2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
61 #include "tests/btrfs-tests.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
72 const char *btrfs_decode_error(int errno)
74 char *errstr = "unknown";
78 errstr = "IO failure";
81 errstr = "Out of memory";
84 errstr = "Readonly filesystem";
87 errstr = "Object already exists";
90 errstr = "No space left";
93 errstr = "No such entry";
100 static void save_error_info(struct btrfs_fs_info *fs_info)
103 * today we only save the error info into ram. Long term we'll
104 * also send it down to the disk
106 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
112 struct super_block *sb = fs_info->sb;
114 if (sb->s_flags & MS_RDONLY)
117 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
118 sb->s_flags |= MS_RDONLY;
119 btrfs_info(fs_info, "forced readonly");
121 * Note that a running device replace operation is not
122 * canceled here although there is no way to update
123 * the progress. It would add the risk of a deadlock,
124 * therefore the canceling is ommited. The only penalty
125 * is that some I/O remains active until the procedure
126 * completes. The next time when the filesystem is
127 * mounted writeable again, the device replace
128 * operation continues.
134 * __btrfs_std_error decodes expected errors from the caller and
135 * invokes the approciate error response.
138 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
139 unsigned int line, int errno, const char *fmt, ...)
141 struct super_block *sb = fs_info->sb;
147 * Special case: if the error is EROFS, and we're already
148 * under MS_RDONLY, then it is safe here.
150 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
154 errstr = btrfs_decode_error(errno);
156 struct va_format vaf;
164 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
165 sb->s_id, function, line, errno, errstr, &vaf);
168 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
169 sb->s_id, function, line, errno, errstr);
173 /* Don't go through full error handling during mount */
174 save_error_info(fs_info);
175 if (sb->s_flags & MS_BORN)
176 btrfs_handle_error(fs_info);
180 static const char * const logtypes[] = {
191 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
193 struct super_block *sb = fs_info->sb;
195 struct va_format vaf;
197 const char *type = logtypes[4];
202 kern_level = printk_get_level(fmt);
204 size_t size = printk_skip_level(fmt) - fmt;
205 memcpy(lvl, fmt, size);
208 type = logtypes[kern_level - '0'];
215 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
222 * We only mark the transaction aborted and then set the file system read-only.
223 * This will prevent new transactions from starting or trying to join this
226 * This means that error recovery at the call site is limited to freeing
227 * any local memory allocations and passing the error code up without
228 * further cleanup. The transaction should complete as it normally would
229 * in the call path but will return -EIO.
231 * We'll complete the cleanup in btrfs_end_transaction and
232 * btrfs_commit_transaction.
235 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
236 struct btrfs_root *root, const char *function,
237 unsigned int line, int errno)
239 trans->aborted = errno;
240 /* Nothing used. The other threads that have joined this
241 * transaction may be able to continue. */
242 if (!trans->blocks_used && list_empty(&trans->new_bgs)) {
245 errstr = btrfs_decode_error(errno);
246 btrfs_warn(root->fs_info,
247 "%s:%d: Aborting unused transaction(%s).",
248 function, line, errstr);
251 ACCESS_ONCE(trans->transaction->aborted) = errno;
252 /* Wake up anybody who may be waiting on this transaction */
253 wake_up(&root->fs_info->transaction_wait);
254 wake_up(&root->fs_info->transaction_blocked_wait);
255 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
258 * __btrfs_panic decodes unexpected, fatal errors from the caller,
259 * issues an alert, and either panics or BUGs, depending on mount options.
262 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
263 unsigned int line, int errno, const char *fmt, ...)
265 char *s_id = "<unknown>";
267 struct va_format vaf = { .fmt = fmt };
271 s_id = fs_info->sb->s_id;
276 errstr = btrfs_decode_error(errno);
277 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
278 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
279 s_id, function, line, &vaf, errno, errstr);
281 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
282 function, line, &vaf, errno, errstr);
284 /* Caller calls BUG() */
287 static void btrfs_put_super(struct super_block *sb)
289 close_ctree(btrfs_sb(sb)->tree_root);
293 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
294 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
295 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
296 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
297 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
298 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
299 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
300 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
301 Opt_skip_balance, Opt_check_integrity,
302 Opt_check_integrity_including_extent_data,
303 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
304 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
305 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
306 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
308 #ifdef CONFIG_BTRFS_DEBUG
309 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
314 static const match_table_t tokens = {
315 {Opt_degraded, "degraded"},
316 {Opt_subvol, "subvol=%s"},
317 {Opt_subvolid, "subvolid=%s"},
318 {Opt_device, "device=%s"},
319 {Opt_nodatasum, "nodatasum"},
320 {Opt_datasum, "datasum"},
321 {Opt_nodatacow, "nodatacow"},
322 {Opt_datacow, "datacow"},
323 {Opt_nobarrier, "nobarrier"},
324 {Opt_barrier, "barrier"},
325 {Opt_max_inline, "max_inline=%s"},
326 {Opt_alloc_start, "alloc_start=%s"},
327 {Opt_thread_pool, "thread_pool=%d"},
328 {Opt_compress, "compress"},
329 {Opt_compress_type, "compress=%s"},
330 {Opt_compress_force, "compress-force"},
331 {Opt_compress_force_type, "compress-force=%s"},
333 {Opt_ssd_spread, "ssd_spread"},
334 {Opt_nossd, "nossd"},
336 {Opt_noacl, "noacl"},
337 {Opt_notreelog, "notreelog"},
338 {Opt_treelog, "treelog"},
339 {Opt_nologreplay, "nologreplay"},
340 {Opt_flushoncommit, "flushoncommit"},
341 {Opt_noflushoncommit, "noflushoncommit"},
342 {Opt_ratio, "metadata_ratio=%d"},
343 {Opt_discard, "discard"},
344 {Opt_nodiscard, "nodiscard"},
345 {Opt_space_cache, "space_cache"},
346 {Opt_space_cache_version, "space_cache=%s"},
347 {Opt_clear_cache, "clear_cache"},
348 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
349 {Opt_enospc_debug, "enospc_debug"},
350 {Opt_noenospc_debug, "noenospc_debug"},
351 {Opt_subvolrootid, "subvolrootid=%d"},
352 {Opt_defrag, "autodefrag"},
353 {Opt_nodefrag, "noautodefrag"},
354 {Opt_inode_cache, "inode_cache"},
355 {Opt_noinode_cache, "noinode_cache"},
356 {Opt_no_space_cache, "nospace_cache"},
357 {Opt_recovery, "recovery"}, /* deprecated */
358 {Opt_usebackuproot, "usebackuproot"},
359 {Opt_skip_balance, "skip_balance"},
360 {Opt_check_integrity, "check_int"},
361 {Opt_check_integrity_including_extent_data, "check_int_data"},
362 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
363 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
364 {Opt_fatal_errors, "fatal_errors=%s"},
365 {Opt_commit_interval, "commit=%d"},
366 #ifdef CONFIG_BTRFS_DEBUG
367 {Opt_fragment_data, "fragment=data"},
368 {Opt_fragment_metadata, "fragment=metadata"},
369 {Opt_fragment_all, "fragment=all"},
375 * Regular mount options parser. Everything that is needed only when
376 * reading in a new superblock is parsed here.
