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/crc32c.h>
45 #include <linux/btrfs.h>
46 #include "delayed-inode.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #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;
70 * Types for mounting the default subvolume and a subvolume explicitly
71 * requested by subvol=/path. That way the callchain is straightforward and we
72 * don't have to play tricks with the mount options and recursive calls to
75 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
77 static struct file_system_type btrfs_fs_type;
78 static struct file_system_type btrfs_root_fs_type;
80 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
82 const char *btrfs_decode_error(int errno)
84 char *errstr = "unknown";
88 errstr = "IO failure";
91 errstr = "Out of memory";
94 errstr = "Readonly filesystem";
97 errstr = "Object already exists";
100 errstr = "No space left";
103 errstr = "No such entry";
111 * __btrfs_handle_fs_error decodes expected errors from the caller and
112 * invokes the approciate error response.
115 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
116 unsigned int line, int errno, const char *fmt, ...)
118 struct super_block *sb = fs_info->sb;
124 * Special case: if the error is EROFS, and we're already
125 * under SB_RDONLY, then it is safe here.
127 if (errno == -EROFS && sb_rdonly(sb))
131 errstr = btrfs_decode_error(errno);
133 struct va_format vaf;
140 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
141 sb->s_id, function, line, errno, errstr, &vaf);
144 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
145 sb->s_id, function, line, errno, errstr);
150 * Today we only save the error info to memory. Long term we'll
151 * also send it down to the disk
153 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
155 /* Don't go through full error handling during mount */
156 if (!(sb->s_flags & SB_BORN))
162 /* btrfs handle error by forcing the filesystem readonly */
163 sb->s_flags |= SB_RDONLY;
164 btrfs_info(fs_info, "forced readonly");
166 * Note that a running device replace operation is not canceled here
167 * although there is no way to update the progress. It would add the
168 * risk of a deadlock, therefore the canceling is omitted. The only
169 * penalty is that some I/O remains active until the procedure
170 * completes. The next time when the filesystem is mounted writeable
171 * again, the device replace operation continues.
176 static const char * const logtypes[] = {
189 * Use one ratelimit state per log level so that a flood of less important
190 * messages doesn't cause more important ones to be dropped.
192 static struct ratelimit_state printk_limits[] = {
193 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
200 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
203 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
206 struct va_format vaf;
209 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit = &printk_limits[4];
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
229 if (__ratelimit(ratelimit))
230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
231 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
238 * We only mark the transaction aborted and then set the file system read-only.
239 * This will prevent new transactions from starting or trying to join this
242 * This means that error recovery at the call site is limited to freeing
243 * any local memory allocations and passing the error code up without
244 * further cleanup. The transaction should complete as it normally would
245 * in the call path but will return -EIO.
247 * We'll complete the cleanup in btrfs_end_transaction and
248 * btrfs_commit_transaction.
251 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
252 const char *function,
253 unsigned int line, int errno)
255 struct btrfs_fs_info *fs_info = trans->fs_info;
257 trans->aborted = errno;
258 /* Nothing used. The other threads that have joined this
259 * transaction may be able to continue. */
260 if (!trans->dirty && list_empty(&trans->new_bgs)) {
263 errstr = btrfs_decode_error(errno);
265 "%s:%d: Aborting unused transaction(%s).",
266 function, line, errstr);
269 WRITE_ONCE(trans->transaction->aborted, errno);
270 /* Wake up anybody who may be waiting on this transaction */
271 wake_up(&fs_info->transaction_wait);
272 wake_up(&fs_info->transaction_blocked_wait);
273 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
276 * __btrfs_panic decodes unexpected, fatal errors from the caller,
277 * issues an alert, and either panics or BUGs, depending on mount options.
280 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
281 unsigned int line, int errno, const char *fmt, ...)
283 char *s_id = "<unknown>";
285 struct va_format vaf = { .fmt = fmt };
289 s_id = fs_info->sb->s_id;
294 errstr = btrfs_decode_error(errno);
295 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
296 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
297 s_id, function, line, &vaf, errno, errstr);
299 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
300 function, line, &vaf, errno, errstr);
302 /* Caller calls BUG() */
305 static void btrfs_put_super(struct super_block *sb)
307 close_ctree(btrfs_sb(sb));
316 Opt_compress_force_type,
321 Opt_flushoncommit, Opt_noflushoncommit,
322 Opt_inode_cache, Opt_noinode_cache,
324 Opt_barrier, Opt_nobarrier,
325 Opt_datacow, Opt_nodatacow,
326 Opt_datasum, Opt_nodatasum,
327 Opt_defrag, Opt_nodefrag,
328 Opt_discard, Opt_nodiscard,
332 Opt_rescan_uuid_tree,
334 Opt_space_cache, Opt_no_space_cache,
335 Opt_space_cache_version,
337 Opt_ssd_spread, Opt_nossd_spread,
341 Opt_treelog, Opt_notreelog,
343 Opt_user_subvol_rm_allowed,
345 /* Deprecated options */
350 /* Debugging options */
352 Opt_check_integrity_including_extent_data,
353 Opt_check_integrity_print_mask,
354 Opt_enospc_debug, Opt_noenospc_debug,
355 #ifdef CONFIG_BTRFS_DEBUG
356 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
358 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
364 static const match_table_t tokens = {
366 {Opt_noacl, "noacl"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_commit_interval, "commit=%u"},
