]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blob - fs/locks.c
Merge tag 'nfsd-4.19-1' of git://linux-nfs.org/~bfields/linux
[mirror_ubuntu-eoan-kernel.git] / fs / locks.c
1 /*
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.
63 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
64 * 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>
119 #include <linux/fdtable.h>
120 #include <linux/fs.h>
121 #include <linux/init.h>
122 #include <linux/security.h>
123 #include <linux/slab.h>
124 #include <linux/syscalls.h>
125 #include <linux/time.h>
126 #include <linux/rcupdate.h>
127 #include <linux/pid_namespace.h>
128 #include <linux/hashtable.h>
129 #include <linux/percpu.h>
130
131 #define CREATE_TRACE_POINTS
132 #include <trace/events/filelock.h>
133
134 #include <linux/uaccess.h>
135
136 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
137 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
138 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
139 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
140 #define IS_REMOTELCK(fl) (fl->fl_pid <= 0)
141
142 static bool lease_breaking(struct file_lock *fl)
143 {
144 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
145 }
146
147 static int target_leasetype(struct file_lock *fl)
148 {
149 if (fl->fl_flags & FL_UNLOCK_PENDING)
150 return F_UNLCK;
151 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
152 return F_RDLCK;
153 return fl->fl_type;
154 }
155
156 int leases_enable = 1;
157 int lease_break_time = 45;
158
159 /*
160 * The global file_lock_list is only used for displaying /proc/locks, so we
161 * keep a list on each CPU, with each list protected by its own spinlock.
162 * Global serialization is done using file_rwsem.
163 *
164 * Note that alterations to the list also require that the relevant flc_lock is
165 * held.
166 */
167 struct file_lock_list_struct {
168 spinlock_t lock;
169 struct hlist_head hlist;
170 };
171 static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list);
172 DEFINE_STATIC_PERCPU_RWSEM(file_rwsem);
173
174 /*
175 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
176 * It is protected by blocked_lock_lock.
177 *
178 * We hash locks by lockowner in order to optimize searching for the lock a
179 * particular lockowner is waiting on.
180 *
181 * FIXME: make this value scale via some heuristic? We generally will want more
182 * buckets when we have more lockowners holding locks, but that's a little
183 * difficult to determine without knowing what the workload will look like.
184 */
185 #define BLOCKED_HASH_BITS 7
186 static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
187
188 /*
189 * This lock protects the blocked_hash. Generally, if you're accessing it, you
190 * want to be holding this lock.
191 *
192 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
193 * pointer for file_lock structures that are acting as lock requests (in
194 * contrast to those that are acting as records of acquired locks).
195 *
196 * Note that when we acquire this lock in order to change the above fields,
197 * we often hold the flc_lock as well. In certain cases, when reading the fields
198 * protected by this lock, we can skip acquiring it iff we already hold the
199 * flc_lock.
200 */
201 static DEFINE_SPINLOCK(blocked_lock_lock);
202
203 static struct kmem_cache *flctx_cache __read_mostly;
204 static struct kmem_cache *filelock_cache __read_mostly;
205
206 static struct file_lock_context *
207 locks_get_lock_context(struct inode *inode, int type)
208 {
209 struct file_lock_context *ctx;
210
211 /* paired with cmpxchg() below */
212 ctx = smp_load_acquire(&inode->i_flctx);
213 if (likely(ctx) || type == F_UNLCK)
214 goto out;
215
216 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
217 if (!ctx)
218 goto out;
219
220 spin_lock_init(&ctx->flc_lock);
221 INIT_LIST_HEAD(&ctx->flc_flock);
222 INIT_LIST_HEAD(&ctx->flc_posix);
223 INIT_LIST_HEAD(&ctx->flc_lease);
224
225 /*
226 * Assign the pointer if it's not already assigned. If it is, then
227 * free the context we just allocated.
228 */
229 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
230 kmem_cache_free(flctx_cache, ctx);
231 ctx = smp_load_acquire(&inode->i_flctx);
232 }
233 out:
234 trace_locks_get_lock_context(inode, type, ctx);
235 return ctx;
236 }
237
238 static void
239 locks_dump_ctx_list(struct list_head *list, char *list_type)
240 {
241 struct file_lock *fl;
242
243 list_for_each_entry(fl, list, fl_list) {
244 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);
245 }
246 }
247
248 static void
249 locks_check_ctx_lists(struct inode *inode)
250 {
251 struct file_lock_context *ctx = inode->i_flctx;
252
253 if (unlikely(!list_empty(&ctx->flc_flock) ||
254 !list_empty(&ctx->flc_posix) ||
255 !list_empty(&ctx->flc_lease))) {
256 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
257 MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev),
258 inode->i_ino);
259 locks_dump_ctx_list(&ctx->flc_flock, "FLOCK");
260 locks_dump_ctx_list(&ctx->flc_posix, "POSIX");
261 locks_dump_ctx_list(&ctx->flc_lease, "LEASE");
262 }
263 }
264
265 static void
266 locks_check_ctx_file_list(struct file *filp, struct list_head *list,
267 char *list_type)
268 {
269 struct file_lock *fl;
270 struct inode *inode = locks_inode(filp);
271
272 list_for_each_entry(fl, list, fl_list)
273 if (fl->fl_file == filp)
274 pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx "
275 " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n",
276 list_type, MAJOR(inode->i_sb->s_dev),
277 MINOR(inode->i_sb->s_dev), inode->i_ino,
278 fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
279 }
280
281 void
282 locks_free_lock_context(struct inode *inode)
283 {
284 struct file_lock_context *ctx = inode->i_flctx;
285
286 if (unlikely(ctx)) {
287 locks_check_ctx_lists(inode);
288 kmem_cache_free(flctx_cache, ctx);
289 }
290 }
291
292 static void locks_init_lock_heads(struct file_lock *fl)
293 {
294 INIT_HLIST_NODE(&fl->fl_link);
295 INIT_LIST_HEAD(&fl->fl_list);
296 INIT_LIST_HEAD(&fl->fl_block);
297 init_waitqueue_head(&fl->fl_wait);
298 }
299
300 /* Allocate an empty lock structure. */
301 struct file_lock *locks_alloc_lock(void)
302 {
303 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
304
305 if (fl)
306 locks_init_lock_heads(fl);
307
308 return fl;
309 }
310 EXPORT_SYMBOL_GPL(locks_alloc_lock);
311
312 void locks_release_private(struct file_lock *fl)
313 {
314 if (fl->fl_ops) {
315 if (fl->fl_ops->fl_release_private)
316 fl->fl_ops->fl_release_private(fl);
317 fl->fl_ops = NULL;
318 }
319
320 if (fl->fl_lmops) {
321 if (fl->fl_lmops->lm_put_owner) {
322 fl->fl_lmops->lm_put_owner(fl->fl_owner);
323 fl->fl_owner = NULL;
324 }
325 fl->fl_lmops = NULL;
326 }
327 }
328 EXPORT_SYMBOL_GPL(locks_release_private);
329
330 /* Free a lock which is not in use. */
331 void locks_free_lock(struct file_lock *fl)
332 {
333 BUG_ON(waitqueue_active(&fl->fl_wait));
334 BUG_ON(!list_empty(&fl->fl_list));
335 BUG_ON(!list_empty(&fl->fl_block));
336 BUG_ON(!hlist_unhashed(&fl->fl_link));
337
338 locks_release_private(fl);
339 kmem_cache_free(filelock_cache, fl);
340 }
341 EXPORT_SYMBOL(locks_free_lock);
342
343 static void
344 locks_dispose_list(struct list_head *dispose)
345 {
346 struct file_lock *fl;
347
348 while (!list_empty(dispose)) {
349 fl = list_first_entry(dispose, struct file_lock, fl_list);
350 list_del_init(&fl->fl_list);
351 locks_free_lock(fl);
352 }
353 }
354
355 void locks_init_lock(struct file_lock *fl)
356 {
357 memset(fl, 0, sizeof(struct file_lock));
358 locks_init_lock_heads(fl);
359 }
360
361 EXPORT_SYMBOL(locks_init_lock);
362
363 /*
364 * Initialize a new lock from an existing file_lock structure.
365 */
366 void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
367 {
368 new->fl_owner = fl->fl_owner;
369 new->fl_pid = fl->fl_pid;
370 new->fl_file = NULL;
371 new->fl_flags = fl->fl_flags;
372 new->fl_type = fl->fl_type;
373 new->fl_start = fl->fl_start;
374 new->fl_end = fl->fl_end;
375 new->fl_lmops = fl->fl_lmops;
376 new->fl_ops = NULL;
377
378 if (fl->fl_lmops) {
379 if (fl->fl_lmops->lm_get_owner)
380 fl->fl_lmops->lm_get_owner(fl->fl_owner);
381 }
382 }
383 EXPORT_SYMBOL(locks_copy_conflock);
384
385 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
386 {
387 /* "new" must be a freshly-initialized lock */
388 WARN_ON_ONCE(new->fl_ops);
389
390 locks_copy_conflock(new, fl);
391
392 new->fl_file = fl->fl_file;
393 new->fl_ops = fl->fl_ops;
394
395 if (fl->fl_ops) {
396 if (fl->fl_ops->fl_copy_lock)
397 fl->fl_ops->fl_copy_lock(new, fl);
398 }
399 }
400
401 EXPORT_SYMBOL(locks_copy_lock);
402
403 static inline int flock_translate_cmd(int cmd) {
404 if (cmd & LOCK_MAND)
405 return cmd & (LOCK_MAND | LOCK_RW);
406 switch (cmd) {
407 case LOCK_SH:
408 return F_RDLCK;
409 case LOCK_EX:
410 return F_WRLCK;
411 case LOCK_UN:
412 return F_UNLCK;
413 }
414 return -EINVAL;
415 }
416
417 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
418 static struct file_lock *
419 flock_make_lock(struct file *filp, unsigned int cmd)
420 {
421 struct file_lock *fl;
422 int type = flock_translate_cmd(cmd);
423
424 if (type < 0)
425 return ERR_PTR(type);
426
427 fl = locks_alloc_lock();
428 if (fl == NULL)
429 return ERR_PTR(-ENOMEM);
430
431 fl->fl_file = filp;
432 fl->fl_owner = filp;
433 fl->fl_pid = current->tgid;
434 fl->fl_flags = FL_FLOCK;
435 fl->fl_type = type;
436 fl->fl_end = OFFSET_MAX;
437
438 return fl;
439 }
440
441 static int assign_type(struct file_lock *fl, long type)
442 {
443 switch (type) {
444 case F_RDLCK:
445 case F_WRLCK:
446 case F_UNLCK:
447 fl->fl_type = type;
448 break;
449 default:
450 return -EINVAL;
451 }
452 return 0;
453 }
454
455 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
456 struct flock64 *l)
457 {
458 switch (l->l_whence) {
459 case SEEK_SET:
460 fl->fl_start = 0;
461 break;
462 case SEEK_CUR:
463 fl->fl_start = filp->f_pos;
464 break;
465 case SEEK_END:
466 fl->fl_start = i_size_read(file_inode(filp));
467 break;
468 default:
469 return -EINVAL;
470 }
471 if (l->l_start > OFFSET_MAX - fl->fl_start)
472 return -EOVERFLOW;
473 fl->fl_start += l->l_start;
474 if (fl->fl_start < 0)
475 return -EINVAL;
476
477 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
478 POSIX-2001 defines it. */
479 if (l->l_len > 0) {
480 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
481 return -EOVERFLOW;
482 fl->fl_end = fl->fl_start + l->l_len - 1;
483
484 } else if (l->l_len < 0) {
485 if (fl->fl_start + l->l_len < 0)
486 return -EINVAL;
487 fl->fl_end = fl->fl_start - 1;
488 fl->fl_start += l->l_len;
489 } else
490 fl->fl_end = OFFSET_MAX;
491
492 fl->fl_owner = current->files;
493 fl->fl_pid = current->tgid;
494 fl->fl_file = filp;
495 fl->fl_flags = FL_POSIX;
496 fl->fl_ops = NULL;
497 fl->fl_lmops = NULL;
498
499 return assign_type(fl, l->l_type);
500 }
501
502 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
503 * style lock.
