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 /* btrfs handle error by forcing the filesystem readonly */
101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
103 struct super_block *sb = fs_info->sb;
105 if (sb->s_flags & MS_RDONLY)
108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
109 sb->s_flags |= MS_RDONLY;
110 btrfs_info(fs_info, "forced readonly");
112 * Note that a running device replace operation is not
113 * canceled here although there is no way to update
114 * the progress. It would add the risk of a deadlock,
115 * therefore the canceling is omitted. The only penalty
116 * is that some I/O remains active until the procedure
117 * completes. The next time when the filesystem is
118 * mounted writeable again, the device replace
119 * operation continues.
125 * __btrfs_handle_fs_error decodes expected errors from the caller and
126 * invokes the approciate error response.
129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
130 unsigned int line, int errno, const char *fmt, ...)
132 struct super_block *sb = fs_info->sb;
138 * Special case: if the error is EROFS, and we're already
139 * under MS_RDONLY, then it is safe here.
141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
145 errstr = btrfs_decode_error(errno);
147 struct va_format vaf;
154 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
155 sb->s_id, function, line, errno, errstr, &vaf);
158 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
164 * Today we only save the error info to memory. Long term we'll
165 * also send it down to the disk
167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
175 static const char * const logtypes[] = {
188 * Use one ratelimit state per log level so that a flood of less important
189 * messages doesn't cause more important ones to be dropped.
191 static struct ratelimit_state printk_limits[] = {
192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
204 struct super_block *sb = fs_info->sb;
206 struct va_format vaf;
208 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit;
214 kern_level = printk_get_level(fmt);
216 size_t size = printk_skip_level(fmt) - fmt;
217 memcpy(lvl, fmt, size);
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
224 /* Default to debug output */
225 ratelimit = &printk_limits[7];
231 if (__ratelimit(ratelimit))
232 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
239 * We only mark the transaction aborted and then set the file system read-only.
240 * This will prevent new transactions from starting or trying to join this
243 * This means that error recovery at the call site is limited to freeing
244 * any local memory allocations and passing the error code up without
245 * further cleanup. The transaction should complete as it normally would
246 * in the call path but will return -EIO.
248 * We'll complete the cleanup in btrfs_end_transaction and
249 * btrfs_commit_transaction.
252 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
253 const char *function,
254 unsigned int line, int errno)
256 struct btrfs_fs_info *fs_info = trans->fs_info;
258 trans->aborted = errno;
259 /* Nothing used. The other threads that have joined this
260 * transaction may be able to continue. */
261 if (!trans->dirty && list_empty(&trans->new_bgs)) {
264 errstr = btrfs_decode_error(errno);
266 "%s:%d: Aborting unused transaction(%s).",
267 function, line, errstr);
270 ACCESS_ONCE(trans->transaction->aborted) = errno;
271 /* Wake up anybody who may be waiting on this transaction */
272 wake_up(&fs_info->transaction_wait);
273 wake_up(&fs_info->transaction_blocked_wait);
274 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
277 * __btrfs_panic decodes unexpected, fatal errors from the caller,
278 * issues an alert, and either panics or BUGs, depending on mount options.
281 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
282 unsigned int line, int errno, const char *fmt, ...)
284 char *s_id = "<unknown>";
286 struct va_format vaf = { .fmt = fmt };
290 s_id = fs_info->sb->s_id;
295 errstr = btrfs_decode_error(errno);
296 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
297 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
298 s_id, function, line, &vaf, errno, errstr);
300 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
301 function, line, &vaf, errno, errstr);
303 /* Caller calls BUG() */
306 static void btrfs_put_super(struct super_block *sb)
308 close_ctree(btrfs_sb(sb)->tree_root);
312 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
313 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
314 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
315 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
316 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
317 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
318 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
319 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
320 Opt_skip_balance, Opt_check_integrity,
321 Opt_check_integrity_including_extent_data,
322 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
323 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
324 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
325 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
326 Opt_nologreplay, Opt_norecovery,
327 #ifdef CONFIG_BTRFS_DEBUG
328 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
333 static const match_table_t tokens = {
334 {Opt_degraded, "degraded"},
335 {Opt_subvol, "subvol=%s"},
336 {Opt_subvolid, "subvolid=%s"},
337 {Opt_device, "device=%s"},
338 {Opt_nodatasum, "nodatasum"},
339 {Opt_datasum, "datasum"},
340 {Opt_nodatacow, "nodatacow"},
341 {Opt_datacow, "datacow"},
342 {Opt_nobarrier, "nobarrier"},
343 {Opt_barrier, "barrier"},
344 {Opt_max_inline, "max_inline=%s"},
345 {Opt_alloc_start, "alloc_start=%s"},
346 {Opt_thread_pool, "thread_pool=%d"},
347 {Opt_compress, "compress"},
348 {Opt_compress_type, "compress=%s"},
349 {Opt_compress_force, "compress-force"},
350 {Opt_compress_force_type, "compress-force=%s"},
352 {Opt_ssd_spread, "ssd_spread"},
353 {Opt_nossd, "nossd"},
355 {Opt_noacl, "noacl"},
356 {Opt_notreelog, "notreelog"},
357 {Opt_treelog, "treelog"},
358 {Opt_nologreplay, "nologreplay"},
359 {Opt_norecovery, "norecovery"},
360 {Opt_flushoncommit, "flushoncommit"},
361 {Opt_noflushoncommit, "noflushoncommit"},
362 {Opt_ratio, "metadata_ratio=%d"},
363 {Opt_discard, "discard"},
364 {Opt_nodiscard, "nodiscard"},
365 {Opt_space_cache, "space_cache"},
366 {Opt_space_cache_version, "space_cache=%s"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug, "enospc_debug"},
370 {Opt_noenospc_debug, "noenospc_debug"},
371 {Opt_subvolrootid, "subvolrootid=%d"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_inode_cache, "inode_cache"},
375 {Opt_noinode_cache, "noinode_cache"},
376 {Opt_no_space_cache, "nospace_cache"},
377 {Opt_recovery, "recovery"}, /* deprecated */
378 {Opt_usebackuproot, "usebackuproot"},
379 {Opt_skip_balance, "skip_balance"},
380 {Opt_check_integrity, "check_int"},
381 {Opt_check_integrity_including_extent_data, "check_int_data"},
382 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
383 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
384 {Opt_fatal_errors, "fatal_errors=%s"},
385 {Opt_commit_interval, "commit=%d"},
386 #ifdef CONFIG_BTRFS_DEBUG
387 {Opt_fragment_data, "fragment=data"},
388 {Opt_fragment_metadata, "fragment=metadata"},
389 {Opt_fragment_all, "fragment=all"},
395 * Regular mount options parser. Everything that is needed only when
396 * reading in a new superblock is parsed here.
397 * XXX JDM: This needs to be cleaned up for remount.
