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_clear_cache, Opt_user_subvol_rm_allowed,
299 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
300 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
301 Opt_check_integrity, Opt_check_integrity_including_extent_data,
302 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
303 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
304 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
305 Opt_datasum, Opt_treelog, Opt_noinode_cache,
306 #ifdef CONFIG_BTRFS_DEBUG
307 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
312 static match_table_t tokens = {
313 {Opt_degraded, "degraded"},
314 {Opt_subvol, "subvol=%s"},
315 {Opt_subvolid, "subvolid=%s"},
316 {Opt_device, "device=%s"},
317 {Opt_nodatasum, "nodatasum"},
318 {Opt_datasum, "datasum"},
319 {Opt_nodatacow, "nodatacow"},
320 {Opt_datacow, "datacow"},
321 {Opt_nobarrier, "nobarrier"},
322 {Opt_barrier, "barrier"},
323 {Opt_max_inline, "max_inline=%s"},
324 {Opt_alloc_start, "alloc_start=%s"},
325 {Opt_thread_pool, "thread_pool=%d"},
326 {Opt_compress, "compress"},
327 {Opt_compress_type, "compress=%s"},
328 {Opt_compress_force, "compress-force"},
329 {Opt_compress_force_type, "compress-force=%s"},
331 {Opt_ssd_spread, "ssd_spread"},
332 {Opt_nossd, "nossd"},
334 {Opt_noacl, "noacl"},
335 {Opt_notreelog, "notreelog"},
336 {Opt_treelog, "treelog"},
337 {Opt_flushoncommit, "flushoncommit"},
338 {Opt_noflushoncommit, "noflushoncommit"},
339 {Opt_ratio, "metadata_ratio=%d"},
340 {Opt_discard, "discard"},
341 {Opt_nodiscard, "nodiscard"},
342 {Opt_space_cache, "space_cache"},
343 {Opt_clear_cache, "clear_cache"},
344 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
345 {Opt_enospc_debug, "enospc_debug"},
346 {Opt_noenospc_debug, "noenospc_debug"},
347 {Opt_subvolrootid, "subvolrootid=%d"},
348 {Opt_defrag, "autodefrag"},
349 {Opt_nodefrag, "noautodefrag"},
350 {Opt_inode_cache, "inode_cache"},
351 {Opt_noinode_cache, "noinode_cache"},
352 {Opt_no_space_cache, "nospace_cache"},
353 {Opt_recovery, "recovery"},
354 {Opt_skip_balance, "skip_balance"},
355 {Opt_check_integrity, "check_int"},
356 {Opt_check_integrity_including_extent_data, "check_int_data"},
357 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
358 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
359 {Opt_fatal_errors, "fatal_errors=%s"},
360 {Opt_commit_interval, "commit=%d"},
361 #ifdef CONFIG_BTRFS_DEBUG
362 {Opt_fragment_data, "fragment=data"},
363 {Opt_fragment_metadata, "fragment=metadata"},
364 {Opt_fragment_all, "fragment=all"},
370 * Regular mount options parser. Everything that is needed only when
371 * reading in a new superblock is parsed here.
372 * XXX JDM: This needs to be cleaned up for remount.
374 int btrfs_parse_options(struct btrfs_root *root, char *options)
376 struct btrfs_fs_info *info = root->fs_info;
377 substring_t args[MAX_OPT_ARGS];
378 char *p, *num, *orig = NULL;
383 bool compress_force = false;
385 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
387 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
393 * strsep changes the string, duplicate it because parse_options
396 options = kstrdup(options, GFP_NOFS);
402 while ((p = strsep(&options, ",")) != NULL) {
407 token = match_token(p, tokens, args);
410 btrfs_info(root->fs_info, "allowing degraded mounts");
411 btrfs_set_opt(info->mount_opt, DEGRADED);
415 case Opt_subvolrootid:
418 * These are parsed by btrfs_parse_early_options
419 * and can be happily ignored here.
423 btrfs_set_and_info(root, NODATASUM,
424 "setting nodatasum");
427 if (btrfs_test_opt(root, NODATASUM)) {
428 if (btrfs_test_opt(root, NODATACOW))
429 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
431 btrfs_info(root->fs_info, "setting datasum");
433 btrfs_clear_opt(info->mount_opt, NODATACOW);
434 btrfs_clear_opt(info->mount_opt, NODATASUM);
437 if (!btrfs_test_opt(root, NODATACOW)) {
438 if (!btrfs_test_opt(root, COMPRESS) ||
439 !btrfs_test_opt(root, FORCE_COMPRESS)) {
440 btrfs_info(root->fs_info,
441 "setting nodatacow, compression disabled");
443 btrfs_info(root->fs_info, "setting nodatacow");
446 btrfs_clear_opt(info->mount_opt, COMPRESS);
447 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
448 btrfs_set_opt(info->mount_opt, NODATACOW);
449 btrfs_set_opt(info->mount_opt, NODATASUM);
452 btrfs_clear_and_info(root, NODATACOW,
455 case Opt_compress_force:
456 case Opt_compress_force_type:
457 compress_force = true;
460 case Opt_compress_type:
461 if (token == Opt_compress ||
462 token == Opt_compress_force ||
463 strcmp(args[0].from, "zlib") == 0) {
464 compress_type = "zlib";
465 info->compress_type = BTRFS_COMPRESS_ZLIB;
466 btrfs_set_opt(info->mount_opt, COMPRESS);
467 btrfs_clear_opt(info->mount_opt, NODATACOW);
468 btrfs_clear_opt(info->mount_opt, NODATASUM);
469 } else if (strcmp(args[0].from, "lzo") == 0) {
470 compress_type = "lzo";
471 info->compress_type = BTRFS_COMPRESS_LZO;
472 btrfs_set_opt(info->mount_opt, COMPRESS);
473 btrfs_clear_opt(info->mount_opt, NODATACOW);
474 btrfs_clear_opt(info->mount_opt, NODATASUM);
475 btrfs_set_fs_incompat(info, COMPRESS_LZO);
476 } else if (strncmp(args[0].from, "no", 2) == 0) {
477 compress_type = "no";
478 btrfs_clear_opt(info->mount_opt, COMPRESS);
479 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
480 compress_force = false;
486 if (compress_force) {
487 btrfs_set_and_info(root, FORCE_COMPRESS,
488 "force %s compression",
491 if (!btrfs_test_opt(root, COMPRESS))
492 btrfs_info(root->fs_info,
493 "btrfs: use %s compression",
496 * If we remount from compress-force=xxx to
497 * compress=xxx, we need clear FORCE_COMPRESS
498 * flag, otherwise, there is no way for users
499 * to disable forcible compression separately.