369 {Opt_compress, "compress"},
370 {Opt_compress_type, "compress=%s"},
371 {Opt_compress_force, "compress-force"},
372 {Opt_compress_force_type, "compress-force=%s"},
373 {Opt_degraded, "degraded"},
374 {Opt_device, "device=%s"},
375 {Opt_fatal_errors, "fatal_errors=%s"},
376 {Opt_flushoncommit, "flushoncommit"},
377 {Opt_noflushoncommit, "noflushoncommit"},
378 {Opt_inode_cache, "inode_cache"},
379 {Opt_noinode_cache, "noinode_cache"},
380 {Opt_max_inline, "max_inline=%s"},
381 {Opt_barrier, "barrier"},
382 {Opt_nobarrier, "nobarrier"},
383 {Opt_datacow, "datacow"},
384 {Opt_nodatacow, "nodatacow"},
385 {Opt_datasum, "datasum"},
386 {Opt_nodatasum, "nodatasum"},
387 {Opt_defrag, "autodefrag"},
388 {Opt_nodefrag, "noautodefrag"},
389 {Opt_discard, "discard"},
390 {Opt_nodiscard, "nodiscard"},
391 {Opt_nologreplay, "nologreplay"},
392 {Opt_norecovery, "norecovery"},
393 {Opt_ratio, "metadata_ratio=%u"},
394 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
395 {Opt_skip_balance, "skip_balance"},
396 {Opt_space_cache, "space_cache"},
397 {Opt_no_space_cache, "nospace_cache"},
398 {Opt_space_cache_version, "space_cache=%s"},
400 {Opt_nossd, "nossd"},
401 {Opt_ssd_spread, "ssd_spread"},
402 {Opt_nossd_spread, "nossd_spread"},
403 {Opt_subvol, "subvol=%s"},
404 {Opt_subvolid, "subvolid=%s"},
405 {Opt_thread_pool, "thread_pool=%u"},
406 {Opt_treelog, "treelog"},
407 {Opt_notreelog, "notreelog"},
408 {Opt_usebackuproot, "usebackuproot"},
409 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
411 /* Deprecated options */
412 {Opt_alloc_start, "alloc_start=%s"},
413 {Opt_recovery, "recovery"},
414 {Opt_subvolrootid, "subvolrootid=%d"},
416 /* Debugging options */
417 {Opt_check_integrity, "check_int"},
418 {Opt_check_integrity_including_extent_data, "check_int_data"},
419 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
420 {Opt_enospc_debug, "enospc_debug"},
421 {Opt_noenospc_debug, "noenospc_debug"},
422 #ifdef CONFIG_BTRFS_DEBUG
423 {Opt_fragment_data, "fragment=data"},
424 {Opt_fragment_metadata, "fragment=metadata"},
425 {Opt_fragment_all, "fragment=all"},
427 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
428 {Opt_ref_verify, "ref_verify"},
434 * Regular mount options parser. Everything that is needed only when
435 * reading in a new superblock is parsed here.
436 * XXX JDM: This needs to be cleaned up for remount.
438 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
439 unsigned long new_flags)
441 substring_t args[MAX_OPT_ARGS];
447 bool compress_force = false;
448 enum btrfs_compression_type saved_compress_type;
449 bool saved_compress_force;
452 cache_gen = btrfs_super_cache_generation(info->super_copy);
453 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
454 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
456 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
459 * Even the options are empty, we still need to do extra check
465 while ((p = strsep(&options, ",")) != NULL) {
470 token = match_token(p, tokens, args);
473 btrfs_info(info, "allowing degraded mounts");
474 btrfs_set_opt(info->mount_opt, DEGRADED);
478 case Opt_subvolrootid:
481 * These are parsed by btrfs_parse_subvol_options
482 * and btrfs_parse_early_options
483 * and can be happily ignored here.
487 btrfs_set_and_info(info, NODATASUM,
488 "setting nodatasum");
491 if (btrfs_test_opt(info, NODATASUM)) {
492 if (btrfs_test_opt(info, NODATACOW))
494 "setting datasum, datacow enabled");
496 btrfs_info(info, "setting datasum");
498 btrfs_clear_opt(info->mount_opt, NODATACOW);
499 btrfs_clear_opt(info->mount_opt, NODATASUM);
502 if (!btrfs_test_opt(info, NODATACOW)) {
503 if (!btrfs_test_opt(info, COMPRESS) ||
504 !btrfs_test_opt(info, FORCE_COMPRESS)) {
506 "setting nodatacow, compression disabled");
508 btrfs_info(info, "setting nodatacow");
511 btrfs_clear_opt(info->mount_opt, COMPRESS);
512 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
513 btrfs_set_opt(info->mount_opt, NODATACOW);
514 btrfs_set_opt(info->mount_opt, NODATASUM);
517 btrfs_clear_and_info(info, NODATACOW,
520 case Opt_compress_force:
521 case Opt_compress_force_type:
522 compress_force = true;
525 case Opt_compress_type:
526 saved_compress_type = btrfs_test_opt(info,
528 info->compress_type : BTRFS_COMPRESS_NONE;
529 saved_compress_force =
530 btrfs_test_opt(info, FORCE_COMPRESS);
531 if (token == Opt_compress ||
532 token == Opt_compress_force ||
533 strncmp(args[0].from, "zlib", 4) == 0) {
534 compress_type = "zlib";
536 info->compress_type = BTRFS_COMPRESS_ZLIB;
537 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
539 * args[0] contains uninitialized data since
540 * for these tokens we don't expect any
543 if (token != Opt_compress &&
544 token != Opt_compress_force)
545 info->compress_level =
546 btrfs_compress_str2level(args[0].from);
547 btrfs_set_opt(info->mount_opt, COMPRESS);
548 btrfs_clear_opt(info->mount_opt, NODATACOW);
549 btrfs_clear_opt(info->mount_opt, NODATASUM);
551 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
552 compress_type = "lzo";
553 info->compress_type = BTRFS_COMPRESS_LZO;
554 btrfs_set_opt(info->mount_opt, COMPRESS);
555 btrfs_clear_opt(info->mount_opt, NODATACOW);
556 btrfs_clear_opt(info->mount_opt, NODATASUM);
557 btrfs_set_fs_incompat(info, COMPRESS_LZO);
559 } else if (strcmp(args[0].from, "zstd") == 0) {
560 compress_type = "zstd";
561 info->compress_type = BTRFS_COMPRESS_ZSTD;
562 btrfs_set_opt(info->mount_opt, COMPRESS);
563 btrfs_clear_opt(info->mount_opt, NODATACOW);
564 btrfs_clear_opt(info->mount_opt, NODATASUM);
565 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
567 } else if (strncmp(args[0].from, "no", 2) == 0) {
568 compress_type = "no";
569 btrfs_clear_opt(info->mount_opt, COMPRESS);
570 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
571 compress_force = false;
578 if (compress_force) {
579 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
582 * If we remount from compress-force=xxx to
583 * compress=xxx, we need clear FORCE_COMPRESS
584 * flag, otherwise, there is no way for users
585 * to disable forcible compression separately.
