blob: 22c5b4aa49611ac46cb50dbf4fc8ef25ec500b34 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122#include <linux/security.h>
123#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124#include <linux/syscalls.h>
125#include <linux/time.h>
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700126#include <linux/rcupdate.h>
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000127#include <linux/pid_namespace.h>
Jeff Layton48f74182013-06-21 08:58:18 -0400128#include <linux/hashtable.h>
Jeff Layton7012b022013-06-21 08:58:22 -0400129#include <linux/percpu.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Jeff Layton62af4f12014-05-09 14:13:05 -0400131#define CREATE_TRACE_POINTS
132#include <trace/events/filelock.h>
133
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134#include <asm/uaccess.h>
135
136#define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
137#define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
Christoph Hellwig11afe9f2015-01-21 19:17:03 +0100138#define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400139#define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140
Miklos Szeredic568d682016-09-16 12:44:20 +0200141static inline bool is_remote_lock(struct file *filp)
142{
143 return likely(!(filp->f_path.dentry->d_sb->s_flags & MS_NOREMOTELOCK));
144}
145
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -0400146static bool lease_breaking(struct file_lock *fl)
147{
J. Bruce Fields778fc542011-07-26 18:25:49 -0400148 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
149}
150
151static int target_leasetype(struct file_lock *fl)
152{
153 if (fl->fl_flags & FL_UNLOCK_PENDING)
154 return F_UNLCK;
155 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
156 return F_RDLCK;
157 return fl->fl_type;
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -0400158}
159
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160int leases_enable = 1;
161int lease_break_time = 45;
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
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200165 * keep a list on each CPU, with each list protected by its own spinlock.
166 * Global serialization is done using file_rwsem.
167 *
168 * Note that alterations to the list also require that the relevant flc_lock is
169 * held.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400170 */
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200171struct file_lock_list_struct {
172 spinlock_t lock;
173 struct hlist_head hlist;
174};
175static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list);
Peter Zijlstraaba37662015-06-22 14:16:33 +0200176DEFINE_STATIC_PERCPU_RWSEM(file_rwsem);
Jeff Layton88974692013-06-21 08:58:14 -0400177
Jeff Layton1c8c6012013-06-21 08:58:15 -0400178/*
Jeff Layton48f74182013-06-21 08:58:18 -0400179 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
Jeff Layton7b2296a2013-06-21 08:58:20 -0400180 * It is protected by blocked_lock_lock.
Jeff Layton48f74182013-06-21 08:58:18 -0400181 *
182 * We hash locks by lockowner in order to optimize searching for the lock a
183 * particular lockowner is waiting on.
184 *
185 * FIXME: make this value scale via some heuristic? We generally will want more
186 * buckets when we have more lockowners holding locks, but that's a little
187 * difficult to determine without knowing what the workload will look like.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400188 */
Jeff Layton48f74182013-06-21 08:58:18 -0400189#define BLOCKED_HASH_BITS 7
190static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
Jeff Layton88974692013-06-21 08:58:14 -0400191
Jeff Layton1c8c6012013-06-21 08:58:15 -0400192/*
Jeff Layton7b2296a2013-06-21 08:58:20 -0400193 * This lock protects the blocked_hash. Generally, if you're accessing it, you
194 * want to be holding this lock.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400195 *
196 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
197 * pointer for file_lock structures that are acting as lock requests (in
198 * contrast to those that are acting as records of acquired locks).
199 *
200 * Note that when we acquire this lock in order to change the above fields,
Jeff Layton6109c852015-01-16 15:05:57 -0500201 * we often hold the flc_lock as well. In certain cases, when reading the fields
Jeff Layton1c8c6012013-06-21 08:58:15 -0400202 * protected by this lock, we can skip acquiring it iff we already hold the
Jeff Layton6109c852015-01-16 15:05:57 -0500203 * flc_lock.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400204 *
205 * In particular, adding an entry to the fl_block list requires that you hold
Jeff Layton6109c852015-01-16 15:05:57 -0500206 * both the flc_lock and the blocked_lock_lock (acquired in that order).
207 * Deleting an entry from the list however only requires the file_lock_lock.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400208 */
Jeff Layton7b2296a2013-06-21 08:58:20 -0400209static DEFINE_SPINLOCK(blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210
Jeff Layton4a075e32015-01-16 15:05:54 -0500211static struct kmem_cache *flctx_cache __read_mostly;
Christoph Lametere18b8902006-12-06 20:33:20 -0800212static struct kmem_cache *filelock_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
Jeff Layton4a075e32015-01-16 15:05:54 -0500214static struct file_lock_context *
Jeff Layton5c1c6692015-04-03 09:04:03 -0400215locks_get_lock_context(struct inode *inode, int type)
Jeff Layton4a075e32015-01-16 15:05:54 -0500216{
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200217 struct file_lock_context *ctx;
Jeff Layton4a075e32015-01-16 15:05:54 -0500218
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200219 /* paired with cmpxchg() below */
220 ctx = smp_load_acquire(&inode->i_flctx);
221 if (likely(ctx) || type == F_UNLCK)
Jeff Layton4a075e32015-01-16 15:05:54 -0500222 goto out;
223
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200224 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
225 if (!ctx)
Jeff Layton4a075e32015-01-16 15:05:54 -0500226 goto out;
227
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200228 spin_lock_init(&ctx->flc_lock);
229 INIT_LIST_HEAD(&ctx->flc_flock);
230 INIT_LIST_HEAD(&ctx->flc_posix);
231 INIT_LIST_HEAD(&ctx->flc_lease);
Jeff Layton4a075e32015-01-16 15:05:54 -0500232
233 /*
234 * Assign the pointer if it's not already assigned. If it is, then
235 * free the context we just allocated.
236 */
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200237 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
238 kmem_cache_free(flctx_cache, ctx);
239 ctx = smp_load_acquire(&inode->i_flctx);
240 }
Jeff Layton4a075e32015-01-16 15:05:54 -0500241out:
Jeff Layton18909102016-01-06 21:26:10 -0500242 trace_locks_get_lock_context(inode, type, ctx);
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200243 return ctx;
Jeff Layton4a075e32015-01-16 15:05:54 -0500244}
245
Jeff Laytone24dada2016-01-06 21:28:41 -0500246static void
247locks_dump_ctx_list(struct list_head *list, char *list_type)
248{
249 struct file_lock *fl;
250
251 list_for_each_entry(fl, list, fl_list) {
252 pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type, fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
253 }
254}
255
256static void
257locks_check_ctx_lists(struct inode *inode)
258{
259 struct file_lock_context *ctx = inode->i_flctx;
260
261 if (unlikely(!list_empty(&ctx->flc_flock) ||
262 !list_empty(&ctx->flc_posix) ||
263 !list_empty(&ctx->flc_lease))) {
264 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
265 MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev),
266 inode->i_ino);
267 locks_dump_ctx_list(&ctx->flc_flock, "FLOCK");
268 locks_dump_ctx_list(&ctx->flc_posix, "POSIX");
269 locks_dump_ctx_list(&ctx->flc_lease, "LEASE");
270 }
271}
272
Jeff Layton4a075e32015-01-16 15:05:54 -0500273void
Jeff Laytonf27a0fe2016-01-07 15:08:51 -0500274locks_free_lock_context(struct inode *inode)
Jeff Layton4a075e32015-01-16 15:05:54 -0500275{
Jeff Laytonf27a0fe2016-01-07 15:08:51 -0500276 struct file_lock_context *ctx = inode->i_flctx;
277
Jeff Laytone24dada2016-01-06 21:28:41 -0500278 if (unlikely(ctx)) {
279 locks_check_ctx_lists(inode);
Jeff Layton4a075e32015-01-16 15:05:54 -0500280 kmem_cache_free(flctx_cache, ctx);
281 }
282}
283
Miklos Szerediee19cc42011-07-07 13:06:09 +0200284static void locks_init_lock_heads(struct file_lock *fl)
Miklos Szeredia51cb912011-07-06 12:33:55 +0200285{
Jeff Layton139ca042013-06-21 08:58:17 -0400286 INIT_HLIST_NODE(&fl->fl_link);
Jeff Layton6dee60f2015-01-16 15:05:54 -0500287 INIT_LIST_HEAD(&fl->fl_list);
Miklos Szerediee19cc42011-07-07 13:06:09 +0200288 INIT_LIST_HEAD(&fl->fl_block);
289 init_waitqueue_head(&fl->fl_wait);
Miklos Szeredia51cb912011-07-06 12:33:55 +0200290}
291
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292/* Allocate an empty lock structure. */
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +0200293struct file_lock *locks_alloc_lock(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294{
Miklos Szerediee19cc42011-07-07 13:06:09 +0200295 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
Miklos Szeredia51cb912011-07-06 12:33:55 +0200296
297 if (fl)
Miklos Szerediee19cc42011-07-07 13:06:09 +0200298 locks_init_lock_heads(fl);
Miklos Szeredia51cb912011-07-06 12:33:55 +0200299
300 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301}
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +0200302EXPORT_SYMBOL_GPL(locks_alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
Felix Blyakhera9e61e22009-03-31 15:12:56 -0500304void locks_release_private(struct file_lock *fl)
Trond Myklebust47831f32006-03-20 13:44:05 -0500305{
306 if (fl->fl_ops) {
307 if (fl->fl_ops->fl_release_private)
308 fl->fl_ops->fl_release_private(fl);
309 fl->fl_ops = NULL;
310 }
Trond Myklebust47831f32006-03-20 13:44:05 -0500311
Kinglong Mee5c97d7b2014-08-22 10:18:43 -0400312 if (fl->fl_lmops) {
Jeff Laytoncae80b32015-04-03 09:04:04 -0400313 if (fl->fl_lmops->lm_put_owner) {
314 fl->fl_lmops->lm_put_owner(fl->fl_owner);
315 fl->fl_owner = NULL;
316 }
Kinglong Mee5c97d7b2014-08-22 10:18:43 -0400317 fl->fl_lmops = NULL;
318 }
Trond Myklebust47831f32006-03-20 13:44:05 -0500319}
Felix Blyakhera9e61e22009-03-31 15:12:56 -0500320EXPORT_SYMBOL_GPL(locks_release_private);
Trond Myklebust47831f32006-03-20 13:44:05 -0500321
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322/* Free a lock which is not in use. */
J. Bruce Fields05fa3132010-10-30 17:31:15 -0400323void locks_free_lock(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324{
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800325 BUG_ON(waitqueue_active(&fl->fl_wait));
Jeff Layton6dee60f2015-01-16 15:05:54 -0500326 BUG_ON(!list_empty(&fl->fl_list));
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800327 BUG_ON(!list_empty(&fl->fl_block));
Jeff Layton139ca042013-06-21 08:58:17 -0400328 BUG_ON(!hlist_unhashed(&fl->fl_link));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
Trond Myklebust47831f32006-03-20 13:44:05 -0500330 locks_release_private(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 kmem_cache_free(filelock_cache, fl);
332}
J. Bruce Fields05fa3132010-10-30 17:31:15 -0400333EXPORT_SYMBOL(locks_free_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
Jeff Laytoned9814d2014-08-11 14:20:31 -0400335static void
336locks_dispose_list(struct list_head *dispose)
337{
338 struct file_lock *fl;
339
340 while (!list_empty(dispose)) {
Jeff Layton6dee60f2015-01-16 15:05:54 -0500341 fl = list_first_entry(dispose, struct file_lock, fl_list);
342 list_del_init(&fl->fl_list);
Jeff Laytoned9814d2014-08-11 14:20:31 -0400343 locks_free_lock(fl);
344 }
345}
346
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347void locks_init_lock(struct file_lock *fl)
348{
Miklos Szerediee19cc42011-07-07 13:06:09 +0200349 memset(fl, 0, sizeof(struct file_lock));
350 locks_init_lock_heads(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351}
352
353EXPORT_SYMBOL(locks_init_lock);
354
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355/*
356 * Initialize a new lock from an existing file_lock structure.