501 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
505 btrfs_set_and_info(root, SSD,
506 "use ssd allocation scheme");
509 btrfs_set_and_info(root, SSD_SPREAD,
510 "use spread ssd allocation scheme");
511 btrfs_set_opt(info->mount_opt, SSD);
514 btrfs_set_and_info(root, NOSSD,
515 "not using ssd allocation scheme");
516 btrfs_clear_opt(info->mount_opt, SSD);
519 btrfs_clear_and_info(root, NOBARRIER,
520 "turning on barriers");
523 btrfs_set_and_info(root, NOBARRIER,
524 "turning off barriers");
526 case Opt_thread_pool:
527 ret = match_int(&args[0], &intarg);
530 } else if (intarg > 0) {
531 info->thread_pool_size = intarg;
538 num = match_strdup(&args[0]);
540 info->max_inline = memparse(num, NULL);
543 if (info->max_inline) {
544 info->max_inline = min_t(u64,
548 btrfs_info(root->fs_info, "max_inline at %llu",
555 case Opt_alloc_start:
556 num = match_strdup(&args[0]);
558 mutex_lock(&info->chunk_mutex);
559 info->alloc_start = memparse(num, NULL);
560 mutex_unlock(&info->chunk_mutex);
562 btrfs_info(root->fs_info, "allocations start at %llu",
570 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
571 root->fs_info->sb->s_flags |= MS_POSIXACL;
574 btrfs_err(root->fs_info,
575 "support for ACL not compiled in!");
580 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
583 btrfs_set_and_info(root, NOTREELOG,
584 "disabling tree log");
587 btrfs_clear_and_info(root, NOTREELOG,
588 "enabling tree log");
590 case Opt_flushoncommit:
591 btrfs_set_and_info(root, FLUSHONCOMMIT,
592 "turning on flush-on-commit");
594 case Opt_noflushoncommit:
595 btrfs_clear_and_info(root, FLUSHONCOMMIT,
596 "turning off flush-on-commit");
599 ret = match_int(&args[0], &intarg);
602 } else if (intarg >= 0) {
603 info->metadata_ratio = intarg;
604 btrfs_info(root->fs_info, "metadata ratio %d",
605 info->metadata_ratio);
612 btrfs_set_and_info(root, DISCARD,
613 "turning on discard");
616 btrfs_clear_and_info(root, DISCARD,
617 "turning off discard");
619 case Opt_space_cache:
620 btrfs_set_and_info(root, SPACE_CACHE,
621 "enabling disk space caching");
623 case Opt_rescan_uuid_tree:
624 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
626 case Opt_no_space_cache:
627 btrfs_clear_and_info(root, SPACE_CACHE,
628 "disabling disk space caching");
630 case Opt_inode_cache:
631 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
632 "enabling inode map caching");
634 case Opt_noinode_cache:
635 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
636 "disabling inode map caching");
638 case Opt_clear_cache:
639 btrfs_set_and_info(root, CLEAR_CACHE,
640 "force clearing of disk cache");
642 case Opt_user_subvol_rm_allowed:
643 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
645 case Opt_enospc_debug:
646 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
648 case Opt_noenospc_debug:
649 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
652 btrfs_set_and_info(root, AUTO_DEFRAG,
653 "enabling auto defrag");
656 btrfs_clear_and_info(root, AUTO_DEFRAG,
657 "disabling auto defrag");
660 btrfs_info(root->fs_info, "enabling auto recovery");
661 btrfs_set_opt(info->mount_opt, RECOVERY);
663 case Opt_skip_balance:
664 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
666 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
667 case Opt_check_integrity_including_extent_data:
668 btrfs_info(root->fs_info,
669 "enabling check integrity including extent data");
670 btrfs_set_opt(info->mount_opt,
671 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
672 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
674 case Opt_check_integrity:
675 btrfs_info(root->fs_info, "enabling check integrity");
676 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
678 case Opt_check_integrity_print_mask:
679 ret = match_int(&args[0], &intarg);
682 } else if (intarg >= 0) {
683 info->check_integrity_print_mask = intarg;
684 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
685 info->check_integrity_print_mask);
692 case Opt_check_integrity_including_extent_data:
693 case Opt_check_integrity:
694 case Opt_check_integrity_print_mask:
695 btrfs_err(root->fs_info,
696 "support for check_integrity* not compiled in!");
700 case Opt_fatal_errors:
701 if (strcmp(args[0].from, "panic") == 0)
702 btrfs_set_opt(info->mount_opt,
703 PANIC_ON_FATAL_ERROR);
704 else if (strcmp(args[0].from, "bug") == 0)
705 btrfs_clear_opt(info->mount_opt,
706 PANIC_ON_FATAL_ERROR);
712 case Opt_commit_interval:
714 ret = match_int(&args[0], &intarg);
716 btrfs_err(root->fs_info, "invalid commit interval");
722 btrfs_warn(root->fs_info, "excessive commit interval %d",
725 info->commit_interval = intarg;
727 btrfs_info(root->fs_info, "using default commit interval %ds",
728 BTRFS_DEFAULT_COMMIT_INTERVAL);
729 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
732 #ifdef CONFIG_BTRFS_DEBUG
733 case Opt_fragment_all:
734 btrfs_info(root->fs_info, "fragmenting all space");
735 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
736 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
738 case Opt_fragment_metadata:
739 btrfs_info(root->fs_info, "fragmenting metadata");
740 btrfs_set_opt(info->mount_opt,
743 case Opt_fragment_data:
744 btrfs_info(root->fs_info, "fragmenting data");
745 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
749 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
757 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
758 btrfs_info(root->fs_info, "disk space caching is enabled");
764 * Parse mount options that are required early in the mount process.
