *inode, struct list_head *dispose)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331{
1332 struct file_lock **before;
1333 struct file_lock *fl;
1334
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001335 lockdep_assert_held(&inode->i_lock);
1336
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 before = &inode->i_flock;
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04001338 while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) {
Jeff Layton62af4f12014-05-09 14:13:05 -04001339 trace_time_out_leases(inode, fl);
J. Bruce Fields778fc542011-07-26 18:25:49 -04001340 if (past_time(fl->fl_downgrade_time))
Jeff Laytonc45198e2014-09-01 07:12:07 -04001341 lease_modify(before, F_RDLCK, dispose);
J. Bruce Fields778fc542011-07-26 18:25:49 -04001342 if (past_time(fl->fl_break_time))
Jeff Laytonc45198e2014-09-01 07:12:07 -04001343 lease_modify(before, F_UNLCK, dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 if (fl == *before) /* lease_modify may have freed fl */
1345 before = &fl->fl_next;
1346 }
1347}
1348
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001349static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1350{
1351 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1352 return false;
1353 return locks_conflict(breaker, lease);
1354}
1355
Jeff Layton03d12dd2014-09-01 14:53:41 -04001356static bool
1357any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1358{
1359 struct file_lock *fl;
1360
1361 lockdep_assert_held(&inode->i_lock);
1362
1363 for (fl = inode->i_flock ; fl && IS_LEASE(fl); fl = fl->fl_next) {
1364 if (leases_conflict(fl, breaker))
1365 return true;
1366 }
1367 return false;
1368}
1369
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370/**
1371 * __break_lease - revoke all outstanding leases on file
1372 * @inode: the inode of the file to return
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001373 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1374 * break all leases
1375 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1376 * only delegations
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 *
david m. richter87250dd2007-05-09 16:10:27 -04001378 * break_lease (inlined for speed) has checked there already is at least
1379 * some kind of lock (maybe a lease) on this file. Leases are broken on
1380 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 * specified %O_NONBLOCK to your open().
1382 */
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001383int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384{
J. Bruce Fields778fc542011-07-26 18:25:49 -04001385 int error = 0;
Jeff Layton03d12dd2014-09-01 14:53:41 -04001386 struct file_lock *new_fl;
Jeff Layton4d01b7f2014-09-01 15:06:54 -04001387 struct file_lock *fl, **before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 unsigned long break_time;
Al Viro8737c932009-12-24 06:47:55 -05001389 int want_write = (mode & O_ACCMODE) != O_RDONLY;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001390 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391
Al Viro8737c932009-12-24 06:47:55 -05001392 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
Linus Torvalds6d4b9e32011-12-26 10:25:26 -08001393 if (IS_ERR(new_fl))
1394 return PTR_ERR(new_fl);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001395 new_fl->fl_flags = type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396
Jeff Layton1c8c6012013-06-21 08:58:15 -04001397 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
Jeff Laytonc45198e2014-09-01 07:12:07 -04001399 time_out_leases(inode, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400
Jeff Layton03d12dd2014-09-01 14:53:41 -04001401 if (!any_leases_conflict(inode, new_fl))
J. Bruce Fields778fc542011-07-26 18:25:49 -04001402 goto out;
1403
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 break_time = 0;
1405 if (lease_break_time > 0) {
1406 break_time = jiffies + lease_break_time * HZ;
1407 if (break_time == 0)
1408 break_time++; /* so that 0 means no break time */
1409 }
1410
Jeff Layton4d01b7f2014-09-01 15:06:54 -04001411 for (before = &inode->i_flock;
1412 ((fl = *before) != NULL) && IS_LEASE(fl);
1413 before = &fl->fl_next) {
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001414 if (!leases_conflict(fl, new_fl))
1415 continue;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001416 if (want_write) {
1417 if (fl->fl_flags & FL_UNLOCK_PENDING)
1418 continue;
1419 fl->fl_flags |= FL_UNLOCK_PENDING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 fl->fl_break_time = break_time;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001421 } else {
Jeff Layton03d12dd2014-09-01 14:53:41 -04001422 if (lease_breaking(inode->i_flock))
J. Bruce Fields778fc542011-07-26 18:25:49 -04001423 continue;
1424 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1425 fl->fl_downgrade_time = break_time;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 }
Jeff Layton4d01b7f2014-09-01 15:06:54 -04001427 if (fl->fl_lmops->lm_break(fl))
1428 locks_delete_lock(before, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 }
1430
Jeff Layton4d01b7f2014-09-01 15:06:54 -04001431 fl = inode->i_flock;
1432 if (!fl || !IS_LEASE(fl))
1433 goto out;
1434
Jeff Layton843c6b22014-09-01 14:27:43 -04001435 if (mode & O_NONBLOCK) {
Jeff Layton62af4f12014-05-09 14:13:05 -04001436 trace_break_lease_noblock(inode, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 error = -EWOULDBLOCK;
1438 goto out;
1439 }
1440
1441restart:
Jeff Layton03d12dd2014-09-01 14:53:41 -04001442 break_time = inode->i_flock->fl_break_time;
Jeff Laytonf1c6bb22014-04-15 06:17:49 -04001443 if (break_time != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 break_time -= jiffies;
Jeff Laytonf1c6bb22014-04-15 06:17:49 -04001445 if (break_time == 0)
1446 break_time++;
Jeff Layton03d12dd2014-09-01 14:53:41 -04001447 locks_insert_block(inode->i_flock, new_fl);
Jeff Layton62af4f12014-05-09 14:13:05 -04001448 trace_break_lease_block(inode, new_fl);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001449 spin_unlock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001450 locks_dispose_list(&dispose);
Matthew Wilcox4321e012008-01-14 21:28:30 -07001451 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1452 !new_fl->fl_next, break_time);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001453 spin_lock(&inode->i_lock);
Jeff Layton62af4f12014-05-09 14:13:05 -04001454 trace_break_lease_unblock(inode, new_fl);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001455 locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 if (error >= 0) {