504 */
505 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
506 struct flock *l)
507 {
508 struct flock64 ll = {
509 .l_type = l->l_type,
510 .l_whence = l->l_whence,
511 .l_start = l->l_start,
512 .l_len = l->l_len,
513 };
514
515 return flock64_to_posix_lock(filp, fl, &ll);
516 }
517
518 /* default lease lock manager operations */
519 static bool
520 lease_break_callback(struct file_lock *fl)
521 {
522 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
523 return false;
524 }
525
526 static void
527 lease_setup(struct file_lock *fl, void **priv)
528 {
529 struct file *filp = fl->fl_file;
530 struct fasync_struct *fa = *priv;
531
532 /*
533 * fasync_insert_entry() returns the old entry if any. If there was no
534 * old entry, then it used "priv" and inserted it into the fasync list.
535 * Clear the pointer to indicate that it shouldn't be freed.
536 */
537 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
538 *priv = NULL;
539
540 __f_setown(filp, task_pid(current), PIDTYPE_TGID, 0);
541 }
542
543 static const struct lock_manager_operations lease_manager_ops = {
544 .lm_break = lease_break_callback,
545 .lm_change = lease_modify,
546 .lm_setup = lease_setup,
547 };
548
549 /*
550 * Initialize a lease, use the default lock manager operations
551 */
552 static int lease_init(struct file *filp, long type, struct file_lock *fl)
553 {
554 if (assign_type(fl, type) != 0)
555 return -EINVAL;
556
557 fl->fl_owner = filp;
558 fl->fl_pid = current->tgid;
559
560 fl->fl_file = filp;
561 fl->fl_flags = FL_LEASE;
562 fl->fl_start = 0;
563 fl->fl_end = OFFSET_MAX;
564 fl->fl_ops = NULL;
565 fl->fl_lmops = &lease_manager_ops;
566 return 0;
567 }
568
569 /* Allocate a file_lock initialised to this type of lease */
570 static struct file_lock *lease_alloc(struct file *filp, long type)
571 {
572 struct file_lock *fl = locks_alloc_lock();
573 int error = -ENOMEM;
574
575 if (fl == NULL)
576 return ERR_PTR(error);
577
578 error = lease_init(filp, type, fl);
579 if (error) {
580 locks_free_lock(fl);
581 return ERR_PTR(error);
582 }
583 return fl;
584 }
585
586 /* Check if two locks overlap each other.
587 */
588 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
589 {
590 return ((fl1->fl_end >= fl2->fl_start) &&
591 (fl2->fl_end >= fl1->fl_start));
592 }
593
594 /*
595 * Check whether two locks have the same owner.
596 */
597 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
598 {
599 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
600 return fl2->fl_lmops == fl1->fl_lmops &&
601 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
602 return fl1->fl_owner == fl2->fl_owner;
603 }
604
605 /* Must be called with the flc_lock held! */
606 static void locks_insert_global_locks(struct file_lock *fl)
607 {
608 struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list);
609
610 percpu_rwsem_assert_held(&file_rwsem);
611
612 spin_lock(&fll->lock);
613 fl->fl_link_cpu = smp_processor_id();
614 hlist_add_head(&fl->fl_link, &fll->hlist);
615 spin_unlock(&fll->lock);
616 }
617
618 /* Must be called with the flc_lock held! */
619 static void locks_delete_global_locks(struct file_lock *fl)
620 {
621 struct file_lock_list_struct *fll;
622
623 percpu_rwsem_assert_held(&file_rwsem);
624
625 /*
626 * Avoid taking lock if already unhashed. This is safe since this check
627 * is done while holding the flc_lock, and new insertions into the list
628 * also require that it be held.
629 */
630 if (hlist_unhashed(&fl->fl_link))
631 return;
632
633 fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu);
634 spin_lock(&fll->lock);
635 hlist_del_init(&fl->fl_link);
636 spin_unlock(&fll->lock);
637 }
638
639 static unsigned long
640 posix_owner_key(struct file_lock *fl)
641 {
642 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
643 return fl->fl_lmops->lm_owner_key(fl);
644 return (unsigned long)fl->fl_owner;
645 }
646
647 static void locks_insert_global_blocked(struct file_lock *waiter)
648 {
649 lockdep_assert_held(&blocked_lock_lock);
650
651 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
652 }
653
654 static void locks_delete_global_blocked(struct file_lock *waiter)
655 {
656 lockdep_assert_held(&blocked_lock_lock);
657
658 hash_del(&waiter->fl_link);
659 }
660
661 /* Remove waiter from blocker's block list.
662 * When blocker ends up pointing to itself then the list is empty.
663 *
664 * Must be called with blocked_lock_lock held.
665 */
666 static void __locks_delete_block(struct file_lock *waiter)
667 {
668 locks_delete_global_blocked(waiter);
669 list_del_init(&waiter->fl_block);
670 waiter->fl_next = NULL;
671 }
672
673 static void locks_delete_block(struct file_lock *waiter)
674 {
675 spin_lock(&blocked_lock_lock);
676 __locks_delete_block(waiter);
677 spin_unlock(&blocked_lock_lock);
678 }
679
680 /* Insert waiter into blocker's block list.
681 * We use a circular list so that processes can be easily woken up in
682 * the order they blocked. The documentation doesn't require this but
683 * it seems like the reasonable thing to do.
684 *
685 * Must be called with both the flc_lock and blocked_lock_lock held. The
686 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
687 * that the flc_lock is also held on insertions we can avoid taking the
688 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
689 */
690 static void __locks_insert_block(struct file_lock *blocker,
691 struct file_lock *waiter)
692 {
693 BUG_ON(!list_empty(&waiter->fl_block));
694 waiter->fl_next = blocker;
695 list_add_tail(&waiter->fl_block, &blocker->fl_block);
696 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
697 locks_insert_global_blocked(waiter);
698 }
699
700 /* Must be called with flc_lock held. */
701 static void locks_insert_block(struct file_lock *blocker,
702 struct file_lock *waiter)
703 {
704 spin_lock(&blocked_lock_lock);
705 __locks_insert_block(blocker, waiter);
706 spin_unlock(&blocked_lock_lock);
707 }
708
709 /*
710 * Wake up processes blocked waiting for blocker.
711 *
712 * Must be called with the inode->flc_lock held!
713 */
714 static void locks_wake_up_blocks(struct file_lock *blocker)
715 {
716 /*
717 * Avoid taking global lock if list is empty. This is safe since new
718 * blocked requests are only added to the list under the flc_lock, and
719 * the flc_lock is always held here. Note that removal from the fl_block
720 * list does not require the flc_lock, so we must recheck list_empty()
721 * after acquiring the blocked_lock_lock.
722 */
723 if (list_empty(&blocker->fl_block))
724 return;
725
726 spin_lock(&blocked_lock_lock);
727 while (!list_empty(&blocker->fl_block)) {
728 struct file_lock *waiter;
729
730 waiter = list_first_entry(&blocker->fl_block,
731 struct file_lock, fl_block);
732 __locks_delete_block(waiter);
733 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
734 waiter->fl_lmops->lm_notify(waiter);
735 else
736 wake_up(&waiter->fl_wait);
737 }
738 spin_unlock(&blocked_lock_lock);
739 }
740
741 static void
742 locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
743 {
744 list_add_tail(&fl->fl_list, before);
745 locks_insert_global_locks(fl);
746 }
747
748 static void
749 locks_unlink_lock_ctx(struct file_lock *fl)
750 {
751 locks_delete_global_locks(fl);
752 list_del_init(&fl->fl_list);
753 locks_wake_up_blocks(fl);
754 }
755
756 static void
757 locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
758 {
759 locks_unlink_lock_ctx(fl);
760 if (dispose)
761 list_add(&fl->fl_list, dispose);
762 else
763 locks_free_lock(fl);
764 }
765
766 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
767 * checks for shared/exclusive status of overlapping locks.
768 */
769 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
770 {
771 if (sys_fl->fl_type == F_WRLCK)
772 return 1;
773 if (caller_fl->fl_type == F_WRLCK)
774 return 1;
775 return 0;
776 }
777
778 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
779 * checking before calling the locks_conflict().
780 */
781 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
782 {
783 /* POSIX locks owned by the same process do not conflict with
784 * each other.
785 */
786 if (posix_same_owner(caller_fl, sys_fl))
787 return (0);
788
789 /* Check whether they overlap */
790 if (!locks_overlap(caller_fl, sys_fl))
791 return 0;
792
793 return (locks_conflict(caller_fl, sys_fl));
794 }
795
796 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
797 * checking before calling the locks_conflict().