766 * All other options will be parsed on much later in the mount process and
767 * only when we need to allocate a new super block.
769 static int btrfs_parse_early_options(const char *options, fmode_t flags,
770 void *holder, char **subvol_name, u64 *subvol_objectid,
771 struct btrfs_fs_devices **fs_devices)
773 substring_t args[MAX_OPT_ARGS];
774 char *device_name, *opts, *orig, *p;
782 * strsep changes the string, duplicate it because parse_options
785 opts = kstrdup(options, GFP_KERNEL);
790 while ((p = strsep(&opts, ",")) != NULL) {
795 token = match_token(p, tokens, args);
799 *subvol_name = match_strdup(&args[0]);
806 num = match_strdup(&args[0]);
808 *subvol_objectid = memparse(num, NULL);
810 /* we want the original fs_tree */
811 if (!*subvol_objectid)
813 BTRFS_FS_TREE_OBJECTID;
819 case Opt_subvolrootid:
821 "BTRFS: 'subvolrootid' mount option is deprecated and has "
825 device_name = match_strdup(&args[0]);
830 error = btrfs_scan_one_device(device_name,
831 flags, holder, fs_devices);
846 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
849 struct btrfs_root *root = fs_info->tree_root;
850 struct btrfs_root *fs_root;
851 struct btrfs_root_ref *root_ref;
852 struct btrfs_inode_ref *inode_ref;
853 struct btrfs_key key;
854 struct btrfs_path *path = NULL;
855 char *name = NULL, *ptr;
860 path = btrfs_alloc_path();
865 path->leave_spinning = 1;
867 name = kmalloc(PATH_MAX, GFP_NOFS);
872 ptr = name + PATH_MAX - 1;
876 * Walk up the subvolume trees in the tree of tree roots by root
877 * backrefs until we hit the top-level subvolume.
879 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
880 key.objectid = subvol_objectid;
881 key.type = BTRFS_ROOT_BACKREF_KEY;
882 key.offset = (u64)-1;
884 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
887 } else if (ret > 0) {
888 ret = btrfs_previous_item(root, path, subvol_objectid,
889 BTRFS_ROOT_BACKREF_KEY);
892 } else if (ret > 0) {
898 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
899 subvol_objectid = key.offset;
901 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
902 struct btrfs_root_ref);
903 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
909 read_extent_buffer(path->nodes[0], ptr + 1,
910 (unsigned long)(root_ref + 1), len);
912 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
913 btrfs_release_path(path);
915 key.objectid = subvol_objectid;
916 key.type = BTRFS_ROOT_ITEM_KEY;
917 key.offset = (u64)-1;
918 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
919 if (IS_ERR(fs_root)) {
920 ret = PTR_ERR(fs_root);
925 * Walk up the filesystem tree by inode refs until we hit the
928 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
929 key.objectid = dirid;
930 key.type = BTRFS_INODE_REF_KEY;
931 key.offset = (u64)-1;
933 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
936 } else if (ret > 0) {
937 ret = btrfs_previous_item(fs_root, path, dirid,
938 BTRFS_INODE_REF_KEY);
941 } else if (ret > 0) {
947 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
950 inode_ref = btrfs_item_ptr(path->nodes[0],
952 struct btrfs_inode_ref);
953 len = btrfs_inode_ref_name_len(path->nodes[0],
960 read_extent_buffer(path->nodes[0], ptr + 1,
961 (unsigned long)(inode_ref + 1), len);
963 btrfs_release_path(path);
967 btrfs_free_path(path);
968 if (ptr == name + PATH_MAX - 1) {
972 memmove(name, ptr, name + PATH_MAX - ptr);
977 btrfs_free_path(path);
982 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
984 struct btrfs_root *root = fs_info->tree_root;
985 struct btrfs_dir_item *di;
986 struct btrfs_path *path;
987 struct btrfs_key location;
990 path = btrfs_alloc_path();
993 path->leave_spinning = 1;
996 * Find the "default" dir item which points to the root item that we
997 * will mount by default if we haven't been given a specific subvolume
1000 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1001 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1003 btrfs_free_path(path);
1008 * Ok the default dir item isn't there. This is weird since
1009 * it's always been there, but don't freak out, just try and
1010 * mount the top-level subvolume.
1012 btrfs_free_path(path);
1013 *objectid = BTRFS_FS_TREE_OBJECTID;
1017 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1018 btrfs_free_path(path);
1019 *objectid = location.objectid;
1023 static int btrfs_fill_super(struct super_block *sb,
1024 struct btrfs_fs_devices *fs_devices,
1025 void *data, int silent)
1027 struct inode *inode;
1028 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1029 struct btrfs_key key;
1032 sb->s_maxbytes = MAX_LFS_FILESIZE;
1033 sb->s_magic = BTRFS_SUPER_MAGIC;
1034 sb->s_op = &btrfs_super_ops;
1035 sb->s_d_op = &btrfs_dentry_operations;
1036 sb->s_export_op = &btrfs_export_ops;
1037 sb->s_xattr = btrfs_xattr_handlers;
1038 sb->s_time_gran = 1;
1039 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1040 sb->s_flags |= MS_POSIXACL;
1042 sb->s_flags |= MS_I_VERSION;
1043 sb->s_iflags |= SB_I_CGROUPWB;
1044 err = open_ctree(sb, fs_devices, (char *)data);
1046 printk(KERN_ERR "BTRFS: open_ctree failed\n");
1050 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1051 key.type = BTRFS_INODE_ITEM_KEY;
1053 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1054 if (IS_ERR(inode)) {
1055 err = PTR_ERR(inode);
1059 sb->s_root = d_make_root(inode);
1065 save_mount_options(sb, data);
1066 cleancache_init_fs(sb);
1067 sb->s_flags |= MS_ACTIVE;
1071 close_ctree(fs_info->tree_root);
1075 int btrfs_sync_fs(struct super_block *sb, int wait)
1077 struct btrfs_trans_handle *trans;
1078 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1079 struct btrfs_root *root = fs_info->tree_root;
1081 trace_btrfs_sync_fs(wait);
1084 filemap_flush(fs_info->btree_inode->i_mapping);
1088 btrfs_wait_ordered_roots(fs_info, -1);
1090 trans = btrfs_attach_transaction_barrier(root);
1091 if (IS_ERR(trans)) {
1092 /* no transaction, don't bother */
1093 if (PTR_ERR(trans) == -ENOENT) {
1095 * Exit unless we have some pending changes
1096 * that need to go through commit
1098 if (fs_info->pending_changes == 0)
1101 * A non-blocking test if the fs is frozen. We must not
1102 * start a new transaction here otherwise a deadlock
1103 * happens. The pending operations are delayed to the
1104 * next commit after thawing.