J. Bruce Fields778fc542011-07-26 18:25:49 -04001457 /*
1458 * Wait for the next conflicting lease that has not been
1459 * broken yet
1460 */
Jeff Layton03d12dd2014-09-01 14:53:41 -04001461 if (error == 0)
1462 time_out_leases(inode, &dispose);
1463 if (any_leases_conflict(inode, new_fl))
1464 goto restart;
1465
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 error = 0;
1467 }
1468
1469out:
Jeff Layton1c8c6012013-06-21 08:58:15 -04001470 spin_unlock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001471 locks_dispose_list(&dispose);
Linus Torvalds6d4b9e32011-12-26 10:25:26 -08001472 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 return error;
1474}
1475
1476EXPORT_SYMBOL(__break_lease);
1477
1478/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001479 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 * @inode: the inode
1481 * @time: pointer to a timespec which will contain the last modified time
1482 *
1483 * This is to force NFS clients to flush their caches for files with
1484 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001485 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 */
1487void lease_get_mtime(struct inode *inode, struct timespec *time)
1488{
Jeff Laytonbfe86022014-08-22 10:18:44 -04001489 bool has_lease = false;
1490 struct file_lock *flock;
1491
1492 if (inode->i_flock) {
1493 spin_lock(&inode->i_lock);
1494 flock = inode->i_flock;
1495 if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK))
1496 has_lease = true;
1497 spin_unlock(&inode->i_lock);
1498 }
1499
1500 if (has_lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 *time = current_fs_time(inode->i_sb);
1502 else
1503 *time = inode->i_mtime;
1504}
1505
1506EXPORT_SYMBOL(lease_get_mtime);
1507
1508/**
1509 * fcntl_getlease - Enquire what lease is currently active
1510 * @filp: the file
1511 *
1512 * The value returned by this function will be one of
1513 * (if no lease break is pending):
1514 *
1515 * %F_RDLCK to indicate a shared lease is held.
1516 *
1517 * %F_WRLCK to indicate an exclusive lease is held.
1518 *
1519 * %F_UNLCK to indicate no lease is held.
1520 *
1521 * (if a lease break is pending):
1522 *
1523 * %F_RDLCK to indicate an exclusive lease needs to be
1524 * changed to a shared lease (or removed).
1525 *
1526 * %F_UNLCK to indicate the lease needs to be removed.
1527 *
1528 * XXX: sfr & willy disagree over whether F_INPROGRESS
1529 * should be returned to userspace.
1530 */
1531int fcntl_getlease(struct file *filp)
1532{
1533 struct file_lock *fl;
Jeff Layton1c8c6012013-06-21 08:58:15 -04001534 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 int type = F_UNLCK;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001536 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537
Jeff Layton1c8c6012013-06-21 08:58:15 -04001538 spin_lock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001539 time_out_leases(file_inode(filp), &dispose);
Al Viro496ad9a2013-01-23 17:07:38 -05001540 for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 fl = fl->fl_next) {
1542 if (fl->fl_file == filp) {
J. Bruce Fields778fc542011-07-26 18:25:49 -04001543 type = target_leasetype(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 break;
1545 }
1546 }
Jeff Layton1c8c6012013-06-21 08:58:15 -04001547 spin_unlock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001548 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 return type;
1550}
1551
Jeff Layton24cbe782014-02-03 12:13:06 -05001552/**
1553 * check_conflicting_open - see if the given dentry points to a file that has
1554 * an existing open that would conflict with the
1555 * desired lease.
1556 * @dentry: dentry to check
1557 * @arg: type of lease that we're trying to acquire
1558 *
1559 * Check to see if there's an existing open fd on this file that would
1560 * conflict with the lease we're trying to set.
1561 */
1562static int
1563check_conflicting_open(const struct dentry *dentry, const long arg)
1564{
1565 int ret = 0;
1566 struct inode *inode = dentry->d_inode;
1567
1568 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1569 return -EAGAIN;
1570
1571 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1572 (atomic_read(&inode->i_count) > 1)))
1573 ret = -EAGAIN;
1574
1575 return ret;
1576}
1577
Jeff Laytone6f5c782014-08-22 10:40:25 -04001578static int
1579generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001581 struct file_lock *fl, **before, **my_before = NULL, *lease;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001582 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 struct inode *inode = dentry->d_inode;
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001584 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001585 int error;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001586 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587
J. Bruce Fields096657b2010-10-30 17:31:14 -04001588 lease = *flp;
Jeff Layton62af4f12014-05-09 14:13:05 -04001589 trace_generic_add_lease(inode, lease);
1590
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001591 /*
1592 * In the delegation case we need mutual exclusion with
1593 * a number of operations that take the i_mutex. We trylock
1594 * because delegations are an optional optimization, and if
1595 * there's some chance of a conflict--we'd rather not
1596 * bother, maybe that's a sign this just isn't a good file to
1597 * hand out a delegation on.
1598 */
1599 if (is_deleg && !mutex_trylock(&inode->i_mutex))
1600 return -EAGAIN;
1601
1602 if (is_deleg && arg == F_WRLCK) {
1603 /* Write delegations are not currently supported: */
Dan Carpenter4fdb7932013-11-13 10:56:27 +03001604 mutex_unlock(&inode->i_mutex);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001605 WARN_ON_ONCE(1);
1606 return -EINVAL;
1607 }
J. Bruce Fields096657b2010-10-30 17:31:14 -04001608
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001609 spin_lock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001610 time_out_leases(inode, &dispose);
Jeff Layton24cbe782014-02-03 12:13:06 -05001611 error = check_conflicting_open(dentry, arg);
1612 if (error)
J. Bruce Fields096657b2010-10-30 17:31:14 -04001613 goto out;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001614
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 /*
1616 * At this point, we know that if there is an exclusive
1617 * lease on this file, then we hold it on this filp
1618 * (otherwise our open of this file would have blocked).