587 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
589 if ((btrfs_test_opt(info, COMPRESS) &&
590 (info->compress_type != saved_compress_type ||
591 compress_force != saved_compress_force)) ||
592 (!btrfs_test_opt(info, COMPRESS) &&
594 btrfs_info(info, "%s %s compression, level %d",
595 (compress_force) ? "force" : "use",
596 compress_type, info->compress_level);
598 compress_force = false;
601 btrfs_set_and_info(info, SSD,
602 "enabling ssd optimizations");
603 btrfs_clear_opt(info->mount_opt, NOSSD);
606 btrfs_set_and_info(info, SSD,
607 "enabling ssd optimizations");
608 btrfs_set_and_info(info, SSD_SPREAD,
609 "using spread ssd allocation scheme");
610 btrfs_clear_opt(info->mount_opt, NOSSD);
613 btrfs_set_opt(info->mount_opt, NOSSD);
614 btrfs_clear_and_info(info, SSD,
615 "not using ssd optimizations");
617 case Opt_nossd_spread:
618 btrfs_clear_and_info(info, SSD_SPREAD,
619 "not using spread ssd allocation scheme");
622 btrfs_clear_and_info(info, NOBARRIER,
623 "turning on barriers");
626 btrfs_set_and_info(info, NOBARRIER,
627 "turning off barriers");
629 case Opt_thread_pool:
630 ret = match_int(&args[0], &intarg);
633 } else if (intarg == 0) {
637 info->thread_pool_size = intarg;
640 num = match_strdup(&args[0]);
642 info->max_inline = memparse(num, NULL);
645 if (info->max_inline) {
646 info->max_inline = min_t(u64,
650 btrfs_info(info, "max_inline at %llu",
657 case Opt_alloc_start:
659 "option alloc_start is obsolete, ignored");
662 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
663 info->sb->s_flags |= SB_POSIXACL;
666 btrfs_err(info, "support for ACL not compiled in!");
671 info->sb->s_flags &= ~SB_POSIXACL;
674 btrfs_set_and_info(info, NOTREELOG,
675 "disabling tree log");
678 btrfs_clear_and_info(info, NOTREELOG,
679 "enabling tree log");
682 case Opt_nologreplay:
683 btrfs_set_and_info(info, NOLOGREPLAY,
684 "disabling log replay at mount time");
686 case Opt_flushoncommit:
687 btrfs_set_and_info(info, FLUSHONCOMMIT,
688 "turning on flush-on-commit");
690 case Opt_noflushoncommit:
691 btrfs_clear_and_info(info, FLUSHONCOMMIT,
692 "turning off flush-on-commit");
695 ret = match_int(&args[0], &intarg);
698 info->metadata_ratio = intarg;
699 btrfs_info(info, "metadata ratio %u",
700 info->metadata_ratio);
703 btrfs_set_and_info(info, DISCARD,
704 "turning on discard");
707 btrfs_clear_and_info(info, DISCARD,
708 "turning off discard");
710 case Opt_space_cache:
711 case Opt_space_cache_version:
712 if (token == Opt_space_cache ||
713 strcmp(args[0].from, "v1") == 0) {
714 btrfs_clear_opt(info->mount_opt,
716 btrfs_set_and_info(info, SPACE_CACHE,
717 "enabling disk space caching");
718 } else if (strcmp(args[0].from, "v2") == 0) {
719 btrfs_clear_opt(info->mount_opt,
721 btrfs_set_and_info(info, FREE_SPACE_TREE,
722 "enabling free space tree");
728 case Opt_rescan_uuid_tree:
729 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
731 case Opt_no_space_cache:
732 if (btrfs_test_opt(info, SPACE_CACHE)) {
733 btrfs_clear_and_info(info, SPACE_CACHE,
734 "disabling disk space caching");
736 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
737 btrfs_clear_and_info(info, FREE_SPACE_TREE,
738 "disabling free space tree");
741 case Opt_inode_cache:
742 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
743 "enabling inode map caching");
745 case Opt_noinode_cache:
746 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
747 "disabling inode map caching");
749 case Opt_clear_cache:
750 btrfs_set_and_info(info, CLEAR_CACHE,
751 "force clearing of disk cache");
753 case Opt_user_subvol_rm_allowed:
754 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
756 case Opt_enospc_debug:
757 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
759 case Opt_noenospc_debug:
760 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
763 btrfs_set_and_info(info, AUTO_DEFRAG,
764 "enabling auto defrag");
767 btrfs_clear_and_info(info, AUTO_DEFRAG,
768 "disabling auto defrag");
772 "'recovery' is deprecated, use 'usebackuproot' instead");
773 case Opt_usebackuproot:
775 "trying to use backup root at mount time");
776 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
778 case Opt_skip_balance:
779 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
781 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
782 case Opt_check_integrity_including_extent_data:
784 "enabling check integrity including extent data");
785 btrfs_set_opt(info->mount_opt,
786 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
787 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