1106 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1107 __sb_end_write(sb, SB_FREEZE_WRITE);
1110 trans = btrfs_start_transaction(root, 0);
1113 return PTR_ERR(trans);
1115 return btrfs_commit_transaction(trans, root);
1118 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1120 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1121 struct btrfs_root *root = info->tree_root;
1122 char *compress_type;
1124 if (btrfs_test_opt(root, DEGRADED))
1125 seq_puts(seq, ",degraded");
1126 if (btrfs_test_opt(root, NODATASUM))
1127 seq_puts(seq, ",nodatasum");
1128 if (btrfs_test_opt(root, NODATACOW))
1129 seq_puts(seq, ",nodatacow");
1130 if (btrfs_test_opt(root, NOBARRIER))
1131 seq_puts(seq, ",nobarrier");
1132 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1133 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1134 if (info->alloc_start != 0)
1135 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1136 if (info->thread_pool_size != min_t(unsigned long,
1137 num_online_cpus() + 2, 8))
1138 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1139 if (btrfs_test_opt(root, COMPRESS)) {
1140 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1141 compress_type = "zlib";
1143 compress_type = "lzo";
1144 if (btrfs_test_opt(root, FORCE_COMPRESS))
1145 seq_printf(seq, ",compress-force=%s", compress_type);
1147 seq_printf(seq, ",compress=%s", compress_type);
1149 if (btrfs_test_opt(root, NOSSD))
1150 seq_puts(seq, ",nossd");
1151 if (btrfs_test_opt(root, SSD_SPREAD))
1152 seq_puts(seq, ",ssd_spread");
1153 else if (btrfs_test_opt(root, SSD))
1154 seq_puts(seq, ",ssd");
1155 if (btrfs_test_opt(root, NOTREELOG))
1156 seq_puts(seq, ",notreelog");
1157 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1158 seq_puts(seq, ",flushoncommit");
1159 if (btrfs_test_opt(root, DISCARD))
1160 seq_puts(seq, ",discard");
1161 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1162 seq_puts(seq, ",noacl");
1163 if (btrfs_test_opt(root, SPACE_CACHE))
1164 seq_puts(seq, ",space_cache");
1166 seq_puts(seq, ",nospace_cache");
1167 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1168 seq_puts(seq, ",rescan_uuid_tree");
1169 if (btrfs_test_opt(root, CLEAR_CACHE))
1170 seq_puts(seq, ",clear_cache");
1171 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1172 seq_puts(seq, ",user_subvol_rm_allowed");
1173 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1174 seq_puts(seq, ",enospc_debug");
1175 if (btrfs_test_opt(root, AUTO_DEFRAG))
1176 seq_puts(seq, ",autodefrag");
1177 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1178 seq_puts(seq, ",inode_cache");
1179 if (btrfs_test_opt(root, SKIP_BALANCE))
1180 seq_puts(seq, ",skip_balance");
1181 if (btrfs_test_opt(root, RECOVERY))
1182 seq_puts(seq, ",recovery");
1183 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1184 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1185 seq_puts(seq, ",check_int_data");
1186 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1187 seq_puts(seq, ",check_int");
1188 if (info->check_integrity_print_mask)
1189 seq_printf(seq, ",check_int_print_mask=%d",
1190 info->check_integrity_print_mask);
1192 if (info->metadata_ratio)
1193 seq_printf(seq, ",metadata_ratio=%d",
1194 info->metadata_ratio);
1195 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1196 seq_puts(seq, ",fatal_errors=panic");
1197 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1198 seq_printf(seq, ",commit=%d", info->commit_interval);
1199 #ifdef CONFIG_BTRFS_DEBUG
1200 if (btrfs_test_opt(root, FRAGMENT_DATA))
1201 seq_puts(seq, ",fragment=data");
1202 if (btrfs_test_opt(root, FRAGMENT_METADATA))
1203 seq_puts(seq, ",fragment=metadata");
1205 seq_printf(seq, ",subvolid=%llu",
1206 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1207 seq_puts(seq, ",subvol=");
1208 seq_dentry(seq, dentry, " \t\n\\");
1212 static int btrfs_test_super(struct super_block *s, void *data)
1214 struct btrfs_fs_info *p = data;
1215 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1217 return fs_info->fs_devices == p->fs_devices;
1220 static int btrfs_set_super(struct super_block *s, void *data)
1222 int err = set_anon_super(s, data);
1224 s->s_fs_info = data;
1229 * subvolumes are identified by ino 256
1231 static inline int is_subvolume_inode(struct inode *inode)
1233 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1239 * This will add subvolid=0 to the argument string while removing any subvol=
1240 * and subvolid= arguments to make sure we get the top-level root for path
1241 * walking to the subvol we want.