399 int btrfs_parse_options(struct btrfs_root *root, char *options,
400 unsigned long new_flags)
402 struct btrfs_fs_info *info = root->fs_info;
403 substring_t args[MAX_OPT_ARGS];
404 char *p, *num, *orig = NULL;
409 bool compress_force = false;
410 enum btrfs_compression_type saved_compress_type;
411 bool saved_compress_force;
414 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
415 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
416 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
418 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
421 * Even the options are empty, we still need to do extra check
428 * strsep changes the string, duplicate it because parse_options
431 options = kstrdup(options, GFP_NOFS);
437 while ((p = strsep(&options, ",")) != NULL) {
442 token = match_token(p, tokens, args);
445 btrfs_info(root->fs_info, "allowing degraded mounts");
446 btrfs_set_opt(info->mount_opt, DEGRADED);
450 case Opt_subvolrootid:
453 * These are parsed by btrfs_parse_early_options
454 * and can be happily ignored here.
458 btrfs_set_and_info(info, NODATASUM,
459 "setting nodatasum");
462 if (btrfs_test_opt(info, NODATASUM)) {
463 if (btrfs_test_opt(info, NODATACOW))
464 btrfs_info(root->fs_info,
465 "setting datasum, datacow enabled");
467 btrfs_info(root->fs_info,
470 btrfs_clear_opt(info->mount_opt, NODATACOW);
471 btrfs_clear_opt(info->mount_opt, NODATASUM);
474 if (!btrfs_test_opt(info, NODATACOW)) {
475 if (!btrfs_test_opt(info, COMPRESS) ||
476 !btrfs_test_opt(info, FORCE_COMPRESS)) {
477 btrfs_info(root->fs_info,
478 "setting nodatacow, compression disabled");
480 btrfs_info(root->fs_info,
481 "setting nodatacow");
484 btrfs_clear_opt(info->mount_opt, COMPRESS);
485 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
486 btrfs_set_opt(info->mount_opt, NODATACOW);
487 btrfs_set_opt(info->mount_opt, NODATASUM);
490 btrfs_clear_and_info(info, NODATACOW,
493 case Opt_compress_force:
494 case Opt_compress_force_type:
495 compress_force = true;
498 case Opt_compress_type:
499 saved_compress_type = btrfs_test_opt(info,
501 info->compress_type : BTRFS_COMPRESS_NONE;
502 saved_compress_force =
503 btrfs_test_opt(info, FORCE_COMPRESS);
504 if (token == Opt_compress ||
505 token == Opt_compress_force ||
506 strcmp(args[0].from, "zlib") == 0) {
507 compress_type = "zlib";
508 info->compress_type = BTRFS_COMPRESS_ZLIB;
509 btrfs_set_opt(info->mount_opt, COMPRESS);
510 btrfs_clear_opt(info->mount_opt, NODATACOW);
511 btrfs_clear_opt(info->mount_opt, NODATASUM);
513 } else if (strcmp(args[0].from, "lzo") == 0) {
514 compress_type = "lzo";
515 info->compress_type = BTRFS_COMPRESS_LZO;
516 btrfs_set_opt(info->mount_opt, COMPRESS);
517 btrfs_clear_opt(info->mount_opt, NODATACOW);
518 btrfs_clear_opt(info->mount_opt, NODATASUM);
519 btrfs_set_fs_incompat(info, COMPRESS_LZO);
521 } else if (strncmp(args[0].from, "no", 2) == 0) {
522 compress_type = "no";
523 btrfs_clear_opt(info->mount_opt, COMPRESS);
524 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
525 compress_force = false;
532 if (compress_force) {
533 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
536 * If we remount from compress-force=xxx to
537 * compress=xxx, we need clear FORCE_COMPRESS
538 * flag, otherwise, there is no way for users
539 * to disable forcible compression separately.
541 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
543 if ((btrfs_test_opt(info, COMPRESS) &&
544 (info->compress_type != saved_compress_type ||
545 compress_force != saved_compress_force)) ||
546 (!btrfs_test_opt(info, COMPRESS) &&
548 btrfs_info(root->fs_info,
550 (compress_force) ? "force" : "use",
553 compress_force = false;
556 btrfs_set_and_info(info, SSD,
557 "use ssd allocation scheme");
560 btrfs_set_and_info(info, SSD_SPREAD,
561 "use spread ssd allocation scheme");
562 btrfs_set_opt(info->mount_opt, SSD);
565 btrfs_set_and_info(info, NOSSD,
566 "not using ssd allocation scheme");
567 btrfs_clear_opt(info->mount_opt, SSD);
570 btrfs_clear_and_info(info, NOBARRIER,
571 "turning on barriers");
574 btrfs_set_and_info(info, NOBARRIER,
575 "turning off barriers");
577 case Opt_thread_pool:
578 ret = match_int(&args[0], &intarg);
581 } else if (intarg > 0) {
582 info->thread_pool_size = intarg;
589 num = match_strdup(&args[0]);
591 info->max_inline = memparse(num, NULL);
594 if (info->max_inline) {
595 info->max_inline = min_t(u64,
599 btrfs_info(root->fs_info, "max_inline at %llu",
606 case Opt_alloc_start:
607 num = match_strdup(&args[0]);
609 mutex_lock(&info->chunk_mutex);
610 info->alloc_start = memparse(num, NULL);
611 mutex_unlock(&info->chunk_mutex);
613 btrfs_info(root->fs_info,
614 "allocations start at %llu",
622 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
623 root->fs_info->sb->s_flags |= MS_POSIXACL;
626 btrfs_err(root->fs_info,
627 "support for ACL not compiled in!");
632 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
635 btrfs_set_and_info(info, NOTREELOG,
636 "disabling tree log");
639 btrfs_clear_and_info(info, NOTREELOG,
640 "enabling tree log");
643 case Opt_nologreplay:
644 btrfs_set_and_info(info, NOLOGREPLAY,
645 "disabling log replay at mount time");
647 case Opt_flushoncommit:
648 btrfs_set_and_info(info, FLUSHONCOMMIT,
649 "turning on flush-on-commit");
651 case Opt_noflushoncommit:
652 btrfs_clear_and_info(info, FLUSHONCOMMIT,
653 "turning off flush-on-commit");
656 ret = match_int(&args[0], &intarg);
659 } else if (intarg >= 0) {
660 info->metadata_ratio = intarg;
661 btrfs_info(root->fs_info, "metadata ratio %d",
662 info->metadata_ratio);
669 btrfs_set_and_info(info, DISCARD,
670 "turning on discard");
673 btrfs_clear_and_info(info, DISCARD,
674 "turning off discard");
676 case Opt_space_cache:
677 case Opt_space_cache_version:
678 if (token == Opt_space_cache ||
679 strcmp(args[0].from, "v1") == 0) {
680 btrfs_clear_opt(root->fs_info->mount_opt,
682 btrfs_set_and_info(info, SPACE_CACHE,
683 "enabling disk space caching");
684 } else if (strcmp(args[0].from, "v2") == 0) {