377 * XXX JDM: This needs to be cleaned up for remount.
379 int btrfs_parse_options(struct btrfs_root *root, char *options,
380 unsigned long new_flags)
382 struct btrfs_fs_info *info = root->fs_info;
383 substring_t args[MAX_OPT_ARGS];
384 char *p, *num, *orig = NULL;
389 bool compress_force = false;
390 enum btrfs_compression_type saved_compress_type;
391 bool saved_compress_force;
394 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
395 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
396 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
398 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
401 * Even the options are empty, we still need to do extra check
408 * strsep changes the string, duplicate it because parse_options
411 options = kstrdup(options, GFP_NOFS);
417 while ((p = strsep(&options, ",")) != NULL) {
422 token = match_token(p, tokens, args);
425 btrfs_info(root->fs_info, "allowing degraded mounts");
426 btrfs_set_opt(info->mount_opt, DEGRADED);
430 case Opt_subvolrootid:
433 * These are parsed by btrfs_parse_early_options
434 * and can be happily ignored here.
438 btrfs_set_and_info(root, NODATASUM,
439 "setting nodatasum");
442 if (btrfs_test_opt(root, NODATASUM)) {
443 if (btrfs_test_opt(root, NODATACOW))
444 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
446 btrfs_info(root->fs_info, "setting datasum");
448 btrfs_clear_opt(info->mount_opt, NODATACOW);
449 btrfs_clear_opt(info->mount_opt, NODATASUM);
452 if (!btrfs_test_opt(root, NODATACOW)) {
453 if (!btrfs_test_opt(root, COMPRESS) ||
454 !btrfs_test_opt(root, FORCE_COMPRESS)) {
455 btrfs_info(root->fs_info,
456 "setting nodatacow, compression disabled");
458 btrfs_info(root->fs_info, "setting nodatacow");
461 btrfs_clear_opt(info->mount_opt, COMPRESS);
462 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
463 btrfs_set_opt(info->mount_opt, NODATACOW);
464 btrfs_set_opt(info->mount_opt, NODATASUM);
467 btrfs_clear_and_info(root, NODATACOW,
470 case Opt_compress_force:
471 case Opt_compress_force_type:
472 compress_force = true;
475 case Opt_compress_type:
476 saved_compress_type = btrfs_test_opt(root, COMPRESS) ?
477 info->compress_type : BTRFS_COMPRESS_NONE;
478 saved_compress_force =
479 btrfs_test_opt(root, FORCE_COMPRESS);
480 if (token == Opt_compress ||
481 token == Opt_compress_force ||
482 strcmp(args[0].from, "zlib") == 0) {
483 compress_type = "zlib";
484 info->compress_type = BTRFS_COMPRESS_ZLIB;
485 btrfs_set_opt(info->mount_opt, COMPRESS);
486 btrfs_clear_opt(info->mount_opt, NODATACOW);
487 btrfs_clear_opt(info->mount_opt, NODATASUM);
489 } else if (strcmp(args[0].from, "lzo") == 0) {
490 compress_type = "lzo";
491 info->compress_type = BTRFS_COMPRESS_LZO;
492 btrfs_set_opt(info->mount_opt, COMPRESS);
493 btrfs_clear_opt(info->mount_opt, NODATACOW);
494 btrfs_clear_opt(info->mount_opt, NODATASUM);
495 btrfs_set_fs_incompat(info, COMPRESS_LZO);
497 } else if (strncmp(args[0].from, "no", 2) == 0) {
498 compress_type = "no";
499 btrfs_clear_opt(info->mount_opt, COMPRESS);
500 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
501 compress_force = false;
508 if (compress_force) {
509 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
512 * If we remount from compress-force=xxx to
513 * compress=xxx, we need clear FORCE_COMPRESS
514 * flag, otherwise, there is no way for users
515 * to disable forcible compression separately.
517 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
519 if ((btrfs_test_opt(root, COMPRESS) &&
520 (info->compress_type != saved_compress_type ||
521 compress_force != saved_compress_force)) ||
522 (!btrfs_test_opt(root, COMPRESS) &&
524 btrfs_info(root->fs_info,
526 (compress_force) ? "force" : "use",
529 compress_force = false;
532 btrfs_set_and_info(root, SSD,
533 "use ssd allocation scheme");
536 btrfs_set_and_info(root, SSD_SPREAD,
537 "use spread ssd allocation scheme");
538 btrfs_set_opt(info->mount_opt, SSD);
541 btrfs_set_and_info(root, NOSSD,
542 "not using ssd allocation scheme");
543 btrfs_clear_opt(info->mount_opt, SSD);
546 btrfs_clear_and_info(root, NOBARRIER,
547 "turning on barriers");
550 btrfs_set_and_info(root, NOBARRIER,
551 "turning off barriers");
553 case Opt_thread_pool:
554 ret = match_int(&args[0], &intarg);
557 } else if (intarg > 0) {
558 info->thread_pool_size = intarg;
565 num = match_strdup(&args[0]);
567 info->max_inline = memparse(num, NULL);
570 if (info->max_inline) {
571 info->max_inline = min_t(u64,
575 btrfs_info(root->fs_info, "max_inline at %llu",
582 case Opt_alloc_start:
583 num = match_strdup(&args[0]);
585 mutex_lock(&info->chunk_mutex);
586 info->alloc_start = memparse(num, NULL);
587 mutex_unlock(&info->chunk_mutex);
589 btrfs_info(root->fs_info, "allocations start at %llu",
597 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
598 root->fs_info->sb->s_flags |= MS_POSIXACL;
601 btrfs_err(root->fs_info,
602 "support for ACL not compiled in!");
607 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
610 btrfs_set_and_info(root, NOTREELOG,
611 "disabling tree log");
614 btrfs_clear_and_info(root, NOTREELOG,
615 "enabling tree log");
617 case Opt_nologreplay:
618 btrfs_set_and_info(root, NOLOGREPLAY,
619 "disabling log replay at mount time");
621 case Opt_flushoncommit:
622 btrfs_set_and_info(root, FLUSHONCOMMIT,
623 "turning on flush-on-commit");
625 case Opt_noflushoncommit:
626 btrfs_clear_and_info(root, FLUSHONCOMMIT,
627 "turning off flush-on-commit");
630 ret = match_int(&args[0], &intarg);
633 } else if (intarg >= 0) {
634 info->metadata_ratio = intarg;
635 btrfs_info(root->fs_info, "metadata ratio %d",
636 info->metadata_ratio);
643 btrfs_set_and_info(root, DISCARD,
644 "turning on discard");
647 btrfs_clear_and_info(root, DISCARD,
648 "turning off discard");
650 case Opt_space_cache:
651 case Opt_space_cache_version:
652 if (token == Opt_space_cache ||
653 strcmp(args[0].from, "v1") == 0) {
654 btrfs_clear_opt(root->fs_info->mount_opt,
656 btrfs_set_and_info(root, SPACE_CACHE,
657 "enabling disk space caching");
658 } else if (strcmp(args[0].from, "v2") == 0) {