798 */
799 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
800 {
801 /* FLOCK locks referring to the same filp do not conflict with
802 * each other.
803 */
804 if (caller_fl->fl_file == sys_fl->fl_file)
805 return (0);
806 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
807 return 0;
808
809 return (locks_conflict(caller_fl, sys_fl));
810 }
811
812 void
813 posix_test_lock(struct file *filp, struct file_lock *fl)
814 {
815 struct file_lock *cfl;
816 struct file_lock_context *ctx;
817 struct inode *inode = locks_inode(filp);
818
819 ctx = smp_load_acquire(&inode->i_flctx);
820 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
821 fl->fl_type = F_UNLCK;
822 return;
823 }
824
825 spin_lock(&ctx->flc_lock);
826 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
827 if (posix_locks_conflict(fl, cfl)) {
828 locks_copy_conflock(fl, cfl);
829 goto out;
830 }
831 }
832 fl->fl_type = F_UNLCK;
833 out:
834 spin_unlock(&ctx->flc_lock);
835 return;
836 }
837 EXPORT_SYMBOL(posix_test_lock);
838
839 /*
840 * Deadlock detection:
841 *
842 * We attempt to detect deadlocks that are due purely to posix file
843 * locks.
844 *
845 * We assume that a task can be waiting for at most one lock at a time.
846 * So for any acquired lock, the process holding that lock may be
847 * waiting on at most one other lock. That lock in turns may be held by
848 * someone waiting for at most one other lock. Given a requested lock
849 * caller_fl which is about to wait for a conflicting lock block_fl, we
850 * follow this chain of waiters to ensure we are not about to create a
851 * cycle.
852 *
853 * Since we do this before we ever put a process to sleep on a lock, we
854 * are ensured that there is never a cycle; that is what guarantees that
855 * the while() loop in posix_locks_deadlock() eventually completes.
856 *
857 * Note: the above assumption may not be true when handling lock
858 * requests from a broken NFS client. It may also fail in the presence
859 * of tasks (such as posix threads) sharing the same open file table.
860 * To handle those cases, we just bail out after a few iterations.
861 *
862 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
863 * Because the owner is not even nominally tied to a thread of
864 * execution, the deadlock detection below can't reasonably work well. Just
865 * skip it for those.
866 *
867 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
868 * locks that just checks for the case where two tasks are attempting to
869 * upgrade from read to write locks on the same inode.
870 */
871
872 #define MAX_DEADLK_ITERATIONS 10
873
874 /* Find a lock that the owner of the given block_fl is blocking on. */
875 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
876 {
877 struct file_lock *fl;
878
879 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
880 if (posix_same_owner(fl, block_fl))
881 return fl->fl_next;
882 }
883 return NULL;
884 }
885
886 /* Must be called with the blocked_lock_lock held! */
887 static int posix_locks_deadlock(struct file_lock *caller_fl,
888 struct file_lock *block_fl)
889 {
890 int i = 0;
891
892 lockdep_assert_held(&blocked_lock_lock);
893
894 /*
895 * This deadlock detector can't reasonably detect deadlocks with
896 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
897 */
898 if (IS_OFDLCK(caller_fl))
899 return 0;
900
901 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
902 if (i++ > MAX_DEADLK_ITERATIONS)
903 return 0;
904 if (posix_same_owner(caller_fl, block_fl))
905 return 1;
906 }
907 return 0;
908 }
909
910 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
911 * after any leases, but before any posix locks.
912 *
913 * Note that if called with an FL_EXISTS argument, the caller may determine
914 * whether or not a lock was successfully freed by testing the return
915 * value for -ENOENT.
916 */
917 static int flock_lock_inode(struct inode *inode, struct file_lock *request)
918 {
919 struct file_lock *new_fl = NULL;
920 struct file_lock *fl;
921 struct file_lock_context *ctx;
922 int error = 0;
923 bool found = false;
924 LIST_HEAD(dispose);
925
926 ctx = locks_get_lock_context(inode, request->fl_type);
927 if (!ctx) {
928 if (request->fl_type != F_UNLCK)
929 return -ENOMEM;
930 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
931 }
932
933 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
934 new_fl = locks_alloc_lock();
935 if (!new_fl)
936 return -ENOMEM;
937 }
938
939 percpu_down_read_preempt_disable(&file_rwsem);
940 spin_lock(&ctx->flc_lock);
941 if (request->fl_flags & FL_ACCESS)
942 goto find_conflict;
943
944 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
945 if (request->fl_file != fl->fl_file)
946 continue;
947 if (request->fl_type == fl->fl_type)
948 goto out;
949 found = true;
950 locks_delete_lock_ctx(fl, &dispose);
951 break;
952 }
953
954 if (request->fl_type == F_UNLCK) {
955 if ((request->fl_flags & FL_EXISTS) && !found)
956 error = -ENOENT;
957 goto out;
958 }
959
960 find_conflict:
961 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
962 if (!flock_locks_conflict(request, fl))
963 continue;
964 error = -EAGAIN;
965 if (!(request->fl_flags & FL_SLEEP))
966 goto out;
967 error = FILE_LOCK_DEFERRED;
968 locks_insert_block(fl, request);
969 goto out;
970 }
971 if (request->fl_flags & FL_ACCESS)
972 goto out;
973 locks_copy_lock(new_fl, request);
974 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
975 new_fl = NULL;
976 error = 0;
977
978 out:
979 spin_unlock(&ctx->flc_lock);
980 percpu_up_read_preempt_enable(&file_rwsem);
981 if (new_fl)
982 locks_free_lock(new_fl);
983 locks_dispose_list(&dispose);
984 trace_flock_lock_inode(inode, request, error);
985 return error;
986 }
987
988 static int posix_lock_inode(struct inode *inode, struct file_lock *request,
989 struct file_lock *conflock)
990 {
991 struct file_lock *fl, *tmp;
992 struct file_lock *new_fl = NULL;
993 struct file_lock *new_fl2 = NULL;
994 struct file_lock *left = NULL;
995 struct file_lock *right = NULL;
996 struct file_lock_context *ctx;
997 int error;
998 bool added = false;
999 LIST_HEAD(dispose);
1000
1001 ctx = locks_get_lock_context(inode, request->fl_type);
1002 if (!ctx)
1003 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
1004
1005 /*
1006 * We may need two file_lock structures for this operation,
1007 * so we get them in advance to avoid races.
1008 *
1009 * In some cases we can be sure, that no new locks will be needed
1010 */
1011 if (!(request->fl_flags & FL_ACCESS) &&
1012 (request->fl_type != F_UNLCK ||
1013 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
1014 new_fl = locks_alloc_lock();
1015 new_fl2 = locks_alloc_lock();
1016 }
1017
1018 percpu_down_read_preempt_disable(&file_rwsem);
1019 spin_lock(&ctx->flc_lock);
1020 /*
1021 * New lock request. Walk all POSIX locks and look for conflicts. If
1022 * there are any, either return error or put the request on the
1023 * blocker's list of waiters and the global blocked_hash.
1024 */
1025 if (request->fl_type != F_UNLCK) {
1026 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1027 if (!posix_locks_conflict(request, fl))
1028 continue;
1029 if (conflock)
1030 locks_copy_conflock(conflock, fl);
1031 error = -EAGAIN;
1032 if (!(request->fl_flags & FL_SLEEP))
1033 goto out;
1034 /*
1035 * Deadlock detection and insertion into the blocked
1036 * locks list must be done while holding the same lock!
1037 */
1038 error = -EDEADLK;
1039 spin_lock(&blocked_lock_lock);
1040 if (likely(!posix_locks_deadlock(request, fl))) {
1041 error = FILE_LOCK_DEFERRED;
1042 __locks_insert_block(fl, request);
1043 }
1044 spin_unlock(&blocked_lock_lock);
1045 goto out;
1046 }
1047 }
1048
1049 /* If we're just looking for a conflict, we're done. */
1050 error = 0;
1051 if (request->fl_flags & FL_ACCESS)
1052 goto out;
1053
1054 /* Find the first old lock with the same owner as the new lock */
1055 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1056 if (posix_same_owner(request, fl))
1057 break;
1058 }
1059
1060 /* Process locks with this owner. */
1061 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1062 if (!posix_same_owner(request, fl))
1063 break;
1064
1065 /* Detect adjacent or overlapping regions (if same lock type) */
1066 if (request->fl_type == fl->fl_type) {
1067 /* In all comparisons of start vs end, use
1068 * "start - 1" rather than "end + 1". If end
1069 * is OFFSET_MAX, end + 1 will become negative.
1070 */
1071 if (fl->fl_end < request->fl_start - 1)
1072 continue;
1073 /* If the next lock in the list has entirely bigger
1074 * addresses than the new one, insert the lock here.
1075 */
1076 if (fl->fl_start - 1 > request->fl_end)
1077 break;
1078
1079 /* If we come here, the new and old lock are of the
1080 * same type and adjacent or overlapping. Make one
1081 * lock yielding from the lower start address of both
1082 * locks to the higher end address.
1083 */
1084 if (fl->fl_start > request->fl_start)
1085 fl->fl_start = request->fl_start;
1086 else
1087 request->fl_start = fl->fl_start;
1088 if (fl->fl_end < request->fl_end)
1089 fl->fl_end = request->fl_end;
1090 else
1091 request->fl_end = fl->fl_end;
1092 if (added) {
1093 locks_delete_lock_ctx(fl, &dispose);
1094 continue;
1095 }
1096 request = fl;
1097 added = true;
1098 } else {
1099 /* Processing for different lock types is a bit
1100 * more complex.
1101 */
1102 if (fl->fl_end < request->fl_start)
1103 continue;
1104 if (fl->fl_start > request->fl_end)
1105 break;
1106 if (request->fl_type == F_UNLCK)
1107 added = true;
1108 if (fl->fl_start < request->fl_start)
1109 left = fl;
1110 /* If the next lock in the list has a higher end
1111 * address than the new one, insert the new one here.
1112 */
1113 if (fl->fl_end > request->fl_end) {
1114 right = fl;
1115 break;
1116 }
1117 if (fl->fl_start >= request->fl_start) {
1118 /* The new lock completely replaces an old
1119 * one (This may happen several times).