1243 static char *setup_root_args(char *args)
1245 char *buf, *dst, *sep;
1248 return kstrdup("subvolid=0", GFP_NOFS);
1250 /* The worst case is that we add ",subvolid=0" to the end. */
1251 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1256 sep = strchrnul(args, ',');
1257 if (!strstarts(args, "subvol=") &&
1258 !strstarts(args, "subvolid=")) {
1259 memcpy(dst, args, sep - args);
1268 strcpy(dst, "subvolid=0");
1273 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1274 int flags, const char *device_name,
1277 struct dentry *root;
1278 struct vfsmount *mnt = NULL;
1282 newargs = setup_root_args(data);
1284 root = ERR_PTR(-ENOMEM);
1288 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1289 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1290 if (flags & MS_RDONLY) {
1291 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1292 device_name, newargs);
1294 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1295 device_name, newargs);
1297 root = ERR_CAST(mnt);
1302 down_write(&mnt->mnt_sb->s_umount);
1303 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1304 up_write(&mnt->mnt_sb->s_umount);
1306 root = ERR_PTR(ret);
1312 root = ERR_CAST(mnt);
1318 if (!subvol_objectid) {
1319 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1322 root = ERR_PTR(ret);
1326 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1328 if (IS_ERR(subvol_name)) {
1329 root = ERR_CAST(subvol_name);
1336 root = mount_subtree(mnt, subvol_name);
1337 /* mount_subtree() drops our reference on the vfsmount. */
1340 if (!IS_ERR(root)) {
1341 struct super_block *s = root->d_sb;
1342 struct inode *root_inode = d_inode(root);
1343 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1346 if (!is_subvolume_inode(root_inode)) {
1347 pr_err("BTRFS: '%s' is not a valid subvolume\n",
1351 if (subvol_objectid && root_objectid != subvol_objectid) {
1353 * This will also catch a race condition where a
1354 * subvolume which was passed by ID is renamed and
1355 * another subvolume is renamed over the old location.
1357 pr_err("BTRFS: subvol '%s' does not match subvolid %llu\n",
1358 subvol_name, subvol_objectid);
1363 root = ERR_PTR(ret);
1364 deactivate_locked_super(s);
1375 static int parse_security_options(char *orig_opts,
1376 struct security_mnt_opts *sec_opts)
1378 char *secdata = NULL;
1381 secdata = alloc_secdata();
1384 ret = security_sb_copy_data(orig_opts, secdata);
1386 free_secdata(secdata);
1389 ret = security_sb_parse_opts_str(secdata, sec_opts);
1390 free_secdata(secdata);
1394 static int setup_security_options(struct btrfs_fs_info *fs_info,
1395 struct super_block *sb,
1396 struct security_mnt_opts *sec_opts)
1401 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1404 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1408 #ifdef CONFIG_SECURITY
1409 if (!fs_info->security_opts.num_mnt_opts) {
1410 /* first time security setup, copy sec_opts to fs_info */
1411 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1414 * Since SELinux(the only one supports security_mnt_opts) does
1415 * NOT support changing context during remount/mount same sb,
1416 * This must be the same or part of the same security options,
1419 security_free_mnt_opts(sec_opts);
1426 * Find a superblock for the given device / mount point.
1428 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1429 * for multiple device setup. Make sure to keep it in sync.
1431 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1432 const char *device_name, void *data)
1434 struct block_device *bdev = NULL;
1435 struct super_block *s;
1436 struct btrfs_fs_devices *fs_devices = NULL;
1437 struct btrfs_fs_info *fs_info = NULL;
1438 struct security_mnt_opts new_sec_opts;
1439 fmode_t mode = FMODE_READ;
1440 char *subvol_name = NULL;
1441 u64 subvol_objectid = 0;
1444 if (!(flags & MS_RDONLY))
1445 mode |= FMODE_WRITE;
1447 error = btrfs_parse_early_options(data, mode, fs_type,
1448 &subvol_name, &subvol_objectid,
1452 return ERR_PTR(error);
1455 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1456 /* mount_subvol() will free subvol_name. */
1457 return mount_subvol(subvol_name, subvol_objectid, flags,
1461 security_init_mnt_opts(&new_sec_opts);
1463 error = parse_security_options(data, &new_sec_opts);
1465 return ERR_PTR(error);
1468 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1470 goto error_sec_opts;
1473 * Setup a dummy root and fs_info for test/set super. This is because
1474 * we don't actually fill this stuff out until open_ctree, but we need
1475 * it for searching for existing supers, so this lets us do that and
1476 * then open_ctree will properly initialize everything later.
1478 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1481 goto error_sec_opts;
1484 fs_info->fs_devices = fs_devices;
1486 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1487 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1488 security_init_mnt_opts(&fs_info->security_opts);
1489 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1494 error = btrfs_open_devices(fs_devices, mode, fs_type);
1498 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1500 goto error_close_devices;
1503 bdev = fs_devices->latest_bdev;
1504 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1508 goto error_close_devices;
1512 btrfs_close_devices(fs_devices);
1513 free_fs_info(fs_info);
1514 if ((flags ^ s->s_flags) & MS_RDONLY)
1517 char b[BDEVNAME_SIZE];
1519 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1520 btrfs_sb(s)->bdev_holder = fs_type;
1521 error = btrfs_fill_super(s, fs_devices, data,
1522 flags & MS_SILENT ? 1 : 0);
1525 deactivate_locked_super(s);
1526 goto error_sec_opts;
1529 fs_info = btrfs_sb(s);
1530 error = setup_security_options(fs_info, s, &new_sec_opts);
1532 deactivate_locked_super(s);
1533 goto error_sec_opts;
1536 return dget(s->s_root);
1538 error_close_devices:
1539 btrfs_close_devices(fs_devices);
1541 free_fs_info(fs_info);
1543 security_free_mnt_opts(&new_sec_opts);
1544 return ERR_PTR(error);
1547 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1548 int new_pool_size, int old_pool_size)
1550 if (new_pool_size == old_pool_size)
1553 fs_info->thread_pool_size = new_pool_size;
1555 btrfs_info(fs_info, "resize thread pool %d -> %d",
1556 old_pool_size, new_pool_size);
1558 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1559 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1560 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1561 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1562 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1563 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1564 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1566 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1567 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1568 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1569 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1570 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1574 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1576 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1579 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1580 unsigned long old_opts, int flags)
1582 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1583 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1584 (flags & MS_RDONLY))) {
1585 /* wait for any defraggers to finish */
1586 wait_event(fs_info->transaction_wait,
1587 (atomic_read(&fs_info->defrag_running) == 0));
1588 if (flags & MS_RDONLY)
1589 sync_filesystem(fs_info->sb);
1593 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1594 unsigned long old_opts)
1597 * We need cleanup all defragable inodes if the autodefragment is
1598 * close or the fs is R/O.