659 btrfs_clear_opt(root->fs_info->mount_opt,
661 btrfs_set_and_info(root, FREE_SPACE_TREE,
662 "enabling free space tree");
668 case Opt_rescan_uuid_tree:
669 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
671 case Opt_no_space_cache:
672 if (btrfs_test_opt(root, SPACE_CACHE)) {
673 btrfs_clear_and_info(root, SPACE_CACHE,
674 "disabling disk space caching");
676 if (btrfs_test_opt(root, FREE_SPACE_TREE)) {
677 btrfs_clear_and_info(root, FREE_SPACE_TREE,
678 "disabling free space tree");
681 case Opt_inode_cache:
682 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
683 "enabling inode map caching");
685 case Opt_noinode_cache:
686 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
687 "disabling inode map caching");
689 case Opt_clear_cache:
690 btrfs_set_and_info(root, CLEAR_CACHE,
691 "force clearing of disk cache");
693 case Opt_user_subvol_rm_allowed:
694 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
696 case Opt_enospc_debug:
697 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
699 case Opt_noenospc_debug:
700 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
703 btrfs_set_and_info(root, AUTO_DEFRAG,
704 "enabling auto defrag");
707 btrfs_clear_and_info(root, AUTO_DEFRAG,
708 "disabling auto defrag");
711 btrfs_warn(root->fs_info,
712 "'recovery' is deprecated, use 'usebackuproot' instead");
713 case Opt_usebackuproot:
714 btrfs_info(root->fs_info,
715 "trying to use backup root at mount time");
716 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
718 case Opt_skip_balance:
719 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
721 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
722 case Opt_check_integrity_including_extent_data:
723 btrfs_info(root->fs_info,
724 "enabling check integrity including extent data");
725 btrfs_set_opt(info->mount_opt,
726 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
727 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
729 case Opt_check_integrity:
730 btrfs_info(root->fs_info, "enabling check integrity");
731 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
733 case Opt_check_integrity_print_mask:
734 ret = match_int(&args[0], &intarg);
737 } else if (intarg >= 0) {
738 info->check_integrity_print_mask = intarg;
739 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
740 info->check_integrity_print_mask);
747 case Opt_check_integrity_including_extent_data:
748 case Opt_check_integrity:
749 case Opt_check_integrity_print_mask:
750 btrfs_err(root->fs_info,
751 "support for check_integrity* not compiled in!");
755 case Opt_fatal_errors:
756 if (strcmp(args[0].from, "panic") == 0)
757 btrfs_set_opt(info->mount_opt,
758 PANIC_ON_FATAL_ERROR);
759 else if (strcmp(args[0].from, "bug") == 0)
760 btrfs_clear_opt(info->mount_opt,
761 PANIC_ON_FATAL_ERROR);
767 case Opt_commit_interval:
769 ret = match_int(&args[0], &intarg);
771 btrfs_err(root->fs_info, "invalid commit interval");
777 btrfs_warn(root->fs_info, "excessive commit interval %d",
780 info->commit_interval = intarg;
782 btrfs_info(root->fs_info, "using default commit interval %ds",
783 BTRFS_DEFAULT_COMMIT_INTERVAL);
784 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
787 #ifdef CONFIG_BTRFS_DEBUG
788 case Opt_fragment_all:
789 btrfs_info(root->fs_info, "fragmenting all space");
790 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
791 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
793 case Opt_fragment_metadata:
794 btrfs_info(root->fs_info, "fragmenting metadata");
795 btrfs_set_opt(info->mount_opt,
798 case Opt_fragment_data:
799 btrfs_info(root->fs_info, "fragmenting data");
800 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
804 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
813 * Extra check for current option against current flag
815 if (btrfs_test_opt(root, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
816 btrfs_err(root->fs_info,
817 "nologreplay must be used with ro mount option");
821 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
822 !btrfs_test_opt(root, FREE_SPACE_TREE) &&
823 !btrfs_test_opt(root, CLEAR_CACHE)) {
824 btrfs_err(root->fs_info, "cannot disable free space tree");
828 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
829 btrfs_info(root->fs_info, "disk space caching is enabled");
830 if (!ret && btrfs_test_opt(root, FREE_SPACE_TREE))
831 btrfs_info(root->fs_info, "using free space tree");
837 * Parse mount options that are required early in the mount process.
839 * All other options will be parsed on much later in the mount process and
840 * only when we need to allocate a new super block.
842 static int btrfs_parse_early_options(const char *options, fmode_t flags,
843 void *holder, char **subvol_name, u64 *subvol_objectid,
844 struct btrfs_fs_devices **fs_devices)
846 substring_t args[MAX_OPT_ARGS];
847 char *device_name, *opts, *orig, *p;
855 * strsep changes the string, duplicate it because parse_options
858 opts = kstrdup(options, GFP_KERNEL);
863 while ((p = strsep(&opts, ",")) != NULL) {
868 token = match_token(p, tokens, args);
872 *subvol_name = match_strdup(&args[0]);
879 num = match_strdup(&args[0]);
881 *subvol_objectid = memparse(num, NULL);
883 /* we want the original fs_tree */
884 if (!*subvol_objectid)
886 BTRFS_FS_TREE_OBJECTID;
892 case Opt_subvolrootid:
894 "BTRFS: 'subvolrootid' mount option is deprecated and has "
898 device_name = match_strdup(&args[0]);
903 error = btrfs_scan_one_device(device_name,
904 flags, holder, fs_devices);
919 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
922 struct btrfs_root *root = fs_info->tree_root;
923 struct btrfs_root *fs_root;
924 struct btrfs_root_ref *root_ref;
925 struct btrfs_inode_ref *inode_ref;
926 struct btrfs_key key;
927 struct btrfs_path *path = NULL;
928 char *name = NULL, *ptr;
933 path = btrfs_alloc_path();
938 path->leave_spinning = 1;
940 name = kmalloc(PATH_MAX, GFP_NOFS);
945 ptr = name + PATH_MAX - 1;
949 * Walk up the subvolume trees in the tree of tree roots by root
950 * backrefs until we hit the top-level subvolume.