685 btrfs_clear_opt(root->fs_info->mount_opt,
687 btrfs_set_and_info(info,
689 "enabling free space tree");
695 case Opt_rescan_uuid_tree:
696 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
698 case Opt_no_space_cache:
699 if (btrfs_test_opt(info, SPACE_CACHE)) {
700 btrfs_clear_and_info(info,
702 "disabling disk space caching");
704 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
705 btrfs_clear_and_info(info,
707 "disabling free space tree");
710 case Opt_inode_cache:
711 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
712 "enabling inode map caching");
714 case Opt_noinode_cache:
715 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
716 "disabling inode map caching");
718 case Opt_clear_cache:
719 btrfs_set_and_info(info, CLEAR_CACHE,
720 "force clearing of disk cache");
722 case Opt_user_subvol_rm_allowed:
723 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
725 case Opt_enospc_debug:
726 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
728 case Opt_noenospc_debug:
729 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
732 btrfs_set_and_info(info, AUTO_DEFRAG,
733 "enabling auto defrag");
736 btrfs_clear_and_info(info, AUTO_DEFRAG,
737 "disabling auto defrag");
740 btrfs_warn(root->fs_info,
741 "'recovery' is deprecated, use 'usebackuproot' instead");
742 case Opt_usebackuproot:
743 btrfs_info(root->fs_info,
744 "trying to use backup root at mount time");
745 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
747 case Opt_skip_balance:
748 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
750 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
751 case Opt_check_integrity_including_extent_data:
752 btrfs_info(root->fs_info,
753 "enabling check integrity including extent data");
754 btrfs_set_opt(info->mount_opt,
755 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
756 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
758 case Opt_check_integrity:
759 btrfs_info(root->fs_info, "enabling check integrity");
760 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
762 case Opt_check_integrity_print_mask:
763 ret = match_int(&args[0], &intarg);
766 } else if (intarg >= 0) {
767 info->check_integrity_print_mask = intarg;
768 btrfs_info(root->fs_info,
769 "check_integrity_print_mask 0x%x",
770 info->check_integrity_print_mask);
777 case Opt_check_integrity_including_extent_data:
778 case Opt_check_integrity:
779 case Opt_check_integrity_print_mask:
780 btrfs_err(root->fs_info,
781 "support for check_integrity* not compiled in!");
785 case Opt_fatal_errors:
786 if (strcmp(args[0].from, "panic") == 0)
787 btrfs_set_opt(info->mount_opt,
788 PANIC_ON_FATAL_ERROR);
789 else if (strcmp(args[0].from, "bug") == 0)
790 btrfs_clear_opt(info->mount_opt,
791 PANIC_ON_FATAL_ERROR);
797 case Opt_commit_interval:
799 ret = match_int(&args[0], &intarg);
801 btrfs_err(root->fs_info,
802 "invalid commit interval");
808 btrfs_warn(root->fs_info,
809 "excessive commit interval %d",
812 info->commit_interval = intarg;
814 btrfs_info(root->fs_info,
815 "using default commit interval %ds",
816 BTRFS_DEFAULT_COMMIT_INTERVAL);
817 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
820 #ifdef CONFIG_BTRFS_DEBUG
821 case Opt_fragment_all:
822 btrfs_info(root->fs_info, "fragmenting all space");
823 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
824 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
826 case Opt_fragment_metadata:
827 btrfs_info(root->fs_info, "fragmenting metadata");
828 btrfs_set_opt(info->mount_opt,
831 case Opt_fragment_data:
832 btrfs_info(root->fs_info, "fragmenting data");
833 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
837 btrfs_info(root->fs_info,
838 "unrecognized mount option '%s'", p);
847 * Extra check for current option against current flag
849 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
850 btrfs_err(root->fs_info,
851 "nologreplay must be used with ro mount option");
855 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
856 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
857 !btrfs_test_opt(info, CLEAR_CACHE)) {
858 btrfs_err(root->fs_info, "cannot disable free space tree");
862 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
863 btrfs_info(root->fs_info, "disk space caching is enabled");
864 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
865 btrfs_info(root->fs_info, "using free space tree");
871 * Parse mount options that are required early in the mount process.
873 * All other options will be parsed on much later in the mount process and
874 * only when we need to allocate a new super block.
876 static int btrfs_parse_early_options(const char *options, fmode_t flags,
877 void *holder, char **subvol_name, u64 *subvol_objectid,
878 struct btrfs_fs_devices **fs_devices)
880 substring_t args[MAX_OPT_ARGS];
881 char *device_name, *opts, *orig, *p;
889 * strsep changes the string, duplicate it because parse_options
892 opts = kstrdup(options, GFP_KERNEL);
897 while ((p = strsep(&opts, ",")) != NULL) {
902 token = match_token(p, tokens, args);
906 *subvol_name = match_strdup(&args[0]);
913 num = match_strdup(&args[0]);
915 *subvol_objectid = memparse(num, NULL);
917 /* we want the original fs_tree */
918 if (!*subvol_objectid)
920 BTRFS_FS_TREE_OBJECTID;
926 case Opt_subvolrootid:
927 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
930 device_name = match_strdup(&args[0]);
935 error = btrfs_scan_one_device(device_name,
936 flags, holder, fs_devices);
951 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
954 struct btrfs_root *root = fs_info->tree_root;
955 struct btrfs_root *fs_root;
956 struct btrfs_root_ref *root_ref;
957 struct btrfs_inode_ref *inode_ref;
958 struct btrfs_key key;
959 struct btrfs_path *path = NULL;
960 char *name = NULL, *ptr;
965 path = btrfs_alloc_path();
970 path->leave_spinning = 1;
972 name = kmalloc(PATH_MAX, GFP_NOFS);
977 ptr = name + PATH_MAX - 1;
981 * Walk up the subvolume trees in the tree of tree roots by root
982 * backrefs until we hit the top-level subvolume.