357 */
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400358void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
360 new->fl_owner = fl->fl_owner;
361 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500362 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 new->fl_flags = fl->fl_flags;
364 new->fl_type = fl->fl_type;
365 new->fl_start = fl->fl_start;
366 new->fl_end = fl->fl_end;
Kinglong Meef3282962014-08-22 10:18:43 -0400367 new->fl_lmops = fl->fl_lmops;
Trond Myklebust09969052006-03-20 13:44:38 -0500368 new->fl_ops = NULL;
Kinglong Meef3282962014-08-22 10:18:43 -0400369
370 if (fl->fl_lmops) {
371 if (fl->fl_lmops->lm_get_owner)
Jeff Laytoncae80b32015-04-03 09:04:04 -0400372 fl->fl_lmops->lm_get_owner(fl->fl_owner);
Kinglong Meef3282962014-08-22 10:18:43 -0400373 }
Trond Myklebust09969052006-03-20 13:44:38 -0500374}
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400375EXPORT_SYMBOL(locks_copy_conflock);
Trond Myklebust09969052006-03-20 13:44:38 -0500376
377void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
378{
Jeff Layton566709b2014-08-11 14:09:35 -0400379 /* "new" must be a freshly-initialized lock */
380 WARN_ON_ONCE(new->fl_ops);
Trond Myklebust09969052006-03-20 13:44:38 -0500381
Kinglong Mee3fe0fff12014-08-22 10:18:42 -0400382 locks_copy_conflock(new, fl);
Kinglong Meef3282962014-08-22 10:18:43 -0400383
Trond Myklebust09969052006-03-20 13:44:38 -0500384 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 new->fl_ops = fl->fl_ops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500386
Kinglong Meef3282962014-08-22 10:18:43 -0400387 if (fl->fl_ops) {
388 if (fl->fl_ops->fl_copy_lock)
389 fl->fl_ops->fl_copy_lock(new, fl);
390 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391}
392
393EXPORT_SYMBOL(locks_copy_lock);
394
395static inline int flock_translate_cmd(int cmd) {
396 if (cmd & LOCK_MAND)
397 return cmd & (LOCK_MAND | LOCK_RW);
398 switch (cmd) {
399 case LOCK_SH:
400 return F_RDLCK;
401 case LOCK_EX:
402 return F_WRLCK;
403 case LOCK_UN:
404 return F_UNLCK;
405 }
406 return -EINVAL;
407}
408
409/* Fill in a file_lock structure with an appropriate FLOCK lock. */
Jeff Layton6e129d002014-09-04 10:25:06 -0400410static struct file_lock *
411flock_make_lock(struct file *filp, unsigned int cmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412{
413 struct file_lock *fl;
414 int type = flock_translate_cmd(cmd);
Jeff Layton6e129d002014-09-04 10:25:06 -0400415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 if (type < 0)
Jeff Layton6e129d002014-09-04 10:25:06 -0400417 return ERR_PTR(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418
419 fl = locks_alloc_lock();
420 if (fl == NULL)
Jeff Layton6e129d002014-09-04 10:25:06 -0400421 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422
423 fl->fl_file = filp;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +0200424 fl->fl_owner = filp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 fl->fl_pid = current->tgid;
426 fl->fl_flags = FL_FLOCK;
427 fl->fl_type = type;
428 fl->fl_end = OFFSET_MAX;
429
Jeff Layton6e129d002014-09-04 10:25:06 -0400430 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431}
432
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400433static int assign_type(struct file_lock *fl, long type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434{
435 switch (type) {
436 case F_RDLCK:
437 case F_WRLCK:
438 case F_UNLCK:
439 fl->fl_type = type;
440 break;
441 default:
442 return -EINVAL;
443 }
444 return 0;
445}
446
J. Bruce Fieldsef12e722014-02-03 12:13:08 -0500447static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
448 struct flock64 *l)
449{
450 switch (l->l_whence) {
451 case SEEK_SET:
452 fl->fl_start = 0;
453 break;
454 case SEEK_CUR:
455 fl->fl_start = filp->f_pos;
456 break;
457 case SEEK_END:
458 fl->fl_start = i_size_read(file_inode(filp));
459 break;
460 default:
461 return -EINVAL;
462 }
463 if (l->l_start > OFFSET_MAX - fl->fl_start)
464 return -EOVERFLOW;
465 fl->fl_start += l->l_start;
466 if (fl->fl_start < 0)
467 return -EINVAL;
468
469 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
470 POSIX-2001 defines it. */
471 if (l->l_len > 0) {
472 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
473 return -EOVERFLOW;
474 fl->fl_end = fl->fl_start + l->l_len - 1;
475
476 } else if (l->l_len < 0) {
477 if (fl->fl_start + l->l_len < 0)
478 return -EINVAL;
479 fl->fl_end = fl->fl_start - 1;
480 fl->fl_start += l->l_len;
481 } else
482 fl->fl_end = OFFSET_MAX;
483
484 fl->fl_owner = current->files;
485 fl->fl_pid = current->tgid;
486 fl->fl_file = filp;
487 fl->fl_flags = FL_POSIX;
488 fl->fl_ops = NULL;
489 fl->fl_lmops = NULL;
490
491 return assign_type(fl, l->l_type);
492}
493
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
495 * style lock.
496 */
497static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
498 struct flock *l)
499{
J. Bruce Fieldsef12e722014-02-03 12:13:08 -0500500 struct flock64 ll = {
501 .l_type = l->l_type,
502 .l_whence = l->l_whence,
503 .l_start = l->l_start,
504 .l_len = l->l_len,
505 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
J. Bruce Fieldsef12e722014-02-03 12:13:08 -0500507 return flock64_to_posix_lock(filp, fl, &ll);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508}
509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510/* default lease lock manager operations */
Jeff Layton4d01b7f2014-09-01 15:06:54 -0400511static bool
512lease_break_callback(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
514 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
Jeff Layton4d01b7f2014-09-01 15:06:54 -0400515 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516}
517
Jeff Layton1c7dd2f2014-08-22 10:55:47 -0400518static void
519lease_setup(struct file_lock *fl, void **priv)
520{
521 struct file *filp = fl->fl_file;
522 struct fasync_struct *fa = *priv;
523
524 /*
525 * fasync_insert_entry() returns the old entry if any. If there was no
526 * old entry, then it used "priv" and inserted it into the fasync list.
527 * Clear the pointer to indicate that it shouldn't be freed.
528 */
529 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
530 *priv = NULL;
531
532 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
533}
534
Alexey Dobriyan7b021962009-09-21 17:01:12 -0700535static const struct lock_manager_operations lease_manager_ops = {
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400536 .lm_break = lease_break_callback,
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400537 .lm_change = lease_modify,
Jeff Layton1c7dd2f2014-08-22 10:55:47 -0400538 .lm_setup = lease_setup,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539};
540
541/*
542 * Initialize a lease, use the default lock manager operations
543 */
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400544static int lease_init(struct file *filp, long type, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400546 if (assign_type(fl, type) != 0)
547 return -EINVAL;
548
Jeff Layton7ca76312014-09-01 19:04:48 -0400549 fl->fl_owner = filp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 fl->fl_pid = current->tgid;
551
552 fl->fl_file = filp;
553 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 fl->fl_start = 0;
555 fl->fl_end = OFFSET_MAX;
556 fl->fl_ops = NULL;
557 fl->fl_lmops = &lease_manager_ops;
558 return 0;
559}
560
561/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fields0ec4f432012-07-23 15:17:17 -0400562static struct file_lock *lease_alloc(struct file *filp, long type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563{
564 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400565 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566
567 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500568 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
570 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400571 if (error) {
572 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500573 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400574 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500575 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576}
577
578/* Check if two locks overlap each other.
579 */
580static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
581{
582 return ((fl1->fl_end >= fl2->fl_start) &&
583 (fl2->fl_end >= fl1->fl_start));
584}
585
586/*
587 * Check whether two locks have the same owner.
588 */
Matt Mackall33443c42006-01-08 01:05:22 -0800589static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590{
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400591 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 return fl2->fl_lmops == fl1->fl_lmops &&
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400593 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 return fl1->fl_owner == fl2->fl_owner;
595}
596
Jeff Layton6109c852015-01-16 15:05:57 -0500597/* Must be called with the flc_lock held! */
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500598static void locks_insert_global_locks(struct file_lock *fl)
Jeff Layton88974692013-06-21 08:58:14 -0400599{
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200600 struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list);
601
Peter Zijlstraaba37662015-06-22 14:16:33 +0200602 percpu_rwsem_assert_held(&file_rwsem);
603
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200604 spin_lock(&fll->lock);
Jeff Layton7012b022013-06-21 08:58:22 -0400605 fl->fl_link_cpu = smp_processor_id();
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200606 hlist_add_head(&fl->fl_link, &fll->hlist);
607 spin_unlock(&fll->lock);
Jeff Layton88974692013-06-21 08:58:14 -0400608}
609
Jeff Layton6109c852015-01-16 15:05:57 -0500610/* Must be called with the flc_lock held! */
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500611static void locks_delete_global_locks(struct file_lock *fl)
Jeff Layton88974692013-06-21 08:58:14 -0400612{
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200613 struct file_lock_list_struct *fll;
614
Peter Zijlstraaba37662015-06-22 14:16:33 +0200615 percpu_rwsem_assert_held(&file_rwsem);
616
Jeff Layton7012b022013-06-21 08:58:22 -0400617 /*
618 * Avoid taking lock if already unhashed. This is safe since this check
Jeff Layton6109c852015-01-16 15:05:57 -0500619 * is done while holding the flc_lock, and new insertions into the list
Jeff Layton7012b022013-06-21 08:58:22 -0400620 * also require that it be held.
621 */
622 if (hlist_unhashed(&fl->fl_link))
623 return;
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200624
625 fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu);
626 spin_lock(&fll->lock);
Jeff Layton139ca042013-06-21 08:58:17 -0400627 hlist_del_init(&fl->fl_link);
Peter Zijlstra7c3f6542015-06-22 14:16:34 +0200628 spin_unlock(&fll->lock);
Jeff Layton88974692013-06-21 08:58:14 -0400629}
630
Jeff Layton3999e492013-06-21 08:58:19 -0400631static unsigned long
632posix_owner_key(struct file_lock *fl)
633{
634 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
635 return fl->fl_lmops->lm_owner_key(fl);
636 return (unsigned long)fl->fl_owner;
637}
638
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500639static void locks_insert_global_blocked(struct file_lock *waiter)
Jeff Layton88974692013-06-21 08:58:14 -0400640{
Daniel Wagner663d5af2015-04-03 09:04:03 -0400641 lockdep_assert_held(&blocked_lock_lock);
642
Jeff Layton3999e492013-06-21 08:58:19 -0400643 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
Jeff Layton88974692013-06-21 08:58:14 -0400644}
645
Jeff Layton6ca10ed2014-02-03 12:13:07 -0500646static void locks_delete_global_blocked(struct file_lock *waiter)
Jeff Layton88974692013-06-21 08:58:14 -0400647{
Daniel Wagner663d5af2015-04-03 09:04:03 -0400648 lockdep_assert_held(&blocked_lock_lock);
649
Jeff Layton48f74182013-06-21 08:58:18 -0400650 hash_del(&waiter->fl_link);
Jeff Layton88974692013-06-21 08:58:14 -0400651}
652
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653/* Remove waiter from blocker's block list.
654 * When blocker ends up pointing to itself then the list is empty.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400655 *
Jeff Layton7b2296a2013-06-21 08:58:20 -0400656 * Must be called with blocked_lock_lock held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 */
Matt Mackall33443c42006-01-08 01:05:22 -0800658static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659{
Jeff Layton88974692013-06-21 08:58:14 -0400660 locks_delete_global_blocked(waiter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 list_del_init(&waiter->fl_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 waiter->fl_next = NULL;
663}
664
Jeff Layton1a9e64a2013-06-21 08:58:10 -0400665static void locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666{
Jeff Layton7b2296a2013-06-21 08:58:20 -0400667 spin_lock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 __locks_delete_block(waiter);
Jeff Layton7b2296a2013-06-21 08:58:20 -0400669 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670}
671
672/* Insert waiter into blocker's block list.
673 * We use a circular list so that processes can be easily woken up in
674 * the order they blocked. The documentation doesn't require this but
675 * it seems like the reasonable thing to do.
Jeff Layton1c8c6012013-06-21 08:58:15 -0400676 *
Jeff Layton6109c852015-01-16 15:05:57 -0500677 * Must be called with both the flc_lock and blocked_lock_lock held. The
678 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
679 * that the flc_lock is also held on insertions we can avoid taking the
680 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 */
Jeff Layton1c8c6012013-06-21 08:58:15 -0400682static void __locks_insert_block(struct file_lock *blocker,
683 struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800685 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 waiter->fl_next = blocker;
Jeff Layton88974692013-06-21 08:58:14 -0400687 list_add_tail(&waiter->fl_block, &blocker->fl_block);
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400688 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
Jeff Layton1c8c6012013-06-21 08:58:15 -0400689 locks_insert_global_blocked(waiter);
690}
691
Jeff Layton6109c852015-01-16 15:05:57 -0500692/* Must be called with flc_lock held. */
Jeff Layton1c8c6012013-06-21 08:58:15 -0400693static void locks_insert_block(struct file_lock *blocker,
694 struct file_lock *waiter)
695{
Jeff Layton7b2296a2013-06-21 08:58:20 -0400696 spin_lock(&blocked_lock_lock);
Jeff Layton1c8c6012013-06-21 08:58:15 -0400697 __locks_insert_block(blocker, waiter);
Jeff Layton7b2296a2013-06-21 08:58:20 -0400698 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699}
700
Jeff Layton1cb36012013-06-21 08:58:12 -0400701/*
702 * Wake up processes blocked waiting for blocker.
703 *
Jeff Layton6109c852015-01-16 15:05:57 -0500704 * Must be called with the inode->flc_lock held!
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 */
706static void locks_wake_up_blocks(struct file_lock *blocker)
707{
Jeff Layton4e8c7652013-06-21 08:58:16 -0400708 /*
709 * Avoid taking global lock if list is empty. This is safe since new
Jeff Layton6109c852015-01-16 15:05:57 -0500710 * blocked requests are only added to the list under the flc_lock, and
711 * the flc_lock is always held here. Note that removal from the fl_block
712 * list does not require the flc_lock, so we must recheck list_empty()
Jeff Layton7b2296a2013-06-21 08:58:20 -0400713 * after acquiring the blocked_lock_lock.
Jeff Layton4e8c7652013-06-21 08:58:16 -0400714 */
715 if (list_empty(&blocker->fl_block))
716 return;
717
Jeff Layton7b2296a2013-06-21 08:58:20 -0400718 spin_lock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400720 struct file_lock *waiter;
721
722 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 struct file_lock, fl_block);
724 __locks_delete_block(waiter);
J. Bruce Fields8fb47a42011-07-20 20:21:59 -0400725 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
726 waiter->fl_lmops->lm_notify(waiter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 else
728 wake_up(&waiter->fl_wait);
729 }
Jeff Layton7b2296a2013-06-21 08:58:20 -0400730 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731}
732
Jeff Layton5263e312015-01-16 15:05:55 -0500733static void
Jeff Laytone084c1b2015-02-16 14:32:03 -0500734locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
Jeff Layton5263e312015-01-16 15:05:55 -0500735{
736 fl->fl_nspid = get_pid(task_tgid(current));
737 list_add_tail(&fl->fl_list, before);
738 locks_insert_global_locks(fl);
739}
740
Jeff Layton8634b512015-01-16 15:05:55 -0500741static void
Jeff Laytone084c1b2015-02-16 14:32:03 -0500742locks_unlink_lock_ctx(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743{
Jeff Layton88974692013-06-21 08:58:14 -0400744 locks_delete_global_locks(fl);
Jeff Layton8634b512015-01-16 15:05:55 -0500745 list_del_init(&fl->fl_list);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000746 if (fl->fl_nspid) {
747 put_pid(fl->fl_nspid);
748 fl->fl_nspid = NULL;
749 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 locks_wake_up_blocks(fl);
Jeff Layton24cbe782014-02-03 12:13:06 -0500751}
752
Jeff Layton5263e312015-01-16 15:05:55 -0500753static void
Jeff Laytone084c1b2015-02-16 14:32:03 -0500754locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
Jeff Layton5263e312015-01-16 15:05:55 -0500755{
Jeff Laytone084c1b2015-02-16 14:32:03 -0500756 locks_unlink_lock_ctx(fl);
Jeff Layton8634b512015-01-16 15:05:55 -0500757 if (dispose)
758 list_add(&fl->fl_list, dispose);
759 else
760 locks_free_lock(fl);
Jeff Layton5263e312015-01-16 15:05:55 -0500761}
762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
764 * checks for shared/exclusive status of overlapping locks.