1619 * And if we are trying to acquire an exclusive lease,
1620 * then the file is not open by anyone (including us)
1621 * except for this filp.
1622 */
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001623 error = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 for (before = &inode->i_flock;
1625 ((fl = *before) != NULL) && IS_LEASE(fl);
1626 before = &fl->fl_next) {
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001627 if (fl->fl_file == filp) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 my_before = before;
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001629 continue;
1630 }
1631 /*
1632 * No exclusive leases if someone else has a lease on
1633 * this file:
1634 */
1635 if (arg == F_WRLCK)
1636 goto out;
1637 /*
1638 * Modifying our existing lease is OK, but no getting a
1639 * new lease if someone else is opening for write:
1640 */
1641 if (fl->fl_flags & FL_UNLOCK_PENDING)
1642 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 }
1644
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 if (my_before != NULL) {
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001646 lease = *my_before;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001647 error = lease->fl_lmops->lm_change(my_before, arg, &dispose);
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001648 if (error)
1649 goto out;
1650 goto out_setup;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 }
1652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 error = -EINVAL;
1654 if (!leases_enable)
1655 goto out;
1656
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +02001657 locks_insert_lock(before, lease);
Jeff Layton24cbe782014-02-03 12:13:06 -05001658 /*
1659 * The check in break_lease() is lockless. It's possible for another
1660 * open to race in after we did the earlier check for a conflicting
1661 * open but before the lease was inserted. Check again for a
1662 * conflicting open and cancel the lease if there is one.
1663 *
1664 * We also add a barrier here to ensure that the insertion of the lock
1665 * precedes these checks.
1666 */
1667 smp_mb();
1668 error = check_conflicting_open(dentry, arg);
1669 if (error)
Jeff Laytone6f5c782014-08-22 10:40:25 -04001670 goto out_unlink;
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001671
1672out_setup:
1673 if (lease->fl_lmops->lm_setup)
1674 lease->fl_lmops->lm_setup(lease, priv);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675out:
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001676 spin_unlock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001677 locks_dispose_list(&dispose);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001678 if (is_deleg)
1679 mutex_unlock(&inode->i_mutex);
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001680 if (!error && !my_before)
1681 *flp = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 return error;
Jeff Laytone6f5c782014-08-22 10:40:25 -04001683out_unlink:
1684 locks_unlink_lock(before);
1685 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686}
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001687
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001688static int generic_delete_lease(struct file *filp)
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001689{
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001690 int error = -EAGAIN;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001691 struct file_lock *fl, **before;
1692 struct dentry *dentry = filp->f_path.dentry;
1693 struct inode *inode = dentry->d_inode;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001694 LIST_HEAD(dispose);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001695
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001696 spin_lock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001697 time_out_leases(inode, &dispose);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001698 for (before = &inode->i_flock;
1699 ((fl = *before) != NULL) && IS_LEASE(fl);
1700 before = &fl->fl_next) {
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001701 if (fl->fl_file == filp)
1702 break;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001703 }
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001704 trace_generic_delete_lease(inode, fl);
1705 if (fl)
Jeff Laytonc45198e2014-09-01 07:12:07 -04001706 error = fl->fl_lmops->lm_change(before, F_UNLCK, &dispose);
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001707 spin_unlock(&inode->i_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001708 locks_dispose_list(&dispose);
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001709 return error;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001710}
1711
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712/**
1713 * generic_setlease - sets a lease on an open file
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001714 * @filp: file pointer
1715 * @arg: type of lease to obtain
1716 * @flp: input - file_lock to use, output - file_lock inserted
1717 * @priv: private data for lm_setup (may be NULL if lm_setup
1718 * doesn't require it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 *
1720 * The (input) flp->fl_lmops->lm_break function is required
1721 * by break_lease().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 */
Jeff Laytone6f5c782014-08-22 10:40:25 -04001723int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1724 void **priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 struct dentry *dentry = filp->f_path.dentry;
1727 struct inode *inode = dentry->d_inode;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001728 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001730 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001731 return -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 if (!S_ISREG(inode->i_mode))
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001733 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 error = security_file_lock(filp, arg);
1735 if (error)
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001736 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001738 switch (arg) {
1739 case F_UNLCK:
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001740 return generic_delete_lease(filp);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001741 case F_RDLCK:
1742 case F_WRLCK:
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001743 if (!(*flp)->fl_lmops->lm_break) {
1744 WARN_ON_ONCE(1);
1745 return -ENOLCK;
1746 }
Jeff Laytone6f5c782014-08-22 10:40:25 -04001747 return generic_add_lease(filp, arg, flp, priv);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001748 default:
Dave Jones8d657eb2012-07-13 13:35:36 -04001749 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001752EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753
Arnd Bergmannb89f4322010-09-18 15:09:31 +02001754/**
Jeff Laytone51673a2014-08-22 18:13:28 -04001755 * vfs_setlease - sets a lease on an open file
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001756 * @filp: file pointer
1757 * @arg: type of lease to obtain
1758 * @lease: file_lock to use when adding a lease
1759 * @priv: private info for lm_setup when adding a lease (may be
1760 * NULL if lm_setup doesn't require it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 *
Jeff Laytone51673a2014-08-22 18:13:28 -04001762 * Call this to establish a lease on the file. The "lease" argument is not
1763 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1764 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1765 * if not, this function will return -ENOLCK (and generate a scary-looking
1766 * stack trace).