984 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
985 key.objectid = subvol_objectid;
986 key.type = BTRFS_ROOT_BACKREF_KEY;
987 key.offset = (u64)-1;
989 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
992 } else if (ret > 0) {
993 ret = btrfs_previous_item(root, path, subvol_objectid,
994 BTRFS_ROOT_BACKREF_KEY);
997 } else if (ret > 0) {
1003 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1004 subvol_objectid = key.offset;
1006 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1007 struct btrfs_root_ref);
1008 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1011 ret = -ENAMETOOLONG;
1014 read_extent_buffer(path->nodes[0], ptr + 1,
1015 (unsigned long)(root_ref + 1), len);
1017 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1018 btrfs_release_path(path);
1020 key.objectid = subvol_objectid;
1021 key.type = BTRFS_ROOT_ITEM_KEY;
1022 key.offset = (u64)-1;
1023 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1024 if (IS_ERR(fs_root)) {
1025 ret = PTR_ERR(fs_root);
1030 * Walk up the filesystem tree by inode refs until we hit the
1033 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1034 key.objectid = dirid;
1035 key.type = BTRFS_INODE_REF_KEY;
1036 key.offset = (u64)-1;
1038 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1041 } else if (ret > 0) {
1042 ret = btrfs_previous_item(fs_root, path, dirid,
1043 BTRFS_INODE_REF_KEY);
1046 } else if (ret > 0) {
1052 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1055 inode_ref = btrfs_item_ptr(path->nodes[0],
1057 struct btrfs_inode_ref);
1058 len = btrfs_inode_ref_name_len(path->nodes[0],
1062 ret = -ENAMETOOLONG;
1065 read_extent_buffer(path->nodes[0], ptr + 1,
1066 (unsigned long)(inode_ref + 1), len);
1068 btrfs_release_path(path);
1072 btrfs_free_path(path);
1073 if (ptr == name + PATH_MAX - 1) {
1077 memmove(name, ptr, name + PATH_MAX - ptr);
1082 btrfs_free_path(path);
1084 return ERR_PTR(ret);
1087 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1089 struct btrfs_root *root = fs_info->tree_root;
1090 struct btrfs_dir_item *di;
1091 struct btrfs_path *path;
1092 struct btrfs_key location;
1095 path = btrfs_alloc_path();
1098 path->leave_spinning = 1;
1101 * Find the "default" dir item which points to the root item that we
1102 * will mount by default if we haven't been given a specific subvolume
1105 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1106 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1108 btrfs_free_path(path);
1113 * Ok the default dir item isn't there. This is weird since
1114 * it's always been there, but don't freak out, just try and
1115 * mount the top-level subvolume.
1117 btrfs_free_path(path);
1118 *objectid = BTRFS_FS_TREE_OBJECTID;
1122 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1123 btrfs_free_path(path);
1124 *objectid = location.objectid;
1128 static int btrfs_fill_super(struct super_block *sb,
1129 struct btrfs_fs_devices *fs_devices,
1130 void *data, int silent)
1132 struct inode *inode;
1133 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1134 struct btrfs_key key;
1137 sb->s_maxbytes = MAX_LFS_FILESIZE;
1138 sb->s_magic = BTRFS_SUPER_MAGIC;
1139 sb->s_op = &btrfs_super_ops;
1140 sb->s_d_op = &btrfs_dentry_operations;
1141 sb->s_export_op = &btrfs_export_ops;
1142 sb->s_xattr = btrfs_xattr_handlers;
1143 sb->s_time_gran = 1;
1144 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1145 sb->s_flags |= MS_POSIXACL;
1147 sb->s_flags |= MS_I_VERSION;
1148 sb->s_iflags |= SB_I_CGROUPWB;
1149 err = open_ctree(sb, fs_devices, (char *)data);
1151 btrfs_err(fs_info, "open_ctree failed");
1155 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1156 key.type = BTRFS_INODE_ITEM_KEY;
1158 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1159 if (IS_ERR(inode)) {
1160 err = PTR_ERR(inode);
1164 sb->s_root = d_make_root(inode);
1170 save_mount_options(sb, data);
1171 cleancache_init_fs(sb);
1172 sb->s_flags |= MS_ACTIVE;
1176 close_ctree(fs_info->tree_root);
1180 int btrfs_sync_fs(struct super_block *sb, int wait)
1182 struct btrfs_trans_handle *trans;
1183 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1184 struct btrfs_root *root = fs_info->tree_root;
1186 trace_btrfs_sync_fs(fs_info, wait);
1189 filemap_flush(fs_info->btree_inode->i_mapping);
1193 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1195 trans = btrfs_attach_transaction_barrier(root);
1196 if (IS_ERR(trans)) {
1197 /* no transaction, don't bother */
1198 if (PTR_ERR(trans) == -ENOENT) {
1200 * Exit unless we have some pending changes
1201 * that need to go through commit
1203 if (fs_info->pending_changes == 0)
1206 * A non-blocking test if the fs is frozen. We must not
1207 * start a new transaction here otherwise a deadlock
1208 * happens. The pending operations are delayed to the
1209 * next commit after thawing.
1211 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1212 __sb_end_write(sb, SB_FREEZE_WRITE);
1215 trans = btrfs_start_transaction(root, 0);
1218 return PTR_ERR(trans);
1220 return btrfs_commit_transaction(trans, root);
1223 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1225 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1226 struct btrfs_root *root = info->tree_root;
1227 char *compress_type;
1228 const char *subvol_name;
1230 if (btrfs_test_opt(info, DEGRADED))
1231 seq_puts(seq, ",degraded");
1232 if (btrfs_test_opt(info, NODATASUM))
1233 seq_puts(seq, ",nodatasum");
1234 if (btrfs_test_opt(info, NODATACOW))
1235 seq_puts(seq, ",nodatacow");
1236 if (btrfs_test_opt(info, NOBARRIER))
1237 seq_puts(seq, ",nobarrier");
1238 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1239 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1240 if (info->alloc_start != 0)
1241 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1242 if (info->thread_pool_size != min_t(unsigned long,
1243 num_online_cpus() + 2, 8))
1244 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1245 if (btrfs_test_opt(info, COMPRESS)) {
1246 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1247 compress_type = "zlib";
1249 compress_type = "lzo";
1250 if (btrfs_test_opt(info, FORCE_COMPRESS))
1251 seq_printf(seq, ",compress-force=%s", compress_type);
1253 seq_printf(seq, ",compress=%s", compress_type);
1255 if (btrfs_test_opt(info, NOSSD))
1256 seq_puts(seq, ",nossd");
1257 if (btrfs_test_opt(info, SSD_SPREAD))
1258 seq_puts(seq, ",ssd_spread");
1259 else if (btrfs_test_opt(info, SSD))
1260 seq_puts(seq, ",ssd");
1261 if (btrfs_test_opt(info, NOTREELOG))
1262 seq_puts(seq, ",notreelog");
1263 if (btrfs_test_opt(info, NOLOGREPLAY))
1264 seq_puts(seq, ",nologreplay");
1265 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1266 seq_puts(seq, ",flushoncommit");
1267 if (btrfs_test_opt(info, DISCARD))
1268 seq_puts(seq, ",discard");
1269 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1270 seq_puts(seq, ",noacl");
1271 if (btrfs_test_opt(info, SPACE_CACHE))
1272 seq_puts(seq, ",space_cache");
1273 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1274 seq_puts(seq, ",space_cache=v2");
1276 seq_puts(seq, ",nospace_cache");
1277 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1278 seq_puts(seq, ",rescan_uuid_tree");
1279 if (btrfs_test_opt(info, CLEAR_CACHE))
1280 seq_puts(seq, ",clear_cache");
1281 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1282 seq_puts(seq, ",user_subvol_rm_allowed");
1283 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1284 seq_puts(seq, ",enospc_debug");
1285 if (btrfs_test_opt(info, AUTO_DEFRAG))
1286 seq_puts(seq, ",autodefrag");
1287 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1288 seq_puts(seq, ",inode_cache");
1289 if (btrfs_test_opt(info, SKIP_BALANCE))
1290 seq_puts(seq, ",skip_balance");
1291 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1292 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1293 seq_puts(seq, ",check_int_data");
1294 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1295 seq_puts(seq, ",check_int");
1296 if (info->check_integrity_print_mask)
1297 seq_printf(seq, ",check_int_print_mask=%d",
1298 info->check_integrity_print_mask);
1300 if (info->metadata_ratio)
1301 seq_printf(seq, ",metadata_ratio=%d",
1302 info->metadata_ratio);
1303 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1304 seq_puts(seq, ",fatal_errors=panic");
1305 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1306 seq_printf(seq, ",commit=%d", info->commit_interval);
1307 #ifdef CONFIG_BTRFS_DEBUG
1308 if (btrfs_test_opt(info, FRAGMENT_DATA))
1309 seq_puts(seq, ",fragment=data");
1310 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1311 seq_puts(seq, ",fragment=metadata");
1313 seq_printf(seq, ",subvolid=%llu",
1314 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1315 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1316 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1317 if (!IS_ERR(subvol_name)) {
1318 seq_puts(seq, ",subvol=");
1319 seq_escape(seq, subvol_name, " \t\n\\");
1325 static int btrfs_test_super(struct super_block *s, void *data)
1327 struct btrfs_fs_info *p = data;
1328 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1330 return fs_info->fs_devices == p->fs_devices;
1333 static int btrfs_set_super(struct super_block *s, void *data)
1335 int err = set_anon_super(s, data);
1337 s->s_fs_info = data;
1342 * subvolumes are identified by ino 256
1344 static inline int is_subvolume_inode(struct inode *inode)
1346 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1352 * This will add subvolid=0 to the argument string while removing any subvol=
1353 * and subvolid= arguments to make sure we get the top-level root for path
1354 * walking to the subvol we want.