765 */
766static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
767{
768 if (sys_fl->fl_type == F_WRLCK)
769 return 1;
770 if (caller_fl->fl_type == F_WRLCK)
771 return 1;
772 return 0;
773}
774
775/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
776 * checking before calling the locks_conflict().
777 */
778static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
779{
780 /* POSIX locks owned by the same process do not conflict with
781 * each other.
782 */
Jeff Layton9b8c8692015-04-03 09:04:02 -0400783 if (posix_same_owner(caller_fl, sys_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 return (0);
785
786 /* Check whether they overlap */
787 if (!locks_overlap(caller_fl, sys_fl))
788 return 0;
789
790 return (locks_conflict(caller_fl, sys_fl));
791}
792
793/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
794 * checking before calling the locks_conflict().
795 */
796static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
797{
798 /* FLOCK locks referring to the same filp do not conflict with
799 * each other.
800 */
Jeff Layton9b8c8692015-04-03 09:04:02 -0400801 if (caller_fl->fl_file == sys_fl->fl_file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 return (0);
803 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
804 return 0;
805
806 return (locks_conflict(caller_fl, sys_fl));
807}
808
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400809void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500810posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811{
812 struct file_lock *cfl;
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500813 struct file_lock_context *ctx;
Miklos Szeredic568d682016-09-16 12:44:20 +0200814 struct inode *inode = locks_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815
Dmitry Vyukov128a3782015-09-21 09:43:06 +0200816 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500817 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400818 fl->fl_type = F_UNLCK;
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500819 return;
820 }
821
Jeff Layton6109c852015-01-16 15:05:57 -0500822 spin_lock(&ctx->flc_lock);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500823 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
824 if (posix_locks_conflict(fl, cfl)) {
825 locks_copy_conflock(fl, cfl);
826 if (cfl->fl_nspid)
827 fl->fl_pid = pid_vnr(cfl->fl_nspid);
828 goto out;
829 }
830 }
831 fl->fl_type = F_UNLCK;
832out:
Jeff Layton6109c852015-01-16 15:05:57 -0500833 spin_unlock(&ctx->flc_lock);
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400834 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836EXPORT_SYMBOL(posix_test_lock);
837
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400838/*
839 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400841 * We attempt to detect deadlocks that are due purely to posix file
842 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400844 * We assume that a task can be waiting for at most one lock at a time.
845 * So for any acquired lock, the process holding that lock may be
846 * waiting on at most one other lock. That lock in turns may be held by
847 * someone waiting for at most one other lock. Given a requested lock
848 * caller_fl which is about to wait for a conflicting lock block_fl, we
849 * follow this chain of waiters to ensure we are not about to create a
850 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400851 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400852 * Since we do this before we ever put a process to sleep on a lock, we
853 * are ensured that there is never a cycle; that is what guarantees that
854 * the while() loop in posix_locks_deadlock() eventually completes.
855 *
856 * Note: the above assumption may not be true when handling lock
857 * requests from a broken NFS client. It may also fail in the presence
858 * of tasks (such as posix threads) sharing the same open file table.
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400859 * To handle those cases, we just bail out after a few iterations.
Jeff Layton57b65322014-02-03 12:13:09 -0500860 *
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400861 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
Jeff Layton57b65322014-02-03 12:13:09 -0500862 * Because the owner is not even nominally tied to a thread of
863 * execution, the deadlock detection below can't reasonably work well. Just
864 * skip it for those.
865 *
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400866 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
Jeff Layton57b65322014-02-03 12:13:09 -0500867 * locks that just checks for the case where two tasks are attempting to
868 * upgrade from read to write locks on the same inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400870
871#define MAX_DEADLK_ITERATIONS 10
872
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400873/* Find a lock that the owner of the given block_fl is blocking on. */
874static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
875{
876 struct file_lock *fl;
877
Jeff Layton3999e492013-06-21 08:58:19 -0400878 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400879 if (posix_same_owner(fl, block_fl))
880 return fl->fl_next;
881 }
882 return NULL;
883}
884
Jeff Layton7b2296a2013-06-21 08:58:20 -0400885/* Must be called with the blocked_lock_lock held! */
Adrian Bunkb0904e12006-06-23 02:05:13 -0700886static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 struct file_lock *block_fl)
888{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400889 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
Daniel Wagner663d5af2015-04-03 09:04:03 -0400891 lockdep_assert_held(&blocked_lock_lock);
892
Jeff Layton57b65322014-02-03 12:13:09 -0500893 /*
894 * This deadlock detector can't reasonably detect deadlocks with
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400895 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
Jeff Layton57b65322014-02-03 12:13:09 -0500896 */
Jeff Laytoncff2fce2014-04-22 08:24:32 -0400897 if (IS_OFDLCK(caller_fl))
Jeff Layton57b65322014-02-03 12:13:09 -0500898 return 0;
899
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400900 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
901 if (i++ > MAX_DEADLK_ITERATIONS)
902 return 0;
903 if (posix_same_owner(caller_fl, block_fl))
904 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 }
906 return 0;
907}
908
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400910 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400911 *
912 * Note that if called with an FL_EXISTS argument, the caller may determine
913 * whether or not a lock was successfully freed by testing the return
914 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 */
Jeff Laytonbcd7f782015-07-11 06:43:02 -0400916static int flock_lock_inode(struct inode *inode, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800918 struct file_lock *new_fl = NULL;
Jeff Layton5263e312015-01-16 15:05:55 -0500919 struct file_lock *fl;
920 struct file_lock_context *ctx;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 int error = 0;
Jeff Layton5263e312015-01-16 15:05:55 -0500922 bool found = false;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400923 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
Jeff Layton5c1c6692015-04-03 09:04:03 -0400925 ctx = locks_get_lock_context(inode, request->fl_type);
926 if (!ctx) {
927 if (request->fl_type != F_UNLCK)
928 return -ENOMEM;
929 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
930 }
Jeff Layton5263e312015-01-16 15:05:55 -0500931
Arnd Bergmannb89f4322010-09-18 15:09:31 +0200932 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
933 new_fl = locks_alloc_lock();
934 if (!new_fl)
935 return -ENOMEM;
936 }
937
Peter Zijlstra87709e22016-05-30 16:48:35 +0200938 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -0500939 spin_lock(&ctx->flc_lock);
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400940 if (request->fl_flags & FL_ACCESS)
941 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400942
Jeff Layton5263e312015-01-16 15:05:55 -0500943 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
Jeff Laytonbcd7f782015-07-11 06:43:02 -0400944 if (request->fl_file != fl->fl_file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800946 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 goto out;
Jeff Layton5263e312015-01-16 15:05:55 -0500948 found = true;
Jeff Laytone084c1b2015-02-16 14:32:03 -0500949 locks_delete_lock_ctx(fl, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 break;
951 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952
Trond Myklebustf475ae92006-06-29 16:38:32 -0400953 if (request->fl_type == F_UNLCK) {
954 if ((request->fl_flags & FL_EXISTS) && !found)
955 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800956 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400957 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400959find_conflict:
Jeff Layton5263e312015-01-16 15:05:55 -0500960 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
Trond Myklebust993dfa82006-03-31 02:30:55 -0800961 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 continue;
963 error = -EAGAIN;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700964 if (!(request->fl_flags & FL_SLEEP))
965 goto out;
966 error = FILE_LOCK_DEFERRED;
967 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 goto out;
969 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400970 if (request->fl_flags & FL_ACCESS)
971 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800972 locks_copy_lock(new_fl, request);
Jeff Laytone084c1b2015-02-16 14:32:03 -0500973 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800974 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400975 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
977out:
Jeff Layton6109c852015-01-16 15:05:57 -0500978 spin_unlock(&ctx->flc_lock);
Peter Zijlstra87709e22016-05-30 16:48:35 +0200979 percpu_up_read_preempt_enable(&file_rwsem);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800980 if (new_fl)
981 locks_free_lock(new_fl);
Jeff Laytoned9814d2014-08-11 14:20:31 -0400982 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 return error;
984}
985
Jeff Laytonb4d629a2016-01-07 18:27:42 -0500986static int posix_lock_inode(struct inode *inode, struct file_lock *request,
987 struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988{
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500989 struct file_lock *fl, *tmp;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700990 struct file_lock *new_fl = NULL;
991 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 struct file_lock *left = NULL;
993 struct file_lock *right = NULL;
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500994 struct file_lock_context *ctx;
Jeff Laytonb9746ef2013-06-21 08:58:13 -0400995 int error;
996 bool added = false;
Jeff Laytoned9814d2014-08-11 14:20:31 -0400997 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
Jeff Layton5c1c6692015-04-03 09:04:03 -0400999 ctx = locks_get_lock_context(inode, request->fl_type);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001000 if (!ctx)
Jeff Layton5c1c6692015-04-03 09:04:03 -04001001 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001002
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 /*
1004 * We may need two file_lock structures for this operation,
1005 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -07001006 *
1007 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 */
Miklos Szeredi39005d02006-06-23 02:05:10 -07001009 if (!(request->fl_flags & FL_ACCESS) &&
1010 (request->fl_type != F_UNLCK ||
1011 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
1012 new_fl = locks_alloc_lock();
1013 new_fl2 = locks_alloc_lock();
1014 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
Peter Zijlstra87709e22016-05-30 16:48:35 +02001016 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05001017 spin_lock(&ctx->flc_lock);
Jeff Layton1cb36012013-06-21 08:58:12 -04001018 /*
1019 * New lock request. Walk all POSIX locks and look for conflicts. If
1020 * there are any, either return error or put the request on the
Jeff Layton48f74182013-06-21 08:58:18 -04001021 * blocker's list of waiters and the global blocked_hash.
Jeff Layton1cb36012013-06-21 08:58:12 -04001022 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 if (request->fl_type != F_UNLCK) {
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001024 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 if (!posix_locks_conflict(request, fl))
1026 continue;
Andy Adamson5842add2006-03-26 01:37:26 -08001027 if (conflock)
Kinglong Mee3fe0fff12014-08-22 10:18:42 -04001028 locks_copy_conflock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 error = -EAGAIN;
1030 if (!(request->fl_flags & FL_SLEEP))
1031 goto out;
Jeff Layton1c8c6012013-06-21 08:58:15 -04001032 /*
1033 * Deadlock detection and insertion into the blocked
1034 * locks list must be done while holding the same lock!
1035 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 error = -EDEADLK;
Jeff Layton7b2296a2013-06-21 08:58:20 -04001037 spin_lock(&blocked_lock_lock);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001038 if (likely(!posix_locks_deadlock(request, fl))) {
1039 error = FILE_LOCK_DEFERRED;
1040 __locks_insert_block(fl, request);
1041 }
Jeff Layton7b2296a2013-06-21 08:58:20 -04001042 spin_unlock(&blocked_lock_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 goto out;
1044 }
1045 }
1046
1047 /* If we're just looking for a conflict, we're done. */
1048 error = 0;
1049 if (request->fl_flags & FL_ACCESS)
1050 goto out;
1051
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001052 /* Find the first old lock with the same owner as the new lock */
1053 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1054 if (posix_same_owner(request, fl))
1055 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 }
1057
Jeff Layton1cb36012013-06-21 08:58:12 -04001058 /* Process locks with this owner. */
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001059 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1060 if (!posix_same_owner(request, fl))
1061 break;
1062
1063 /* Detect adjacent or overlapping regions (if same lock type) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -07001065 /* In all comparisons of start vs end, use
1066 * "start - 1" rather than "end + 1". If end
1067 * is OFFSET_MAX, end + 1 will become negative.
1068 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 if (fl->fl_end < request->fl_start - 1)
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001070 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 /* If the next lock in the list has entirely bigger
1072 * addresses than the new one, insert the lock here.
1073 */
Olaf Kirch449231d2005-08-25 16:25:35 -07001074 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 break;
1076
1077 /* If we come here, the new and old lock are of the
1078 * same type and adjacent or overlapping. Make one
1079 * lock yielding from the lower start address of both
1080 * locks to the higher end address.
1081 */
1082 if (fl->fl_start > request->fl_start)
1083 fl->fl_start = request->fl_start;
1084 else
1085 request->fl_start = fl->fl_start;
1086 if (fl->fl_end < request->fl_end)
1087 fl->fl_end = request->fl_end;
1088 else
1089 request->fl_end = fl->fl_end;
1090 if (added) {
Jeff Laytone084c1b2015-02-16 14:32:03 -05001091 locks_delete_lock_ctx(fl, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 continue;
1093 }
1094 request = fl;
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001095 added = true;
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001096 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 /* Processing for different lock types is a bit
1098 * more complex.
1099 */
1100 if (fl->fl_end < request->fl_start)
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001101 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 if (fl->fl_start > request->fl_end)
1103 break;
1104 if (request->fl_type == F_UNLCK)
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001105 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 if (fl->fl_start < request->fl_start)
1107 left = fl;
1108 /* If the next lock in the list has a higher end
1109 * address than the new one, insert the new one here.
1110 */
1111 if (fl->fl_end > request->fl_end) {
1112 right = fl;
1113 break;
1114 }
1115 if (fl->fl_start >= request->fl_start) {
1116 /* The new lock completely replaces an old
1117 * one (This may happen several times).
1118 */
1119 if (added) {
Jeff Laytone084c1b2015-02-16 14:32:03 -05001120 locks_delete_lock_ctx(fl, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 continue;
1122 }
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001123 /*
1124 * Replace the old lock with new_fl, and
1125 * remove the old one. It's safe to do the
1126 * insert here since we know that we won't be
1127 * using new_fl later, and that the lock is
1128 * just replacing an existing lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 */
Jeff Laytonb84d49f2014-08-12 08:03:49 -04001130 error = -ENOLCK;
1131 if (!new_fl)
1132 goto out;
1133 locks_copy_lock(new_fl, request);
1134 request = new_fl;
1135 new_fl = NULL;
Jeff Laytone084c1b2015-02-16 14:32:03 -05001136 locks_insert_lock_ctx(request, &fl->fl_list);
1137 locks_delete_lock_ctx(fl, &dispose);
Jeff Laytonb9746ef2013-06-21 08:58:13 -04001138 added = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 }
1140 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 }
1142
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001143 /*
Jeff Layton1cb36012013-06-21 08:58:12 -04001144 * The above code only modifies existing locks in case of merging or
1145 * replacing. If new lock(s) need to be inserted all modifications are
1146 * done below this, so it's safe yet to bail out.