789 case Opt_check_integrity:
790 btrfs_info(info, "enabling check integrity");
791 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
793 case Opt_check_integrity_print_mask:
794 ret = match_int(&args[0], &intarg);
797 info->check_integrity_print_mask = intarg;
798 btrfs_info(info, "check_integrity_print_mask 0x%x",
799 info->check_integrity_print_mask);
802 case Opt_check_integrity_including_extent_data:
803 case Opt_check_integrity:
804 case Opt_check_integrity_print_mask:
806 "support for check_integrity* not compiled in!");
810 case Opt_fatal_errors:
811 if (strcmp(args[0].from, "panic") == 0)
812 btrfs_set_opt(info->mount_opt,
813 PANIC_ON_FATAL_ERROR);
814 else if (strcmp(args[0].from, "bug") == 0)
815 btrfs_clear_opt(info->mount_opt,
816 PANIC_ON_FATAL_ERROR);
822 case Opt_commit_interval:
824 ret = match_int(&args[0], &intarg);
829 "using default commit interval %us",
830 BTRFS_DEFAULT_COMMIT_INTERVAL);
831 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
832 } else if (intarg > 300) {
833 btrfs_warn(info, "excessive commit interval %d",
836 info->commit_interval = intarg;
838 #ifdef CONFIG_BTRFS_DEBUG
839 case Opt_fragment_all:
840 btrfs_info(info, "fragmenting all space");
841 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
842 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
844 case Opt_fragment_metadata:
845 btrfs_info(info, "fragmenting metadata");
846 btrfs_set_opt(info->mount_opt,
849 case Opt_fragment_data:
850 btrfs_info(info, "fragmenting data");
851 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
854 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
856 btrfs_info(info, "doing ref verification");
857 btrfs_set_opt(info->mount_opt, REF_VERIFY);
861 btrfs_info(info, "unrecognized mount option '%s'", p);
870 * Extra check for current option against current flag
872 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
874 "nologreplay must be used with ro mount option");
878 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
879 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
880 !btrfs_test_opt(info, CLEAR_CACHE)) {
881 btrfs_err(info, "cannot disable free space tree");
885 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
886 btrfs_info(info, "disk space caching is enabled");
887 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
888 btrfs_info(info, "using free space tree");
893 * Parse mount options that are required early in the mount process.
895 * All other options will be parsed on much later in the mount process and
896 * only when we need to allocate a new super block.
898 static int btrfs_parse_early_options(const char *options, fmode_t flags,
899 void *holder, struct btrfs_fs_devices **fs_devices)
901 substring_t args[MAX_OPT_ARGS];
902 char *device_name, *opts, *orig, *p;
909 * strsep changes the string, duplicate it because btrfs_parse_options
912 opts = kstrdup(options, GFP_KERNEL);
917 while ((p = strsep(&opts, ",")) != NULL) {
923 token = match_token(p, tokens, args);
924 if (token == Opt_device) {
925 device_name = match_strdup(&args[0]);
930 error = btrfs_scan_one_device(device_name,
931 flags, holder, fs_devices);
944 * Parse mount options that are related to subvolume id
946 * The value is later passed to mount_subvol()
948 static int btrfs_parse_subvol_options(const char *options, fmode_t flags,
949 char **subvol_name, u64 *subvol_objectid)
951 substring_t args[MAX_OPT_ARGS];
952 char *opts, *orig, *p;
960 * strsep changes the string, duplicate it because
961 * btrfs_parse_early_options gets called later
963 opts = kstrdup(options, GFP_KERNEL);
968 while ((p = strsep(&opts, ",")) != NULL) {
973 token = match_token(p, tokens, args);
977 *subvol_name = match_strdup(&args[0]);
984 error = match_u64(&args[0], &subvolid);
988 /* we want the original fs_tree */
990 subvolid = BTRFS_FS_TREE_OBJECTID;
992 *subvol_objectid = subvolid;
994 case Opt_subvolrootid:
995 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
1007 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1008 u64 subvol_objectid)
1010 struct btrfs_root *root = fs_info->tree_root;
1011 struct btrfs_root *fs_root;
1012 struct btrfs_root_ref *root_ref;
1013 struct btrfs_inode_ref *inode_ref;
1014 struct btrfs_key key;
1015 struct btrfs_path *path = NULL;
1016 char *name = NULL, *ptr;
1021 path = btrfs_alloc_path();
1026 path->leave_spinning = 1;
1028 name = kmalloc(PATH_MAX, GFP_KERNEL);
1033 ptr = name + PATH_MAX - 1;
1037 * Walk up the subvolume trees in the tree of tree roots by root
1038 * backrefs until we hit the top-level subvolume.