1120 */
1121 if (added) {
1122 locks_delete_lock_ctx(fl, &dispose);
1123 continue;
1124 }
1125 /*
1126 * Replace the old lock with new_fl, and
1127 * remove the old one. It's safe to do the
1128 * insert here since we know that we won't be
1129 * using new_fl later, and that the lock is
1130 * just replacing an existing lock.
1131 */
1132 error = -ENOLCK;
1133 if (!new_fl)
1134 goto out;
1135 locks_copy_lock(new_fl, request);
1136 request = new_fl;
1137 new_fl = NULL;
1138 locks_insert_lock_ctx(request, &fl->fl_list);
1139 locks_delete_lock_ctx(fl, &dispose);
1140 added = true;
1141 }
1142 }
1143 }
1144
1145 /*
1146 * The above code only modifies existing locks in case of merging or
1147 * replacing. If new lock(s) need to be inserted all modifications are
1148 * done below this, so it's safe yet to bail out.
1149 */
1150 error = -ENOLCK; /* "no luck" */
1151 if (right && left == right && !new_fl2)
1152 goto out;
1153
1154 error = 0;
1155 if (!added) {
1156 if (request->fl_type == F_UNLCK) {
1157 if (request->fl_flags & FL_EXISTS)
1158 error = -ENOENT;
1159 goto out;
1160 }
1161
1162 if (!new_fl) {
1163 error = -ENOLCK;
1164 goto out;
1165 }
1166 locks_copy_lock(new_fl, request);
1167 locks_insert_lock_ctx(new_fl, &fl->fl_list);
1168 fl = new_fl;
1169 new_fl = NULL;
1170 }
1171 if (right) {
1172 if (left == right) {
1173 /* The new lock breaks the old one in two pieces,
1174 * so we have to use the second new lock.
1175 */
1176 left = new_fl2;
1177 new_fl2 = NULL;
1178 locks_copy_lock(left, right);
1179 locks_insert_lock_ctx(left, &fl->fl_list);
1180 }
1181 right->fl_start = request->fl_end + 1;
1182 locks_wake_up_blocks(right);
1183 }
1184 if (left) {
1185 left->fl_end = request->fl_start - 1;
1186 locks_wake_up_blocks(left);
1187 }
1188 out:
1189 spin_unlock(&ctx->flc_lock);
1190 percpu_up_read_preempt_enable(&file_rwsem);
1191 /*
1192 * Free any unused locks.
1193 */
1194 if (new_fl)
1195 locks_free_lock(new_fl);
1196 if (new_fl2)
1197 locks_free_lock(new_fl2);
1198 locks_dispose_list(&dispose);
1199 trace_posix_lock_inode(inode, request, error);
1200
1201 return error;
1202 }
1203
1204 /**
1205 * posix_lock_file - Apply a POSIX-style lock to a file
1206 * @filp: The file to apply the lock to
1207 * @fl: The lock to be applied
1208 * @conflock: Place to return a copy of the conflicting lock, if found.
1209 *
1210 * Add a POSIX style lock to a file.
1211 * We merge adjacent & overlapping locks whenever possible.
1212 * POSIX locks are sorted by owner task, then by starting address
1213 *
1214 * Note that if called with an FL_EXISTS argument, the caller may determine
1215 * whether or not a lock was successfully freed by testing the return
1216 * value for -ENOENT.
1217 */
1218 int posix_lock_file(struct file *filp, struct file_lock *fl,
1219 struct file_lock *conflock)
1220 {
1221 return posix_lock_inode(locks_inode(filp), fl, conflock);
1222 }
1223 EXPORT_SYMBOL(posix_lock_file);
1224
1225 /**
1226 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1227 * @inode: inode of file to which lock request should be applied
1228 * @fl: The lock to be applied
1229 *
1230 * Apply a POSIX style lock request to an inode.
1231 */
1232 static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1233 {
1234 int error;
1235 might_sleep ();
1236 for (;;) {
1237 error = posix_lock_inode(inode, fl, NULL);
1238 if (error != FILE_LOCK_DEFERRED)
1239 break;
1240 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1241 if (!error)
1242 continue;
1243
1244 locks_delete_block(fl);
1245 break;
1246 }
1247 return error;
1248 }
1249
1250 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1251 /**
1252 * locks_mandatory_locked - Check for an active lock
1253 * @file: the file to check
1254 *
1255 * Searches the inode's list of locks to find any POSIX locks which conflict.
1256 * This function is called from locks_verify_locked() only.
1257 */
1258 int locks_mandatory_locked(struct file *file)
1259 {
1260 int ret;
1261 struct inode *inode = locks_inode(file);
1262 struct file_lock_context *ctx;
1263 struct file_lock *fl;
1264
1265 ctx = smp_load_acquire(&inode->i_flctx);
1266 if (!ctx || list_empty_careful(&ctx->flc_posix))
1267 return 0;
1268
1269 /*
1270 * Search the lock list for this inode for any POSIX locks.
1271 */
1272 spin_lock(&ctx->flc_lock);
1273 ret = 0;
1274 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1275 if (fl->fl_owner != current->files &&
1276 fl->fl_owner != file) {
1277 ret = -EAGAIN;
1278 break;
1279 }
1280 }
1281 spin_unlock(&ctx->flc_lock);
1282 return ret;
1283 }
1284
1285 /**
1286 * locks_mandatory_area - Check for a conflicting lock
1287 * @inode: the file to check
1288 * @filp: how the file was opened (if it was)
1289 * @start: first byte in the file to check
1290 * @end: lastbyte in the file to check
1291 * @type: %F_WRLCK for a write lock, else %F_RDLCK
1292 *
1293 * Searches the inode's list of locks to find any POSIX locks which conflict.
1294 */
1295 int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start,
1296 loff_t end, unsigned char type)
1297 {
1298 struct file_lock fl;
1299 int error;
1300 bool sleep = false;
1301
1302 locks_init_lock(&fl);
1303 fl.fl_pid = current->tgid;
1304 fl.fl_file = filp;
1305 fl.fl_flags = FL_POSIX | FL_ACCESS;
1306 if (filp && !(filp->f_flags & O_NONBLOCK))
1307 sleep = true;
1308 fl.fl_type = type;
1309 fl.fl_start = start;
1310 fl.fl_end = end;
1311
1312 for (;;) {
1313 if (filp) {
1314 fl.fl_owner = filp;
1315 fl.fl_flags &= ~FL_SLEEP;
1316 error = posix_lock_inode(inode, &fl, NULL);
1317 if (!error)
1318 break;
1319 }
1320
1321 if (sleep)
1322 fl.fl_flags |= FL_SLEEP;
1323 fl.fl_owner = current->files;
1324 error = posix_lock_inode(inode, &fl, NULL);
1325 if (error != FILE_LOCK_DEFERRED)
1326 break;
1327 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1328 if (!error) {
1329 /*
1330 * If we've been sleeping someone might have
1331 * changed the permissions behind our back.
1332 */
1333 if (__mandatory_lock(inode))
1334 continue;
1335 }
1336
1337 locks_delete_block(&fl);
1338 break;
1339 }
1340
1341 return error;
1342 }
1343
1344 EXPORT_SYMBOL(locks_mandatory_area);
1345 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1346
1347 static void lease_clear_pending(struct file_lock *fl, int arg)
1348 {
1349 switch (arg) {
1350 case F_UNLCK:
1351 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1352 /* fall through: */
1353 case F_RDLCK:
1354 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1355 }
1356 }
1357
1358 /* We already had a lease on this file; just change its type */
1359 int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1360 {
1361 int error = assign_type(fl, arg);
1362
1363 if (error)
1364 return error;
1365 lease_clear_pending(fl, arg);
1366 locks_wake_up_blocks(fl);
1367 if (arg == F_UNLCK) {
1368 struct file *filp = fl->fl_file;
1369
1370 f_delown(filp);
1371 filp->f_owner.signum = 0;
1372 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1373 if (fl->fl_fasync != NULL) {
1374 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1375 fl->fl_fasync = NULL;
1376 }
1377 locks_delete_lock_ctx(fl, dispose);
1378 }
1379 return 0;
1380 }
1381 EXPORT_SYMBOL(lease_modify);
1382
1383 static bool past_time(unsigned long then)
1384 {
1385 if (!then)
1386 /* 0 is a special value meaning "this never expires": */
1387 return false;
1388 return time_after(jiffies, then);
1389 }
1390
1391 static void time_out_leases(struct inode *inode, struct list_head *dispose)
1392 {
1393 struct file_lock_context *ctx = inode->i_flctx;
1394 struct file_lock *fl, *tmp;
1395
1396 lockdep_assert_held(&ctx->flc_lock);
1397
1398 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1399 trace_time_out_leases(inode, fl);
1400 if (past_time(fl->fl_downgrade_time))
1401 lease_modify(fl, F_RDLCK, dispose);
1402 if (past_time(fl->fl_break_time))
1403 lease_modify(fl, F_UNLCK, dispose);
1404 }
1405 }
1406
1407 static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1408 {
1409 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1410 return false;
1411 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1412 return false;
1413 return locks_conflict(breaker, lease);
1414 }
1415
1416 static bool
1417 any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1418 {
1419 struct file_lock_context *ctx = inode->i_flctx;
1420 struct file_lock *fl;
1421
1422 lockdep_assert_held(&ctx->flc_lock);
1423
1424 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1425 if (leases_conflict(fl, breaker))
1426 return true;
1427 }
1428 return false;
1429 }
1430
1431 /**
1432 * __break_lease - revoke all outstanding leases on file
1433 * @inode: the inode of the file to return
1434 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1435 * break all leases
1436 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1437 * only delegations
1438 *
1439 * break_lease (inlined for speed) has checked there already is at least
1440 * some kind of lock (maybe a lease) on this file. Leases are broken on
1441 * a call to open() or truncate(). This function can sleep unless you
1442 * specified %O_NONBLOCK to your open().