1356 static char *setup_root_args(char *args)
1358 char *buf, *dst, *sep;
1361 return kstrdup("subvolid=0", GFP_NOFS);
1363 /* The worst case is that we add ",subvolid=0" to the end. */
1364 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1369 sep = strchrnul(args, ',');
1370 if (!strstarts(args, "subvol=") &&
1371 !strstarts(args, "subvolid=")) {
1372 memcpy(dst, args, sep - args);
1381 strcpy(dst, "subvolid=0");
1386 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1387 int flags, const char *device_name,
1390 struct dentry *root;
1391 struct vfsmount *mnt = NULL;
1395 newargs = setup_root_args(data);
1397 root = ERR_PTR(-ENOMEM);
1401 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1402 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1403 if (flags & MS_RDONLY) {
1404 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1405 device_name, newargs);
1407 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1408 device_name, newargs);
1410 root = ERR_CAST(mnt);
1415 down_write(&mnt->mnt_sb->s_umount);
1416 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1417 up_write(&mnt->mnt_sb->s_umount);
1419 root = ERR_PTR(ret);
1425 root = ERR_CAST(mnt);
1431 if (!subvol_objectid) {
1432 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1435 root = ERR_PTR(ret);
1439 subvol_name = btrfs_get_subvol_name_from_objectid(
1440 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1441 if (IS_ERR(subvol_name)) {
1442 root = ERR_CAST(subvol_name);
1449 root = mount_subtree(mnt, subvol_name);
1450 /* mount_subtree() drops our reference on the vfsmount. */
1453 if (!IS_ERR(root)) {
1454 struct super_block *s = root->d_sb;
1455 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1456 struct inode *root_inode = d_inode(root);
1457 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1460 if (!is_subvolume_inode(root_inode)) {
1461 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1465 if (subvol_objectid && root_objectid != subvol_objectid) {
1467 * This will also catch a race condition where a
1468 * subvolume which was passed by ID is renamed and
1469 * another subvolume is renamed over the old location.
1472 "subvol '%s' does not match subvolid %llu",
1473 subvol_name, subvol_objectid);
1478 root = ERR_PTR(ret);
1479 deactivate_locked_super(s);
1490 static int parse_security_options(char *orig_opts,
1491 struct security_mnt_opts *sec_opts)
1493 char *secdata = NULL;
1496 secdata = alloc_secdata();
1499 ret = security_sb_copy_data(orig_opts, secdata);
1501 free_secdata(secdata);
1504 ret = security_sb_parse_opts_str(secdata, sec_opts);
1505 free_secdata(secdata);
1509 static int setup_security_options(struct btrfs_fs_info *fs_info,
1510 struct super_block *sb,
1511 struct security_mnt_opts *sec_opts)
1516 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1519 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1523 #ifdef CONFIG_SECURITY
1524 if (!fs_info->security_opts.num_mnt_opts) {
1525 /* first time security setup, copy sec_opts to fs_info */
1526 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1529 * Since SELinux (the only one supporting security_mnt_opts)
1530 * does NOT support changing context during remount/mount of
1531 * the same sb, this must be the same or part of the same
1532 * security options, just free it.
1534 security_free_mnt_opts(sec_opts);
1541 * Find a superblock for the given device / mount point.
1543 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1544 * for multiple device setup. Make sure to keep it in sync.
1546 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1547 const char *device_name, void *data)
1549 struct block_device *bdev = NULL;
1550 struct super_block *s;
1551 struct btrfs_fs_devices *fs_devices = NULL;
1552 struct btrfs_fs_info *fs_info = NULL;
1553 struct security_mnt_opts new_sec_opts;
1554 fmode_t mode = FMODE_READ;
1555 char *subvol_name = NULL;
1556 u64 subvol_objectid = 0;
1559 if (!(flags & MS_RDONLY))
1560 mode |= FMODE_WRITE;
1562 error = btrfs_parse_early_options(data, mode, fs_type,
1563 &subvol_name, &subvol_objectid,
1567 return ERR_PTR(error);
1570 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1571 /* mount_subvol() will free subvol_name. */
1572 return mount_subvol(subvol_name, subvol_objectid, flags,
1576 security_init_mnt_opts(&new_sec_opts);
1578 error = parse_security_options(data, &new_sec_opts);
1580 return ERR_PTR(error);
1583 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1585 goto error_sec_opts;
1588 * Setup a dummy root and fs_info for test/set super. This is because
1589 * we don't actually fill this stuff out until open_ctree, but we need
1590 * it for searching for existing supers, so this lets us do that and
1591 * then open_ctree will properly initialize everything later.