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001147 */
1148 error = -ENOLCK; /* "no luck" */
1149 if (right && left == right && !new_fl2)
1150 goto out;
1151
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 error = 0;
1153 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -04001154 if (request->fl_type == F_UNLCK) {
1155 if (request->fl_flags & FL_EXISTS)
1156 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -04001158 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -07001159
1160 if (!new_fl) {
1161 error = -ENOLCK;
1162 goto out;
1163 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 locks_copy_lock(new_fl, request);
Jeff Laytone084c1b2015-02-16 14:32:03 -05001165 locks_insert_lock_ctx(new_fl, &fl->fl_list);
Jeff Layton2e2f756f2015-02-17 17:08:23 -05001166 fl = new_fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 new_fl = NULL;
1168 }
1169 if (right) {
1170 if (left == right) {
1171 /* The new lock breaks the old one in two pieces,
1172 * so we have to use the second new lock.
1173 */
1174 left = new_fl2;
1175 new_fl2 = NULL;
1176 locks_copy_lock(left, right);
Jeff Laytone084c1b2015-02-16 14:32:03 -05001177 locks_insert_lock_ctx(left, &fl->fl_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 }
1179 right->fl_start = request->fl_end + 1;
1180 locks_wake_up_blocks(right);
1181 }
1182 if (left) {
1183 left->fl_end = request->fl_start - 1;
1184 locks_wake_up_blocks(left);
1185 }
1186 out:
Jeff Layton6109c852015-01-16 15:05:57 -05001187 spin_unlock(&ctx->flc_lock);
Peter Zijlstra87709e22016-05-30 16:48:35 +02001188 percpu_up_read_preempt_enable(&file_rwsem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 /*
1190 * Free any unused locks.
1191 */
1192 if (new_fl)
1193 locks_free_lock(new_fl);
1194 if (new_fl2)
1195 locks_free_lock(new_fl2);
Jeff Laytoned9814d2014-08-11 14:20:31 -04001196 locks_dispose_list(&dispose);
Jeff Layton18909102016-01-06 21:26:10 -05001197 trace_posix_lock_inode(inode, request, error);
1198
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 return error;
1200}
1201
1202/**
1203 * posix_lock_file - Apply a POSIX-style lock to a file
1204 * @filp: The file to apply the lock to
1205 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001206 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 *
1208 * Add a POSIX style lock to a file.
1209 * We merge adjacent & overlapping locks whenever possible.
1210 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001211 *
1212 * Note that if called with an FL_EXISTS argument, the caller may determine
1213 * whether or not a lock was successfully freed by testing the return
1214 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 */
Marc Eshel150b3932007-01-18 16:15:35 -05001216int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001217 struct file_lock *conflock)
1218{
Miklos Szeredic568d682016-09-16 12:44:20 +02001219 return posix_lock_inode(locks_inode(filp), fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001220}
Marc Eshel150b3932007-01-18 16:15:35 -05001221EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222
1223/**
Jeff Layton29d01b22015-07-11 06:43:02 -04001224 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1225 * @inode: inode of file to which lock request should be applied
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 * @fl: The lock to be applied
1227 *
Benjamin Coddington616fb382015-10-22 13:38:15 -04001228 * Apply a POSIX style lock request to an inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 */
Benjamin Coddington616fb382015-10-22 13:38:15 -04001230static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231{
1232 int error;
1233 might_sleep ();
1234 for (;;) {
Jeff Laytonb4d629a2016-01-07 18:27:42 -05001235 error = posix_lock_inode(inode, fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001236 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 break;
1238 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1239 if (!error)
1240 continue;
1241
1242 locks_delete_block(fl);
1243 break;
1244 }
1245 return error;
1246}
Jeff Layton29d01b22015-07-11 06:43:02 -04001247
Jeff Layton9e8925b2015-11-16 09:49:34 -05001248#ifdef CONFIG_MANDATORY_FILE_LOCKING
Jeff Layton29d01b22015-07-11 06:43:02 -04001249/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 * locks_mandatory_locked - Check for an active lock
Jeff Laytond7a06982014-03-10 09:54:15 -04001251 * @file: the file to check
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 *
1253 * Searches the inode's list of locks to find any POSIX locks which conflict.
1254 * This function is called from locks_verify_locked() only.
1255 */
Jeff Laytond7a06982014-03-10 09:54:15 -04001256int locks_mandatory_locked(struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257{
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001258 int ret;
Miklos Szeredic568d682016-09-16 12:44:20 +02001259 struct inode *inode = locks_inode(file);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001260 struct file_lock_context *ctx;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 struct file_lock *fl;
1262
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001263 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001264 if (!ctx || list_empty_careful(&ctx->flc_posix))
1265 return 0;
1266
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 /*
1268 * Search the lock list for this inode for any POSIX locks.
1269 */
Jeff Layton6109c852015-01-16 15:05:57 -05001270 spin_lock(&ctx->flc_lock);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001271 ret = 0;
1272 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001273 if (fl->fl_owner != current->files &&
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001274 fl->fl_owner != file) {
1275 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 break;
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001277 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 }
Jeff Layton6109c852015-01-16 15:05:57 -05001279 spin_unlock(&ctx->flc_lock);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05001280 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281}
1282
1283/**
1284 * locks_mandatory_area - Check for a conflicting lock
Christoph Hellwigacc15572015-12-03 12:59:49 +01001285 * @inode: the file to check
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 * @filp: how the file was opened (if it was)
Christoph Hellwigacc15572015-12-03 12:59:49 +01001287 * @start: first byte in the file to check
1288 * @end: lastbyte in the file to check
1289 * @type: %F_WRLCK for a write lock, else %F_RDLCK
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 *
1291 * Searches the inode's list of locks to find any POSIX locks which conflict.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 */
Christoph Hellwigacc15572015-12-03 12:59:49 +01001293int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start,
1294 loff_t end, unsigned char type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295{
1296 struct file_lock fl;
1297 int error;
Jeff Layton29723ad2014-03-10 09:54:19 -04001298 bool sleep = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
1300 locks_init_lock(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 fl.fl_pid = current->tgid;
1302 fl.fl_file = filp;
1303 fl.fl_flags = FL_POSIX | FL_ACCESS;
1304 if (filp && !(filp->f_flags & O_NONBLOCK))
Jeff Layton29723ad2014-03-10 09:54:19 -04001305 sleep = true;
Christoph Hellwigacc15572015-12-03 12:59:49 +01001306 fl.fl_type = type;
1307 fl.fl_start = start;
1308 fl.fl_end = end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309
1310 for (;;) {
Jeff Layton29723ad2014-03-10 09:54:19 -04001311 if (filp) {
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02001312 fl.fl_owner = filp;
Jeff Layton29723ad2014-03-10 09:54:19 -04001313 fl.fl_flags &= ~FL_SLEEP;
Jeff Laytonb4d629a2016-01-07 18:27:42 -05001314 error = posix_lock_inode(inode, &fl, NULL);
Jeff Layton29723ad2014-03-10 09:54:19 -04001315 if (!error)
1316 break;
1317 }
1318
1319 if (sleep)
1320 fl.fl_flags |= FL_SLEEP;
1321 fl.fl_owner = current->files;
Jeff Laytonb4d629a2016-01-07 18:27:42 -05001322 error = posix_lock_inode(inode, &fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001323 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 break;
1325 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1326 if (!error) {
1327 /*
1328 * If we've been sleeping someone might have
1329 * changed the permissions behind our back.
1330 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001331 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 continue;
1333 }
1334
1335 locks_delete_block(&fl);
1336 break;
1337 }
1338
1339 return error;
1340}
1341
1342EXPORT_SYMBOL(locks_mandatory_area);
Jeff Layton9e8925b2015-11-16 09:49:34 -05001343#endif /* CONFIG_MANDATORY_FILE_LOCKING */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344
J. Bruce Fields778fc542011-07-26 18:25:49 -04001345static void lease_clear_pending(struct file_lock *fl, int arg)
1346{
1347 switch (arg) {
1348 case F_UNLCK:
1349 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1350 /* fall through: */
1351 case F_RDLCK:
1352 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1353 }
1354}
1355
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356/* We already had a lease on this file; just change its type */
Jeff Layton7448cc32015-01-16 15:05:57 -05001357int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 int error = assign_type(fl, arg);
1360
1361 if (error)
1362 return error;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001363 lease_clear_pending(fl, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 locks_wake_up_blocks(fl);
Filipe Brandenburger3b6e2722012-07-27 00:42:52 -04001365 if (arg == F_UNLCK) {
1366 struct file *filp = fl->fl_file;
1367
1368 f_delown(filp);
1369 filp->f_owner.signum = 0;
J. Bruce Fields96d6d592012-07-27 16:18:00 -04001370 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1371 if (fl->fl_fasync != NULL) {
1372 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1373 fl->fl_fasync = NULL;
1374 }
Jeff Laytone084c1b2015-02-16 14:32:03 -05001375 locks_delete_lock_ctx(fl, dispose);
Filipe Brandenburger3b6e2722012-07-27 00:42:52 -04001376 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 return 0;
1378}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379EXPORT_SYMBOL(lease_modify);
1380
J. Bruce Fields778fc542011-07-26 18:25:49 -04001381static bool past_time(unsigned long then)
1382{
1383 if (!then)
1384 /* 0 is a special value meaning "this never expires": */
1385 return false;
1386 return time_after(jiffies, then);
1387}
1388
Jeff Laytonc45198e2014-09-01 07:12:07 -04001389static void time_out_leases(struct inode *inode, struct list_head *dispose)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390{
Jeff Layton8634b512015-01-16 15:05:55 -05001391 struct file_lock_context *ctx = inode->i_flctx;
1392 struct file_lock *fl, *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
Jeff Layton6109c852015-01-16 15:05:57 -05001394 lockdep_assert_held(&ctx->flc_lock);
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001395
Jeff Layton8634b512015-01-16 15:05:55 -05001396 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
Jeff Layton62af4f12014-05-09 14:13:05 -04001397 trace_time_out_leases(inode, fl);
J. Bruce Fields778fc542011-07-26 18:25:49 -04001398 if (past_time(fl->fl_downgrade_time))
Jeff Layton7448cc32015-01-16 15:05:57 -05001399 lease_modify(fl, F_RDLCK, dispose);
J. Bruce Fields778fc542011-07-26 18:25:49 -04001400 if (past_time(fl->fl_break_time))
Jeff Layton7448cc32015-01-16 15:05:57 -05001401 lease_modify(fl, F_UNLCK, dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 }
1403}
1404
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001405static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1406{
Christoph Hellwig11afe9f2015-01-21 19:17:03 +01001407 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1408 return false;
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001409 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1410 return false;
1411 return locks_conflict(breaker, lease);
1412}
1413
Jeff Layton03d12dd2014-09-01 14:53:41 -04001414static bool
1415any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1416{
Jeff Layton8634b512015-01-16 15:05:55 -05001417 struct file_lock_context *ctx = inode->i_flctx;
Jeff Layton03d12dd2014-09-01 14:53:41 -04001418 struct file_lock *fl;
1419
Jeff Layton6109c852015-01-16 15:05:57 -05001420 lockdep_assert_held(&ctx->flc_lock);
Jeff Layton03d12dd2014-09-01 14:53:41 -04001421
Jeff Layton8634b512015-01-16 15:05:55 -05001422 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
Jeff Layton03d12dd2014-09-01 14:53:41 -04001423 if (leases_conflict(fl, breaker))
1424 return true;
1425 }
1426 return false;
1427}
1428
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429/**
1430 * __break_lease - revoke all outstanding leases on file
1431 * @inode: the inode of the file to return
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001432 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1433 * break all leases
1434 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1435 * only delegations
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 *
david m. richter87250dd2007-05-09 16:10:27 -04001437 * break_lease (inlined for speed) has checked there already is at least
1438 * some kind of lock (maybe a lease) on this file. Leases are broken on
1439 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 * specified %O_NONBLOCK to your open().