1600 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1601 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1602 (fs_info->sb->s_flags & MS_RDONLY))) {
1603 btrfs_cleanup_defrag_inodes(fs_info);
1606 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1609 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1611 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1612 struct btrfs_root *root = fs_info->tree_root;
1613 unsigned old_flags = sb->s_flags;
1614 unsigned long old_opts = fs_info->mount_opt;
1615 unsigned long old_compress_type = fs_info->compress_type;
1616 u64 old_max_inline = fs_info->max_inline;
1617 u64 old_alloc_start = fs_info->alloc_start;
1618 int old_thread_pool_size = fs_info->thread_pool_size;
1619 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1622 sync_filesystem(sb);
1623 btrfs_remount_prepare(fs_info);
1626 struct security_mnt_opts new_sec_opts;
1628 security_init_mnt_opts(&new_sec_opts);
1629 ret = parse_security_options(data, &new_sec_opts);
1632 ret = setup_security_options(fs_info, sb,
1635 security_free_mnt_opts(&new_sec_opts);
1640 ret = btrfs_parse_options(root, data);
1646 btrfs_remount_begin(fs_info, old_opts, *flags);
1647 btrfs_resize_thread_pool(fs_info,
1648 fs_info->thread_pool_size, old_thread_pool_size);
1650 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1653 if (*flags & MS_RDONLY) {
1655 * this also happens on 'umount -rf' or on shutdown, when
1656 * the filesystem is busy.
1658 cancel_work_sync(&fs_info->async_reclaim_work);
1660 /* wait for the uuid_scan task to finish */
1661 down(&fs_info->uuid_tree_rescan_sem);
1662 /* avoid complains from lockdep et al. */
1663 up(&fs_info->uuid_tree_rescan_sem);
1665 sb->s_flags |= MS_RDONLY;
1668 * Setting MS_RDONLY will put the cleaner thread to
1669 * sleep at the next loop if it's already active.
1670 * If it's already asleep, we'll leave unused block
1671 * groups on disk until we're mounted read-write again
1672 * unless we clean them up here.
1674 btrfs_delete_unused_bgs(fs_info);
1676 btrfs_dev_replace_suspend_for_unmount(fs_info);
1677 btrfs_scrub_cancel(fs_info);
1678 btrfs_pause_balance(fs_info);
1680 ret = btrfs_commit_super(root);
1684 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1686 "Remounting read-write after error is not allowed");
1690 if (fs_info->fs_devices->rw_devices == 0) {
1695 if (fs_info->fs_devices->missing_devices >
1696 fs_info->num_tolerated_disk_barrier_failures &&
1697 !(*flags & MS_RDONLY)) {
1699 "too many missing devices, writeable remount is not allowed");
1704 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1709 ret = btrfs_cleanup_fs_roots(fs_info);
1713 /* recover relocation */
1714 mutex_lock(&fs_info->cleaner_mutex);
1715 ret = btrfs_recover_relocation(root);
1716 mutex_unlock(&fs_info->cleaner_mutex);
1720 ret = btrfs_resume_balance_async(fs_info);
1724 ret = btrfs_resume_dev_replace_async(fs_info);
1726 btrfs_warn(fs_info, "failed to resume dev_replace");
1730 if (!fs_info->uuid_root) {
1731 btrfs_info(fs_info, "creating UUID tree");
1732 ret = btrfs_create_uuid_tree(fs_info);
1734 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1738 sb->s_flags &= ~MS_RDONLY;
1741 wake_up_process(fs_info->transaction_kthread);
1742 btrfs_remount_cleanup(fs_info, old_opts);
1746 /* We've hit an error - don't reset MS_RDONLY */
1747 if (sb->s_flags & MS_RDONLY)
1748 old_flags |= MS_RDONLY;
1749 sb->s_flags = old_flags;
1750 fs_info->mount_opt = old_opts;
1751 fs_info->compress_type = old_compress_type;
1752 fs_info->max_inline = old_max_inline;
1753 mutex_lock(&fs_info->chunk_mutex);
1754 fs_info->alloc_start = old_alloc_start;
1755 mutex_unlock(&fs_info->chunk_mutex);
1756 btrfs_resize_thread_pool(fs_info,
1757 old_thread_pool_size, fs_info->thread_pool_size);
1758 fs_info->metadata_ratio = old_metadata_ratio;
1759 btrfs_remount_cleanup(fs_info, old_opts);
1763 /* Used to sort the devices by max_avail(descending sort) */
1764 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1765 const void *dev_info2)
1767 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1768 ((struct btrfs_device_info *)dev_info2)->max_avail)
1770 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1771 ((struct btrfs_device_info *)dev_info2)->max_avail)
1778 * sort the devices by max_avail, in which max free extent size of each device
1779 * is stored.(Descending Sort)
1781 static inline void btrfs_descending_sort_devices(
1782 struct btrfs_device_info *devices,
1785 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1786 btrfs_cmp_device_free_bytes, NULL);
1790 * The helper to calc the free space on the devices that can be used to store
1793 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1795 struct btrfs_fs_info *fs_info = root->fs_info;
1796 struct btrfs_device_info *devices_info;
1797 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1798 struct btrfs_device *device;
1803 u64 min_stripe_size;
1804 int min_stripes = 1, num_stripes = 1;
1805 int i = 0, nr_devices;
1809 * We aren't under the device list lock, so this is racey-ish, but good
1810 * enough for our purposes.