1593 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1596 goto error_sec_opts;
1599 fs_info->fs_devices = fs_devices;
1601 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1602 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1603 security_init_mnt_opts(&fs_info->security_opts);
1604 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1609 error = btrfs_open_devices(fs_devices, mode, fs_type);
1613 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1615 goto error_close_devices;
1618 bdev = fs_devices->latest_bdev;
1619 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1623 goto error_close_devices;
1627 btrfs_close_devices(fs_devices);
1628 free_fs_info(fs_info);
1629 if ((flags ^ s->s_flags) & MS_RDONLY)
1632 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1633 btrfs_sb(s)->bdev_holder = fs_type;
1634 error = btrfs_fill_super(s, fs_devices, data,
1635 flags & MS_SILENT ? 1 : 0);
1638 deactivate_locked_super(s);
1639 goto error_sec_opts;
1642 fs_info = btrfs_sb(s);
1643 error = setup_security_options(fs_info, s, &new_sec_opts);
1645 deactivate_locked_super(s);
1646 goto error_sec_opts;
1649 return dget(s->s_root);
1651 error_close_devices:
1652 btrfs_close_devices(fs_devices);
1654 free_fs_info(fs_info);
1656 security_free_mnt_opts(&new_sec_opts);
1657 return ERR_PTR(error);
1660 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1661 int new_pool_size, int old_pool_size)
1663 if (new_pool_size == old_pool_size)
1666 fs_info->thread_pool_size = new_pool_size;
1668 btrfs_info(fs_info, "resize thread pool %d -> %d",
1669 old_pool_size, new_pool_size);
1671 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1672 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1673 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1674 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1675 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1676 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1677 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1679 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1680 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1681 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1682 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1683 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1687 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1689 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1692 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1693 unsigned long old_opts, int flags)
1695 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1696 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1697 (flags & MS_RDONLY))) {
1698 /* wait for any defraggers to finish */
1699 wait_event(fs_info->transaction_wait,
1700 (atomic_read(&fs_info->defrag_running) == 0));
1701 if (flags & MS_RDONLY)
1702 sync_filesystem(fs_info->sb);
1706 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1707 unsigned long old_opts)
1710 * We need to cleanup all defragable inodes if the autodefragment is
1711 * close or the filesystem is read only.
1713 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1714 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1715 (fs_info->sb->s_flags & MS_RDONLY))) {
1716 btrfs_cleanup_defrag_inodes(fs_info);
1719 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1722 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1724 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1725 struct btrfs_root *root = fs_info->tree_root;
1726 unsigned old_flags = sb->s_flags;
1727 unsigned long old_opts = fs_info->mount_opt;
1728 unsigned long old_compress_type = fs_info->compress_type;
1729 u64 old_max_inline = fs_info->max_inline;
1730 u64 old_alloc_start = fs_info->alloc_start;
1731 int old_thread_pool_size = fs_info->thread_pool_size;
1732 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1735 sync_filesystem(sb);
1736 btrfs_remount_prepare(fs_info);
1739 struct security_mnt_opts new_sec_opts;
1741 security_init_mnt_opts(&new_sec_opts);
1742 ret = parse_security_options(data, &new_sec_opts);
1745 ret = setup_security_options(fs_info, sb,
1748 security_free_mnt_opts(&new_sec_opts);
1753 ret = btrfs_parse_options(root, data, *flags);
1759 btrfs_remount_begin(fs_info, old_opts, *flags);
1760 btrfs_resize_thread_pool(fs_info,
1761 fs_info->thread_pool_size, old_thread_pool_size);
1763 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1766 if (*flags & MS_RDONLY) {
1768 * this also happens on 'umount -rf' or on shutdown, when
1769 * the filesystem is busy.
1771 cancel_work_sync(&fs_info->async_reclaim_work);
1773 /* wait for the uuid_scan task to finish */
1774 down(&fs_info->uuid_tree_rescan_sem);
1775 /* avoid complains from lockdep et al. */
1776 up(&fs_info->uuid_tree_rescan_sem);
1778 sb->s_flags |= MS_RDONLY;
1781 * Setting MS_RDONLY will put the cleaner thread to
1782 * sleep at the next loop if it's already active.
1783 * If it's already asleep, we'll leave unused block
1784 * groups on disk until we're mounted read-write again
1785 * unless we clean them up here.
1787 btrfs_delete_unused_bgs(fs_info);
1789 btrfs_dev_replace_suspend_for_unmount(fs_info);
1790 btrfs_scrub_cancel(fs_info);
1791 btrfs_pause_balance(fs_info);
1793 ret = btrfs_commit_super(root);
1797 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1799 "Remounting read-write after error is not allowed");
1803 if (fs_info->fs_devices->rw_devices == 0) {
1808 if (fs_info->fs_devices->missing_devices >
1809 fs_info->num_tolerated_disk_barrier_failures &&
1810 !(*flags & MS_RDONLY)) {
1812 "too many missing devices, writeable remount is not allowed");
1817 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1819 "mount required to replay tree-log, cannot remount read-write");
1824 ret = btrfs_cleanup_fs_roots(fs_info);
1828 /* recover relocation */
1829 mutex_lock(&fs_info->cleaner_mutex);
1830 ret = btrfs_recover_relocation(root);
1831 mutex_unlock(&fs_info->cleaner_mutex);
1835 ret = btrfs_resume_balance_async(fs_info);
1839 ret = btrfs_resume_dev_replace_async(fs_info);
1841 btrfs_warn(fs_info, "failed to resume dev_replace");
1845 btrfs_qgroup_rescan_resume(fs_info);
1847 if (!fs_info->uuid_root) {
1848 btrfs_info(fs_info, "creating UUID tree");
1849 ret = btrfs_create_uuid_tree(fs_info);
1852 "failed to create the UUID tree %d",
1857 sb->s_flags &= ~MS_RDONLY;
1859 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1862 wake_up_process(fs_info->transaction_kthread);
1863 btrfs_remount_cleanup(fs_info, old_opts);
1867 /* We've hit an error - don't reset MS_RDONLY */
1868 if (sb->s_flags & MS_RDONLY)
1869 old_flags |= MS_RDONLY;
1870 sb->s_flags = old_flags;
1871 fs_info->mount_opt = old_opts;
1872 fs_info->compress_type = old_compress_type;
1873 fs_info->max_inline = old_max_inline;
1874 mutex_lock(&fs_info->chunk_mutex);
1875 fs_info->alloc_start = old_alloc_start;
1876 mutex_unlock(&fs_info->chunk_mutex);
1877 btrfs_resize_thread_pool(fs_info,
1878 old_thread_pool_size, fs_info->thread_pool_size);
1879 fs_info->metadata_ratio = old_metadata_ratio;
1880 btrfs_remount_cleanup(fs_info, old_opts);
1884 /* Used to sort the devices by max_avail(descending sort) */
1885 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1886 const void *dev_info2)
1888 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1889 ((struct btrfs_device_info *)dev_info2)->max_avail)
1891 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1892 ((struct btrfs_device_info *)dev_info2)->max_avail)
1899 * sort the devices by max_avail, in which max free extent size of each device
1900 * is stored.(Descending Sort)
1902 static inline void btrfs_descending_sort_devices(
1903 struct btrfs_device_info *devices,
1906 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1907 btrfs_cmp_device_free_bytes, NULL);
1911 * The helper to calc the free space on the devices that can be used to store
1914 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1916 struct btrfs_fs_info *fs_info = root->fs_info;
1917 struct btrfs_device_info *devices_info;
1918 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1919 struct btrfs_device *device;
1924 u64 min_stripe_size;
1925 int min_stripes = 1, num_stripes = 1;
1926 int i = 0, nr_devices;
1930 * We aren't under the device list lock, so this is racy-ish, but good
1931 * enough for our purposes.