1441 */
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001442int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443{
J. Bruce Fields778fc542011-07-26 18:25:49 -04001444 int error = 0;
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001445 struct file_lock_context *ctx;
Yan, Zhenga9011252015-03-27 10:34:20 +08001446 struct file_lock *new_fl, *fl, *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 unsigned long break_time;
Al Viro8737c932009-12-24 06:47:55 -05001448 int want_write = (mode & O_ACCMODE) != O_RDONLY;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001449 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450
Al Viro8737c932009-12-24 06:47:55 -05001451 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
Linus Torvalds6d4b9e32011-12-26 10:25:26 -08001452 if (IS_ERR(new_fl))
1453 return PTR_ERR(new_fl);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001454 new_fl->fl_flags = type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455
Jeff Layton8634b512015-01-16 15:05:55 -05001456 /* typically we will check that ctx is non-NULL before calling */
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001457 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Layton8634b512015-01-16 15:05:55 -05001458 if (!ctx) {
1459 WARN_ON_ONCE(1);
1460 return error;
1461 }
1462
Peter Zijlstra87709e22016-05-30 16:48:35 +02001463 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05001464 spin_lock(&ctx->flc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
Jeff Laytonc45198e2014-09-01 07:12:07 -04001466 time_out_leases(inode, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467
Jeff Layton03d12dd2014-09-01 14:53:41 -04001468 if (!any_leases_conflict(inode, new_fl))
J. Bruce Fields778fc542011-07-26 18:25:49 -04001469 goto out;
1470
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 break_time = 0;
1472 if (lease_break_time > 0) {
1473 break_time = jiffies + lease_break_time * HZ;
1474 if (break_time == 0)
1475 break_time++; /* so that 0 means no break time */
1476 }
1477
Yan, Zhenga9011252015-03-27 10:34:20 +08001478 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001479 if (!leases_conflict(fl, new_fl))
1480 continue;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001481 if (want_write) {
1482 if (fl->fl_flags & FL_UNLOCK_PENDING)
1483 continue;
1484 fl->fl_flags |= FL_UNLOCK_PENDING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 fl->fl_break_time = break_time;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001486 } else {
Jeff Layton8634b512015-01-16 15:05:55 -05001487 if (lease_breaking(fl))
J. Bruce Fields778fc542011-07-26 18:25:49 -04001488 continue;
1489 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1490 fl->fl_downgrade_time = break_time;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 }
Jeff Layton4d01b7f2014-09-01 15:06:54 -04001492 if (fl->fl_lmops->lm_break(fl))
Jeff Laytone084c1b2015-02-16 14:32:03 -05001493 locks_delete_lock_ctx(fl, &dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 }
1495
Jeff Layton8634b512015-01-16 15:05:55 -05001496 if (list_empty(&ctx->flc_lease))
Jeff Layton4d01b7f2014-09-01 15:06:54 -04001497 goto out;
1498
Jeff Layton843c6b22014-09-01 14:27:43 -04001499 if (mode & O_NONBLOCK) {
Jeff Layton62af4f12014-05-09 14:13:05 -04001500 trace_break_lease_noblock(inode, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 error = -EWOULDBLOCK;
1502 goto out;
1503 }
1504
1505restart:
Jeff Layton8634b512015-01-16 15:05:55 -05001506 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1507 break_time = fl->fl_break_time;
Jeff Laytonf1c6bb22014-04-15 06:17:49 -04001508 if (break_time != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 break_time -= jiffies;
Jeff Laytonf1c6bb22014-04-15 06:17:49 -04001510 if (break_time == 0)
1511 break_time++;
Jeff Layton8634b512015-01-16 15:05:55 -05001512 locks_insert_block(fl, new_fl);
Jeff Layton62af4f12014-05-09 14:13:05 -04001513 trace_break_lease_block(inode, new_fl);
Jeff Layton6109c852015-01-16 15:05:57 -05001514 spin_unlock(&ctx->flc_lock);
Peter Zijlstra87709e22016-05-30 16:48:35 +02001515 percpu_up_read_preempt_enable(&file_rwsem);
Peter Zijlstraaba37662015-06-22 14:16:33 +02001516
Jeff Laytonc45198e2014-09-01 07:12:07 -04001517 locks_dispose_list(&dispose);
Matthew Wilcox4321e012008-01-14 21:28:30 -07001518 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1519 !new_fl->fl_next, break_time);
Peter Zijlstraaba37662015-06-22 14:16:33 +02001520
Peter Zijlstra87709e22016-05-30 16:48:35 +02001521 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05001522 spin_lock(&ctx->flc_lock);
Jeff Layton62af4f12014-05-09 14:13:05 -04001523 trace_break_lease_unblock(inode, new_fl);
Jeff Layton1c8c6012013-06-21 08:58:15 -04001524 locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 if (error >= 0) {
J. Bruce Fields778fc542011-07-26 18:25:49 -04001526 /*
1527 * Wait for the next conflicting lease that has not been
1528 * broken yet
1529 */
Jeff Layton03d12dd2014-09-01 14:53:41 -04001530 if (error == 0)
1531 time_out_leases(inode, &dispose);
1532 if (any_leases_conflict(inode, new_fl))
1533 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 error = 0;
1535 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536out:
Jeff Layton6109c852015-01-16 15:05:57 -05001537 spin_unlock(&ctx->flc_lock);
Peter Zijlstra87709e22016-05-30 16:48:35 +02001538 percpu_up_read_preempt_enable(&file_rwsem);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001539 locks_dispose_list(&dispose);
Linus Torvalds6d4b9e32011-12-26 10:25:26 -08001540 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 return error;
1542}
1543
1544EXPORT_SYMBOL(__break_lease);
1545
1546/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001547 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 * @inode: the inode
1549 * @time: pointer to a timespec which will contain the last modified time
1550 *
1551 * This is to force NFS clients to flush their caches for files with
1552 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001553 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 */
1555void lease_get_mtime(struct inode *inode, struct timespec *time)
1556{
Jeff Laytonbfe86022014-08-22 10:18:44 -04001557 bool has_lease = false;
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001558 struct file_lock_context *ctx;
Jeff Layton8634b512015-01-16 15:05:55 -05001559 struct file_lock *fl;
Jeff Laytonbfe86022014-08-22 10:18:44 -04001560
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001561 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Layton8634b512015-01-16 15:05:55 -05001562 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
Jeff Layton6109c852015-01-16 15:05:57 -05001563 spin_lock(&ctx->flc_lock);
Geliang Tang8ace5df2015-11-18 21:40:33 +08001564 fl = list_first_entry_or_null(&ctx->flc_lease,
1565 struct file_lock, fl_list);
1566 if (fl && (fl->fl_type == F_WRLCK))
1567 has_lease = true;
Jeff Layton6109c852015-01-16 15:05:57 -05001568 spin_unlock(&ctx->flc_lock);
Jeff Laytonbfe86022014-08-22 10:18:44 -04001569 }
1570
1571 if (has_lease)
Deepa Dinamanic2050a42016-09-14 07:48:06 -07001572 *time = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 else
1574 *time = inode->i_mtime;
1575}
1576
1577EXPORT_SYMBOL(lease_get_mtime);
1578
1579/**
1580 * fcntl_getlease - Enquire what lease is currently active
1581 * @filp: the file
1582 *
1583 * The value returned by this function will be one of
1584 * (if no lease break is pending):
1585 *
1586 * %F_RDLCK to indicate a shared lease is held.
1587 *
1588 * %F_WRLCK to indicate an exclusive lease is held.
1589 *
1590 * %F_UNLCK to indicate no lease is held.
1591 *
1592 * (if a lease break is pending):
1593 *
1594 * %F_RDLCK to indicate an exclusive lease needs to be
1595 * changed to a shared lease (or removed).
1596 *
1597 * %F_UNLCK to indicate the lease needs to be removed.
1598 *
1599 * XXX: sfr & willy disagree over whether F_INPROGRESS
1600 * should be returned to userspace.
1601 */
1602int fcntl_getlease(struct file *filp)
1603{
1604 struct file_lock *fl;
Miklos Szeredic568d682016-09-16 12:44:20 +02001605 struct inode *inode = locks_inode(filp);
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001606 struct file_lock_context *ctx;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 int type = F_UNLCK;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001608 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001610 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Layton8634b512015-01-16 15:05:55 -05001611 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
Peter Zijlstra5f430862016-10-08 10:12:28 +02001612 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05001613 spin_lock(&ctx->flc_lock);
Miklos Szeredic568d682016-09-16 12:44:20 +02001614 time_out_leases(inode, &dispose);
Jeff Layton8634b512015-01-16 15:05:55 -05001615 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1616 if (fl->fl_file != filp)
1617 continue;
J. Bruce Fields778fc542011-07-26 18:25:49 -04001618 type = target_leasetype(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 break;
1620 }
Jeff Layton6109c852015-01-16 15:05:57 -05001621 spin_unlock(&ctx->flc_lock);
Peter Zijlstra5f430862016-10-08 10:12:28 +02001622 percpu_up_read_preempt_enable(&file_rwsem);
1623
Jeff Layton8634b512015-01-16 15:05:55 -05001624 locks_dispose_list(&dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 return type;
1627}
1628
Jeff Layton24cbe782014-02-03 12:13:06 -05001629/**
1630 * check_conflicting_open - see if the given dentry points to a file that has
1631 * an existing open that would conflict with the
1632 * desired lease.
1633 * @dentry: dentry to check
1634 * @arg: type of lease that we're trying to acquire
Randy Dunlap7fadc592015-08-09 18:43:17 -07001635 * @flags: current lock flags
Jeff Layton24cbe782014-02-03 12:13:06 -05001636 *
1637 * Check to see if there's an existing open fd on this file that would
1638 * conflict with the lease we're trying to set.
1639 */
1640static int
Christoph Hellwig11afe9f2015-01-21 19:17:03 +01001641check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
Jeff Layton24cbe782014-02-03 12:13:06 -05001642{
1643 int ret = 0;
1644 struct inode *inode = dentry->d_inode;
1645
Christoph Hellwig11afe9f2015-01-21 19:17:03 +01001646 if (flags & FL_LAYOUT)
1647 return 0;
1648
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +02001649 if ((arg == F_RDLCK) &&
1650 (atomic_read(&d_real_inode(dentry)->i_writecount) > 0))
Jeff Layton24cbe782014-02-03 12:13:06 -05001651 return -EAGAIN;
1652
1653 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1654 (atomic_read(&inode->i_count) > 1)))
1655 ret = -EAGAIN;
1656
1657 return ret;
1658}
1659
Jeff Laytone6f5c782014-08-22 10:40:25 -04001660static int
1661generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662{
Jeff Layton8634b512015-01-16 15:05:55 -05001663 struct file_lock *fl, *my_fl = NULL, *lease;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001664 struct dentry *dentry = filp->f_path.dentry;
Miklos Szeredic568d682016-09-16 12:44:20 +02001665 struct inode *inode = dentry->d_inode;
Jeff Layton8634b512015-01-16 15:05:55 -05001666 struct file_lock_context *ctx;
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001667 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001668 int error;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001669 LIST_HEAD(dispose);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670
J. Bruce Fields096657b2010-10-30 17:31:14 -04001671 lease = *flp;
Jeff Layton62af4f12014-05-09 14:13:05 -04001672 trace_generic_add_lease(inode, lease);
1673
Jeff Layton5c1c6692015-04-03 09:04:03 -04001674 /* Note that arg is never F_UNLCK here */
1675 ctx = locks_get_lock_context(inode, arg);
Jeff Layton8634b512015-01-16 15:05:55 -05001676 if (!ctx)
1677 return -ENOMEM;
1678
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001679 /*
1680 * In the delegation case we need mutual exclusion with
1681 * a number of operations that take the i_mutex. We trylock
1682 * because delegations are an optional optimization, and if
1683 * there's some chance of a conflict--we'd rather not
1684 * bother, maybe that's a sign this just isn't a good file to
1685 * hand out a delegation on.
1686 */
Al Viro59551022016-01-22 15:40:57 -05001687 if (is_deleg && !inode_trylock(inode))
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001688 return -EAGAIN;
1689
1690 if (is_deleg && arg == F_WRLCK) {
1691 /* Write delegations are not currently supported: */
Al Viro59551022016-01-22 15:40:57 -05001692 inode_unlock(inode);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001693 WARN_ON_ONCE(1);
1694 return -EINVAL;
1695 }
J. Bruce Fields096657b2010-10-30 17:31:14 -04001696
Peter Zijlstra87709e22016-05-30 16:48:35 +02001697 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05001698 spin_lock(&ctx->flc_lock);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001699 time_out_leases(inode, &dispose);
Christoph Hellwig11afe9f2015-01-21 19:17:03 +01001700 error = check_conflicting_open(dentry, arg, lease->fl_flags);
Jeff Layton24cbe782014-02-03 12:13:06 -05001701 if (error)
J. Bruce Fields096657b2010-10-30 17:31:14 -04001702 goto out;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001703
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 /*
1705 * At this point, we know that if there is an exclusive
1706 * lease on this file, then we hold it on this filp
1707 * (otherwise our open of this file would have blocked).
1708 * And if we are trying to acquire an exclusive lease,
1709 * then the file is not open by anyone (including us)
1710 * except for this filp.
1711 */
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001712 error = -EAGAIN;
Jeff Layton8634b512015-01-16 15:05:55 -05001713 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
Christoph Hellwig2ab99ee2015-01-21 19:14:02 +01001714 if (fl->fl_file == filp &&
1715 fl->fl_owner == lease->fl_owner) {
Jeff Layton8634b512015-01-16 15:05:55 -05001716 my_fl = fl;
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001717 continue;
1718 }
Jeff Layton8634b512015-01-16 15:05:55 -05001719
J. Bruce Fieldsc1f24ef2011-08-19 10:59:49 -04001720 /*
1721 * No exclusive leases if someone else has a lease on
1722 * this file:
1723 */
1724 if (arg == F_WRLCK)
1725 goto out;
1726 /*
1727 * Modifying our existing lease is OK, but no getting a
1728 * new lease if someone else is opening for write:
1729 */
1730 if (fl->fl_flags & FL_UNLOCK_PENDING)
1731 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 }
1733
Jeff Layton8634b512015-01-16 15:05:55 -05001734 if (my_fl != NULL) {
Jeff Layton0164bf02015-03-04 17:34:32 -05001735 lease = my_fl;
1736 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001737 if (error)
1738 goto out;
1739 goto out_setup;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 }
1741
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 error = -EINVAL;
1743 if (!leases_enable)
1744 goto out;
1745
Jeff Laytone084c1b2015-02-16 14:32:03 -05001746 locks_insert_lock_ctx(lease, &ctx->flc_lease);
Jeff Layton24cbe782014-02-03 12:13:06 -05001747 /*
1748 * The check in break_lease() is lockless. It's possible for another
1749 * open to race in after we did the earlier check for a conflicting
1750 * open but before the lease was inserted. Check again for a
1751 * conflicting open and cancel the lease if there is one.
1752 *
1753 * We also add a barrier here to ensure that the insertion of the lock
1754 * precedes these checks.