1443 */
1444 int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1445 {
1446 int error = 0;
1447 struct file_lock_context *ctx;
1448 struct file_lock *new_fl, *fl, *tmp;
1449 unsigned long break_time;
1450 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1451 LIST_HEAD(dispose);
1452
1453 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1454 if (IS_ERR(new_fl))
1455 return PTR_ERR(new_fl);
1456 new_fl->fl_flags = type;
1457
1458 /* typically we will check that ctx is non-NULL before calling */
1459 ctx = smp_load_acquire(&inode->i_flctx);
1460 if (!ctx) {
1461 WARN_ON_ONCE(1);
1462 return error;
1463 }
1464
1465 percpu_down_read_preempt_disable(&file_rwsem);
1466 spin_lock(&ctx->flc_lock);
1467
1468 time_out_leases(inode, &dispose);
1469
1470 if (!any_leases_conflict(inode, new_fl))
1471 goto out;
1472
1473 break_time = 0;
1474 if (lease_break_time > 0) {
1475 break_time = jiffies + lease_break_time * HZ;
1476 if (break_time == 0)
1477 break_time++; /* so that 0 means no break time */
1478 }
1479
1480 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1481 if (!leases_conflict(fl, new_fl))
1482 continue;
1483 if (want_write) {
1484 if (fl->fl_flags & FL_UNLOCK_PENDING)
1485 continue;
1486 fl->fl_flags |= FL_UNLOCK_PENDING;
1487 fl->fl_break_time = break_time;
1488 } else {
1489 if (lease_breaking(fl))
1490 continue;
1491 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1492 fl->fl_downgrade_time = break_time;
1493 }
1494 if (fl->fl_lmops->lm_break(fl))
1495 locks_delete_lock_ctx(fl, &dispose);
1496 }
1497
1498 if (list_empty(&ctx->flc_lease))
1499 goto out;
1500
1501 if (mode & O_NONBLOCK) {
1502 trace_break_lease_noblock(inode, new_fl);
1503 error = -EWOULDBLOCK;
1504 goto out;
1505 }
1506
1507 restart:
1508 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1509 break_time = fl->fl_break_time;
1510 if (break_time != 0)
1511 break_time -= jiffies;
1512 if (break_time == 0)
1513 break_time++;
1514 locks_insert_block(fl, new_fl);
1515 trace_break_lease_block(inode, new_fl);
1516 spin_unlock(&ctx->flc_lock);
1517 percpu_up_read_preempt_enable(&file_rwsem);
1518
1519 locks_dispose_list(&dispose);
1520 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1521 !new_fl->fl_next, break_time);
1522
1523 percpu_down_read_preempt_disable(&file_rwsem);
1524 spin_lock(&ctx->flc_lock);
1525 trace_break_lease_unblock(inode, new_fl);
1526 locks_delete_block(new_fl);
1527 if (error >= 0) {
1528 /*
1529 * Wait for the next conflicting lease that has not been
1530 * broken yet
1531 */
1532 if (error == 0)
1533 time_out_leases(inode, &dispose);
1534 if (any_leases_conflict(inode, new_fl))
1535 goto restart;
1536 error = 0;
1537 }
1538 out:
1539 spin_unlock(&ctx->flc_lock);
1540 percpu_up_read_preempt_enable(&file_rwsem);
1541 locks_dispose_list(&dispose);
1542 locks_free_lock(new_fl);
1543 return error;
1544 }
1545
1546 EXPORT_SYMBOL(__break_lease);
1547
1548 /**
1549 * lease_get_mtime - update modified time of an inode with exclusive lease
1550 * @inode: the inode
1551 * @time: pointer to a timespec which contains the last modified time
1552 *
1553 * This is to force NFS clients to flush their caches for files with
1554 * exclusive leases. The justification is that if someone has an
1555 * exclusive lease, then they could be modifying it.
1556 */
1557 void lease_get_mtime(struct inode *inode, struct timespec64 *time)
1558 {
1559 bool has_lease = false;
1560 struct file_lock_context *ctx;
1561 struct file_lock *fl;
1562
1563 ctx = smp_load_acquire(&inode->i_flctx);
1564 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1565 spin_lock(&ctx->flc_lock);
1566 fl = list_first_entry_or_null(&ctx->flc_lease,
1567 struct file_lock, fl_list);
1568 if (fl && (fl->fl_type == F_WRLCK))
1569 has_lease = true;
1570 spin_unlock(&ctx->flc_lock);
1571 }
1572
1573 if (has_lease)
1574 *time = current_time(inode);
1575 }
1576
1577 EXPORT_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 */
1602 int fcntl_getlease(struct file *filp)
1603 {
1604 struct file_lock *fl;
1605 struct inode *inode = locks_inode(filp);
1606 struct file_lock_context *ctx;
1607 int type = F_UNLCK;
1608 LIST_HEAD(dispose);
1609
1610 ctx = smp_load_acquire(&inode->i_flctx);
1611 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1612 percpu_down_read_preempt_disable(&file_rwsem);
1613 spin_lock(&ctx->flc_lock);
1614 time_out_leases(inode, &dispose);
1615 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1616 if (fl->fl_file != filp)
1617 continue;
1618 type = target_leasetype(fl);
1619 break;
1620 }
1621 spin_unlock(&ctx->flc_lock);
1622 percpu_up_read_preempt_enable(&file_rwsem);
1623
1624 locks_dispose_list(&dispose);
1625 }
1626 return type;
1627 }
1628
1629 /**
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
1635 * @flags: current lock flags
1636 *
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 */
1640 static int
1641 check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
1642 {
1643 int ret = 0;
1644 struct inode *inode = dentry->d_inode;
1645
1646 if (flags & FL_LAYOUT)
1647 return 0;
1648
1649 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1650 return -EAGAIN;
1651
1652 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1653 (atomic_read(&inode->i_count) > 1)))
1654 ret = -EAGAIN;
1655
1656 return ret;
1657 }
1658
1659 static int
1660 generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1661 {
1662 struct file_lock *fl, *my_fl = NULL, *lease;
1663 struct dentry *dentry = filp->f_path.dentry;
1664 struct inode *inode = dentry->d_inode;
1665 struct file_lock_context *ctx;
1666 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
1667 int error;
1668 LIST_HEAD(dispose);
1669
1670 lease = *flp;
1671 trace_generic_add_lease(inode, lease);
1672
1673 /* Note that arg is never F_UNLCK here */
1674 ctx = locks_get_lock_context(inode, arg);
1675 if (!ctx)
1676 return -ENOMEM;
1677
1678 /*
1679 * In the delegation case we need mutual exclusion with
1680 * a number of operations that take the i_mutex. We trylock
1681 * because delegations are an optional optimization, and if
1682 * there's some chance of a conflict--we'd rather not
1683 * bother, maybe that's a sign this just isn't a good file to
1684 * hand out a delegation on.
1685 */
1686 if (is_deleg && !inode_trylock(inode))
1687 return -EAGAIN;
1688
1689 if (is_deleg && arg == F_WRLCK) {
1690 /* Write delegations are not currently supported: */
1691 inode_unlock(inode);
1692 WARN_ON_ONCE(1);
1693 return -EINVAL;
1694 }
1695
1696 percpu_down_read_preempt_disable(&file_rwsem);
1697 spin_lock(&ctx->flc_lock);
1698 time_out_leases(inode, &dispose);
1699 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1700 if (error)
1701 goto out;
1702
1703 /*
1704 * At this point, we know that if there is an exclusive
1705 * lease on this file, then we hold it on this filp
1706 * (otherwise our open of this file would have blocked).
1707 * And if we are trying to acquire an exclusive lease,
1708 * then the file is not open by anyone (including us)
1709 * except for this filp.
1710 */
1711 error = -EAGAIN;
1712 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1713 if (fl->fl_file == filp &&
1714 fl->fl_owner == lease->fl_owner) {
1715 my_fl = fl;
1716 continue;
1717 }
1718
1719 /*
1720 * No exclusive leases if someone else has a lease on
1721 * this file:
1722 */
1723 if (arg == F_WRLCK)
1724 goto out;
1725 /*
1726 * Modifying our existing lease is OK, but no getting a
1727 * new lease if someone else is opening for write:
1728 */
1729 if (fl->fl_flags & FL_UNLOCK_PENDING)
1730 goto out;
1731 }
1732
1733 if (my_fl != NULL) {
1734 lease = my_fl;
1735 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1736 if (error)
1737 goto out;
1738 goto out_setup;
1739 }
1740
1741 error = -EINVAL;
1742 if (!leases_enable)
1743 goto out;
1744
1745 locks_insert_lock_ctx(lease, &ctx->flc_lease);
1746 /*
1747 * The check in break_lease() is lockless. It's possible for another
1748 * open to race in after we did the earlier check for a conflicting
1749 * open but before the lease was inserted. Check again for a
1750 * conflicting open and cancel the lease if there is one.
1751 *
1752 * We also add a barrier here to ensure that the insertion of the lock
1753 * precedes these checks.
1754 */
1755 smp_mb();
1756 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1757 if (error) {
1758 locks_unlink_lock_ctx(lease);
1759 goto out;
1760 }
1761
1762 out_setup:
1763 if (lease->fl_lmops->lm_setup)
1764 lease->fl_lmops->lm_setup(lease, priv);
1765 out:
1766 spin_unlock(&ctx->flc_lock);
1767 percpu_up_read_preempt_enable(&file_rwsem);
1768 locks_dispose_list(&dispose);
1769 if (is_deleg)
1770 inode_unlock(inode);
1771 if (!error && !my_fl)
1772 *flp = NULL;
1773 return error;
1774 }
1775
1776 static int generic_delete_lease(struct file *filp, void *owner)
1777 {
1778 int error = -EAGAIN;
1779 struct file_lock *fl, *victim = NULL;
1780 struct inode *inode = locks_inode(filp);
1781 struct file_lock_context *ctx;
1782 LIST_HEAD(dispose);
1783
1784 ctx = smp_load_acquire(&inode->i_flctx);
1785 if (!ctx) {
1786 trace_generic_delete_lease(inode, NULL);
1787 return error;
1788 }
1789
1790 percpu_down_read_preempt_disable(&file_rwsem);
1791 spin_lock(&ctx->flc_lock);
1792 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1793 if (fl->fl_file == filp &&
1794 fl->fl_owner == owner) {
1795 victim = fl;
1796 break;
1797 }
1798 }
1799 trace_generic_delete_lease(inode, victim);
1800 if (victim)
1801 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
1802 spin_unlock(&ctx->flc_lock);
1803 percpu_up_read_preempt_enable(&file_rwsem);
1804 locks_dispose_list(&dispose);
1805 return error;
1806 }
1807
1808 /**
1809 * generic_setlease - sets a lease on an open file
1810 * @filp: file pointer
1811 * @arg: type of lease to obtain
1812 * @flp: input - file_lock to use, output - file_lock inserted
1813 * @priv: private data for lm_setup (may be NULL if lm_setup
1814 * doesn't require it)
1815 *
1816 * The (input) flp->fl_lmops->lm_break function is required
1817 * by break_lease().