1933 nr_devices = fs_info->fs_devices->open_devices;
1936 nr_devices = fs_info->fs_devices->open_devices;
1944 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1949 /* calc min stripe number for data space allocation */
1950 type = btrfs_get_alloc_profile(root, 1);
1951 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1953 num_stripes = nr_devices;
1954 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1957 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1962 if (type & BTRFS_BLOCK_GROUP_DUP)
1963 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1965 min_stripe_size = BTRFS_STRIPE_LEN;
1967 if (fs_info->alloc_start)
1968 mutex_lock(&fs_devices->device_list_mutex);
1970 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1971 if (!device->in_fs_metadata || !device->bdev ||
1972 device->is_tgtdev_for_dev_replace)
1975 if (i >= nr_devices)
1978 avail_space = device->total_bytes - device->bytes_used;
1980 /* align with stripe_len */
1981 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1982 avail_space *= BTRFS_STRIPE_LEN;
1985 * In order to avoid overwriting the superblock on the drive,
1986 * btrfs starts at an offset of at least 1MB when doing chunk
1991 /* user can set the offset in fs_info->alloc_start. */
1992 if (fs_info->alloc_start &&
1993 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1994 device->total_bytes) {
1996 skip_space = max(fs_info->alloc_start, skip_space);
1999 * btrfs can not use the free space in
2000 * [0, skip_space - 1], we must subtract it from the
2001 * total. In order to implement it, we account the used
2002 * space in this range first.
2004 ret = btrfs_account_dev_extents_size(device, 0,
2008 kfree(devices_info);
2009 mutex_unlock(&fs_devices->device_list_mutex);
2015 /* calc the free space in [0, skip_space - 1] */
2016 skip_space -= used_space;
2020 * we can use the free space in [0, skip_space - 1], subtract
2021 * it from the total.
2023 if (avail_space && avail_space >= skip_space)
2024 avail_space -= skip_space;
2028 if (avail_space < min_stripe_size)
2031 devices_info[i].dev = device;
2032 devices_info[i].max_avail = avail_space;
2037 if (fs_info->alloc_start)
2038 mutex_unlock(&fs_devices->device_list_mutex);
2042 btrfs_descending_sort_devices(devices_info, nr_devices);
2046 while (nr_devices >= min_stripes) {
2047 if (num_stripes > nr_devices)
2048 num_stripes = nr_devices;
2050 if (devices_info[i].max_avail >= min_stripe_size) {
2054 avail_space += devices_info[i].max_avail * num_stripes;
2055 alloc_size = devices_info[i].max_avail;
2056 for (j = i + 1 - num_stripes; j <= i; j++)
2057 devices_info[j].max_avail -= alloc_size;
2063 kfree(devices_info);
2064 *free_bytes = avail_space;
2069 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2071 * If there's a redundant raid level at DATA block groups, use the respective
2072 * multiplier to scale the sizes.
2074 * Unused device space usage is based on simulating the chunk allocator
2075 * algorithm that respects the device sizes, order of allocations and the
2076 * 'alloc_start' value, this is a close approximation of the actual use but
2077 * there are other factors that may change the result (like a new metadata
2080 * If metadata is exhausted, f_bavail will be 0.
2082 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2084 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2085 struct btrfs_super_block *disk_super = fs_info->super_copy;
2086 struct list_head *head = &fs_info->space_info;
2087 struct btrfs_space_info *found;
2089 u64 total_free_data = 0;
2090 u64 total_free_meta = 0;
2091 int bits = dentry->d_sb->s_blocksize_bits;
2092 __be32 *fsid = (__be32 *)fs_info->fsid;
2093 unsigned factor = 1;
2094 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2100 * holding chunk_mutex to avoid allocating new chunks, holding
2101 * device_list_mutex to avoid the device being removed
2104 list_for_each_entry_rcu(found, head, list) {
2105 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2108 total_free_data += found->disk_total - found->disk_used;
2110 btrfs_account_ro_block_groups_free_space(found);
2112 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2113 if (!list_empty(&found->block_groups[i])) {
2115 case BTRFS_RAID_DUP:
2116 case BTRFS_RAID_RAID1:
2117 case BTRFS_RAID_RAID10:
2125 * Metadata in mixed block goup profiles are accounted in data
2127 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2128 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2131 total_free_meta += found->disk_total -
2135 total_used += found->disk_used;
2140 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2141 buf->f_blocks >>= bits;
2142 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2144 /* Account global block reserve as used, it's in logical size already */
2145 spin_lock(&block_rsv->lock);
2146 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2147 if (buf->f_bfree >= block_rsv->size >> bits)
2148 buf->f_bfree -= block_rsv->size >> bits;
2151 spin_unlock(&block_rsv->lock);
2153 buf->f_bavail = div_u64(total_free_data, factor);
2154 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2157 buf->f_bavail += div_u64(total_free_data, factor);
2158 buf->f_bavail = buf->f_bavail >> bits;
2161 * We calculate the remaining metadata space minus global reserve. If
2162 * this is (supposedly) smaller than zero, there's no space. But this
2163 * does not hold in practice, the exhausted state happens where's still
2164 * some positive delta. So we apply some guesswork and compare the
2165 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2167 * We probably cannot calculate the exact threshold value because this
2168 * depends on the internal reservations requested by various
2169 * operations, so some operations that consume a few metadata will
2170 * succeed even if the Avail is zero. But this is better than the other
2173 thresh = 4 * 1024 * 1024;
2176 * We only want to claim there's no available space if we can no longer
2177 * allocate chunks for our metadata profile and our global reserve will
2178 * not fit in the free metadata space. If we aren't ->full then we
2179 * still can allocate chunks and thus are fine using the currently
2180 * calculated f_bavail.