1812 nr_devices = fs_info->fs_devices->open_devices;
1815 nr_devices = fs_info->fs_devices->open_devices;
1823 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1828 /* calc min stripe number for data space alloction */
1829 type = btrfs_get_alloc_profile(root, 1);
1830 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1832 num_stripes = nr_devices;
1833 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1836 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1841 if (type & BTRFS_BLOCK_GROUP_DUP)
1842 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1844 min_stripe_size = BTRFS_STRIPE_LEN;
1846 if (fs_info->alloc_start)
1847 mutex_lock(&fs_devices->device_list_mutex);
1849 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1850 if (!device->in_fs_metadata || !device->bdev ||
1851 device->is_tgtdev_for_dev_replace)
1854 if (i >= nr_devices)
1857 avail_space = device->total_bytes - device->bytes_used;
1859 /* align with stripe_len */
1860 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1861 avail_space *= BTRFS_STRIPE_LEN;
1864 * In order to avoid overwritting the superblock on the drive,
1865 * btrfs starts at an offset of at least 1MB when doing chunk
1868 skip_space = 1024 * 1024;
1870 /* user can set the offset in fs_info->alloc_start. */
1871 if (fs_info->alloc_start &&
1872 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1873 device->total_bytes) {
1875 skip_space = max(fs_info->alloc_start, skip_space);
1878 * btrfs can not use the free space in
1879 * [0, skip_space - 1], we must subtract it from the
1880 * total. In order to implement it, we account the used
1881 * space in this range first.
1883 ret = btrfs_account_dev_extents_size(device, 0,
1887 kfree(devices_info);
1888 mutex_unlock(&fs_devices->device_list_mutex);
1894 /* calc the free space in [0, skip_space - 1] */
1895 skip_space -= used_space;
1899 * we can use the free space in [0, skip_space - 1], subtract
1900 * it from the total.
1902 if (avail_space && avail_space >= skip_space)
1903 avail_space -= skip_space;
1907 if (avail_space < min_stripe_size)
1910 devices_info[i].dev = device;
1911 devices_info[i].max_avail = avail_space;
1916 if (fs_info->alloc_start)
1917 mutex_unlock(&fs_devices->device_list_mutex);
1921 btrfs_descending_sort_devices(devices_info, nr_devices);
1925 while (nr_devices >= min_stripes) {
1926 if (num_stripes > nr_devices)
1927 num_stripes = nr_devices;
1929 if (devices_info[i].max_avail >= min_stripe_size) {
1933 avail_space += devices_info[i].max_avail * num_stripes;
1934 alloc_size = devices_info[i].max_avail;
1935 for (j = i + 1 - num_stripes; j <= i; j++)
1936 devices_info[j].max_avail -= alloc_size;
1942 kfree(devices_info);
1943 *free_bytes = avail_space;
1948 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1950 * If there's a redundant raid level at DATA block groups, use the respective
1951 * multiplier to scale the sizes.
1953 * Unused device space usage is based on simulating the chunk allocator
1954 * algorithm that respects the device sizes, order of allocations and the
1955 * 'alloc_start' value, this is a close approximation of the actual use but
1956 * there are other factors that may change the result (like a new metadata
1959 * FIXME: not accurate for mixed block groups, total and free/used are ok,
1960 * available appears slightly larger.
1962 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1964 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1965 struct btrfs_super_block *disk_super = fs_info->super_copy;
1966 struct list_head *head = &fs_info->space_info;
1967 struct btrfs_space_info *found;
1969 u64 total_free_data = 0;
1970 int bits = dentry->d_sb->s_blocksize_bits;
1971 __be32 *fsid = (__be32 *)fs_info->fsid;
1972 unsigned factor = 1;
1973 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1977 * holding chunk_muext to avoid allocating new chunks, holding
1978 * device_list_mutex to avoid the device being removed
1981 list_for_each_entry_rcu(found, head, list) {
1982 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1985 total_free_data += found->disk_total - found->disk_used;
1987 btrfs_account_ro_block_groups_free_space(found);
1989 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1990 if (!list_empty(&found->block_groups[i])) {
1992 case BTRFS_RAID_DUP:
1993 case BTRFS_RAID_RAID1:
1994 case BTRFS_RAID_RAID10:
2001 total_used += found->disk_used;
2006 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2007 buf->f_blocks >>= bits;
2008 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2010 /* Account global block reserve as used, it's in logical size already */
2011 spin_lock(&block_rsv->lock);
2012 buf->f_bfree -= block_rsv->size >> bits;
2013 spin_unlock(&block_rsv->lock);
2015 buf->f_bavail = div_u64(total_free_data, factor);
2016 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2019 buf->f_bavail += div_u64(total_free_data, factor);
2020 buf->f_bavail = buf->f_bavail >> bits;
2022 buf->f_type = BTRFS_SUPER_MAGIC;
2023 buf->f_bsize = dentry->d_sb->s_blocksize;
2024 buf->f_namelen = BTRFS_NAME_LEN;
2026 /* We treat it as constant endianness (it doesn't matter _which_)
2027 because we want the fsid to come out the same whether mounted
2028 on a big-endian or little-endian host */
2029 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2030 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2031 /* Mask in the root object ID too, to disambiguate subvols */
2032 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2033 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2038 static void btrfs_kill_super(struct super_block *sb)
2040 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2041 kill_anon_super(sb);
2042 free_fs_info(fs_info);
2045 static struct file_system_type btrfs_fs_type = {
2046 .owner = THIS_MODULE,
2048 .mount = btrfs_mount,
2049 .kill_sb = btrfs_kill_super,
2050 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2052 MODULE_ALIAS_FS("btrfs");
2054 static int btrfs_control_open(struct inode *inode, struct file *file)
2057 * The control file's private_data is used to hold the
2058 * transaction when it is started and is used to keep
2059 * track of whether a transaction is already in progress.