1818 */
1819 int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1820 void **priv)
1821 {
1822 struct inode *inode = locks_inode(filp);
1823 int error;
1824
1825 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1826 return -EACCES;
1827 if (!S_ISREG(inode->i_mode))
1828 return -EINVAL;
1829 error = security_file_lock(filp, arg);
1830 if (error)
1831 return error;
1832
1833 switch (arg) {
1834 case F_UNLCK:
1835 return generic_delete_lease(filp, *priv);
1836 case F_RDLCK:
1837 case F_WRLCK:
1838 if (!(*flp)->fl_lmops->lm_break) {
1839 WARN_ON_ONCE(1);
1840 return -ENOLCK;
1841 }
1842
1843 return generic_add_lease(filp, arg, flp, priv);
1844 default:
1845 return -EINVAL;
1846 }
1847 }
1848 EXPORT_SYMBOL(generic_setlease);
1849
1850 /**
1851 * vfs_setlease - sets a lease on an open file
1852 * @filp: file pointer
1853 * @arg: type of lease to obtain
1854 * @lease: file_lock to use when adding a lease
1855 * @priv: private info for lm_setup when adding a lease (may be
1856 * NULL if lm_setup doesn't require it)
1857 *
1858 * Call this to establish a lease on the file. The "lease" argument is not
1859 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1860 * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be
1861 * set; if not, this function will return -ENOLCK (and generate a scary-looking
1862 * stack trace).
1863 *
1864 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1865 * may be NULL if the lm_setup operation doesn't require it.
1866 */
1867 int
1868 vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1869 {
1870 if (filp->f_op->setlease)
1871 return filp->f_op->setlease(filp, arg, lease, priv);
1872 else
1873 return generic_setlease(filp, arg, lease, priv);
1874 }
1875 EXPORT_SYMBOL_GPL(vfs_setlease);
1876
1877 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1878 {
1879 struct file_lock *fl;
1880 struct fasync_struct *new;
1881 int error;
1882
1883 fl = lease_alloc(filp, arg);
1884 if (IS_ERR(fl))
1885 return PTR_ERR(fl);
1886
1887 new = fasync_alloc();
1888 if (!new) {
1889 locks_free_lock(fl);
1890 return -ENOMEM;
1891 }
1892 new->fa_fd = fd;
1893
1894 error = vfs_setlease(filp, arg, &fl, (void **)&new);
1895 if (fl)
1896 locks_free_lock(fl);
1897 if (new)
1898 fasync_free(new);
1899 return error;
1900 }
1901
1902 /**
1903 * fcntl_setlease - sets a lease on an open file
1904 * @fd: open file descriptor
1905 * @filp: file pointer
1906 * @arg: type of lease to obtain
1907 *
1908 * Call this fcntl to establish a lease on the file.
1909 * Note that you also need to call %F_SETSIG to
1910 * receive a signal when the lease is broken.
1911 */
1912 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1913 {
1914 if (arg == F_UNLCK)
1915 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
1916 return do_fcntl_add_lease(fd, filp, arg);
1917 }
1918
1919 /**
1920 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1921 * @inode: inode of the file to apply to
1922 * @fl: The lock to be applied
1923 *
1924 * Apply a FLOCK style lock request to an inode.
1925 */
1926 static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1927 {
1928 int error;
1929 might_sleep();
1930 for (;;) {
1931 error = flock_lock_inode(inode, fl);
1932 if (error != FILE_LOCK_DEFERRED)
1933 break;
1934 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1935 if (!error)
1936 continue;
1937
1938 locks_delete_block(fl);
1939 break;
1940 }
1941 return error;
1942 }
1943
1944 /**
1945 * locks_lock_inode_wait - Apply a lock to an inode
1946 * @inode: inode of the file to apply to
1947 * @fl: The lock to be applied
1948 *
1949 * Apply a POSIX or FLOCK style lock request to an inode.
1950 */
1951 int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1952 {
1953 int res = 0;
1954 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1955 case FL_POSIX:
1956 res = posix_lock_inode_wait(inode, fl);
1957 break;
1958 case FL_FLOCK:
1959 res = flock_lock_inode_wait(inode, fl);
1960 break;
1961 default:
1962 BUG();
1963 }
1964 return res;
1965 }
1966 EXPORT_SYMBOL(locks_lock_inode_wait);
1967
1968 /**
1969 * sys_flock: - flock() system call.
1970 * @fd: the file descriptor to lock.
1971 * @cmd: the type of lock to apply.
1972 *
1973 * Apply a %FL_FLOCK style lock to an open file descriptor.
1974 * The @cmd can be one of:
1975 *
1976 * - %LOCK_SH -- a shared lock.
1977 * - %LOCK_EX -- an exclusive lock.
1978 * - %LOCK_UN -- remove an existing lock.
1979 * - %LOCK_MAND -- a 'mandatory' flock.
1980 * This exists to emulate Windows Share Modes.
1981 *
1982 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1983 * processes read and write access respectively.
1984 */
1985 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1986 {
1987 struct fd f = fdget(fd);
1988 struct file_lock *lock;
1989 int can_sleep, unlock;
1990 int error;
1991
1992 error = -EBADF;
1993 if (!f.file)
1994 goto out;
1995
1996 can_sleep = !(cmd & LOCK_NB);
1997 cmd &= ~LOCK_NB;
1998 unlock = (cmd == LOCK_UN);
1999
2000 if (!unlock && !(cmd & LOCK_MAND) &&
2001 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
2002 goto out_putf;
2003
2004 lock = flock_make_lock(f.file, cmd);
2005 if (IS_ERR(lock)) {
2006 error = PTR_ERR(lock);
2007 goto out_putf;
2008 }
2009
2010 if (can_sleep)
2011 lock->fl_flags |= FL_SLEEP;
2012
2013 error = security_file_lock(f.file, lock->fl_type);
2014 if (error)
2015 goto out_free;
2016
2017 if (f.file->f_op->flock)
2018 error = f.file->f_op->flock(f.file,
2019 (can_sleep) ? F_SETLKW : F_SETLK,
2020 lock);
2021 else
2022 error = locks_lock_file_wait(f.file, lock);
2023
2024 out_free:
2025 locks_free_lock(lock);
2026
2027 out_putf:
2028 fdput(f);
2029 out:
2030 return error;
2031 }
2032
2033 /**
2034 * vfs_test_lock - test file byte range lock
2035 * @filp: The file to test lock for
2036 * @fl: The lock to test; also used to hold result
2037 *
2038 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2039 * setting conf->fl_type to something other than F_UNLCK.
2040 */
2041 int vfs_test_lock(struct file *filp, struct file_lock *fl)
2042 {
2043 if (filp->f_op->lock)
2044 return filp->f_op->lock(filp, F_GETLK, fl);
2045 posix_test_lock(filp, fl);
2046 return 0;
2047 }
2048 EXPORT_SYMBOL_GPL(vfs_test_lock);
2049
2050 /**
2051 * locks_translate_pid - translate a file_lock's fl_pid number into a namespace
2052 * @fl: The file_lock who's fl_pid should be translated
2053 * @ns: The namespace into which the pid should be translated
2054 *
2055 * Used to tranlate a fl_pid into a namespace virtual pid number
2056 */
2057 static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns)
2058 {
2059 pid_t vnr;
2060 struct pid *pid;
2061
2062 if (IS_OFDLCK(fl))
2063 return -1;
2064 if (IS_REMOTELCK(fl))
2065 return fl->fl_pid;
2066 /*
2067 * If the flock owner process is dead and its pid has been already
2068 * freed, the translation below won't work, but we still want to show
2069 * flock owner pid number in init pidns.
2070 */
2071 if (ns == &init_pid_ns)
2072 return (pid_t)fl->fl_pid;
2073
2074 rcu_read_lock();
2075 pid = find_pid_ns(fl->fl_pid, &init_pid_ns);
2076 vnr = pid_nr_ns(pid, ns);
2077 rcu_read_unlock();
2078 return vnr;
2079 }
2080
2081 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
2082 {
2083 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
2084 #if BITS_PER_LONG == 32
2085 /*
2086 * Make sure we can represent the posix lock via
2087 * legacy 32bit flock.
2088 */
2089 if (fl->fl_start > OFFT_OFFSET_MAX)
2090 return -EOVERFLOW;
2091 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2092 return -EOVERFLOW;
2093 #endif
2094 flock->l_start = fl->fl_start;
2095 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2096 fl->fl_end - fl->fl_start + 1;
2097 flock->l_whence = 0;
2098 flock->l_type = fl->fl_type;
2099 return 0;
2100 }
2101
2102 #if BITS_PER_LONG == 32
2103 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2104 {
2105 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
2106 flock->l_start = fl->fl_start;
2107 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2108 fl->fl_end - fl->fl_start + 1;
2109 flock->l_whence = 0;
2110 flock->l_type = fl->fl_type;
2111 }
2112 #endif
2113
2114 /* Report the first existing lock that would conflict with l.
2115 * This implements the F_GETLK command of fcntl().
2116 */
2117 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock)
2118 {
2119 struct file_lock *fl;
2120 int error;
2121
2122 fl = locks_alloc_lock();
2123 if (fl == NULL)
2124 return -ENOMEM;
2125 error = -EINVAL;
2126 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2127 goto out;
2128
2129 error = flock_to_posix_lock(filp, fl, flock);
2130 if (error)
2131 goto out;
2132
2133 if (cmd == F_OFD_GETLK) {
2134 error = -EINVAL;
2135 if (flock->l_pid != 0)
2136 goto out;
2137
2138 cmd = F_GETLK;
2139 fl->fl_flags |= FL_OFDLCK;
2140 fl->fl_owner = filp;
2141 }
2142
2143 error = vfs_test_lock(filp, fl);
2144 if (error)
2145 goto out;
2146
2147 flock->l_type = fl->fl_type;
2148 if (fl->fl_type != F_UNLCK) {
2149 error = posix_lock_to_flock(flock, fl);
2150 if (error)
2151 goto out;
2152 }
2153 out:
2154 locks_free_lock(fl);
2155 return error;
2156 }
2157
2158 /**
2159 * vfs_lock_file - file byte range lock
2160 * @filp: The file to apply the lock to
2161 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2162 * @fl: The lock to be applied
2163 * @conf: Place to return a copy of the conflicting lock, if found.