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001767 *
1768 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1769 * may be NULL if the lm_setup operation doesn't require it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 */
Jeff Laytone6f5c782014-08-22 10:40:25 -04001771int
1772vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773{
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001774 if (filp->f_op->setlease)
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001775 return filp->f_op->setlease(filp, arg, lease, priv);
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001776 else
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001777 return generic_setlease(filp, arg, lease, priv);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001779EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001781static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782{
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001783 struct file_lock *fl;
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001784 struct fasync_struct *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 int error;
1786
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +02001787 fl = lease_alloc(filp, arg);
1788 if (IS_ERR(fl))
1789 return PTR_ERR(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001791 new = fasync_alloc();
1792 if (!new) {
1793 locks_free_lock(fl);
1794 return -ENOMEM;
1795 }
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001796 new->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001798 error = vfs_setlease(filp, arg, &fl, (void **)&new);
Jeff Layton2dfb9282014-08-11 18:14:12 -04001799 if (fl)
1800 locks_free_lock(fl);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001801 if (new)
1802 fasync_free(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 return error;
1804}
1805
1806/**
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001807 * fcntl_setlease - sets a lease on an open file
1808 * @fd: open file descriptor
1809 * @filp: file pointer
1810 * @arg: type of lease to obtain
1811 *
1812 * Call this fcntl to establish a lease on the file.
1813 * Note that you also need to call %F_SETSIG to
1814 * receive a signal when the lease is broken.
1815 */
1816int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1817{
1818 if (arg == F_UNLCK)
Jeff Laytone6f5c782014-08-22 10:40:25 -04001819 return vfs_setlease(filp, F_UNLCK, NULL, NULL);
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001820 return do_fcntl_add_lease(fd, filp, arg);
1821}
1822
1823/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1825 * @filp: The file to apply the lock to
1826 * @fl: The lock to be applied
1827 *
1828 * Add a FLOCK style lock to a file.
1829 */
1830int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1831{
1832 int error;
1833 might_sleep();
1834 for (;;) {
1835 error = flock_lock_file(filp, fl);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001836 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 break;
1838 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1839 if (!error)
1840 continue;
1841
1842 locks_delete_block(fl);
1843 break;
1844 }
1845 return error;
1846}
1847
1848EXPORT_SYMBOL(flock_lock_file_wait);
1849
1850/**
1851 * sys_flock: - flock() system call.
1852 * @fd: the file descriptor to lock.
1853 * @cmd: the type of lock to apply.
1854 *
1855 * Apply a %FL_FLOCK style lock to an open file descriptor.
1856 * The @cmd can be one of
1857 *
1858 * %LOCK_SH -- a shared lock.
1859 *
1860 * %LOCK_EX -- an exclusive lock.
1861 *
1862 * %LOCK_UN -- remove an existing lock.
1863 *
1864 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1865 *
1866 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1867 * processes read and write access respectively.
1868 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001869SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870{
Al Viro2903ff02012-08-28 12:52:22 -04001871 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 struct file_lock *lock;
1873 int can_sleep, unlock;
1874 int error;
1875
1876 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001877 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 goto out;
1879
1880 can_sleep = !(cmd & LOCK_NB);
1881 cmd &= ~LOCK_NB;
1882 unlock = (cmd == LOCK_UN);
1883
Al Viroaeb5d722008-09-02 15:28:45 -04001884 if (!unlock && !(cmd & LOCK_MAND) &&
Al Viro2903ff02012-08-28 12:52:22 -04001885 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 goto out_putf;
1887
Al Viro2903ff02012-08-28 12:52:22 -04001888 error = flock_make_lock(f.file, &lock, cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 if (error)
1890 goto out_putf;
1891 if (can_sleep)
1892 lock->fl_flags |= FL_SLEEP;
1893
Al Viro2903ff02012-08-28 12:52:22 -04001894 error = security_file_lock(f.file, lock->fl_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 if (error)
1896 goto out_free;
1897
Al Viro72c2d532013-09-22 16:27:52 -04001898 if (f.file->f_op->flock)
Al Viro2903ff02012-08-28 12:52:22 -04001899 error = f.file->f_op->flock(f.file,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 (can_sleep) ? F_SETLKW : F_SETLK,
1901 lock);
1902 else
Al Viro2903ff02012-08-28 12:52:22 -04001903 error = flock_lock_file_wait(f.file, lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904
1905 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001906 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907
1908 out_putf:
Al Viro2903ff02012-08-28 12:52:22 -04001909 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 out:
1911 return error;
1912}
1913
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001914/**
1915 * vfs_test_lock - test file byte range lock
1916 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001917 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001918 *
1919 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1920 * setting conf->fl_type to something other than F_UNLCK.
1921 */
1922int vfs_test_lock(struct file *filp, struct file_lock *fl)
1923{
Al Viro72c2d532013-09-22 16:27:52 -04001924 if (filp->f_op->lock)
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001925 return filp->f_op->lock(filp, F_GETLK, fl);
1926 posix_test_lock(filp, fl);
1927 return 0;
1928}
1929EXPORT_SYMBOL_GPL(vfs_test_lock);
1930
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001931static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1932{
Jeff Laytoncff2fce2014-04-22 08:24:32 -04001933 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001934#if BITS_PER_LONG == 32
1935 /*
1936 * Make sure we can represent the posix lock via
1937 * legacy 32bit flock.
1938 */
1939 if (fl->fl_start > OFFT_OFFSET_MAX)
1940 return -EOVERFLOW;
1941 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1942 return -EOVERFLOW;
1943#endif
1944 flock->l_start = fl->fl_start;
1945 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1946 fl->fl_end - fl->fl_start + 1;
1947 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001948 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001949 return 0;
1950}
1951
1952#if BITS_PER_LONG == 32
1953static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1954{
Jeff Laytoncff2fce2014-04-22 08:24:32 -04001955 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001956 flock->l_start = fl->fl_start;
1957 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1958 fl->fl_end - fl->fl_start + 1;
1959 flock->l_whence = 0;
1960 flock->l_type = fl->fl_type;
1961}
1962#endif
1963
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964/* Report the first existing lock that would conflict with l.
1965 * This implements the F_GETLK command of fcntl().