1040 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1041 key.objectid = subvol_objectid;
1042 key.type = BTRFS_ROOT_BACKREF_KEY;
1043 key.offset = (u64)-1;
1045 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1048 } else if (ret > 0) {
1049 ret = btrfs_previous_item(root, path, subvol_objectid,
1050 BTRFS_ROOT_BACKREF_KEY);
1053 } else if (ret > 0) {
1059 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1060 subvol_objectid = key.offset;
1062 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1063 struct btrfs_root_ref);
1064 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1067 ret = -ENAMETOOLONG;
1070 read_extent_buffer(path->nodes[0], ptr + 1,
1071 (unsigned long)(root_ref + 1), len);
1073 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1074 btrfs_release_path(path);
1076 key.objectid = subvol_objectid;
1077 key.type = BTRFS_ROOT_ITEM_KEY;
1078 key.offset = (u64)-1;
1079 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1080 if (IS_ERR(fs_root)) {
1081 ret = PTR_ERR(fs_root);
1086 * Walk up the filesystem tree by inode refs until we hit the
1089 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1090 key.objectid = dirid;
1091 key.type = BTRFS_INODE_REF_KEY;
1092 key.offset = (u64)-1;
1094 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1097 } else if (ret > 0) {
1098 ret = btrfs_previous_item(fs_root, path, dirid,
1099 BTRFS_INODE_REF_KEY);
1102 } else if (ret > 0) {
1108 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1111 inode_ref = btrfs_item_ptr(path->nodes[0],
1113 struct btrfs_inode_ref);
1114 len = btrfs_inode_ref_name_len(path->nodes[0],
1118 ret = -ENAMETOOLONG;
1121 read_extent_buffer(path->nodes[0], ptr + 1,
1122 (unsigned long)(inode_ref + 1), len);
1124 btrfs_release_path(path);
1128 btrfs_free_path(path);
1129 if (ptr == name + PATH_MAX - 1) {
1133 memmove(name, ptr, name + PATH_MAX - ptr);
1138 btrfs_free_path(path);
1140 return ERR_PTR(ret);
1143 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1145 struct btrfs_root *root = fs_info->tree_root;
1146 struct btrfs_dir_item *di;
1147 struct btrfs_path *path;
1148 struct btrfs_key location;
1151 path = btrfs_alloc_path();
1154 path->leave_spinning = 1;
1157 * Find the "default" dir item which points to the root item that we
1158 * will mount by default if we haven't been given a specific subvolume
1161 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1162 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1164 btrfs_free_path(path);
1169 * Ok the default dir item isn't there. This is weird since
1170 * it's always been there, but don't freak out, just try and
1171 * mount the top-level subvolume.
1173 btrfs_free_path(path);
1174 *objectid = BTRFS_FS_TREE_OBJECTID;
1178 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1179 btrfs_free_path(path);
1180 *objectid = location.objectid;
1184 static int btrfs_fill_super(struct super_block *sb,
1185 struct btrfs_fs_devices *fs_devices,
1188 struct inode *inode;
1189 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1190 struct btrfs_key key;
1193 sb->s_maxbytes = MAX_LFS_FILESIZE;
1194 sb->s_magic = BTRFS_SUPER_MAGIC;
1195 sb->s_op = &btrfs_super_ops;
1196 sb->s_d_op = &btrfs_dentry_operations;
1197 sb->s_export_op = &btrfs_export_ops;
1198 sb->s_xattr = btrfs_xattr_handlers;
1199 sb->s_time_gran = 1;
1200 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1201 sb->s_flags |= SB_POSIXACL;
1203 sb->s_flags |= SB_I_VERSION;
1204 sb->s_iflags |= SB_I_CGROUPWB;
1206 err = super_setup_bdi(sb);
1208 btrfs_err(fs_info, "super_setup_bdi failed");
1212 err = open_ctree(sb, fs_devices, (char *)data);
1214 btrfs_err(fs_info, "open_ctree failed");
1218 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1219 key.type = BTRFS_INODE_ITEM_KEY;
1221 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1222 if (IS_ERR(inode)) {
1223 err = PTR_ERR(inode);
1227 sb->s_root = d_make_root(inode);
1233 cleancache_init_fs(sb);
1234 sb->s_flags |= SB_ACTIVE;
1238 close_ctree(fs_info);
1242 int btrfs_sync_fs(struct super_block *sb, int wait)
1244 struct btrfs_trans_handle *trans;
1245 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1246 struct btrfs_root *root = fs_info->tree_root;
1248 trace_btrfs_sync_fs(fs_info, wait);
1251 filemap_flush(fs_info->btree_inode->i_mapping);
1255 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1257 trans = btrfs_attach_transaction_barrier(root);
1258 if (IS_ERR(trans)) {
1259 /* no transaction, don't bother */
1260 if (PTR_ERR(trans) == -ENOENT) {
1262 * Exit unless we have some pending changes
1263 * that need to go through commit
1265 if (fs_info->pending_changes == 0)
1268 * A non-blocking test if the fs is frozen. We must not
1269 * start a new transaction here otherwise a deadlock
1270 * happens. The pending operations are delayed to the
1271 * next commit after thawing.
1273 if (sb_start_write_trylock(sb))
1277 trans = btrfs_start_transaction(root, 0);
1280 return PTR_ERR(trans);
1282 return btrfs_commit_transaction(trans);
1285 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1287 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1288 const char *compress_type;
1290 if (btrfs_test_opt(info, DEGRADED))
1291 seq_puts(seq, ",degraded");
1292 if (btrfs_test_opt(info, NODATASUM))
1293 seq_puts(seq, ",nodatasum");
1294 if (btrfs_test_opt(info, NODATACOW))
1295 seq_puts(seq, ",nodatacow");
1296 if (btrfs_test_opt(info, NOBARRIER))
1297 seq_puts(seq, ",nobarrier");
1298 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1299 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1300 if (info->thread_pool_size != min_t(unsigned long,
1301 num_online_cpus() + 2, 8))
1302 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1303 if (btrfs_test_opt(info, COMPRESS)) {
1304 compress_type = btrfs_compress_type2str(info->compress_type);
1305 if (btrfs_test_opt(info, FORCE_COMPRESS))
1306 seq_printf(seq, ",compress-force=%s", compress_type);
1308 seq_printf(seq, ",compress=%s", compress_type);
1309 if (info->compress_level)
1310 seq_printf(seq, ":%d", info->compress_level);
1312 if (btrfs_test_opt(info, NOSSD))
1313 seq_puts(seq, ",nossd");
1314 if (btrfs_test_opt(info, SSD_SPREAD))
1315 seq_puts(seq, ",ssd_spread");
1316 else if (btrfs_test_opt(info, SSD))
1317 seq_puts(seq, ",ssd");
1318 if (btrfs_test_opt(info, NOTREELOG))