2061 file->private_data = NULL;
2066 * used by btrfsctl to scan devices when no FS is mounted
2068 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2071 struct btrfs_ioctl_vol_args *vol;
2072 struct btrfs_fs_devices *fs_devices;
2075 if (!capable(CAP_SYS_ADMIN))
2078 vol = memdup_user((void __user *)arg, sizeof(*vol));
2080 return PTR_ERR(vol);
2083 case BTRFS_IOC_SCAN_DEV:
2084 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2085 &btrfs_fs_type, &fs_devices);
2087 case BTRFS_IOC_DEVICES_READY:
2088 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2089 &btrfs_fs_type, &fs_devices);
2092 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2100 static int btrfs_freeze(struct super_block *sb)
2102 struct btrfs_trans_handle *trans;
2103 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2105 trans = btrfs_attach_transaction_barrier(root);
2106 if (IS_ERR(trans)) {
2107 /* no transaction, don't bother */
2108 if (PTR_ERR(trans) == -ENOENT)
2110 return PTR_ERR(trans);
2112 return btrfs_commit_transaction(trans, root);
2115 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2117 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2118 struct btrfs_fs_devices *cur_devices;
2119 struct btrfs_device *dev, *first_dev = NULL;
2120 struct list_head *head;
2121 struct rcu_string *name;
2123 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2124 cur_devices = fs_info->fs_devices;
2125 while (cur_devices) {
2126 head = &cur_devices->devices;
2127 list_for_each_entry(dev, head, dev_list) {
2132 if (!first_dev || dev->devid < first_dev->devid)
2135 cur_devices = cur_devices->seed;
2140 name = rcu_dereference(first_dev->name);
2141 seq_escape(m, name->str, " \t\n\\");
2146 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2150 static const struct super_operations btrfs_super_ops = {
2151 .drop_inode = btrfs_drop_inode,
2152 .evict_inode = btrfs_evict_inode,
2153 .put_super = btrfs_put_super,
2154 .sync_fs = btrfs_sync_fs,
2155 .show_options = btrfs_show_options,
2156 .show_devname = btrfs_show_devname,
2157 .write_inode = btrfs_write_inode,
2158 .alloc_inode = btrfs_alloc_inode,
2159 .destroy_inode = btrfs_destroy_inode,
2160 .statfs = btrfs_statfs,
2161 .remount_fs = btrfs_remount,
2162 .freeze_fs = btrfs_freeze,
2165 static const struct file_operations btrfs_ctl_fops = {
2166 .open = btrfs_control_open,
2167 .unlocked_ioctl = btrfs_control_ioctl,
2168 .compat_ioctl = btrfs_control_ioctl,
2169 .owner = THIS_MODULE,
2170 .llseek = noop_llseek,
2173 static struct miscdevice btrfs_misc = {
2174 .minor = BTRFS_MINOR,
2175 .name = "btrfs-control",
2176 .fops = &btrfs_ctl_fops
2179 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2180 MODULE_ALIAS("devname:btrfs-control");
2182 static int btrfs_interface_init(void)
2184 return misc_register(&btrfs_misc);
2187 static void btrfs_interface_exit(void)
2189 misc_deregister(&btrfs_misc);
2192 static void btrfs_print_info(void)
2194 printk(KERN_INFO "Btrfs loaded"
2195 #ifdef CONFIG_BTRFS_DEBUG
2198 #ifdef CONFIG_BTRFS_ASSERT
2201 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2202 ", integrity-checker=on"
2207 static int btrfs_run_sanity_tests(void)
2211 ret = btrfs_init_test_fs();
2215 ret = btrfs_test_free_space_cache();
2218 ret = btrfs_test_extent_buffer_operations();
2221 ret = btrfs_test_extent_io();
2224 ret = btrfs_test_inodes();
2227 ret = btrfs_test_qgroups();
2229 btrfs_destroy_test_fs();
2233 static int __init init_btrfs_fs(void)
2237 err = btrfs_hash_init();
2243 err = btrfs_init_sysfs();
2247 btrfs_init_compress();
2249 err = btrfs_init_cachep();
2253 err = extent_io_init();
2257 err = extent_map_init();
2259 goto free_extent_io;
2261 err = ordered_data_init();
2263 goto free_extent_map;
2265 err = btrfs_delayed_inode_init();
2267 goto free_ordered_data;
2269 err = btrfs_auto_defrag_init();
2271 goto free_delayed_inode;
2273 err = btrfs_delayed_ref_init();
2275 goto free_auto_defrag;
2277 err = btrfs_prelim_ref_init();
2279 goto free_delayed_ref;
2281 err = btrfs_end_io_wq_init();
2283 goto free_prelim_ref;
2285 err = btrfs_interface_init();
2287 goto free_end_io_wq;
2289 btrfs_init_lockdep();
2293 err = btrfs_run_sanity_tests();
2295 goto unregister_ioctl;
2297 err = register_filesystem(&btrfs_fs_type);
2299 goto unregister_ioctl;
2304 btrfs_interface_exit();
2306 btrfs_end_io_wq_exit();
2308 btrfs_prelim_ref_exit();
2310 btrfs_delayed_ref_exit();
2312 btrfs_auto_defrag_exit();
2314 btrfs_delayed_inode_exit();
2316 ordered_data_exit();
2322 btrfs_destroy_cachep();
2324 btrfs_exit_compress();
2331 static void __exit exit_btrfs_fs(void)
2333 btrfs_destroy_cachep();
2334 btrfs_delayed_ref_exit();
2335 btrfs_auto_defrag_exit();
2336 btrfs_delayed_inode_exit();
2337 btrfs_prelim_ref_exit();
2338 ordered_data_exit();
2341 btrfs_interface_exit();
2342 btrfs_end_io_wq_exit();
2343 unregister_filesystem(&btrfs_fs_type);
2345 btrfs_cleanup_fs_uuids();
2346 btrfs_exit_compress();
2350 late_initcall(init_btrfs_fs);
2351 module_exit(exit_btrfs_fs)
2353 MODULE_LICENSE("GPL");