2164 *
2165 * A caller that doesn't care about the conflicting lock may pass NULL
2166 * as the final argument.
2167 *
2168 * If the filesystem defines a private ->lock() method, then @conf will
2169 * be left unchanged; so a caller that cares should initialize it to
2170 * some acceptable default.
2171 *
2172 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2173 * locks, the ->lock() interface may return asynchronously, before the lock has
2174 * been granted or denied by the underlying filesystem, if (and only if)
2175 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2176 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2177 * the request is for a blocking lock. When ->lock() does return asynchronously,
2178 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2179 * request completes.
2180 * If the request is for non-blocking lock the file system should return
2181 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2182 * with the result. If the request timed out the callback routine will return a
2183 * nonzero return code and the file system should release the lock. The file
2184 * system is also responsible to keep a corresponding posix lock when it
2185 * grants a lock so the VFS can find out which locks are locally held and do
2186 * the correct lock cleanup when required.
2187 * The underlying filesystem must not drop the kernel lock or call
2188 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2189 * return code.
2190 */
2191 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
2192 {
2193 if (filp->f_op->lock)
2194 return filp->f_op->lock(filp, cmd, fl);
2195 else
2196 return posix_lock_file(filp, fl, conf);
2197 }
2198 EXPORT_SYMBOL_GPL(vfs_lock_file);
2199
2200 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2201 struct file_lock *fl)
2202 {
2203 int error;
2204
2205 error = security_file_lock(filp, fl->fl_type);
2206 if (error)
2207 return error;
2208
2209 for (;;) {
2210 error = vfs_lock_file(filp, cmd, fl, NULL);
2211 if (error != FILE_LOCK_DEFERRED)
2212 break;
2213 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2214 if (!error)
2215 continue;
2216
2217 locks_delete_block(fl);
2218 break;
2219 }
2220
2221 return error;
2222 }
2223
2224 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2225 static int
2226 check_fmode_for_setlk(struct file_lock *fl)
2227 {
2228 switch (fl->fl_type) {
2229 case F_RDLCK:
2230 if (!(fl->fl_file->f_mode & FMODE_READ))
2231 return -EBADF;
2232 break;
2233 case F_WRLCK:
2234 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2235 return -EBADF;
2236 }
2237 return 0;
2238 }
2239
2240 /* Apply the lock described by l to an open file descriptor.
2241 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2242 */
2243 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2244 struct flock *flock)
2245 {
2246 struct file_lock *file_lock = locks_alloc_lock();
2247 struct inode *inode = locks_inode(filp);
2248 struct file *f;
2249 int error;
2250
2251 if (file_lock == NULL)
2252 return -ENOLCK;
2253
2254 /* Don't allow mandatory locks on files that may be memory mapped
2255 * and shared.
2256 */
2257 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2258 error = -EAGAIN;
2259 goto out;
2260 }
2261
2262 error = flock_to_posix_lock(filp, file_lock, flock);
2263 if (error)
2264 goto out;
2265
2266 error = check_fmode_for_setlk(file_lock);
2267 if (error)
2268 goto out;
2269
2270 /*
2271 * If the cmd is requesting file-private locks, then set the
2272 * FL_OFDLCK flag and override the owner.
2273 */
2274 switch (cmd) {
2275 case F_OFD_SETLK:
2276 error = -EINVAL;
2277 if (flock->l_pid != 0)
2278 goto out;
2279
2280 cmd = F_SETLK;
2281 file_lock->fl_flags |= FL_OFDLCK;
2282 file_lock->fl_owner = filp;
2283 break;
2284 case F_OFD_SETLKW:
2285 error = -EINVAL;
2286 if (flock->l_pid != 0)
2287 goto out;
2288
2289 cmd = F_SETLKW;
2290 file_lock->fl_flags |= FL_OFDLCK;
2291 file_lock->fl_owner = filp;
2292 /* Fallthrough */
2293 case F_SETLKW:
2294 file_lock->fl_flags |= FL_SLEEP;
2295 }
2296
2297 error = do_lock_file_wait(filp, cmd, file_lock);
2298
2299 /*
2300 * Attempt to detect a close/fcntl race and recover by releasing the
2301 * lock that was just acquired. There is no need to do that when we're
2302 * unlocking though, or for OFD locks.
2303 */
2304 if (!error && file_lock->fl_type != F_UNLCK &&
2305 !(file_lock->fl_flags & FL_OFDLCK)) {
2306 /*
2307 * We need that spin_lock here - it prevents reordering between
2308 * update of i_flctx->flc_posix and check for it done in
2309 * close(). rcu_read_lock() wouldn't do.
2310 */
2311 spin_lock(&current->files->file_lock);
2312 f = fcheck(fd);
2313 spin_unlock(&current->files->file_lock);
2314 if (f != filp) {
2315 file_lock->fl_type = F_UNLCK;
2316 error = do_lock_file_wait(filp, cmd, file_lock);
2317 WARN_ON_ONCE(error);
2318 error = -EBADF;
2319 }
2320 }
2321 out:
2322 trace_fcntl_setlk(inode, file_lock, error);
2323 locks_free_lock(file_lock);
2324 return error;
2325 }
2326
2327 #if BITS_PER_LONG == 32
2328 /* Report the first existing lock that would conflict with l.
2329 * This implements the F_GETLK command of fcntl().
2330 */
2331 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock)
2332 {
2333 struct file_lock *fl;
2334 int error;
2335
2336 fl = locks_alloc_lock();
2337 if (fl == NULL)
2338 return -ENOMEM;
2339
2340 error = -EINVAL;
2341 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2342 goto out;
2343
2344 error = flock64_to_posix_lock(filp, fl, flock);
2345 if (error)
2346 goto out;
2347
2348 if (cmd == F_OFD_GETLK) {
2349 error = -EINVAL;
2350 if (flock->l_pid != 0)
2351 goto out;
2352
2353 cmd = F_GETLK64;
2354 fl->fl_flags |= FL_OFDLCK;
2355 fl->fl_owner = filp;
2356 }
2357
2358 error = vfs_test_lock(filp, fl);
2359 if (error)
2360 goto out;
2361
2362 flock->l_type = fl->fl_type;
2363 if (fl->fl_type != F_UNLCK)
2364 posix_lock_to_flock64(flock, fl);
2365
2366 out:
2367 locks_free_lock(fl);
2368 return error;
2369 }
2370
2371 /* Apply the lock described by l to an open file descriptor.
2372 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2373 */
2374 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2375 struct flock64 *flock)
2376 {
2377 struct file_lock *file_lock = locks_alloc_lock();
2378 struct inode *inode = locks_inode(filp);
2379 struct file *f;
2380 int error;
2381
2382 if (file_lock == NULL)
2383 return -ENOLCK;
2384
2385 /* Don't allow mandatory locks on files that may be memory mapped
2386 * and shared.
2387 */
2388 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2389 error = -EAGAIN;
2390 goto out;
2391 }
2392
2393 error = flock64_to_posix_lock(filp, file_lock, flock);
2394 if (error)
2395 goto out;
2396
2397 error = check_fmode_for_setlk(file_lock);
2398 if (error)
2399 goto out;
2400
2401 /*
2402 * If the cmd is requesting file-private locks, then set the
2403 * FL_OFDLCK flag and override the owner.
2404 */
2405 switch (cmd) {
2406 case F_OFD_SETLK:
2407 error = -EINVAL;
2408 if (flock->l_pid != 0)
2409 goto out;
2410
2411 cmd = F_SETLK64;
2412 file_lock->fl_flags |= FL_OFDLCK;
2413 file_lock->fl_owner = filp;
2414 break;
2415 case F_OFD_SETLKW:
2416 error = -EINVAL;
2417 if (flock->l_pid != 0)
2418 goto out;
2419
2420 cmd = F_SETLKW64;
2421 file_lock->fl_flags |= FL_OFDLCK;
2422 file_lock->fl_owner = filp;
2423 /* Fallthrough */
2424 case F_SETLKW64:
2425 file_lock->fl_flags |= FL_SLEEP;
2426 }
2427
2428 error = do_lock_file_wait(filp, cmd, file_lock);
2429
2430 /*
2431 * Attempt to detect a close/fcntl race and recover by releasing the
2432 * lock that was just acquired. There is no need to do that when we're
2433 * unlocking though, or for OFD locks.
2434 */
2435 if (!error && file_lock->fl_type != F_UNLCK &&
2436 !(file_lock->fl_flags & FL_OFDLCK)) {
2437 /*
2438 * We need that spin_lock here - it prevents reordering between
2439 * update of i_flctx->flc_posix and check for it done in
2440 * close(). rcu_read_lock() wouldn't do.
2441 */
2442 spin_lock(&current->files->file_lock);
2443 f = fcheck(fd);
2444 spin_unlock(&current->files->file_lock);
2445 if (f != filp) {
2446 file_lock->fl_type = F_UNLCK;
2447 error = do_lock_file_wait(filp, cmd, file_lock);
2448 WARN_ON_ONCE(error);
2449 error = -EBADF;
2450 }
2451 }
2452 out:
2453 locks_free_lock(file_lock);
2454 return error;
2455 }
2456 #endif /* BITS_PER_LONG == 32 */
2457
2458 /*
2459 * This function is called when the file is being removed
2460 * from the task's fd array. POSIX locks belonging to this task
2461 * are deleted at this time.
2462 */
2463 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2464 {
2465 int error;
2466 struct inode *inode = locks_inode(filp);
2467 struct file_lock lock;
2468 struct file_lock_context *ctx;
2469
2470 /*
2471 * If there are no locks held on this file, we don't need to call
2472 * posix_lock_file(). Another process could be setting a lock on this
2473 * file at the same time, but we wouldn't remove that lock anyway.