1319 seq_puts(seq, ",notreelog");
1320 if (btrfs_test_opt(info, NOLOGREPLAY))
1321 seq_puts(seq, ",nologreplay");
1322 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1323 seq_puts(seq, ",flushoncommit");
1324 if (btrfs_test_opt(info, DISCARD))
1325 seq_puts(seq, ",discard");
1326 if (!(info->sb->s_flags & SB_POSIXACL))
1327 seq_puts(seq, ",noacl");
1328 if (btrfs_test_opt(info, SPACE_CACHE))
1329 seq_puts(seq, ",space_cache");
1330 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1331 seq_puts(seq, ",space_cache=v2");
1333 seq_puts(seq, ",nospace_cache");
1334 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1335 seq_puts(seq, ",rescan_uuid_tree");
1336 if (btrfs_test_opt(info, CLEAR_CACHE))
1337 seq_puts(seq, ",clear_cache");
1338 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1339 seq_puts(seq, ",user_subvol_rm_allowed");
1340 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1341 seq_puts(seq, ",enospc_debug");
1342 if (btrfs_test_opt(info, AUTO_DEFRAG))
1343 seq_puts(seq, ",autodefrag");
1344 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1345 seq_puts(seq, ",inode_cache");
1346 if (btrfs_test_opt(info, SKIP_BALANCE))
1347 seq_puts(seq, ",skip_balance");
1348 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1349 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1350 seq_puts(seq, ",check_int_data");
1351 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1352 seq_puts(seq, ",check_int");
1353 if (info->check_integrity_print_mask)
1354 seq_printf(seq, ",check_int_print_mask=%d",
1355 info->check_integrity_print_mask);
1357 if (info->metadata_ratio)
1358 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1359 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1360 seq_puts(seq, ",fatal_errors=panic");
1361 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1362 seq_printf(seq, ",commit=%u", info->commit_interval);
1363 #ifdef CONFIG_BTRFS_DEBUG
1364 if (btrfs_test_opt(info, FRAGMENT_DATA))
1365 seq_puts(seq, ",fragment=data");
1366 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1367 seq_puts(seq, ",fragment=metadata");
1369 if (btrfs_test_opt(info, REF_VERIFY))
1370 seq_puts(seq, ",ref_verify");
1371 seq_printf(seq, ",subvolid=%llu",
1372 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1373 seq_puts(seq, ",subvol=");
1374 seq_dentry(seq, dentry, " \t\n\\");
1378 static int btrfs_test_super(struct super_block *s, void *data)
1380 struct btrfs_fs_info *p = data;
1381 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1383 return fs_info->fs_devices == p->fs_devices;
1386 static int btrfs_set_super(struct super_block *s, void *data)
1388 int err = set_anon_super(s, data);
1390 s->s_fs_info = data;
1395 * subvolumes are identified by ino 256
1397 static inline int is_subvolume_inode(struct inode *inode)
1399 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1404 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1405 const char *device_name, struct vfsmount *mnt)
1407 struct dentry *root;
1411 if (!subvol_objectid) {
1412 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1415 root = ERR_PTR(ret);
1419 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1421 if (IS_ERR(subvol_name)) {
1422 root = ERR_CAST(subvol_name);
1429 root = mount_subtree(mnt, subvol_name);
1430 /* mount_subtree() drops our reference on the vfsmount. */
1433 if (!IS_ERR(root)) {
1434 struct super_block *s = root->d_sb;
1435 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1436 struct inode *root_inode = d_inode(root);
1437 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1440 if (!is_subvolume_inode(root_inode)) {
1441 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1445 if (subvol_objectid && root_objectid != subvol_objectid) {
1447 * This will also catch a race condition where a
1448 * subvolume which was passed by ID is renamed and
1449 * another subvolume is renamed over the old location.
1452 "subvol '%s' does not match subvolid %llu",
1453 subvol_name, subvol_objectid);
1458 root = ERR_PTR(ret);
1459 deactivate_locked_super(s);
1469 static int parse_security_options(char *orig_opts,
1470 struct security_mnt_opts *sec_opts)
1472 char *secdata = NULL;
1475 secdata = alloc_secdata();
1478 ret = security_sb_copy_data(orig_opts, secdata);
1480 free_secdata(secdata);
1483 ret = security_sb_parse_opts_str(secdata, sec_opts);
1484 free_secdata(secdata);
1488 static int setup_security_options(struct btrfs_fs_info *fs_info,
1489 struct super_block *sb,
1490 struct security_mnt_opts *sec_opts)
1495 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1498 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1502 #ifdef CONFIG_SECURITY
1503 if (!fs_info->security_opts.num_mnt_opts) {
1504 /* first time security setup, copy sec_opts to fs_info */
1505 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1508 * Since SELinux (the only one supporting security_mnt_opts)
1509 * does NOT support changing context during remount/mount of
1510 * the same sb, this must be the same or part of the same
1511 * security options, just free it.
1513 security_free_mnt_opts(sec_opts);
1520 * Find a superblock for the given device / mount point.
1522 * Note: This is based on mount_bdev from fs/super.c with a few additions
1523 * for multiple device setup. Make sure to keep it in sync.
1525 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1526 int flags, const char *device_name, void *data)
1528 struct block_device *bdev = NULL;
1529 struct super_block *s;
1530 struct btrfs_fs_devices *fs_devices = NULL;
1531 struct btrfs_fs_info *fs_info = NULL;
1532 struct security_mnt_opts new_sec_opts;
1533 fmode_t mode = FMODE_READ;
1536 if (!(flags & SB_RDONLY))
1537 mode |= FMODE_WRITE;
1539 error = btrfs_parse_early_options(data, mode, fs_type,
1542 return ERR_PTR(error);
1545 security_init_mnt_opts(&new_sec_opts);
1547 error = parse_security_options(data, &new_sec_opts);
1549 return ERR_PTR(error);
1552 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1554 goto error_sec_opts;
1557 * Setup a dummy root and fs_info for test/set super. This is because
1558 * we don't actually fill this stuff out until open_ctree, but we need
1559 * it for searching for existing supers, so this lets us do that and
1560 * then open_ctree will properly initialize everything later.