952 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
953 key.objectid = subvol_objectid;
954 key.type = BTRFS_ROOT_BACKREF_KEY;
955 key.offset = (u64)-1;
957 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
960 } else if (ret > 0) {
961 ret = btrfs_previous_item(root, path, subvol_objectid,
962 BTRFS_ROOT_BACKREF_KEY);
965 } else if (ret > 0) {
971 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
972 subvol_objectid = key.offset;
974 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
975 struct btrfs_root_ref);
976 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
982 read_extent_buffer(path->nodes[0], ptr + 1,
983 (unsigned long)(root_ref + 1), len);
985 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
986 btrfs_release_path(path);
988 key.objectid = subvol_objectid;
989 key.type = BTRFS_ROOT_ITEM_KEY;
990 key.offset = (u64)-1;
991 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
992 if (IS_ERR(fs_root)) {
993 ret = PTR_ERR(fs_root);
998 * Walk up the filesystem tree by inode refs until we hit the
1001 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1002 key.objectid = dirid;
1003 key.type = BTRFS_INODE_REF_KEY;
1004 key.offset = (u64)-1;
1006 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1009 } else if (ret > 0) {
1010 ret = btrfs_previous_item(fs_root, path, dirid,
1011 BTRFS_INODE_REF_KEY);
1014 } else if (ret > 0) {
1020 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1023 inode_ref = btrfs_item_ptr(path->nodes[0],
1025 struct btrfs_inode_ref);
1026 len = btrfs_inode_ref_name_len(path->nodes[0],
1030 ret = -ENAMETOOLONG;
1033 read_extent_buffer(path->nodes[0], ptr + 1,
1034 (unsigned long)(inode_ref + 1), len);
1036 btrfs_release_path(path);
1040 btrfs_free_path(path);
1041 if (ptr == name + PATH_MAX - 1) {
1045 memmove(name, ptr, name + PATH_MAX - ptr);
1050 btrfs_free_path(path);
1052 return ERR_PTR(ret);
1055 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1057 struct btrfs_root *root = fs_info->tree_root;
1058 struct btrfs_dir_item *di;
1059 struct btrfs_path *path;
1060 struct btrfs_key location;
1063 path = btrfs_alloc_path();
1066 path->leave_spinning = 1;
1069 * Find the "default" dir item which points to the root item that we
1070 * will mount by default if we haven't been given a specific subvolume
1073 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1074 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1076 btrfs_free_path(path);
1081 * Ok the default dir item isn't there. This is weird since
1082 * it's always been there, but don't freak out, just try and
1083 * mount the top-level subvolume.
1085 btrfs_free_path(path);
1086 *objectid = BTRFS_FS_TREE_OBJECTID;
1090 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1091 btrfs_free_path(path);
1092 *objectid = location.objectid;
1096 static int btrfs_fill_super(struct super_block *sb,
1097 struct btrfs_fs_devices *fs_devices,
1098 void *data, int silent)
1100 struct inode *inode;
1101 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1102 struct btrfs_key key;
1105 sb->s_maxbytes = MAX_LFS_FILESIZE;
1106 sb->s_magic = BTRFS_SUPER_MAGIC;
1107 sb->s_op = &btrfs_super_ops;
1108 sb->s_d_op = &btrfs_dentry_operations;
1109 sb->s_export_op = &btrfs_export_ops;
1110 sb->s_xattr = btrfs_xattr_handlers;
1111 sb->s_time_gran = 1;
1112 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1113 sb->s_flags |= MS_POSIXACL;
1115 sb->s_flags |= MS_I_VERSION;
1116 sb->s_iflags |= SB_I_CGROUPWB;
1117 err = open_ctree(sb, fs_devices, (char *)data);
1119 printk(KERN_ERR "BTRFS: open_ctree failed\n");
1123 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1124 key.type = BTRFS_INODE_ITEM_KEY;
1126 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1127 if (IS_ERR(inode)) {
1128 err = PTR_ERR(inode);
1132 sb->s_root = d_make_root(inode);
1138 save_mount_options(sb, data);
1139 cleancache_init_fs(sb);
1140 sb->s_flags |= MS_ACTIVE;
1144 close_ctree(fs_info->tree_root);
1148 int btrfs_sync_fs(struct super_block *sb, int wait)
1150 struct btrfs_trans_handle *trans;
1151 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1152 struct btrfs_root *root = fs_info->tree_root;
1154 trace_btrfs_sync_fs(wait);
1157 filemap_flush(fs_info->btree_inode->i_mapping);
1161 btrfs_wait_ordered_roots(fs_info, -1);
1163 trans = btrfs_attach_transaction_barrier(root);
1164 if (IS_ERR(trans)) {
1165 /* no transaction, don't bother */
1166 if (PTR_ERR(trans) == -ENOENT) {
1168 * Exit unless we have some pending changes
1169 * that need to go through commit
1171 if (fs_info->pending_changes == 0)
1174 * A non-blocking test if the fs is frozen. We must not
1175 * start a new transaction here otherwise a deadlock
1176 * happens. The pending operations are delayed to the
1177 * next commit after thawing.
1179 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1180 __sb_end_write(sb, SB_FREEZE_WRITE);
1183 trans = btrfs_start_transaction(root, 0);
1186 return PTR_ERR(trans);
1188 return btrfs_commit_transaction(trans, root);
1191 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1193 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1194 struct btrfs_root *root = info->tree_root;
1195 char *compress_type;
1197 if (btrfs_test_opt(root, DEGRADED))
1198 seq_puts(seq, ",degraded");
1199 if (btrfs_test_opt(root, NODATASUM))
1200 seq_puts(seq, ",nodatasum");
1201 if (btrfs_test_opt(root, NODATACOW))
1202 seq_puts(seq, ",nodatacow");
1203 if (btrfs_test_opt(root, NOBARRIER))
1204 seq_puts(seq, ",nobarrier");
1205 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1206 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1207 if (info->alloc_start != 0)
1208 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1209 if (info->thread_pool_size != min_t(unsigned long,
1210 num_online_cpus() + 2, 8))
1211 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1212 if (btrfs_test_opt(root, COMPRESS)) {
1213 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1214 compress_type = "zlib";
1216 compress_type = "lzo";
1217 if (btrfs_test_opt(root, FORCE_COMPRESS))
1218 seq_printf(seq, ",compress-force=%s", compress_type);
1220 seq_printf(seq, ",compress=%s", compress_type);
1222 if (btrfs_test_opt(root, NOSSD))
1223 seq_puts(seq, ",nossd");
1224 if (btrfs_test_opt(root, SSD_SPREAD))
1225 seq_puts(seq, ",ssd_spread");
1226 else if (btrfs_test_opt(root, SSD))
1227 seq_puts(seq, ",ssd");
1228 if (btrfs_test_opt(root, NOTREELOG))
1229 seq_puts(seq, ",notreelog");
1230 if (btrfs_test_opt(root, NOLOGREPLAY))
1231 seq_puts(seq, ",nologreplay");
1232 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1233 seq_puts(seq, ",flushoncommit");
1234 if (btrfs_test_opt(root, DISCARD))
1235 seq_puts(seq, ",discard");
1236 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1237 seq_puts(seq, ",noacl");
1238 if (btrfs_test_opt(root, SPACE_CACHE))
1239 seq_puts(seq, ",space_cache");
1240 else if (btrfs_test_opt(root, FREE_SPACE_TREE))
1241 seq_puts(seq, ",space_cache=v2");
1243 seq_puts(seq, ",nospace_cache");