1966 */
Jeff Laytonc1e62b82014-02-03 12:13:09 -05001967int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001969 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 struct flock flock;
1971 int error;
1972
1973 error = -EFAULT;
1974 if (copy_from_user(&flock, l, sizeof(flock)))
1975 goto out;
1976 error = -EINVAL;
1977 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1978 goto out;
1979
1980 error = flock_to_posix_lock(filp, &file_lock, &flock);
1981 if (error)
1982 goto out;
1983
Jeff Layton0d3f7a22014-04-22 08:23:58 -04001984 if (cmd == F_OFD_GETLK) {
Jeff Layton90478932014-03-04 10:30:23 -05001985 error = -EINVAL;
1986 if (flock.l_pid != 0)
1987 goto out;
1988
Jeff Layton5d50ffd2014-02-03 12:13:10 -05001989 cmd = F_GETLK;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04001990 file_lock.fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001991 file_lock.fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05001992 }
1993
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001994 error = vfs_test_lock(filp, &file_lock);
1995 if (error)
1996 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001998 flock.l_type = file_lock.fl_type;
1999 if (file_lock.fl_type != F_UNLCK) {
2000 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05002001 if (error)
Kinglong Meef3282962014-08-22 10:18:43 -04002002 goto rel_priv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 }
2004 error = -EFAULT;
2005 if (!copy_to_user(l, &flock, sizeof(flock)))
2006 error = 0;
Kinglong Meef3282962014-08-22 10:18:43 -04002007rel_priv:
2008 locks_release_private(&file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009out:
2010 return error;
2011}
2012
Marc Eshel7723ec92007-01-18 15:08:55 -05002013/**
2014 * vfs_lock_file - file byte range lock
2015 * @filp: The file to apply the lock to
2016 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2017 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05002018 * @conf: Place to return a copy of the conflicting lock, if found.
2019 *
2020 * A caller that doesn't care about the conflicting lock may pass NULL
2021 * as the final argument.
2022 *
2023 * If the filesystem defines a private ->lock() method, then @conf will
2024 * be left unchanged; so a caller that cares should initialize it to
2025 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05002026 *
2027 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2028 * locks, the ->lock() interface may return asynchronously, before the lock has
2029 * been granted or denied by the underlying filesystem, if (and only if)
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002030 * lm_grant is set. Callers expecting ->lock() to return asynchronously
Marc Eshel2beb6612006-12-05 23:31:28 -05002031 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2032 * the request is for a blocking lock. When ->lock() does return asynchronously,
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002033 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
Marc Eshel2beb6612006-12-05 23:31:28 -05002034 * request completes.
2035 * If the request is for non-blocking lock the file system should return
Miklos Szeredibde74e42008-07-25 01:48:57 -07002036 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2037 * with the result. If the request timed out the callback routine will return a
Marc Eshel2beb6612006-12-05 23:31:28 -05002038 * nonzero return code and the file system should release the lock. The file
2039 * system is also responsible to keep a corresponding posix lock when it
2040 * grants a lock so the VFS can find out which locks are locally held and do
2041 * the correct lock cleanup when required.
2042 * The underlying filesystem must not drop the kernel lock or call
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002043 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
Marc Eshel2beb6612006-12-05 23:31:28 -05002044 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05002045 */
Marc Eshel150b3932007-01-18 16:15:35 -05002046int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05002047{
Al Viro72c2d532013-09-22 16:27:52 -04002048 if (filp->f_op->lock)
Marc Eshel7723ec92007-01-18 15:08:55 -05002049 return filp->f_op->lock(filp, cmd, fl);
2050 else
Marc Eshel150b3932007-01-18 16:15:35 -05002051 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05002052}
2053EXPORT_SYMBOL_GPL(vfs_lock_file);
2054
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002055static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2056 struct file_lock *fl)
2057{
2058 int error;
2059
2060 error = security_file_lock(filp, fl->fl_type);
2061 if (error)
2062 return error;
2063
Miklos Szeredi764c76b2008-07-25 01:48:58 -07002064 for (;;) {
2065 error = vfs_lock_file(filp, cmd, fl, NULL);
2066 if (error != FILE_LOCK_DEFERRED)
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002067 break;
Miklos Szeredi764c76b2008-07-25 01:48:58 -07002068 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2069 if (!error)
2070 continue;
2071
2072 locks_delete_block(fl);
2073 break;
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002074 }
2075
2076 return error;
2077}
2078
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002079/* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2080static int
2081check_fmode_for_setlk(struct file_lock *fl)
2082{
2083 switch (fl->fl_type) {
2084 case F_RDLCK:
2085 if (!(fl->fl_file->f_mode & FMODE_READ))
2086 return -EBADF;
2087 break;
2088 case F_WRLCK:
2089 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2090 return -EBADF;
2091 }
2092 return 0;
2093}
2094
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095/* Apply the lock described by l to an open file descriptor.
2096 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2097 */
Peter Staubachc2936212005-07-27 11:45:09 -07002098int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2099 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100{
2101 struct file_lock *file_lock = locks_alloc_lock();
2102 struct flock flock;
2103 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04002104 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 int error;
2106
2107 if (file_lock == NULL)
2108 return -ENOLCK;
2109
2110 /*
2111 * This might block, so we do it before checking the inode.
2112 */
2113 error = -EFAULT;
2114 if (copy_from_user(&flock, l, sizeof(flock)))
2115 goto out;
2116
Al Viro496ad9a2013-01-23 17:07:38 -05002117 inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118
2119 /* Don't allow mandatory locks on files that may be memory mapped
2120 * and shared.
2121 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002122 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 error = -EAGAIN;
2124 goto out;
2125 }
2126
Peter Staubachc2936212005-07-27 11:45:09 -07002127again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 error = flock_to_posix_lock(filp, file_lock, &flock);
2129 if (error)
2130 goto out;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002131
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002132 error = check_fmode_for_setlk(file_lock);
2133 if (error)
2134 goto out;
2135
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002136 /*
2137 * If the cmd is requesting file-private locks, then set the
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002138 * FL_OFDLCK flag and override the owner.