1562 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1565 goto error_sec_opts;
1568 fs_info->fs_devices = fs_devices;
1570 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1571 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1572 security_init_mnt_opts(&fs_info->security_opts);
1573 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1578 error = btrfs_open_devices(fs_devices, mode, fs_type);
1582 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1584 goto error_close_devices;
1587 bdev = fs_devices->latest_bdev;
1588 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1592 goto error_close_devices;
1596 btrfs_close_devices(fs_devices);
1597 free_fs_info(fs_info);
1598 if ((flags ^ s->s_flags) & SB_RDONLY)
1601 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1602 btrfs_sb(s)->bdev_holder = fs_type;
1603 error = btrfs_fill_super(s, fs_devices, data);
1606 deactivate_locked_super(s);
1607 goto error_sec_opts;
1610 fs_info = btrfs_sb(s);
1611 error = setup_security_options(fs_info, s, &new_sec_opts);
1613 deactivate_locked_super(s);
1614 goto error_sec_opts;
1617 return dget(s->s_root);
1619 error_close_devices:
1620 btrfs_close_devices(fs_devices);
1622 free_fs_info(fs_info);
1624 security_free_mnt_opts(&new_sec_opts);
1625 return ERR_PTR(error);
1629 * Mount function which is called by VFS layer.
1631 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1632 * which needs vfsmount* of device's root (/). This means device's root has to
1633 * be mounted internally in any case.
1636 * 1. Parse subvol id related options for later use in mount_subvol().
1638 * 2. Mount device's root (/) by calling vfs_kern_mount().
1640 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1641 * first place. In order to avoid calling btrfs_mount() again, we use
1642 * different file_system_type which is not registered to VFS by
1643 * register_filesystem() (btrfs_root_fs_type). As a result,
1644 * btrfs_mount_root() is called. The return value will be used by
1645 * mount_subtree() in mount_subvol().
1647 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1648 * "btrfs subvolume set-default", mount_subvol() is called always.
1650 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1651 const char *device_name, void *data)
1653 struct vfsmount *mnt_root;
1654 struct dentry *root;
1655 fmode_t mode = FMODE_READ;
1656 char *subvol_name = NULL;
1657 u64 subvol_objectid = 0;
1660 if (!(flags & SB_RDONLY))
1661 mode |= FMODE_WRITE;
1663 error = btrfs_parse_subvol_options(data, mode,
1664 &subvol_name, &subvol_objectid);
1667 return ERR_PTR(error);
1670 /* mount device's root (/) */
1671 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1672 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1673 if (flags & SB_RDONLY) {
1674 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1675 flags & ~SB_RDONLY, device_name, data);
1677 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1678 flags | SB_RDONLY, device_name, data);
1679 if (IS_ERR(mnt_root)) {
1680 root = ERR_CAST(mnt_root);
1684 down_write(&mnt_root->mnt_sb->s_umount);
1685 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1686 up_write(&mnt_root->mnt_sb->s_umount);
1688 root = ERR_PTR(error);
1694 if (IS_ERR(mnt_root)) {
1695 root = ERR_CAST(mnt_root);
1699 /* mount_subvol() will free subvol_name and mnt_root */
1700 root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1706 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1707 u32 new_pool_size, u32 old_pool_size)
1709 if (new_pool_size == old_pool_size)
1712 fs_info->thread_pool_size = new_pool_size;
1714 btrfs_info(fs_info, "resize thread pool %d -> %d",
1715 old_pool_size, new_pool_size);
1717 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1718 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1719 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1720 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1721 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1722 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1723 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1725 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1726 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1727 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1728 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1729 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1733 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1735 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1738 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1739 unsigned long old_opts, int flags)
1741 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1742 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1743 (flags & SB_RDONLY))) {
1744 /* wait for any defraggers to finish */
1745 wait_event(fs_info->transaction_wait,
1746 (atomic_read(&fs_info->defrag_running) == 0));
1747 if (flags & SB_RDONLY)
1748 sync_filesystem(fs_info->sb);
1752 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1753 unsigned long old_opts)
1756 * We need to cleanup all defragable inodes if the autodefragment is
1757 * close or the filesystem is read only.
1759 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1760 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1761 btrfs_cleanup_defrag_inodes(fs_info);
1764 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1767 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1769 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1770 struct btrfs_root *root = fs_info->tree_root;
1771 unsigned old_flags = sb->s_flags;
1772 unsigned long old_opts = fs_info->mount_opt;
1773 unsigned long old_compress_type = fs_info->compress_type;
1774 u64 old_max_inline = fs_info->max_inline;
1775 u32 old_thread_pool_size = fs_info->thread_pool_size;
1776 u32 old_metadata_ratio = fs_info->metadata_ratio;
1779 sync_filesystem(sb);
1780 btrfs_remount_prepare(fs_info);
1783 struct security_mnt_opts new_sec_opts;
1785 security_init_mnt_opts(&new_sec_opts);
1786 ret = parse_security_options(data, &new_sec_opts);
1789 ret = setup_security_options(fs_info, sb,
1792 security_free_mnt_opts(&new_sec_opts);
1797 ret = btrfs_parse_options(fs_info, data, *flags);
1803 btrfs_remount_begin(fs_info, old_opts, *flags);
1804 btrfs_resize_thread_pool(fs_info,
1805 fs_info->thread_pool_size, old_thread_pool_size);
1807 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1810 if (*flags & SB_RDONLY) {
1812 * this also happens on 'umount -rf' or on shutdown, when
1813 * the filesystem is busy.
1815 cancel_work_sync(&fs_info->async_reclaim_work);
1817 /* wait for the uuid_scan task to finish */
1818 down(&fs_info->uuid_tree_rescan_sem);
1819 /* avoid complains from lockdep et al. */
1820 up(&fs_info->uuid_tree_rescan_sem);
1822 sb->s_flags |= SB_RDONLY;
1825 * Setting SB_RDONLY will put the cleaner thread to
1826 * sleep at the next loop if it's already active.