1244 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1245 seq_puts(seq, ",rescan_uuid_tree");
1246 if (btrfs_test_opt(root, CLEAR_CACHE))
1247 seq_puts(seq, ",clear_cache");
1248 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1249 seq_puts(seq, ",user_subvol_rm_allowed");
1250 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1251 seq_puts(seq, ",enospc_debug");
1252 if (btrfs_test_opt(root, AUTO_DEFRAG))
1253 seq_puts(seq, ",autodefrag");
1254 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1255 seq_puts(seq, ",inode_cache");
1256 if (btrfs_test_opt(root, SKIP_BALANCE))
1257 seq_puts(seq, ",skip_balance");
1258 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1259 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1260 seq_puts(seq, ",check_int_data");
1261 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1262 seq_puts(seq, ",check_int");
1263 if (info->check_integrity_print_mask)
1264 seq_printf(seq, ",check_int_print_mask=%d",
1265 info->check_integrity_print_mask);
1267 if (info->metadata_ratio)
1268 seq_printf(seq, ",metadata_ratio=%d",
1269 info->metadata_ratio);
1270 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1271 seq_puts(seq, ",fatal_errors=panic");
1272 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1273 seq_printf(seq, ",commit=%d", info->commit_interval);
1274 #ifdef CONFIG_BTRFS_DEBUG
1275 if (btrfs_test_opt(root, FRAGMENT_DATA))
1276 seq_puts(seq, ",fragment=data");
1277 if (btrfs_test_opt(root, FRAGMENT_METADATA))
1278 seq_puts(seq, ",fragment=metadata");
1280 seq_printf(seq, ",subvolid=%llu",
1281 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1282 seq_puts(seq, ",subvol=");
1283 seq_dentry(seq, dentry, " \t\n\\");
1287 static int btrfs_test_super(struct super_block *s, void *data)
1289 struct btrfs_fs_info *p = data;
1290 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1292 return fs_info->fs_devices == p->fs_devices;
1295 static int btrfs_set_super(struct super_block *s, void *data)
1297 int err = set_anon_super(s, data);
1299 s->s_fs_info = data;
1304 * subvolumes are identified by ino 256
1306 static inline int is_subvolume_inode(struct inode *inode)
1308 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1314 * This will add subvolid=0 to the argument string while removing any subvol=
1315 * and subvolid= arguments to make sure we get the top-level root for path
1316 * walking to the subvol we want.
1318 static char *setup_root_args(char *args)
1320 char *buf, *dst, *sep;
1323 return kstrdup("subvolid=0", GFP_NOFS);
1325 /* The worst case is that we add ",subvolid=0" to the end. */
1326 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1331 sep = strchrnul(args, ',');
1332 if (!strstarts(args, "subvol=") &&
1333 !strstarts(args, "subvolid=")) {
1334 memcpy(dst, args, sep - args);
1343 strcpy(dst, "subvolid=0");
1348 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1349 int flags, const char *device_name,
1352 struct dentry *root;
1353 struct vfsmount *mnt = NULL;
1357 newargs = setup_root_args(data);
1359 root = ERR_PTR(-ENOMEM);
1363 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1364 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1365 if (flags & MS_RDONLY) {
1366 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1367 device_name, newargs);
1369 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1370 device_name, newargs);
1372 root = ERR_CAST(mnt);
1377 down_write(&mnt->mnt_sb->s_umount);
1378 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1379 up_write(&mnt->mnt_sb->s_umount);
1381 root = ERR_PTR(ret);
1387 root = ERR_CAST(mnt);
1393 if (!subvol_objectid) {
1394 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1397 root = ERR_PTR(ret);
1401 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1403 if (IS_ERR(subvol_name)) {
1404 root = ERR_CAST(subvol_name);
1411 root = mount_subtree(mnt, subvol_name);
1412 /* mount_subtree() drops our reference on the vfsmount. */
1415 if (!IS_ERR(root)) {
1416 struct super_block *s = root->d_sb;
1417 struct inode *root_inode = d_inode(root);
1418 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1421 if (!is_subvolume_inode(root_inode)) {
1422 pr_err("BTRFS: '%s' is not a valid subvolume\n",
1426 if (subvol_objectid && root_objectid != subvol_objectid) {
1428 * This will also catch a race condition where a
1429 * subvolume which was passed by ID is renamed and
1430 * another subvolume is renamed over the old location.
1432 pr_err("BTRFS: subvol '%s' does not match subvolid %llu\n",
1433 subvol_name, subvol_objectid);
1438 root = ERR_PTR(ret);
1439 deactivate_locked_super(s);
1450 static int parse_security_options(char *orig_opts,
1451 struct security_mnt_opts *sec_opts)
1453 char *secdata = NULL;
1456 secdata = alloc_secdata();
1459 ret = security_sb_copy_data(orig_opts, secdata);
1461 free_secdata(secdata);
1464 ret = security_sb_parse_opts_str(secdata, sec_opts);
1465 free_secdata(secdata);
1469 static int setup_security_options(struct btrfs_fs_info *fs_info,
1470 struct super_block *sb,
1471 struct security_mnt_opts *sec_opts)
1476 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1479 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1483 #ifdef CONFIG_SECURITY
1484 if (!fs_info->security_opts.num_mnt_opts) {
1485 /* first time security setup, copy sec_opts to fs_info */
1486 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1489 * Since SELinux(the only one supports security_mnt_opts) does
1490 * NOT support changing context during remount/mount same sb,
1491 * This must be the same or part of the same security options,
1494 security_free_mnt_opts(sec_opts);
1501 * Find a superblock for the given device / mount point.
1503 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1504 * for multiple device setup. Make sure to keep it in sync.
1506 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1507 const char *device_name, void *data)
1509 struct block_device *bdev = NULL;
1510 struct super_block *s;
1511 struct btrfs_fs_devices *fs_devices = NULL;
1512 struct btrfs_fs_info *fs_info = NULL;
1513 struct security_mnt_opts new_sec_opts;
1514 fmode_t mode = FMODE_READ;
1515 char *subvol_name = NULL;
1516 u64 subvol_objectid = 0;
1519 if (!(flags & MS_RDONLY))
1520 mode |= FMODE_WRITE;
1522 error = btrfs_parse_early_options(data, mode, fs_type,
1523 &subvol_name, &subvol_objectid,
1527 return ERR_PTR(error);
1530 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1531 /* mount_subvol() will free subvol_name. */
1532 return mount_subvol(subvol_name, subvol_objectid, flags,
1536 security_init_mnt_opts(&new_sec_opts);
1538 error = parse_security_options(data, &new_sec_opts);
1540 return ERR_PTR(error);
1543 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1545 goto error_sec_opts;
1548 * Setup a dummy root and fs_info for test/set super. This is because
1549 * we don't actually fill this stuff out until open_ctree, but we need
1550 * it for searching for existing supers, so this lets us do that and
1551 * then open_ctree will properly initialize everything later.