1755 */
1756 smp_mb();
Christoph Hellwig11afe9f2015-01-21 19:17:03 +01001757 error = check_conflicting_open(dentry, arg, lease->fl_flags);
Jeff Layton8634b512015-01-16 15:05:55 -05001758 if (error) {
Jeff Laytone084c1b2015-02-16 14:32:03 -05001759 locks_unlink_lock_ctx(lease);
Jeff Layton8634b512015-01-16 15:05:55 -05001760 goto out;
1761 }
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001762
1763out_setup:
1764 if (lease->fl_lmops->lm_setup)
1765 lease->fl_lmops->lm_setup(lease, priv);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766out:
Jeff Layton6109c852015-01-16 15:05:57 -05001767 spin_unlock(&ctx->flc_lock);
Peter Zijlstra87709e22016-05-30 16:48:35 +02001768 percpu_up_read_preempt_enable(&file_rwsem);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001769 locks_dispose_list(&dispose);
J. Bruce Fieldsdf4e8d22012-03-05 13:18:59 -05001770 if (is_deleg)
Al Viro59551022016-01-22 15:40:57 -05001771 inode_unlock(inode);
Jeff Layton8634b512015-01-16 15:05:55 -05001772 if (!error && !my_fl)
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001773 *flp = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 return error;
1775}
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001776
Christoph Hellwig2ab99ee2015-01-21 19:14:02 +01001777static int generic_delete_lease(struct file *filp, void *owner)
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001778{
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001779 int error = -EAGAIN;
Jeff Layton8634b512015-01-16 15:05:55 -05001780 struct file_lock *fl, *victim = NULL;
Miklos Szeredic568d682016-09-16 12:44:20 +02001781 struct inode *inode = locks_inode(filp);
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001782 struct file_lock_context *ctx;
Jeff Laytonc45198e2014-09-01 07:12:07 -04001783 LIST_HEAD(dispose);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001784
Dmitry Vyukov128a3782015-09-21 09:43:06 +02001785 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Layton8634b512015-01-16 15:05:55 -05001786 if (!ctx) {
1787 trace_generic_delete_lease(inode, NULL);
1788 return error;
1789 }
1790
Peter Zijlstra87709e22016-05-30 16:48:35 +02001791 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05001792 spin_lock(&ctx->flc_lock);
Jeff Layton8634b512015-01-16 15:05:55 -05001793 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
Christoph Hellwig2ab99ee2015-01-21 19:14:02 +01001794 if (fl->fl_file == filp &&
1795 fl->fl_owner == owner) {
Jeff Layton8634b512015-01-16 15:05:55 -05001796 victim = fl;
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001797 break;
Jeff Layton8634b512015-01-16 15:05:55 -05001798 }
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001799 }
Jeff Laytona9b1b452015-03-14 09:45:35 -04001800 trace_generic_delete_lease(inode, victim);
Jeff Layton8634b512015-01-16 15:05:55 -05001801 if (victim)
Jeff Layton7448cc32015-01-16 15:05:57 -05001802 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
Jeff Layton6109c852015-01-16 15:05:57 -05001803 spin_unlock(&ctx->flc_lock);
Peter Zijlstra87709e22016-05-30 16:48:35 +02001804 percpu_up_read_preempt_enable(&file_rwsem);
Jeff Laytonc45198e2014-09-01 07:12:07 -04001805 locks_dispose_list(&dispose);
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001806 return error;
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001807}
1808
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809/**
1810 * generic_setlease - sets a lease on an open file
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001811 * @filp: file pointer
1812 * @arg: type of lease to obtain
1813 * @flp: input - file_lock to use, output - file_lock inserted
1814 * @priv: private data for lm_setup (may be NULL if lm_setup
1815 * doesn't require it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 *
1817 * The (input) flp->fl_lmops->lm_break function is required
1818 * by break_lease().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 */
Jeff Laytone6f5c782014-08-22 10:40:25 -04001820int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1821 void **priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822{
Miklos Szeredic568d682016-09-16 12:44:20 +02001823 struct inode *inode = locks_inode(filp);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001824 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001826 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001827 return -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 if (!S_ISREG(inode->i_mode))
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001829 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 error = security_file_lock(filp, arg);
1831 if (error)
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001832 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001834 switch (arg) {
1835 case F_UNLCK:
Christoph Hellwig2ab99ee2015-01-21 19:14:02 +01001836 return generic_delete_lease(filp, *priv);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001837 case F_RDLCK:
1838 case F_WRLCK:
Jeff Layton0efaa7e2014-08-22 10:18:45 -04001839 if (!(*flp)->fl_lmops->lm_break) {
1840 WARN_ON_ONCE(1);
1841 return -ENOLCK;
1842 }
Christoph Hellwig11afe9f2015-01-21 19:17:03 +01001843
Jeff Laytone6f5c782014-08-22 10:40:25 -04001844 return generic_add_lease(filp, arg, flp, priv);
J. Bruce Fields8335ebd2011-09-21 08:34:32 -04001845 default:
Dave Jones8d657eb2012-07-13 13:35:36 -04001846 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001849EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
Arnd Bergmannb89f4322010-09-18 15:09:31 +02001851/**
Jeff Laytone51673a2014-08-22 18:13:28 -04001852 * vfs_setlease - sets a lease on an open file
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001853 * @filp: file pointer
1854 * @arg: type of lease to obtain
1855 * @lease: file_lock to use when adding a lease
1856 * @priv: private info for lm_setup when adding a lease (may be
1857 * NULL if lm_setup doesn't require it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 *
Jeff Laytone51673a2014-08-22 18:13:28 -04001859 * Call this to establish a lease on the file. The "lease" argument is not
1860 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1861 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1862 * if not, this function will return -ENOLCK (and generate a scary-looking
1863 * stack trace).
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001864 *
1865 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1866 * may be NULL if the lm_setup operation doesn't require it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 */
Jeff Laytone6f5c782014-08-22 10:40:25 -04001868int
1869vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870{
Miklos Szeredic568d682016-09-16 12:44:20 +02001871 if (filp->f_op->setlease && is_remote_lock(filp))
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001872 return filp->f_op->setlease(filp, arg, lease, priv);
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001873 else
Jeff Laytonf82b4b62014-08-22 18:50:48 -04001874 return generic_setlease(filp, arg, lease, priv);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001876EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001878static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879{
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001880 struct file_lock *fl;
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001881 struct fasync_struct *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 int error;
1883
Arnd Bergmannc5b1f0d2010-10-27 15:46:08 +02001884 fl = lease_alloc(filp, arg);
1885 if (IS_ERR(fl))
1886 return PTR_ERR(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001888 new = fasync_alloc();
1889 if (!new) {
1890 locks_free_lock(fl);
1891 return -ENOMEM;
1892 }
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001893 new->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894
Jeff Layton1c7dd2f2014-08-22 10:55:47 -04001895 error = vfs_setlease(filp, arg, &fl, (void **)&new);
Jeff Layton2dfb9282014-08-11 18:14:12 -04001896 if (fl)
1897 locks_free_lock(fl);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -04001898 if (new)
1899 fasync_free(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 return error;
1901}
1902
1903/**
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001904 * fcntl_setlease - sets a lease on an open file
1905 * @fd: open file descriptor
1906 * @filp: file pointer
1907 * @arg: type of lease to obtain
1908 *
1909 * Call this fcntl to establish a lease on the file.
1910 * Note that you also need to call %F_SETSIG to
1911 * receive a signal when the lease is broken.
1912 */
1913int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1914{
1915 if (arg == F_UNLCK)
Christoph Hellwig2ab99ee2015-01-21 19:14:02 +01001916 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
J. Bruce Fields0ceaf6c2010-10-30 17:31:13 -04001917 return do_fcntl_add_lease(fd, filp, arg);
1918}
1919
1920/**
Jeff Layton29d01b22015-07-11 06:43:02 -04001921 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1922 * @inode: inode of the file to apply to
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 * @fl: The lock to be applied
1924 *
Jeff Layton29d01b22015-07-11 06:43:02 -04001925 * Apply a FLOCK style lock request to an inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 */
Benjamin Coddington616fb382015-10-22 13:38:15 -04001927static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928{
1929 int error;
1930 might_sleep();
1931 for (;;) {
Jeff Layton29d01b22015-07-11 06:43:02 -04001932 error = flock_lock_inode(inode, fl);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001933 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 break;
1935 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1936 if (!error)
1937 continue;
1938
1939 locks_delete_block(fl);
1940 break;
1941 }
1942 return error;
1943}
1944
Jeff Layton29d01b22015-07-11 06:43:02 -04001945/**
Benjamin Coddingtone55c34a2015-10-22 13:38:13 -04001946 * locks_lock_inode_wait - Apply a lock to an inode
1947 * @inode: inode of the file to apply to
1948 * @fl: The lock to be applied
1949 *
1950 * Apply a POSIX or FLOCK style lock request to an inode.
1951 */
1952int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1953{
1954 int res = 0;
1955 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1956 case FL_POSIX:
1957 res = posix_lock_inode_wait(inode, fl);
1958 break;
1959 case FL_FLOCK:
1960 res = flock_lock_inode_wait(inode, fl);
1961 break;
1962 default:
1963 BUG();
1964 }
1965 return res;
1966}
1967EXPORT_SYMBOL(locks_lock_inode_wait);
1968
1969/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 * sys_flock: - flock() system call.
1971 * @fd: the file descriptor to lock.
1972 * @cmd: the type of lock to apply.
1973 *
1974 * Apply a %FL_FLOCK style lock to an open file descriptor.
1975 * The @cmd can be one of
1976 *
1977 * %LOCK_SH -- a shared lock.
1978 *
1979 * %LOCK_EX -- an exclusive lock.
1980 *
1981 * %LOCK_UN -- remove an existing lock.
1982 *
1983 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1984 *
1985 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1986 * processes read and write access respectively.
1987 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001988SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989{
Al Viro2903ff02012-08-28 12:52:22 -04001990 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991 struct file_lock *lock;
1992 int can_sleep, unlock;
1993 int error;
1994
1995 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001996 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 goto out;
1998
1999 can_sleep = !(cmd & LOCK_NB);
2000 cmd &= ~LOCK_NB;
2001 unlock = (cmd == LOCK_UN);
2002
Al Viroaeb5d722008-09-02 15:28:45 -04002003 if (!unlock && !(cmd & LOCK_MAND) &&
Al Viro2903ff02012-08-28 12:52:22 -04002004 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 goto out_putf;
2006
Jeff Layton6e129d002014-09-04 10:25:06 -04002007 lock = flock_make_lock(f.file, cmd);
2008 if (IS_ERR(lock)) {
2009 error = PTR_ERR(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010 goto out_putf;
Jeff Layton6e129d002014-09-04 10:25:06 -04002011 }
2012
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 if (can_sleep)
2014 lock->fl_flags |= FL_SLEEP;
2015
Al Viro2903ff02012-08-28 12:52:22 -04002016 error = security_file_lock(f.file, lock->fl_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 if (error)
2018 goto out_free;
2019
Miklos Szeredic568d682016-09-16 12:44:20 +02002020 if (f.file->f_op->flock && is_remote_lock(f.file))
Al Viro2903ff02012-08-28 12:52:22 -04002021 error = f.file->f_op->flock(f.file,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 (can_sleep) ? F_SETLKW : F_SETLK,
2023 lock);
2024 else
Benjamin Coddington4f656362015-10-22 13:38:14 -04002025 error = locks_lock_file_wait(f.file, lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
2027 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08002028 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029
2030 out_putf:
Al Viro2903ff02012-08-28 12:52:22 -04002031 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 out:
2033 return error;
2034}
2035
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002036/**
2037 * vfs_test_lock - test file byte range lock
2038 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04002039 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002040 *
2041 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2042 * setting conf->fl_type to something other than F_UNLCK.
2043 */
2044int vfs_test_lock(struct file *filp, struct file_lock *fl)
2045{
Miklos Szeredic568d682016-09-16 12:44:20 +02002046 if (filp->f_op->lock && is_remote_lock(filp))
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002047 return filp->f_op->lock(filp, F_GETLK, fl);
2048 posix_test_lock(filp, fl);
2049 return 0;
2050}
2051EXPORT_SYMBOL_GPL(vfs_test_lock);
2052
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05002053static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
2054{
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002055 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05002056#if BITS_PER_LONG == 32
2057 /*
2058 * Make sure we can represent the posix lock via
2059 * legacy 32bit flock.
2060 */
2061 if (fl->fl_start > OFFT_OFFSET_MAX)
2062 return -EOVERFLOW;
2063 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2064 return -EOVERFLOW;
2065#endif
2066 flock->l_start = fl->fl_start;
2067 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2068 fl->fl_end - fl->fl_start + 1;
2069 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04002070 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05002071 return 0;
2072}
2073
2074#if BITS_PER_LONG == 32
2075static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2076{
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002077 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05002078 flock->l_start = fl->fl_start;
2079 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2080 fl->fl_end - fl->fl_start + 1;
2081 flock->l_whence = 0;
2082 flock->l_type = fl->fl_type;
2083}
2084#endif
2085
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086/* Report the first existing lock that would conflict with l.
2087 * This implements the F_GETLK command of fcntl().
2088 */
Jeff Laytonc1e62b82014-02-03 12:13:09 -05002089int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002091 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 struct flock flock;
2093 int error;
2094
2095 error = -EFAULT;
2096 if (copy_from_user(&flock, l, sizeof(flock)))
2097 goto out;
2098 error = -EINVAL;
2099 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2100 goto out;
2101
2102 error = flock_to_posix_lock(filp, &file_lock, &flock);
2103 if (error)
2104 goto out;
2105
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002106 if (cmd == F_OFD_GETLK) {
Jeff Layton90478932014-03-04 10:30:23 -05002107 error = -EINVAL;
2108 if (flock.l_pid != 0)
2109 goto out;
2110
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002111 cmd = F_GETLK;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002112 file_lock.fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002113 file_lock.fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002114 }
2115
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002116 error = vfs_test_lock(filp, &file_lock);
2117 if (error)
2118 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002120 flock.l_type = file_lock.fl_type;
2121 if (file_lock.fl_type != F_UNLCK) {
2122 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05002123 if (error)
Kinglong Meef3282962014-08-22 10:18:43 -04002124 goto rel_priv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 }
2126 error = -EFAULT;
2127 if (!copy_to_user(l, &flock, sizeof(flock)))
2128 error = 0;
Kinglong Meef3282962014-08-22 10:18:43 -04002129rel_priv:
2130 locks_release_private(&file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131out:
2132 return error;
2133}
2134
Marc Eshel7723ec92007-01-18 15:08:55 -05002135/**
2136 * vfs_lock_file - file byte range lock
2137 * @filp: The file to apply the lock to
2138 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2139 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05002140 * @conf: Place to return a copy of the conflicting lock, if found.
2141 *
2142 * A caller that doesn't care about the conflicting lock may pass NULL
2143 * as the final argument.
2144 *
2145 * If the filesystem defines a private ->lock() method, then @conf will
2146 * be left unchanged; so a caller that cares should initialize it to
2147 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05002148 *
2149 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2150 * locks, the ->lock() interface may return asynchronously, before the lock has
2151 * been granted or denied by the underlying filesystem, if (and only if)
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002152 * lm_grant is set. Callers expecting ->lock() to return asynchronously
Marc Eshel2beb6612006-12-05 23:31:28 -05002153 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2154 * the request is for a blocking lock. When ->lock() does return asynchronously,
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002155 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
Marc Eshel2beb6612006-12-05 23:31:28 -05002156 * request completes.