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002139 */
2140 switch (cmd) {
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002141 case F_OFD_SETLK:
Jeff Layton90478932014-03-04 10:30:23 -05002142 error = -EINVAL;
2143 if (flock.l_pid != 0)
2144 goto out;
2145
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002146 cmd = F_SETLK;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002147 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002148 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002149 break;
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002150 case F_OFD_SETLKW:
Jeff Layton90478932014-03-04 10:30:23 -05002151 error = -EINVAL;
2152 if (flock.l_pid != 0)
2153 goto out;
2154
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002155 cmd = F_SETLKW;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002156 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002157 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002158 /* Fallthrough */
2159 case F_SETLKW:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160 file_lock->fl_flags |= FL_SLEEP;
2161 }
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002162
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002163 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164
Peter Staubachc2936212005-07-27 11:45:09 -07002165 /*
2166 * Attempt to detect a close/fcntl race and recover by
2167 * releasing the lock that was just acquired.
2168 */
Al Viro0b2bac22008-05-06 13:58:34 -04002169 /*
2170 * we need that spin_lock here - it prevents reordering between
2171 * update of inode->i_flock and check for it done in close().
2172 * rcu_read_lock() wouldn't do.
2173 */
2174 spin_lock(&current->files->file_lock);
2175 f = fcheck(fd);
2176 spin_unlock(&current->files->file_lock);
2177 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07002178 flock.l_type = F_UNLCK;
2179 goto again;
2180 }
2181
2182out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 locks_free_lock(file_lock);
2184 return error;
2185}
2186
2187#if BITS_PER_LONG == 32
2188/* Report the first existing lock that would conflict with l.
2189 * This implements the F_GETLK command of fcntl().
2190 */
Jeff Laytonc1e62b82014-02-03 12:13:09 -05002191int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002193 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 struct flock64 flock;
2195 int error;
2196
2197 error = -EFAULT;
2198 if (copy_from_user(&flock, l, sizeof(flock)))
2199 goto out;
2200 error = -EINVAL;
2201 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2202 goto out;
2203
2204 error = flock64_to_posix_lock(filp, &file_lock, &flock);
2205 if (error)
2206 goto out;
2207
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002208 if (cmd == F_OFD_GETLK) {
Jeff Layton90478932014-03-04 10:30:23 -05002209 error = -EINVAL;
2210 if (flock.l_pid != 0)
2211 goto out;
2212
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002213 cmd = F_GETLK64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002214 file_lock.fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002215 file_lock.fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002216 }
2217
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002218 error = vfs_test_lock(filp, &file_lock);
2219 if (error)
2220 goto out;
2221
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002222 flock.l_type = file_lock.fl_type;
2223 if (file_lock.fl_type != F_UNLCK)
2224 posix_lock_to_flock64(&flock, &file_lock);
2225
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226 error = -EFAULT;
2227 if (!copy_to_user(l, &flock, sizeof(flock)))
2228 error = 0;
Kinglong Meef3282962014-08-22 10:18:43 -04002229
2230 locks_release_private(&file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231out:
2232 return error;
2233}
2234
2235/* Apply the lock described by l to an open file descriptor.
2236 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2237 */
Peter Staubachc2936212005-07-27 11:45:09 -07002238int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2239 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240{
2241 struct file_lock *file_lock = locks_alloc_lock();
2242 struct flock64 flock;
2243 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04002244 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245 int error;
2246
2247 if (file_lock == NULL)
2248 return -ENOLCK;
2249
2250 /*
2251 * This might block, so we do it before checking the inode.
2252 */
2253 error = -EFAULT;
2254 if (copy_from_user(&flock, l, sizeof(flock)))
2255 goto out;
2256
Al Viro496ad9a2013-01-23 17:07:38 -05002257 inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258
2259 /* Don't allow mandatory locks on files that may be memory mapped
2260 * and shared.
2261 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002262 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263 error = -EAGAIN;
2264 goto out;
2265 }
2266
Peter Staubachc2936212005-07-27 11:45:09 -07002267again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268 error = flock64_to_posix_lock(filp, file_lock, &flock);
2269 if (error)
2270 goto out;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002271
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002272 error = check_fmode_for_setlk(file_lock);
2273 if (error)
2274 goto out;
2275
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002276 /*
2277 * If the cmd is requesting file-private locks, then set the
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002278 * FL_OFDLCK flag and override the owner.
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002279 */
2280 switch (cmd) {
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002281 case F_OFD_SETLK:
Jeff Layton90478932014-03-04 10:30:23 -05002282 error = -EINVAL;
2283 if (flock.l_pid != 0)
2284 goto out;
2285
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002286 cmd = F_SETLK64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002287 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002288 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002289 break;
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002290 case F_OFD_SETLKW:
Jeff Layton90478932014-03-04 10:30:23 -05002291 error = -EINVAL;
2292 if (flock.l_pid != 0)
2293 goto out;
2294
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002295 cmd = F_SETLKW64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002296 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002297 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002298 /* Fallthrough */
2299 case F_SETLKW64:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002300 file_lock->fl_flags |= FL_SLEEP;
2301 }
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002302
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002303 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304
Peter Staubachc2936212005-07-27 11:45:09 -07002305 /*
2306 * Attempt to detect a close/fcntl race and recover by
2307 * releasing the lock that was just acquired.
2308 */
Al Viro0b2bac22008-05-06 13:58:34 -04002309 spin_lock(&current->files->file_lock);
2310 f = fcheck(fd);
2311 spin_unlock(&current->files->file_lock);
2312 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07002313 flock.l_type = F_UNLCK;
2314 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315 }
2316
2317out:
2318 locks_free_lock(file_lock);
2319 return error;
2320}
2321#endif /* BITS_PER_LONG == 32 */
2322
2323/*
2324 * This function is called when the file is being removed
2325 * from the task's fd array. POSIX locks belonging to this task
2326 * are deleted at this time.
2327 */
2328void locks_remove_posix(struct file *filp, fl_owner_t owner)
2329{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07002330 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331
2332 /*
2333 * If there are no locks held on this file, we don't need to call
2334 * posix_lock_file(). Another process could be setting a lock on this
2335 * file at the same time, but we wouldn't remove that lock anyway.