1553 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1556 goto error_sec_opts;
1559 fs_info->fs_devices = fs_devices;
1561 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1562 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1563 security_init_mnt_opts(&fs_info->security_opts);
1564 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1569 error = btrfs_open_devices(fs_devices, mode, fs_type);
1573 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1575 goto error_close_devices;
1578 bdev = fs_devices->latest_bdev;
1579 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1583 goto error_close_devices;
1587 btrfs_close_devices(fs_devices);
1588 free_fs_info(fs_info);
1589 if ((flags ^ s->s_flags) & MS_RDONLY)
1592 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1593 btrfs_sb(s)->bdev_holder = fs_type;
1594 error = btrfs_fill_super(s, fs_devices, data,
1595 flags & MS_SILENT ? 1 : 0);
1598 deactivate_locked_super(s);
1599 goto error_sec_opts;
1602 fs_info = btrfs_sb(s);
1603 error = setup_security_options(fs_info, s, &new_sec_opts);
1605 deactivate_locked_super(s);
1606 goto error_sec_opts;
1609 return dget(s->s_root);
1611 error_close_devices:
1612 btrfs_close_devices(fs_devices);
1614 free_fs_info(fs_info);
1616 security_free_mnt_opts(&new_sec_opts);
1617 return ERR_PTR(error);
1620 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1621 int new_pool_size, int old_pool_size)
1623 if (new_pool_size == old_pool_size)
1626 fs_info->thread_pool_size = new_pool_size;
1628 btrfs_info(fs_info, "resize thread pool %d -> %d",
1629 old_pool_size, new_pool_size);
1631 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1632 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1633 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1634 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1635 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1636 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1637 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1639 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1640 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1641 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1642 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1643 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1647 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1649 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1652 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1653 unsigned long old_opts, int flags)
1655 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1656 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1657 (flags & MS_RDONLY))) {
1658 /* wait for any defraggers to finish */
1659 wait_event(fs_info->transaction_wait,
1660 (atomic_read(&fs_info->defrag_running) == 0));
1661 if (flags & MS_RDONLY)
1662 sync_filesystem(fs_info->sb);
1666 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1667 unsigned long old_opts)
1670 * We need cleanup all defragable inodes if the autodefragment is
1671 * close or the fs is R/O.
1673 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1674 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1675 (fs_info->sb->s_flags & MS_RDONLY))) {
1676 btrfs_cleanup_defrag_inodes(fs_info);
1679 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1682 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1684 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1685 struct btrfs_root *root = fs_info->tree_root;
1686 unsigned old_flags = sb->s_flags;
1687 unsigned long old_opts = fs_info->mount_opt;
1688 unsigned long old_compress_type = fs_info->compress_type;
1689 u64 old_max_inline = fs_info->max_inline;
1690 u64 old_alloc_start = fs_info->alloc_start;
1691 int old_thread_pool_size = fs_info->thread_pool_size;
1692 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1695 sync_filesystem(sb);
1696 btrfs_remount_prepare(fs_info);
1699 struct security_mnt_opts new_sec_opts;
1701 security_init_mnt_opts(&new_sec_opts);
1702 ret = parse_security_options(data, &new_sec_opts);
1705 ret = setup_security_options(fs_info, sb,
1708 security_free_mnt_opts(&new_sec_opts);
1713 ret = btrfs_parse_options(root, data, *flags);
1719 btrfs_remount_begin(fs_info, old_opts, *flags);
1720 btrfs_resize_thread_pool(fs_info,
1721 fs_info->thread_pool_size, old_thread_pool_size);
1723 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1726 if (*flags & MS_RDONLY) {
1728 * this also happens on 'umount -rf' or on shutdown, when
1729 * the filesystem is busy.
1731 cancel_work_sync(&fs_info->async_reclaim_work);
1733 /* wait for the uuid_scan task to finish */
1734 down(&fs_info->uuid_tree_rescan_sem);
1735 /* avoid complains from lockdep et al. */
1736 up(&fs_info->uuid_tree_rescan_sem);
1738 sb->s_flags |= MS_RDONLY;
1741 * Setting MS_RDONLY will put the cleaner thread to
1742 * sleep at the next loop if it's already active.
1743 * If it's already asleep, we'll leave unused block
1744 * groups on disk until we're mounted read-write again
1745 * unless we clean them up here.
1747 btrfs_delete_unused_bgs(fs_info);
1749 btrfs_dev_replace_suspend_for_unmount(fs_info);
1750 btrfs_scrub_cancel(fs_info);
1751 btrfs_pause_balance(fs_info);
1753 ret = btrfs_commit_super(root);
1757 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1759 "Remounting read-write after error is not allowed");
1763 if (fs_info->fs_devices->rw_devices == 0) {
1768 if (fs_info->fs_devices->missing_devices >
1769 fs_info->num_tolerated_disk_barrier_failures &&
1770 !(*flags & MS_RDONLY)) {
1772 "too many missing devices, writeable remount is not allowed");
1777 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1782 ret = btrfs_cleanup_fs_roots(fs_info);
1786 /* recover relocation */
1787 mutex_lock(&fs_info->cleaner_mutex);
1788 ret = btrfs_recover_relocation(root);
1789 mutex_unlock(&fs_info->cleaner_mutex);
1793 ret = btrfs_resume_balance_async(fs_info);
1797 ret = btrfs_resume_dev_replace_async(fs_info);
1799 btrfs_warn(fs_info, "failed to resume dev_replace");
1803 if (!fs_info->uuid_root) {
1804 btrfs_info(fs_info, "creating UUID tree");
1805 ret = btrfs_create_uuid_tree(fs_info);
1807 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1811 sb->s_flags &= ~MS_RDONLY;
1814 wake_up_process(fs_info->transaction_kthread);
1815 btrfs_remount_cleanup(fs_info, old_opts);
1819 /* We've hit an error - don't reset MS_RDONLY */
1820 if (sb->s_flags & MS_RDONLY)
1821 old_flags |= MS_RDONLY;
1822 sb->s_flags = old_flags;
1823 fs_info->mount_opt = old_opts;
1824 fs_info->compress_type = old_compress_type;
1825 fs_info->max_inline = old_max_inline;
1826 mutex_lock(&fs_info->chunk_mutex);
1827 fs_info->alloc_start = old_alloc_start;
1828 mutex_unlock(&fs_info->chunk_mutex);
1829 btrfs_resize_thread_pool(fs_info,
1830 old_thread_pool_size, fs_info->thread_pool_size);
1831 fs_info->metadata_ratio = old_metadata_ratio;
1832 btrfs_remount_cleanup(fs_info, old_opts);
1836 /* Used to sort the devices by max_avail(descending sort) */
1837 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1838 const void *dev_info2)
1840 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1841 ((struct btrfs_device_info *)dev_info2)->max_avail)
1843 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1844 ((struct btrfs_device_info *)dev_info2)->max_avail)
1851 * sort the devices by max_avail, in which max free extent size of each device
1852 * is stored.(Descending Sort)
1854 static inline void btrfs_descending_sort_devices(
1855 struct btrfs_device_info *devices,
1858 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1859 btrfs_cmp_device_free_bytes, NULL);
1863 * The helper to calc the free space on the devices that can be used to store
1866 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1868 struct btrfs_fs_info *fs_info = root->fs_info;
1869 struct btrfs_device_info *devices_info;
1870 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1871 struct btrfs_device *device;
1876 u64 min_stripe_size;
1877 int min_stripes = 1, num_stripes = 1;
1878 int i = 0, nr_devices;
1882 * We aren't under the device list lock, so this is racey-ish, but good
1883 * enough for our purposes.