2157 * If the request is for non-blocking lock the file system should return
Miklos Szeredibde74e42008-07-25 01:48:57 -07002158 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2159 * with the result. If the request timed out the callback routine will return a
Marc Eshel2beb6612006-12-05 23:31:28 -05002160 * nonzero return code and the file system should release the lock. The file
2161 * system is also responsible to keep a corresponding posix lock when it
2162 * grants a lock so the VFS can find out which locks are locally held and do
2163 * the correct lock cleanup when required.
2164 * The underlying filesystem must not drop the kernel lock or call
J. Bruce Fields8fb47a42011-07-20 20:21:59 -04002165 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
Marc Eshel2beb6612006-12-05 23:31:28 -05002166 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05002167 */
Marc Eshel150b3932007-01-18 16:15:35 -05002168int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05002169{
Miklos Szeredic568d682016-09-16 12:44:20 +02002170 if (filp->f_op->lock && is_remote_lock(filp))
Marc Eshel7723ec92007-01-18 15:08:55 -05002171 return filp->f_op->lock(filp, cmd, fl);
2172 else
Marc Eshel150b3932007-01-18 16:15:35 -05002173 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05002174}
2175EXPORT_SYMBOL_GPL(vfs_lock_file);
2176
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002177static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2178 struct file_lock *fl)
2179{
2180 int error;
2181
2182 error = security_file_lock(filp, fl->fl_type);
2183 if (error)
2184 return error;
2185
Miklos Szeredi764c76b2008-07-25 01:48:58 -07002186 for (;;) {
2187 error = vfs_lock_file(filp, cmd, fl, NULL);
2188 if (error != FILE_LOCK_DEFERRED)
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002189 break;
Miklos Szeredi764c76b2008-07-25 01:48:58 -07002190 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2191 if (!error)
2192 continue;
2193
2194 locks_delete_block(fl);
2195 break;
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002196 }
2197
2198 return error;
2199}
2200
Benjamin Coddington6ca7d912015-10-15 09:07:07 -04002201/* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002202static int
2203check_fmode_for_setlk(struct file_lock *fl)
2204{
2205 switch (fl->fl_type) {
2206 case F_RDLCK:
2207 if (!(fl->fl_file->f_mode & FMODE_READ))
2208 return -EBADF;
2209 break;
2210 case F_WRLCK:
2211 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2212 return -EBADF;
2213 }
2214 return 0;
2215}
2216
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217/* Apply the lock described by l to an open file descriptor.
2218 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2219 */
Peter Staubachc2936212005-07-27 11:45:09 -07002220int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2221 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222{
2223 struct file_lock *file_lock = locks_alloc_lock();
2224 struct flock flock;
2225 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04002226 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227 int error;
2228
2229 if (file_lock == NULL)
2230 return -ENOLCK;
2231
Miklos Szeredic568d682016-09-16 12:44:20 +02002232 inode = locks_inode(filp);
Jeff Layton18909102016-01-06 21:26:10 -05002233
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 /*
2235 * This might block, so we do it before checking the inode.
2236 */
2237 error = -EFAULT;
2238 if (copy_from_user(&flock, l, sizeof(flock)))
2239 goto out;
2240
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 /* Don't allow mandatory locks on files that may be memory mapped
2242 * and shared.
2243 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002244 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245 error = -EAGAIN;
2246 goto out;
2247 }
2248
2249 error = flock_to_posix_lock(filp, file_lock, &flock);
2250 if (error)
2251 goto out;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002252
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002253 error = check_fmode_for_setlk(file_lock);
2254 if (error)
2255 goto out;
2256
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002257 /*
2258 * If the cmd is requesting file-private locks, then set the
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002259 * FL_OFDLCK flag and override the owner.
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002260 */
2261 switch (cmd) {
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002262 case F_OFD_SETLK:
Jeff Layton90478932014-03-04 10:30:23 -05002263 error = -EINVAL;
2264 if (flock.l_pid != 0)
2265 goto out;
2266
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002267 cmd = F_SETLK;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002268 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002269 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002270 break;
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002271 case F_OFD_SETLKW:
Jeff Layton90478932014-03-04 10:30:23 -05002272 error = -EINVAL;
2273 if (flock.l_pid != 0)
2274 goto out;
2275
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002276 cmd = F_SETLKW;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002277 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002278 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002279 /* Fallthrough */
2280 case F_SETLKW:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281 file_lock->fl_flags |= FL_SLEEP;
2282 }
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002283
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002284 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002285
Peter Staubachc2936212005-07-27 11:45:09 -07002286 /*
Jeff Layton0752ba82016-01-08 07:30:43 -05002287 * Attempt to detect a close/fcntl race and recover by releasing the
2288 * lock that was just acquired. There is no need to do that when we're
2289 * unlocking though, or for OFD locks.
Peter Staubachc2936212005-07-27 11:45:09 -07002290 */
Jeff Layton0752ba82016-01-08 07:30:43 -05002291 if (!error && file_lock->fl_type != F_UNLCK &&
2292 !(file_lock->fl_flags & FL_OFDLCK)) {
Jeff Layton7f3697e2016-01-07 16:38:10 -05002293 /*
2294 * We need that spin_lock here - it prevents reordering between
2295 * update of i_flctx->flc_posix and check for it done in
2296 * close(). rcu_read_lock() wouldn't do.
2297 */
2298 spin_lock(&current->files->file_lock);
2299 f = fcheck(fd);
2300 spin_unlock(&current->files->file_lock);
2301 if (f != filp) {
2302 file_lock->fl_type = F_UNLCK;
2303 error = do_lock_file_wait(filp, cmd, file_lock);
2304 WARN_ON_ONCE(error);
2305 error = -EBADF;
2306 }
Peter Staubachc2936212005-07-27 11:45:09 -07002307 }
Peter Staubachc2936212005-07-27 11:45:09 -07002308out:
Jeff Layton18909102016-01-06 21:26:10 -05002309 trace_fcntl_setlk(inode, file_lock, error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310 locks_free_lock(file_lock);
2311 return error;
2312}
2313
2314#if BITS_PER_LONG == 32
2315/* Report the first existing lock that would conflict with l.
2316 * This implements the F_GETLK command of fcntl().
2317 */
Jeff Laytonc1e62b82014-02-03 12:13:09 -05002318int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002320 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321 struct flock64 flock;
2322 int error;
2323
2324 error = -EFAULT;
2325 if (copy_from_user(&flock, l, sizeof(flock)))
2326 goto out;
2327 error = -EINVAL;
2328 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2329 goto out;
2330
2331 error = flock64_to_posix_lock(filp, &file_lock, &flock);
2332 if (error)
2333 goto out;
2334
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002335 if (cmd == F_OFD_GETLK) {
Jeff Layton90478932014-03-04 10:30:23 -05002336 error = -EINVAL;
2337 if (flock.l_pid != 0)
2338 goto out;
2339
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002340 cmd = F_GETLK64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002341 file_lock.fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002342 file_lock.fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002343 }
2344
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05002345 error = vfs_test_lock(filp, &file_lock);
2346 if (error)
2347 goto out;
2348
Marc Eshel9d6a8c52007-02-21 00:55:18 -05002349 flock.l_type = file_lock.fl_type;
2350 if (file_lock.fl_type != F_UNLCK)
2351 posix_lock_to_flock64(&flock, &file_lock);
2352
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353 error = -EFAULT;
2354 if (!copy_to_user(l, &flock, sizeof(flock)))
2355 error = 0;
Kinglong Meef3282962014-08-22 10:18:43 -04002356
2357 locks_release_private(&file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358out:
2359 return error;
2360}
2361
2362/* Apply the lock described by l to an open file descriptor.
2363 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2364 */
Peter Staubachc2936212005-07-27 11:45:09 -07002365int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2366 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367{
2368 struct file_lock *file_lock = locks_alloc_lock();
2369 struct flock64 flock;
2370 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04002371 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372 int error;
2373
2374 if (file_lock == NULL)
2375 return -ENOLCK;
2376
2377 /*
2378 * This might block, so we do it before checking the inode.
2379 */
2380 error = -EFAULT;
2381 if (copy_from_user(&flock, l, sizeof(flock)))
2382 goto out;
2383
Miklos Szeredic568d682016-09-16 12:44:20 +02002384 inode = locks_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385
2386 /* Don't allow mandatory locks on files that may be memory mapped
2387 * and shared.
2388 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002389 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 error = -EAGAIN;
2391 goto out;
2392 }
2393
2394 error = flock64_to_posix_lock(filp, file_lock, &flock);
2395 if (error)
2396 goto out;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002397
Jeff Laytoncf01f4e2014-05-09 11:41:54 -04002398 error = check_fmode_for_setlk(file_lock);
2399 if (error)
2400 goto out;
2401
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002402 /*
2403 * If the cmd is requesting file-private locks, then set the
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002404 * FL_OFDLCK flag and override the owner.
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002405 */
2406 switch (cmd) {
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002407 case F_OFD_SETLK:
Jeff Layton90478932014-03-04 10:30:23 -05002408 error = -EINVAL;
2409 if (flock.l_pid != 0)
2410 goto out;
2411
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002412 cmd = F_SETLK64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002413 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002414 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002415 break;
Jeff Layton0d3f7a22014-04-22 08:23:58 -04002416 case F_OFD_SETLKW:
Jeff Layton90478932014-03-04 10:30:23 -05002417 error = -EINVAL;
2418 if (flock.l_pid != 0)
2419 goto out;
2420
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002421 cmd = F_SETLKW64;
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002422 file_lock->fl_flags |= FL_OFDLCK;
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002423 file_lock->fl_owner = filp;
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002424 /* Fallthrough */
2425 case F_SETLKW64:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 file_lock->fl_flags |= FL_SLEEP;
2427 }
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002428
Miklos Szeredib648a6d2008-07-25 01:48:57 -07002429 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430
Peter Staubachc2936212005-07-27 11:45:09 -07002431 /*
Jeff Layton0752ba82016-01-08 07:30:43 -05002432 * Attempt to detect a close/fcntl race and recover by releasing the
2433 * lock that was just acquired. There is no need to do that when we're
2434 * unlocking though, or for OFD locks.
Peter Staubachc2936212005-07-27 11:45:09 -07002435 */
Jeff Layton0752ba82016-01-08 07:30:43 -05002436 if (!error && file_lock->fl_type != F_UNLCK &&
2437 !(file_lock->fl_flags & FL_OFDLCK)) {
Jeff Layton7f3697e2016-01-07 16:38:10 -05002438 /*
2439 * We need that spin_lock here - it prevents reordering between
2440 * update of i_flctx->flc_posix and check for it done in
2441 * close(). rcu_read_lock() wouldn't do.
2442 */
2443 spin_lock(&current->files->file_lock);
2444 f = fcheck(fd);
2445 spin_unlock(&current->files->file_lock);
2446 if (f != filp) {
2447 file_lock->fl_type = F_UNLCK;
2448 error = do_lock_file_wait(filp, cmd, file_lock);
2449 WARN_ON_ONCE(error);
2450 error = -EBADF;
2451 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453out:
2454 locks_free_lock(file_lock);
2455 return error;
2456}
2457#endif /* BITS_PER_LONG == 32 */
2458
2459/*
2460 * This function is called when the file is being removed
2461 * from the task's fd array. POSIX locks belonging to this task
2462 * are deleted at this time.
2463 */
2464void locks_remove_posix(struct file *filp, fl_owner_t owner)
2465{
Jeff Layton18909102016-01-06 21:26:10 -05002466 int error;
Miklos Szeredic568d682016-09-16 12:44:20 +02002467 struct inode *inode = locks_inode(filp);
Miklos Szerediff7b86b2006-06-23 02:05:11 -07002468 struct file_lock lock;
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002469 struct file_lock_context *ctx;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470
2471 /*
2472 * If there are no locks held on this file, we don't need to call
2473 * posix_lock_file(). Another process could be setting a lock on this
2474 * file at the same time, but we wouldn't remove that lock anyway.
2475 */
Miklos Szeredic568d682016-09-16 12:44:20 +02002476 ctx = smp_load_acquire(&inode->i_flctx);
Jeff Laytonbd61e0a2015-01-16 15:05:55 -05002477 if (!ctx || list_empty(&ctx->flc_posix))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002478 return;
2479
2480 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07002481 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 lock.fl_start = 0;
2483 lock.fl_end = OFFSET_MAX;
2484 lock.fl_owner = owner;
2485 lock.fl_pid = current->tgid;
2486 lock.fl_file = filp;
2487 lock.fl_ops = NULL;
2488 lock.fl_lmops = NULL;
2489
Jeff Layton18909102016-01-06 21:26:10 -05002490 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491
Linus Torvalds1da177e2005-04-16 15:20:36 -07002492 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2493 lock.fl_ops->fl_release_private(&lock);
Miklos Szeredic568d682016-09-16 12:44:20 +02002494 trace_locks_remove_posix(inode, &lock, error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495}
2496
2497EXPORT_SYMBOL(locks_remove_posix);
2498
Jeff Layton3d8e5602015-01-16 15:05:58 -05002499/* The i_flctx must be valid when calling into here */
Jeff Laytondd459bb2015-01-16 15:05:54 -05002500static void
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002501locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
Jeff Laytondd459bb2015-01-16 15:05:54 -05002502{
2503 struct file_lock fl = {
2504 .fl_owner = filp,
2505 .fl_pid = current->tgid,
2506 .fl_file = filp,
2507 .fl_flags = FL_FLOCK,
2508 .fl_type = F_UNLCK,
2509 .fl_end = OFFSET_MAX,
2510 };
Miklos Szeredic568d682016-09-16 12:44:20 +02002511 struct inode *inode = locks_inode(filp);
Jeff Laytondd459bb2015-01-16 15:05:54 -05002512
Jeff Layton3d8e5602015-01-16 15:05:58 -05002513 if (list_empty(&flctx->flc_flock))
Jeff Laytondd459bb2015-01-16 15:05:54 -05002514 return;
2515
Miklos Szeredic568d682016-09-16 12:44:20 +02002516 if (filp->f_op->flock && is_remote_lock(filp))
Jeff Laytondd459bb2015-01-16 15:05:54 -05002517 filp->f_op->flock(filp, F_SETLKW, &fl);
2518 else
Jeff Laytonbcd7f782015-07-11 06:43:02 -04002519 flock_lock_inode(inode, &fl);
Jeff Laytondd459bb2015-01-16 15:05:54 -05002520
2521 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2522 fl.fl_ops->fl_release_private(&fl);
2523}
2524
Jeff Layton3d8e5602015-01-16 15:05:58 -05002525/* The i_flctx must be valid when calling into here */
Jeff Layton8634b512015-01-16 15:05:55 -05002526static void
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002527locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
Jeff Layton8634b512015-01-16 15:05:55 -05002528{
Jeff Layton8634b512015-01-16 15:05:55 -05002529 struct file_lock *fl, *tmp;
2530 LIST_HEAD(dispose);
2531
Jeff Layton3d8e5602015-01-16 15:05:58 -05002532 if (list_empty(&ctx->flc_lease))
Jeff Layton8634b512015-01-16 15:05:55 -05002533 return;
2534
Peter Zijlstra5f430862016-10-08 10:12:28 +02002535 percpu_down_read_preempt_disable(&file_rwsem);
Jeff Layton6109c852015-01-16 15:05:57 -05002536 spin_lock(&ctx->flc_lock);
Jeff Layton8634b512015-01-16 15:05:55 -05002537 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
Jeff Laytonc4e136c2015-02-16 19:37:42 -05002538 if (filp == fl->fl_file)
2539 lease_modify(fl, F_UNLCK, &dispose);
Jeff Layton6109c852015-01-16 15:05:57 -05002540 spin_unlock(&ctx->flc_lock);
Peter Zijlstra5f430862016-10-08 10:12:28 +02002541 percpu_up_read_preempt_enable(&file_rwsem);
2542
Jeff Layton8634b512015-01-16 15:05:55 -05002543 locks_dispose_list(&dispose);
2544}
2545
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546/*
2547 * This function is called on the last close of an open file.