2182 if (!mixed && block_rsv->space_info->full &&
2183 total_free_meta - thresh < block_rsv->size)
2186 buf->f_type = BTRFS_SUPER_MAGIC;
2187 buf->f_bsize = dentry->d_sb->s_blocksize;
2188 buf->f_namelen = BTRFS_NAME_LEN;
2190 /* We treat it as constant endianness (it doesn't matter _which_)
2191 because we want the fsid to come out the same whether mounted
2192 on a big-endian or little-endian host */
2193 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2194 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2195 /* Mask in the root object ID too, to disambiguate subvols */
2196 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2197 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2202 static void btrfs_kill_super(struct super_block *sb)
2204 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2205 kill_anon_super(sb);
2206 free_fs_info(fs_info);
2209 static struct file_system_type btrfs_fs_type = {
2210 .owner = THIS_MODULE,
2212 .mount = btrfs_mount,
2213 .kill_sb = btrfs_kill_super,
2214 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2216 MODULE_ALIAS_FS("btrfs");
2218 static int btrfs_control_open(struct inode *inode, struct file *file)
2221 * The control file's private_data is used to hold the
2222 * transaction when it is started and is used to keep
2223 * track of whether a transaction is already in progress.
2225 file->private_data = NULL;
2230 * used by btrfsctl to scan devices when no FS is mounted
2232 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2235 struct btrfs_ioctl_vol_args *vol;
2236 struct btrfs_fs_devices *fs_devices;
2239 if (!capable(CAP_SYS_ADMIN))
2242 vol = memdup_user((void __user *)arg, sizeof(*vol));
2244 return PTR_ERR(vol);
2245 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2248 case BTRFS_IOC_SCAN_DEV:
2249 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2250 &btrfs_fs_type, &fs_devices);
2252 case BTRFS_IOC_DEVICES_READY:
2253 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2254 &btrfs_fs_type, &fs_devices);
2257 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2259 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2260 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2268 static int btrfs_freeze(struct super_block *sb)
2270 struct btrfs_trans_handle *trans;
2271 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2273 root->fs_info->fs_frozen = 1;
2275 * We don't need a barrier here, we'll wait for any transaction that
2276 * could be in progress on other threads (and do delayed iputs that
2277 * we want to avoid on a frozen filesystem), or do the commit
2280 trans = btrfs_attach_transaction_barrier(root);
2281 if (IS_ERR(trans)) {
2282 /* no transaction, don't bother */
2283 if (PTR_ERR(trans) == -ENOENT)
2285 return PTR_ERR(trans);
2287 return btrfs_commit_transaction(trans, root);
2290 static int btrfs_unfreeze(struct super_block *sb)
2292 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2294 root->fs_info->fs_frozen = 0;
2298 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2300 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2301 struct btrfs_fs_devices *cur_devices;
2302 struct btrfs_device *dev, *first_dev = NULL;
2303 struct list_head *head;
2304 struct rcu_string *name;
2306 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2307 cur_devices = fs_info->fs_devices;
2308 while (cur_devices) {
2309 head = &cur_devices->devices;
2310 list_for_each_entry(dev, head, dev_list) {
2315 if (!first_dev || dev->devid < first_dev->devid)
2318 cur_devices = cur_devices->seed;
2323 name = rcu_dereference(first_dev->name);
2324 seq_escape(m, name->str, " \t\n\\");
2329 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2333 static const struct super_operations btrfs_super_ops = {
2334 .drop_inode = btrfs_drop_inode,
2335 .evict_inode = btrfs_evict_inode,
2336 .put_super = btrfs_put_super,
2337 .sync_fs = btrfs_sync_fs,
2338 .show_options = btrfs_show_options,
2339 .show_devname = btrfs_show_devname,
2340 .write_inode = btrfs_write_inode,
2341 .alloc_inode = btrfs_alloc_inode,
2342 .destroy_inode = btrfs_destroy_inode,
2343 .statfs = btrfs_statfs,
2344 .remount_fs = btrfs_remount,
2345 .freeze_fs = btrfs_freeze,
2346 .unfreeze_fs = btrfs_unfreeze,
2349 static const struct file_operations btrfs_ctl_fops = {
2350 .open = btrfs_control_open,
2351 .unlocked_ioctl = btrfs_control_ioctl,
2352 .compat_ioctl = btrfs_control_ioctl,
2353 .owner = THIS_MODULE,
2354 .llseek = noop_llseek,
2357 static struct miscdevice btrfs_misc = {
2358 .minor = BTRFS_MINOR,
2359 .name = "btrfs-control",
2360 .fops = &btrfs_ctl_fops
2363 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2364 MODULE_ALIAS("devname:btrfs-control");
2366 static int btrfs_interface_init(void)
2368 return misc_register(&btrfs_misc);
2371 static void btrfs_interface_exit(void)
2373 misc_deregister(&btrfs_misc);
2376 static void btrfs_print_mod_info(void)
2378 pr_info("Btrfs loaded, crc32c=%s"
2379 #ifdef CONFIG_BTRFS_DEBUG
2382 #ifdef CONFIG_BTRFS_ASSERT
2385 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2386 ", integrity-checker=on"
2389 btrfs_crc32c_impl());
2392 static int __init init_btrfs_fs(void)
2396 err = btrfs_hash_init();
2402 err = btrfs_init_sysfs();
2406 btrfs_init_compress();
2408 err = btrfs_init_cachep();
2412 err = extent_io_init();
2416 err = extent_map_init();
2418 goto free_extent_io;
2420 err = ordered_data_init();
2422 goto free_extent_map;
2424 err = btrfs_delayed_inode_init();
2426 goto free_ordered_data;
2428 err = btrfs_auto_defrag_init();
2430 goto free_delayed_inode;
2432 err = btrfs_delayed_ref_init();
2434 goto free_auto_defrag;
2436 err = btrfs_prelim_ref_init();
2438 goto free_delayed_ref;
2440 err = btrfs_end_io_wq_init();
2442 goto free_prelim_ref;
2444 err = btrfs_interface_init();
2446 goto free_end_io_wq;
2448 btrfs_init_lockdep();
2450 btrfs_print_mod_info();
2452 err = btrfs_run_sanity_tests();
2454 goto unregister_ioctl;
2456 err = register_filesystem(&btrfs_fs_type);
2458 goto unregister_ioctl;
2463 btrfs_interface_exit();
2465 btrfs_end_io_wq_exit();
2467 btrfs_prelim_ref_exit();
2469 btrfs_delayed_ref_exit();
2471 btrfs_auto_defrag_exit();
2473 btrfs_delayed_inode_exit();
2475 ordered_data_exit();
2481 btrfs_destroy_cachep();
2483 btrfs_exit_compress();
2490 static void __exit exit_btrfs_fs(void)
2492 btrfs_destroy_cachep();
2493 btrfs_delayed_ref_exit();
2494 btrfs_auto_defrag_exit();
2495 btrfs_delayed_inode_exit();
2496 btrfs_prelim_ref_exit();
2497 ordered_data_exit();
2500 btrfs_interface_exit();
2501 btrfs_end_io_wq_exit();
2502 unregister_filesystem(&btrfs_fs_type);
2504 btrfs_cleanup_fs_uuids();
2505 btrfs_exit_compress();
2509 late_initcall(init_btrfs_fs);
2510 module_exit(exit_btrfs_fs)
2512 MODULE_LICENSE("GPL");