1885 nr_devices = fs_info->fs_devices->open_devices;
1888 nr_devices = fs_info->fs_devices->open_devices;
1896 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1901 /* calc min stripe number for data space alloction */
1902 type = btrfs_get_alloc_profile(root, 1);
1903 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1905 num_stripes = nr_devices;
1906 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1909 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1914 if (type & BTRFS_BLOCK_GROUP_DUP)
1915 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1917 min_stripe_size = BTRFS_STRIPE_LEN;
1919 if (fs_info->alloc_start)
1920 mutex_lock(&fs_devices->device_list_mutex);
1922 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1923 if (!device->in_fs_metadata || !device->bdev ||
1924 device->is_tgtdev_for_dev_replace)
1927 if (i >= nr_devices)
1930 avail_space = device->total_bytes - device->bytes_used;
1932 /* align with stripe_len */
1933 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1934 avail_space *= BTRFS_STRIPE_LEN;
1937 * In order to avoid overwritting the superblock on the drive,
1938 * btrfs starts at an offset of at least 1MB when doing chunk
1943 /* user can set the offset in fs_info->alloc_start. */
1944 if (fs_info->alloc_start &&
1945 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1946 device->total_bytes) {
1948 skip_space = max(fs_info->alloc_start, skip_space);
1951 * btrfs can not use the free space in
1952 * [0, skip_space - 1], we must subtract it from the
1953 * total. In order to implement it, we account the used
1954 * space in this range first.
1956 ret = btrfs_account_dev_extents_size(device, 0,
1960 kfree(devices_info);
1961 mutex_unlock(&fs_devices->device_list_mutex);
1967 /* calc the free space in [0, skip_space - 1] */
1968 skip_space -= used_space;
1972 * we can use the free space in [0, skip_space - 1], subtract
1973 * it from the total.
1975 if (avail_space && avail_space >= skip_space)
1976 avail_space -= skip_space;
1980 if (avail_space < min_stripe_size)
1983 devices_info[i].dev = device;
1984 devices_info[i].max_avail = avail_space;
1989 if (fs_info->alloc_start)
1990 mutex_unlock(&fs_devices->device_list_mutex);
1994 btrfs_descending_sort_devices(devices_info, nr_devices);
1998 while (nr_devices >= min_stripes) {
1999 if (num_stripes > nr_devices)
2000 num_stripes = nr_devices;
2002 if (devices_info[i].max_avail >= min_stripe_size) {
2006 avail_space += devices_info[i].max_avail * num_stripes;
2007 alloc_size = devices_info[i].max_avail;
2008 for (j = i + 1 - num_stripes; j <= i; j++)
2009 devices_info[j].max_avail -= alloc_size;
2015 kfree(devices_info);
2016 *free_bytes = avail_space;
2021 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2023 * If there's a redundant raid level at DATA block groups, use the respective
2024 * multiplier to scale the sizes.
2026 * Unused device space usage is based on simulating the chunk allocator
2027 * algorithm that respects the device sizes, order of allocations and the
2028 * 'alloc_start' value, this is a close approximation of the actual use but
2029 * there are other factors that may change the result (like a new metadata
2032 * If metadata is exhausted, f_bavail will be 0.
2034 * FIXME: not accurate for mixed block groups, total and free/used are ok,
2035 * available appears slightly larger.
2037 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2039 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2040 struct btrfs_super_block *disk_super = fs_info->super_copy;
2041 struct list_head *head = &fs_info->space_info;
2042 struct btrfs_space_info *found;
2044 u64 total_free_data = 0;
2045 u64 total_free_meta = 0;
2046 int bits = dentry->d_sb->s_blocksize_bits;
2047 __be32 *fsid = (__be32 *)fs_info->fsid;
2048 unsigned factor = 1;
2049 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2054 * holding chunk_muext to avoid allocating new chunks, holding
2055 * device_list_mutex to avoid the device being removed
2058 list_for_each_entry_rcu(found, head, list) {
2059 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2062 total_free_data += found->disk_total - found->disk_used;
2064 btrfs_account_ro_block_groups_free_space(found);
2066 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2067 if (!list_empty(&found->block_groups[i])) {
2069 case BTRFS_RAID_DUP:
2070 case BTRFS_RAID_RAID1:
2071 case BTRFS_RAID_RAID10:
2077 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
2078 total_free_meta += found->disk_total - found->disk_used;
2080 total_used += found->disk_used;
2085 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2086 buf->f_blocks >>= bits;
2087 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2089 /* Account global block reserve as used, it's in logical size already */
2090 spin_lock(&block_rsv->lock);
2091 buf->f_bfree -= block_rsv->size >> bits;
2092 spin_unlock(&block_rsv->lock);
2094 buf->f_bavail = div_u64(total_free_data, factor);
2095 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2098 buf->f_bavail += div_u64(total_free_data, factor);
2099 buf->f_bavail = buf->f_bavail >> bits;
2102 * We calculate the remaining metadata space minus global reserve. If
2103 * this is (supposedly) smaller than zero, there's no space. But this
2104 * does not hold in practice, the exhausted state happens where's still
2105 * some positive delta. So we apply some guesswork and compare the
2106 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2108 * We probably cannot calculate the exact threshold value because this
2109 * depends on the internal reservations requested by various
2110 * operations, so some operations that consume a few metadata will
2111 * succeed even if the Avail is zero. But this is better than the other
2114 thresh = 4 * 1024 * 1024;
2116 if (total_free_meta - thresh < block_rsv->size)
2119 buf->f_type = BTRFS_SUPER_MAGIC;
2120 buf->f_bsize = dentry->d_sb->s_blocksize;
2121 buf->f_namelen = BTRFS_NAME_LEN;
2123 /* We treat it as constant endianness (it doesn't matter _which_)
2124 because we want the fsid to come out the same whether mounted
2125 on a big-endian or little-endian host */
2126 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2127 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2128 /* Mask in the root object ID too, to disambiguate subvols */
2129 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2130 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2135 static void btrfs_kill_super(struct super_block *sb)
2137 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2138 kill_anon_super(sb);
2139 free_fs_info(fs_info);
2142 static struct file_system_type btrfs_fs_type = {
2143 .owner = THIS_MODULE,
2145 .mount = btrfs_mount,
2146 .kill_sb = btrfs_kill_super,
2147 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2149 MODULE_ALIAS_FS("btrfs");
2151 static int btrfs_control_open(struct inode *inode, struct file *file)
2154 * The control file's private_data is used to hold the
2155 * transaction when it is started and is used to keep
2156 * track of whether a transaction is already in progress.