2336 */
Al Viro496ad9a2013-01-23 17:07:38 -05002337 if (!file_inode(filp)->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 return;
2339
2340 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07002341 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342 lock.fl_start = 0;
2343 lock.fl_end = OFFSET_MAX;
2344 lock.fl_owner = owner;
2345 lock.fl_pid = current->tgid;
2346 lock.fl_file = filp;
2347 lock.fl_ops = NULL;
2348 lock.fl_lmops = NULL;
2349
Marc Eshel150b3932007-01-18 16:15:35 -05002350 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2353 lock.fl_ops->fl_release_private(&lock);
2354}
2355
2356EXPORT_SYMBOL(locks_remove_posix);
2357
2358/*
2359 * This function is called on the last close of an open file.
2360 */
Jeff Layton78ed8a12014-02-03 12:13:08 -05002361void locks_remove_file(struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362{
Al Viro496ad9a2013-01-23 17:07:38 -05002363 struct inode * inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364 struct file_lock *fl;
2365 struct file_lock **before;
Jeff Laytoned9814d2014-08-11 14:20:31 -04002366 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367
2368 if (!inode->i_flock)
2369 return;
2370
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002371 locks_remove_posix(filp, filp);
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002372
Al Viro72c2d532013-09-22 16:27:52 -04002373 if (filp->f_op->flock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 struct file_lock fl = {
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002375 .fl_owner = filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 .fl_pid = current->tgid,
2377 .fl_file = filp,
2378 .fl_flags = FL_FLOCK,
2379 .fl_type = F_UNLCK,
2380 .fl_end = OFFSET_MAX,
2381 };
2382 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002383 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2384 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385 }
2386
Jeff Layton1c8c6012013-06-21 08:58:15 -04002387 spin_lock(&inode->i_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388 before = &inode->i_flock;
2389
2390 while ((fl = *before) != NULL) {
2391 if (fl->fl_file == filp) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392 if (IS_LEASE(fl)) {
Jeff Laytonc45198e2014-09-01 07:12:07 -04002393 lease_modify(before, F_UNLCK, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394 continue;
2395 }
Jeff Layton8c3cac52014-02-03 12:13:07 -05002396
2397 /*
2398 * There's a leftover lock on the list of a type that
2399 * we didn't expect to see. Most likely a classic
2400 * POSIX lock that ended up not getting released
2401 * properly, or that raced onto the list somehow. Log
2402 * some info about it and then just remove it from
2403 * the list.
2404 */
2405 WARN(!IS_FLOCK(fl),
2406 "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n",
2407 MAJOR(inode->i_sb->s_dev),
2408 MINOR(inode->i_sb->s_dev), inode->i_ino,
2409 fl->fl_type, fl->fl_flags,
2410 fl->fl_start, fl->fl_end);
2411
Jeff Laytoned9814d2014-08-11 14:20:31 -04002412 locks_delete_lock(before, &dispose);
Jeff Layton8c3cac52014-02-03 12:13:07 -05002413 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414 }
2415 before = &fl->fl_next;
2416 }
Jeff Layton1c8c6012013-06-21 08:58:15 -04002417 spin_unlock(&inode->i_lock);
Jeff Laytoned9814d2014-08-11 14:20:31 -04002418 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419}
2420
2421/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422 * posix_unblock_lock - stop waiting for a file lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423 * @waiter: the lock which was waiting
2424 *
2425 * lockd needs to block waiting for locks.
2426 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002427int
Jeff Laytonf891a292013-06-21 08:58:09 -04002428posix_unblock_lock(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002430 int status = 0;
2431
Jeff Layton7b2296a2013-06-21 08:58:20 -04002432 spin_lock(&blocked_lock_lock);
J. Bruce Fields5996a292006-01-03 09:55:44 +01002433 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002435 else
2436 status = -ENOENT;
Jeff Layton7b2296a2013-06-21 08:58:20 -04002437 spin_unlock(&blocked_lock_lock);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002438 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440EXPORT_SYMBOL(posix_unblock_lock);
2441
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002442/**
2443 * vfs_cancel_lock - file byte range unblock lock
2444 * @filp: The file to apply the unblock to
2445 * @fl: The lock to be unblocked
2446 *
2447 * Used by lock managers to cancel blocked requests
2448 */
2449int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2450{
Al Viro72c2d532013-09-22 16:27:52 -04002451 if (filp->f_op->lock)
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002452 return filp->f_op->lock(filp, F_CANCELLK, fl);
2453 return 0;
2454}
2455
2456EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2457
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002458#ifdef CONFIG_PROC_FS
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002459#include <linux/proc_fs.h>
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002460#include <linux/seq_file.h>
2461
Jeff Layton7012b022013-06-21 08:58:22 -04002462struct locks_iterator {
2463 int li_cpu;
2464 loff_t li_pos;
2465};
2466
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002467static void lock_get_status(struct seq_file *f, struct file_lock *fl,
Jerome Marchand99dc8292010-10-26 14:22:33 -07002468 loff_t id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469{
2470 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002471 unsigned int fl_pid;
2472
2473 if (fl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08002474 fl_pid = pid_vnr(fl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002475 else
2476 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477
2478 if (fl->fl_file != NULL)
Al Viro496ad9a2013-01-23 17:07:38 -05002479 inode = file_inode(fl->fl_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480
Jerome Marchand99dc8292010-10-26 14:22:33 -07002481 seq_printf(f, "%lld:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 if (IS_POSIX(fl)) {
Jeff Laytonc918d422014-02-03 12:13:09 -05002483 if (fl->fl_flags & FL_ACCESS)
Fabian Frederick5315c262014-05-09 14:13:05 -04002484 seq_puts(f, "ACCESS");
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002485 else if (IS_OFDLCK(fl))
Fabian Frederick5315c262014-05-09 14:13:05 -04002486 seq_puts(f, "OFDLCK");
Jeff Laytonc918d422014-02-03 12:13:09 -05002487 else
Fabian Frederick5315c262014-05-09 14:13:05 -04002488 seq_puts(f, "POSIX ");
Jeff Laytonc918d422014-02-03 12:13:09 -05002489
2490 seq_printf(f, " %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002492 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493 } else if (IS_FLOCK(fl)) {