2474 */
2475 ctx = smp_load_acquire(&inode->i_flctx);
2476 if (!ctx || list_empty(&ctx->flc_posix))
2477 return;
2478
2479 lock.fl_type = F_UNLCK;
2480 lock.fl_flags = FL_POSIX | FL_CLOSE;
2481 lock.fl_start = 0;
2482 lock.fl_end = OFFSET_MAX;
2483 lock.fl_owner = owner;
2484 lock.fl_pid = current->tgid;
2485 lock.fl_file = filp;
2486 lock.fl_ops = NULL;
2487 lock.fl_lmops = NULL;
2488
2489 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
2490
2491 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2492 lock.fl_ops->fl_release_private(&lock);
2493 trace_locks_remove_posix(inode, &lock, error);
2494 }
2495
2496 EXPORT_SYMBOL(locks_remove_posix);
2497
2498 /* The i_flctx must be valid when calling into here */
2499 static void
2500 locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
2501 {
2502 struct file_lock fl = {
2503 .fl_owner = filp,
2504 .fl_pid = current->tgid,
2505 .fl_file = filp,
2506 .fl_flags = FL_FLOCK | FL_CLOSE,
2507 .fl_type = F_UNLCK,
2508 .fl_end = OFFSET_MAX,
2509 };
2510 struct inode *inode = locks_inode(filp);
2511
2512 if (list_empty(&flctx->flc_flock))
2513 return;
2514
2515 if (filp->f_op->flock)
2516 filp->f_op->flock(filp, F_SETLKW, &fl);
2517 else
2518 flock_lock_inode(inode, &fl);
2519
2520 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2521 fl.fl_ops->fl_release_private(&fl);
2522 }
2523
2524 /* The i_flctx must be valid when calling into here */
2525 static void
2526 locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
2527 {
2528 struct file_lock *fl, *tmp;
2529 LIST_HEAD(dispose);
2530
2531 if (list_empty(&ctx->flc_lease))
2532 return;
2533
2534 percpu_down_read_preempt_disable(&file_rwsem);
2535 spin_lock(&ctx->flc_lock);
2536 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
2537 if (filp == fl->fl_file)
2538 lease_modify(fl, F_UNLCK, &dispose);
2539 spin_unlock(&ctx->flc_lock);
2540 percpu_up_read_preempt_enable(&file_rwsem);
2541
2542 locks_dispose_list(&dispose);
2543 }
2544
2545 /*
2546 * This function is called on the last close of an open file.
2547 */
2548 void locks_remove_file(struct file *filp)
2549 {
2550 struct file_lock_context *ctx;
2551
2552 ctx = smp_load_acquire(&locks_inode(filp)->i_flctx);
2553 if (!ctx)
2554 return;
2555
2556 /* remove any OFD locks */
2557 locks_remove_posix(filp, filp);
2558
2559 /* remove flock locks */
2560 locks_remove_flock(filp, ctx);
2561
2562 /* remove any leases */
2563 locks_remove_lease(filp, ctx);
2564
2565 spin_lock(&ctx->flc_lock);
2566 locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX");
2567 locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK");
2568 locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE");
2569 spin_unlock(&ctx->flc_lock);
2570 }
2571
2572 /**
2573 * posix_unblock_lock - stop waiting for a file lock
2574 * @waiter: the lock which was waiting
2575 *
2576 * lockd needs to block waiting for locks.
2577 */
2578 int
2579 posix_unblock_lock(struct file_lock *waiter)
2580 {
2581 int status = 0;
2582
2583 spin_lock(&blocked_lock_lock);
2584 if (waiter->fl_next)
2585 __locks_delete_block(waiter);
2586 else
2587 status = -ENOENT;
2588 spin_unlock(&blocked_lock_lock);
2589 return status;
2590 }
2591 EXPORT_SYMBOL(posix_unblock_lock);
2592
2593 /**
2594 * vfs_cancel_lock - file byte range unblock lock
2595 * @filp: The file to apply the unblock to
2596 * @fl: The lock to be unblocked
2597 *
2598 * Used by lock managers to cancel blocked requests
2599 */
2600 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2601 {
2602 if (filp->f_op->lock)
2603 return filp->f_op->lock(filp, F_CANCELLK, fl);
2604 return 0;
2605 }
2606
2607 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2608
2609 #ifdef CONFIG_PROC_FS
2610 #include <linux/proc_fs.h>
2611 #include <linux/seq_file.h>
2612
2613 struct locks_iterator {
2614 int li_cpu;
2615 loff_t li_pos;
2616 };
2617
2618 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2619 loff_t id, char *pfx)
2620 {
2621 struct inode *inode = NULL;
2622 unsigned int fl_pid;
2623 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2624
2625 fl_pid = locks_translate_pid(fl, proc_pidns);
2626 /*
2627 * If lock owner is dead (and pid is freed) or not visible in current
2628 * pidns, zero is shown as a pid value. Check lock info from
2629 * init_pid_ns to get saved lock pid value.
2630 */
2631
2632 if (fl->fl_file != NULL)
2633 inode = locks_inode(fl->fl_file);
2634
2635 seq_printf(f, "%lld:%s ", id, pfx);
2636 if (IS_POSIX(fl)) {
2637 if (fl->fl_flags & FL_ACCESS)
2638 seq_puts(f, "ACCESS");
2639 else if (IS_OFDLCK(fl))
2640 seq_puts(f, "OFDLCK");
2641 else
2642 seq_puts(f, "POSIX ");
2643
2644 seq_printf(f, " %s ",
2645 (inode == NULL) ? "*NOINODE*" :
2646 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2647 } else if (IS_FLOCK(fl)) {
2648 if (fl->fl_type & LOCK_MAND) {
2649 seq_puts(f, "FLOCK MSNFS ");
2650 } else {
2651 seq_puts(f, "FLOCK ADVISORY ");
2652 }
2653 } else if (IS_LEASE(fl)) {
2654 if (fl->fl_flags & FL_DELEG)
2655 seq_puts(f, "DELEG ");
2656 else
2657 seq_puts(f, "LEASE ");
2658
2659 if (lease_breaking(fl))
2660 seq_puts(f, "BREAKING ");
2661 else if (fl->fl_file)
2662 seq_puts(f, "ACTIVE ");
2663 else
2664 seq_puts(f, "BREAKER ");
2665 } else {
2666 seq_puts(f, "UNKNOWN UNKNOWN ");
2667 }
2668 if (fl->fl_type & LOCK_MAND) {
2669 seq_printf(f, "%s ",
2670 (fl->fl_type & LOCK_READ)
2671 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2672 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2673 } else {
2674 seq_printf(f, "%s ",
2675 (lease_breaking(fl))
2676 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2677 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2678 }
2679 if (inode) {
2680 /* userspace relies on this representation of dev_t */
2681 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2682 MAJOR(inode->i_sb->s_dev),
2683 MINOR(inode->i_sb->s_dev), inode->i_ino);
2684 } else {
2685 seq_printf(f, "%d <none>:0 ", fl_pid);
2686 }
2687 if (IS_POSIX(fl)) {
2688 if (fl->fl_end == OFFSET_MAX)
2689 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2690 else
2691 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2692 } else {
2693 seq_puts(f, "0 EOF\n");
2694 }
2695 }
2696
2697 static int locks_show(struct seq_file *f, void *v)
2698 {
2699 struct locks_iterator *iter = f->private;
2700 struct file_lock *fl, *bfl;
2701 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2702
2703 fl = hlist_entry(v, struct file_lock, fl_link);
2704
2705 if (locks_translate_pid(fl, proc_pidns) == 0)
2706 return 0;
2707
2708 lock_get_status(f, fl, iter->li_pos, "");
2709
2710 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2711 lock_get_status(f, bfl, iter->li_pos, " ->");
2712
2713 return 0;
2714 }
2715
2716 static void __show_fd_locks(struct seq_file *f,
2717 struct list_head *head, int *id,
2718 struct file *filp, struct files_struct *files)
2719 {
2720 struct file_lock *fl;
2721
2722 list_for_each_entry(fl, head, fl_list) {
2723
2724 if (filp != fl->fl_file)
2725 continue;
2726 if (fl->fl_owner != files &&
2727 fl->fl_owner != filp)
2728 continue;
2729
2730 (*id)++;
2731 seq_puts(f, "lock:\t");
2732 lock_get_status(f, fl, *id, "");
2733 }
2734 }
2735
2736 void show_fd_locks(struct seq_file *f,
2737 struct file *filp, struct files_struct *files)
2738 {
2739 struct inode *inode = locks_inode(filp);
2740 struct file_lock_context *ctx;
2741 int id = 0;
2742
2743 ctx = smp_load_acquire(&inode->i_flctx);
2744 if (!ctx)
2745 return;
2746
2747 spin_lock(&ctx->flc_lock);
2748 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2749 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2750 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2751 spin_unlock(&ctx->flc_lock);
2752 }
2753
2754 static void *locks_start(struct seq_file *f, loff_t *pos)
2755 __acquires(&blocked_lock_lock)
2756 {
2757 struct locks_iterator *iter = f->private;
2758
2759 iter->li_pos = *pos + 1;
2760 percpu_down_write(&file_rwsem);
2761 spin_lock(&blocked_lock_lock);
2762 return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos);
2763 }
2764
2765 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2766 {
2767 struct locks_iterator *iter = f->private;
2768
2769 ++iter->li_pos;
2770 return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos);
2771 }
2772
2773 static void locks_stop(struct seq_file *f, void *v)
2774 __releases(&blocked_lock_lock)
2775 {
2776 spin_unlock(&blocked_lock_lock);
2777 percpu_up_write(&file_rwsem);
2778 }
2779
2780 static const struct seq_operations locks_seq_operations = {
2781 .start = locks_start,
2782 .next = locks_next,
2783 .stop = locks_stop,
2784 .show = locks_show,
2785 };
2786
2787 static int __init proc_locks_init(void)
2788 {
2789 proc_create_seq_private("locks", 0, NULL, &locks_seq_operations,
2790 sizeof(struct locks_iterator), NULL);
2791 return 0;
2792 }
2793 fs_initcall(proc_locks_init);
2794 #endif
2795
2796 static int __init filelock_init(void)
2797 {
2798 int i;
2799
2800 flctx_cache = kmem_cache_create("file_lock_ctx",
2801 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2802
2803 filelock_cache = kmem_cache_create("file_lock_cache",
2804 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2805
2806
2807 for_each_possible_cpu(i) {
2808 struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i);
2809
2810 spin_lock_init(&fll->lock);
2811 INIT_HLIST_HEAD(&fll->hlist);
2812 }
2813
2814 return 0;
2815 }
2816
2817 core_initcall(filelock_init);