2158 file->private_data = NULL;
2163 * used by btrfsctl to scan devices when no FS is mounted
2165 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2168 struct btrfs_ioctl_vol_args *vol;
2169 struct btrfs_fs_devices *fs_devices;
2172 if (!capable(CAP_SYS_ADMIN))
2175 vol = memdup_user((void __user *)arg, sizeof(*vol));
2177 return PTR_ERR(vol);
2180 case BTRFS_IOC_SCAN_DEV:
2181 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2182 &btrfs_fs_type, &fs_devices);
2184 case BTRFS_IOC_DEVICES_READY:
2185 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2186 &btrfs_fs_type, &fs_devices);
2189 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2197 static int btrfs_freeze(struct super_block *sb)
2199 struct btrfs_trans_handle *trans;
2200 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2202 trans = btrfs_attach_transaction_barrier(root);
2203 if (IS_ERR(trans)) {
2204 /* no transaction, don't bother */
2205 if (PTR_ERR(trans) == -ENOENT)
2207 return PTR_ERR(trans);
2209 return btrfs_commit_transaction(trans, root);
2212 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2214 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2215 struct btrfs_fs_devices *cur_devices;
2216 struct btrfs_device *dev, *first_dev = NULL;
2217 struct list_head *head;
2218 struct rcu_string *name;
2220 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2221 cur_devices = fs_info->fs_devices;
2222 while (cur_devices) {
2223 head = &cur_devices->devices;
2224 list_for_each_entry(dev, head, dev_list) {
2229 if (!first_dev || dev->devid < first_dev->devid)
2232 cur_devices = cur_devices->seed;
2237 name = rcu_dereference(first_dev->name);
2238 seq_escape(m, name->str, " \t\n\\");
2243 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2247 static const struct super_operations btrfs_super_ops = {
2248 .drop_inode = btrfs_drop_inode,
2249 .evict_inode = btrfs_evict_inode,
2250 .put_super = btrfs_put_super,
2251 .sync_fs = btrfs_sync_fs,
2252 .show_options = btrfs_show_options,
2253 .show_devname = btrfs_show_devname,
2254 .write_inode = btrfs_write_inode,
2255 .alloc_inode = btrfs_alloc_inode,
2256 .destroy_inode = btrfs_destroy_inode,
2257 .statfs = btrfs_statfs,
2258 .remount_fs = btrfs_remount,
2259 .freeze_fs = btrfs_freeze,
2262 static const struct file_operations btrfs_ctl_fops = {
2263 .open = btrfs_control_open,
2264 .unlocked_ioctl = btrfs_control_ioctl,
2265 .compat_ioctl = btrfs_control_ioctl,
2266 .owner = THIS_MODULE,
2267 .llseek = noop_llseek,
2270 static struct miscdevice btrfs_misc = {
2271 .minor = BTRFS_MINOR,
2272 .name = "btrfs-control",
2273 .fops = &btrfs_ctl_fops
2276 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2277 MODULE_ALIAS("devname:btrfs-control");
2279 static int btrfs_interface_init(void)
2281 return misc_register(&btrfs_misc);
2284 static void btrfs_interface_exit(void)
2286 misc_deregister(&btrfs_misc);
2289 static void btrfs_print_info(void)
2291 printk(KERN_INFO "Btrfs loaded"
2292 #ifdef CONFIG_BTRFS_DEBUG
2295 #ifdef CONFIG_BTRFS_ASSERT
2298 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2299 ", integrity-checker=on"
2304 static int btrfs_run_sanity_tests(void)
2308 ret = btrfs_init_test_fs();
2312 ret = btrfs_test_free_space_cache();
2315 ret = btrfs_test_extent_buffer_operations();
2318 ret = btrfs_test_extent_io();
2321 ret = btrfs_test_inodes();
2324 ret = btrfs_test_qgroups();
2327 ret = btrfs_test_free_space_tree();
2329 btrfs_destroy_test_fs();
2333 static int __init init_btrfs_fs(void)
2337 err = btrfs_hash_init();
2343 err = btrfs_init_sysfs();
2347 btrfs_init_compress();
2349 err = btrfs_init_cachep();
2353 err = extent_io_init();
2357 err = extent_map_init();
2359 goto free_extent_io;
2361 err = ordered_data_init();
2363 goto free_extent_map;
2365 err = btrfs_delayed_inode_init();
2367 goto free_ordered_data;
2369 err = btrfs_auto_defrag_init();
2371 goto free_delayed_inode;
2373 err = btrfs_delayed_ref_init();
2375 goto free_auto_defrag;
2377 err = btrfs_prelim_ref_init();
2379 goto free_delayed_ref;
2381 err = btrfs_end_io_wq_init();
2383 goto free_prelim_ref;
2385 err = btrfs_interface_init();
2387 goto free_end_io_wq;
2389 btrfs_init_lockdep();
2393 err = btrfs_run_sanity_tests();
2395 goto unregister_ioctl;
2397 err = register_filesystem(&btrfs_fs_type);
2399 goto unregister_ioctl;
2404 btrfs_interface_exit();
2406 btrfs_end_io_wq_exit();
2408 btrfs_prelim_ref_exit();
2410 btrfs_delayed_ref_exit();
2412 btrfs_auto_defrag_exit();
2414 btrfs_delayed_inode_exit();
2416 ordered_data_exit();
2422 btrfs_destroy_cachep();
2424 btrfs_exit_compress();
2431 static void __exit exit_btrfs_fs(void)
2433 btrfs_destroy_cachep();
2434 btrfs_delayed_ref_exit();
2435 btrfs_auto_defrag_exit();
2436 btrfs_delayed_inode_exit();
2437 btrfs_prelim_ref_exit();
2438 ordered_data_exit();
2441 btrfs_interface_exit();
2442 btrfs_end_io_wq_exit();
2443 unregister_filesystem(&btrfs_fs_type);
2445 btrfs_cleanup_fs_uuids();
2446 btrfs_exit_compress();
2450 late_initcall(init_btrfs_fs);
2451 module_exit(exit_btrfs_fs)
2453 MODULE_LICENSE("GPL");