2494 if (fl->fl_type & LOCK_MAND) {
Fabian Frederick5315c262014-05-09 14:13:05 -04002495 seq_puts(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002497 seq_puts(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002498 }
2499 } else if (IS_LEASE(fl)) {
Jeff Layton8144f1f2014-08-11 13:36:54 -04002500 if (fl->fl_flags & FL_DELEG)
2501 seq_puts(f, "DELEG ");
2502 else
2503 seq_puts(f, "LEASE ");
2504
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04002505 if (lease_breaking(fl))
Fabian Frederick5315c262014-05-09 14:13:05 -04002506 seq_puts(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002507 else if (fl->fl_file)
Fabian Frederick5315c262014-05-09 14:13:05 -04002508 seq_puts(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509 else
Fabian Frederick5315c262014-05-09 14:13:05 -04002510 seq_puts(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002512 seq_puts(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513 }
2514 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002515 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 (fl->fl_type & LOCK_READ)
2517 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2518 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2519 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002520 seq_printf(f, "%s ",
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04002521 (lease_breaking(fl))
Jeff Layton0ee5c6d2012-08-02 15:46:30 -04002522 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2523 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524 }
2525 if (inode) {
2526#ifdef WE_CAN_BREAK_LSLK_NOW
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002527 seq_printf(f, "%d %s:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528 inode->i_sb->s_id, inode->i_ino);
2529#else
2530 /* userspace relies on this representation of dev_t ;-( */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002531 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002532 MAJOR(inode->i_sb->s_dev),
2533 MINOR(inode->i_sb->s_dev), inode->i_ino);
2534#endif
2535 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002536 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537 }
2538 if (IS_POSIX(fl)) {
2539 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002540 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002542 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002543 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002544 seq_puts(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545 }
2546}
2547
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002548static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549{
Jeff Layton7012b022013-06-21 08:58:22 -04002550 struct locks_iterator *iter = f->private;
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002551 struct file_lock *fl, *bfl;
2552
Jeff Layton139ca042013-06-21 08:58:17 -04002553 fl = hlist_entry(v, struct file_lock, fl_link);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002554
Jeff Layton7012b022013-06-21 08:58:22 -04002555 lock_get_status(f, fl, iter->li_pos, "");
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002556
2557 list_for_each_entry(bfl, &fl->fl_block, fl_block)
Jeff Layton7012b022013-06-21 08:58:22 -04002558 lock_get_status(f, bfl, iter->li_pos, " ->");
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002559
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002560 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561}
2562
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002563static void *locks_start(struct seq_file *f, loff_t *pos)
Jeff Laytonb03dfdec2014-02-03 12:13:07 -05002564 __acquires(&blocked_lock_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565{
Jeff Layton7012b022013-06-21 08:58:22 -04002566 struct locks_iterator *iter = f->private;
Jerome Marchand99dc8292010-10-26 14:22:33 -07002567
Jeff Layton7012b022013-06-21 08:58:22 -04002568 iter->li_pos = *pos + 1;
2569 lg_global_lock(&file_lock_lglock);
Jeff Layton7b2296a2013-06-21 08:58:20 -04002570 spin_lock(&blocked_lock_lock);
Jeff Layton7012b022013-06-21 08:58:22 -04002571 return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572}
2573
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002574static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2575{
Jeff Layton7012b022013-06-21 08:58:22 -04002576 struct locks_iterator *iter = f->private;
2577
2578 ++iter->li_pos;
2579 return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002580}
2581
2582static void locks_stop(struct seq_file *f, void *v)
Jeff Laytonb03dfdec2014-02-03 12:13:07 -05002583 __releases(&blocked_lock_lock)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002584{
Jeff Layton7b2296a2013-06-21 08:58:20 -04002585 spin_unlock(&blocked_lock_lock);
Jeff Layton7012b022013-06-21 08:58:22 -04002586 lg_global_unlock(&file_lock_lglock);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002587}
2588
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002589static const struct seq_operations locks_seq_operations = {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002590 .start = locks_start,
2591 .next = locks_next,
2592 .stop = locks_stop,
2593 .show = locks_show,
2594};
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002595
2596static int locks_open(struct inode *inode, struct file *filp)
2597{
Jeff Layton7012b022013-06-21 08:58:22 -04002598 return seq_open_private(filp, &locks_seq_operations,
2599 sizeof(struct locks_iterator));
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002600}
2601
2602static const struct file_operations proc_locks_operations = {
2603 .open = locks_open,
2604 .read = seq_read,
2605 .llseek = seq_lseek,
Jerome Marchand99dc8292010-10-26 14:22:33 -07002606 .release = seq_release_private,
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002607};
2608
2609static int __init proc_locks_init(void)
2610{
2611 proc_create("locks", 0, NULL, &proc_locks_operations);
2612 return 0;
2613}
2614module_init(proc_locks_init);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002615#endif
2616
Linus Torvalds1da177e2005-04-16 15:20:36 -07002617static int __init filelock_init(void)
2618{
Jeff Layton7012b022013-06-21 08:58:22 -04002619 int i;
2620
Linus Torvalds1da177e2005-04-16 15:20:36 -07002621 filelock_cache = kmem_cache_create("file_lock_cache",
Miklos Szerediee19cc42011-07-07 13:06:09 +02002622 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2623
Jeff Layton7012b022013-06-21 08:58:22 -04002624 lg_lock_init(&file_lock_lglock, "file_lock_lglock");
2625
2626 for_each_possible_cpu(i)
2627 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i));
2628
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 return 0;
2630}
2631
2632core_initcall(filelock_init);