2548 */
Jeff Layton78ed8a12014-02-03 12:13:08 -05002549void locks_remove_file(struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550{
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002551 struct file_lock_context *ctx;
2552
Miklos Szeredic568d682016-09-16 12:44:20 +02002553 ctx = smp_load_acquire(&locks_inode(filp)->i_flctx);
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002554 if (!ctx)
Jeff Layton3d8e5602015-01-16 15:05:58 -05002555 return;
2556
Jeff Laytondd459bb2015-01-16 15:05:54 -05002557 /* remove any OFD locks */
Christoph Hellwig73a8f5f2014-07-13 17:00:38 +02002558 locks_remove_posix(filp, filp);
Jeff Layton5d50ffd2014-02-03 12:13:10 -05002559
Jeff Laytondd459bb2015-01-16 15:05:54 -05002560 /* remove flock locks */
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002561 locks_remove_flock(filp, ctx);
Jeff Laytondd459bb2015-01-16 15:05:54 -05002562
Jeff Layton8634b512015-01-16 15:05:55 -05002563 /* remove any leases */
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002564 locks_remove_lease(filp, ctx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565}
2566
2567/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002568 * posix_unblock_lock - stop waiting for a file lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07002569 * @waiter: the lock which was waiting
2570 *
2571 * lockd needs to block waiting for locks.
2572 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002573int
Jeff Laytonf891a292013-06-21 08:58:09 -04002574posix_unblock_lock(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002576 int status = 0;
2577
Jeff Layton7b2296a2013-06-21 08:58:20 -04002578 spin_lock(&blocked_lock_lock);
J. Bruce Fields5996a292006-01-03 09:55:44 +01002579 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002581 else
2582 status = -ENOENT;
Jeff Layton7b2296a2013-06-21 08:58:20 -04002583 spin_unlock(&blocked_lock_lock);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002584 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002585}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002586EXPORT_SYMBOL(posix_unblock_lock);
2587
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002588/**
2589 * vfs_cancel_lock - file byte range unblock lock
2590 * @filp: The file to apply the unblock to
2591 * @fl: The lock to be unblocked
2592 *
2593 * Used by lock managers to cancel blocked requests
2594 */
2595int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2596{
Miklos Szeredic568d682016-09-16 12:44:20 +02002597 if (filp->f_op->lock && is_remote_lock(filp))
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002598 return filp->f_op->lock(filp, F_CANCELLK, fl);
2599 return 0;
2600}
2601
2602EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2603
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002604#ifdef CONFIG_PROC_FS
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002605#include <linux/proc_fs.h>
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002606#include <linux/seq_file.h>
2607
Jeff Layton7012b022013-06-21 08:58:22 -04002608struct locks_iterator {
2609 int li_cpu;
2610 loff_t li_pos;
2611};
2612
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002613static void lock_get_status(struct seq_file *f, struct file_lock *fl,
Jerome Marchand99dc8292010-10-26 14:22:33 -07002614 loff_t id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002615{
2616 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002617 unsigned int fl_pid;
2618
Nikolay Borisovd67fd442016-08-17 16:18:46 -04002619 if (fl->fl_nspid) {
2620 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2621
2622 /* Don't let fl_pid change based on who is reading the file */
2623 fl_pid = pid_nr_ns(fl->fl_nspid, proc_pidns);
2624
2625 /*
2626 * If there isn't a fl_pid don't display who is waiting on
2627 * the lock if we are called from locks_show, or if we are
2628 * called from __show_fd_info - skip lock entirely
2629 */
2630 if (fl_pid == 0)
2631 return;
2632 } else
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002633 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634
2635 if (fl->fl_file != NULL)
Miklos Szeredic568d682016-09-16 12:44:20 +02002636 inode = locks_inode(fl->fl_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002637
Jerome Marchand99dc8292010-10-26 14:22:33 -07002638 seq_printf(f, "%lld:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002639 if (IS_POSIX(fl)) {
Jeff Laytonc918d422014-02-03 12:13:09 -05002640 if (fl->fl_flags & FL_ACCESS)
Fabian Frederick5315c262014-05-09 14:13:05 -04002641 seq_puts(f, "ACCESS");
Jeff Laytoncff2fce2014-04-22 08:24:32 -04002642 else if (IS_OFDLCK(fl))
Fabian Frederick5315c262014-05-09 14:13:05 -04002643 seq_puts(f, "OFDLCK");
Jeff Laytonc918d422014-02-03 12:13:09 -05002644 else
Fabian Frederick5315c262014-05-09 14:13:05 -04002645 seq_puts(f, "POSIX ");
Jeff Laytonc918d422014-02-03 12:13:09 -05002646
2647 seq_printf(f, " %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002648 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002649 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002650 } else if (IS_FLOCK(fl)) {
2651 if (fl->fl_type & LOCK_MAND) {
Fabian Frederick5315c262014-05-09 14:13:05 -04002652 seq_puts(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002654 seq_puts(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655 }
2656 } else if (IS_LEASE(fl)) {
Jeff Layton8144f1f2014-08-11 13:36:54 -04002657 if (fl->fl_flags & FL_DELEG)
2658 seq_puts(f, "DELEG ");
2659 else
2660 seq_puts(f, "LEASE ");
2661
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04002662 if (lease_breaking(fl))
Fabian Frederick5315c262014-05-09 14:13:05 -04002663 seq_puts(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002664 else if (fl->fl_file)
Fabian Frederick5315c262014-05-09 14:13:05 -04002665 seq_puts(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002666 else
Fabian Frederick5315c262014-05-09 14:13:05 -04002667 seq_puts(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002668 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002669 seq_puts(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670 }
2671 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002672 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002673 (fl->fl_type & LOCK_READ)
2674 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2675 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2676 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002677 seq_printf(f, "%s ",
J. Bruce Fieldsab83fa42011-07-26 20:10:51 -04002678 (lease_breaking(fl))
Jeff Layton0ee5c6d2012-08-02 15:46:30 -04002679 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2680 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002681 }
2682 if (inode) {
Jeff Layton36488882015-04-03 09:04:04 -04002683 /* userspace relies on this representation of dev_t */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002684 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002685 MAJOR(inode->i_sb->s_dev),
2686 MINOR(inode->i_sb->s_dev), inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002687 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002688 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002689 }
2690 if (IS_POSIX(fl)) {
2691 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002692 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002693 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002694 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002695 } else {
Fabian Frederick5315c262014-05-09 14:13:05 -04002696 seq_puts(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697 }
2698}
2699
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002700static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002701{
Jeff Layton7012b022013-06-21 08:58:22 -04002702 struct locks_iterator *iter = f->private;
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002703 struct file_lock *fl, *bfl;
Nikolay Borisovd67fd442016-08-17 16:18:46 -04002704 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002705
Jeff Layton139ca042013-06-21 08:58:17 -04002706 fl = hlist_entry(v, struct file_lock, fl_link);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002707
Nikolay Borisovd67fd442016-08-17 16:18:46 -04002708 if (fl->fl_nspid && !pid_nr_ns(fl->fl_nspid, proc_pidns))
2709 return 0;
2710
Jeff Layton7012b022013-06-21 08:58:22 -04002711 lock_get_status(f, fl, iter->li_pos, "");
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002712
2713 list_for_each_entry(bfl, &fl->fl_block, fl_block)
Jeff Layton7012b022013-06-21 08:58:22 -04002714 lock_get_status(f, bfl, iter->li_pos, " ->");
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002715
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002716 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002717}
2718
Andrey Vagin6c8c9032015-04-16 12:49:38 -07002719static void __show_fd_locks(struct seq_file *f,
2720 struct list_head *head, int *id,
2721 struct file *filp, struct files_struct *files)
2722{
2723 struct file_lock *fl;
2724
2725 list_for_each_entry(fl, head, fl_list) {
2726
2727 if (filp != fl->fl_file)
2728 continue;
2729 if (fl->fl_owner != files &&
2730 fl->fl_owner != filp)
2731 continue;
2732
2733 (*id)++;
2734 seq_puts(f, "lock:\t");
2735 lock_get_status(f, fl, *id, "");
2736 }
2737}
2738
2739void show_fd_locks(struct seq_file *f,
2740 struct file *filp, struct files_struct *files)
2741{
Miklos Szeredic568d682016-09-16 12:44:20 +02002742 struct inode *inode = locks_inode(filp);
Andrey Vagin6c8c9032015-04-16 12:49:38 -07002743 struct file_lock_context *ctx;
2744 int id = 0;
2745
Dmitry Vyukov128a3782015-09-21 09:43:06 +02002746 ctx = smp_load_acquire(&inode->i_flctx);
Andrey Vagin6c8c9032015-04-16 12:49:38 -07002747 if (!ctx)
2748 return;
2749
2750 spin_lock(&ctx->flc_lock);
2751 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2752 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2753 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2754 spin_unlock(&ctx->flc_lock);
2755}
2756
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002757static void *locks_start(struct seq_file *f, loff_t *pos)
Jeff Laytonb03dfdec2014-02-03 12:13:07 -05002758 __acquires(&blocked_lock_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002759{
Jeff Layton7012b022013-06-21 08:58:22 -04002760 struct locks_iterator *iter = f->private;
Jerome Marchand99dc8292010-10-26 14:22:33 -07002761
Jeff Layton7012b022013-06-21 08:58:22 -04002762 iter->li_pos = *pos + 1;
Peter Zijlstraaba37662015-06-22 14:16:33 +02002763 percpu_down_write(&file_rwsem);
Jeff Layton7b2296a2013-06-21 08:58:20 -04002764 spin_lock(&blocked_lock_lock);
Peter Zijlstra7c3f6542015-06-22 14:16:34 +02002765 return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766}
2767
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002768static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2769{
Jeff Layton7012b022013-06-21 08:58:22 -04002770 struct locks_iterator *iter = f->private;
2771
2772 ++iter->li_pos;
Peter Zijlstra7c3f6542015-06-22 14:16:34 +02002773 return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002774}
2775
2776static void locks_stop(struct seq_file *f, void *v)
Jeff Laytonb03dfdec2014-02-03 12:13:07 -05002777 __releases(&blocked_lock_lock)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002778{
Jeff Layton7b2296a2013-06-21 08:58:20 -04002779 spin_unlock(&blocked_lock_lock);
Peter Zijlstraaba37662015-06-22 14:16:33 +02002780 percpu_up_write(&file_rwsem);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002781}
2782
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002783static const struct seq_operations locks_seq_operations = {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002784 .start = locks_start,
2785 .next = locks_next,
2786 .stop = locks_stop,
2787 .show = locks_show,
2788};
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002789
2790static int locks_open(struct inode *inode, struct file *filp)
2791{
Jeff Layton7012b022013-06-21 08:58:22 -04002792 return seq_open_private(filp, &locks_seq_operations,
2793 sizeof(struct locks_iterator));
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002794}
2795
2796static const struct file_operations proc_locks_operations = {
2797 .open = locks_open,
2798 .read = seq_read,
2799 .llseek = seq_lseek,
Jerome Marchand99dc8292010-10-26 14:22:33 -07002800 .release = seq_release_private,
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002801};
2802
2803static int __init proc_locks_init(void)
2804{
2805 proc_create("locks", 0, NULL, &proc_locks_operations);
2806 return 0;
2807}
Paul Gortmaker91899222015-12-17 14:11:03 -05002808fs_initcall(proc_locks_init);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002809#endif
2810
Linus Torvalds1da177e2005-04-16 15:20:36 -07002811static int __init filelock_init(void)
2812{
Jeff Layton7012b022013-06-21 08:58:22 -04002813 int i;
2814
Jeff Layton4a075e32015-01-16 15:05:54 -05002815 flctx_cache = kmem_cache_create("file_lock_ctx",
2816 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2817
Linus Torvalds1da177e2005-04-16 15:20:36 -07002818 filelock_cache = kmem_cache_create("file_lock_cache",
Miklos Szerediee19cc42011-07-07 13:06:09 +02002819 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2820
Jeff Layton7012b022013-06-21 08:58:22 -04002821
Peter Zijlstra7c3f6542015-06-22 14:16:34 +02002822 for_each_possible_cpu(i) {
2823 struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i);
2824
2825 spin_lock_init(&fll->lock);
2826 INIT_HLIST_HEAD(&fll->hlist);
2827 }
Jeff Layton7012b022013-06-21 08:58:22 -04002828
Linus Torvalds1da177e2005-04-16 15:20:36 -07002829 return 0;
2830}
2831
2832core_initcall(filelock_init);