1827 * If it's already asleep, we'll leave unused block
1828 * groups on disk until we're mounted read-write again
1829 * unless we clean them up here.
1831 btrfs_delete_unused_bgs(fs_info);
1833 btrfs_dev_replace_suspend_for_unmount(fs_info);
1834 btrfs_scrub_cancel(fs_info);
1835 btrfs_pause_balance(fs_info);
1837 ret = btrfs_commit_super(fs_info);
1841 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1843 "Remounting read-write after error is not allowed");
1847 if (fs_info->fs_devices->rw_devices == 0) {
1852 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1854 "too many missing devices, writeable remount is not allowed");
1859 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1864 ret = btrfs_cleanup_fs_roots(fs_info);
1868 /* recover relocation */
1869 mutex_lock(&fs_info->cleaner_mutex);
1870 ret = btrfs_recover_relocation(root);
1871 mutex_unlock(&fs_info->cleaner_mutex);
1875 ret = btrfs_resume_balance_async(fs_info);
1879 ret = btrfs_resume_dev_replace_async(fs_info);
1881 btrfs_warn(fs_info, "failed to resume dev_replace");
1885 btrfs_qgroup_rescan_resume(fs_info);
1887 if (!fs_info->uuid_root) {
1888 btrfs_info(fs_info, "creating UUID tree");
1889 ret = btrfs_create_uuid_tree(fs_info);
1892 "failed to create the UUID tree %d",
1897 sb->s_flags &= ~SB_RDONLY;
1899 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1902 wake_up_process(fs_info->transaction_kthread);
1903 btrfs_remount_cleanup(fs_info, old_opts);
1907 /* We've hit an error - don't reset SB_RDONLY */
1909 old_flags |= SB_RDONLY;
1910 sb->s_flags = old_flags;
1911 fs_info->mount_opt = old_opts;
1912 fs_info->compress_type = old_compress_type;
1913 fs_info->max_inline = old_max_inline;
1914 btrfs_resize_thread_pool(fs_info,
1915 old_thread_pool_size, fs_info->thread_pool_size);
1916 fs_info->metadata_ratio = old_metadata_ratio;
1917 btrfs_remount_cleanup(fs_info, old_opts);
1921 /* Used to sort the devices by max_avail(descending sort) */
1922 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1923 const void *dev_info2)
1925 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1926 ((struct btrfs_device_info *)dev_info2)->max_avail)
1928 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1929 ((struct btrfs_device_info *)dev_info2)->max_avail)
1936 * sort the devices by max_avail, in which max free extent size of each device
1937 * is stored.(Descending Sort)
1939 static inline void btrfs_descending_sort_devices(
1940 struct btrfs_device_info *devices,
1943 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1944 btrfs_cmp_device_free_bytes, NULL);
1948 * The helper to calc the free space on the devices that can be used to store
1951 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1954 struct btrfs_device_info *devices_info;
1955 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1956 struct btrfs_device *device;
1960 u64 min_stripe_size;
1961 int min_stripes = 1, num_stripes = 1;
1962 int i = 0, nr_devices;
1965 * We aren't under the device list lock, so this is racy-ish, but good
1966 * enough for our purposes.
1968 nr_devices = fs_info->fs_devices->open_devices;
1971 nr_devices = fs_info->fs_devices->open_devices;
1979 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1984 /* calc min stripe number for data space allocation */
1985 type = btrfs_data_alloc_profile(fs_info);
1986 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1988 num_stripes = nr_devices;
1989 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1992 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1997 if (type & BTRFS_BLOCK_GROUP_DUP)
1998 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
2000 min_stripe_size = BTRFS_STRIPE_LEN;
2003 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2004 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2005 &device->dev_state) ||
2007 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2010 if (i >= nr_devices)
2013 avail_space = device->total_bytes - device->bytes_used;
2015 /* align with stripe_len */
2016 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
2017 avail_space *= BTRFS_STRIPE_LEN;
2020 * In order to avoid overwriting the superblock on the drive,
2021 * btrfs starts at an offset of at least 1MB when doing chunk
2027 * we can use the free space in [0, skip_space - 1], subtract
2028 * it from the total.
2030 if (avail_space && avail_space >= skip_space)
2031 avail_space -= skip_space;
2035 if (avail_space < min_stripe_size)
2038 devices_info[i].dev = device;
2039 devices_info[i].max_avail = avail_space;
2047 btrfs_descending_sort_devices(devices_info, nr_devices);
2051 while (nr_devices >= min_stripes) {
2052 if (num_stripes > nr_devices)
2053 num_stripes = nr_devices;
2055 if (devices_info[i].max_avail >= min_stripe_size) {
2059 avail_space += devices_info[i].max_avail * num_stripes;
2060 alloc_size = devices_info[i].max_avail;
2061 for (j = i + 1 - num_stripes; j <= i; j++)
2062 devices_info[j].max_avail -= alloc_size;
2068 kfree(devices_info);
2069 *free_bytes = avail_space;
2074 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2076 * If there's a redundant raid level at DATA block groups, use the respective
2077 * multiplier to scale the sizes.
2079 * Unused device space usage is based on simulating the chunk allocator
2080 * algorithm that respects the device sizes and order of allocations. This is
2081 * a close approximation of the actual use but there are other factors that may
2082 * change the result (like a new metadata chunk).
2084 * If metadata is exhausted, f_bavail will be 0.
2086 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2088 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2089 struct btrfs_super_block *disk_super = fs_info->super_copy;
2090 struct list_head *head = &fs_info->space_info;
2091 struct btrfs_space_info *found;
2093 u64 total_free_data = 0;
2094 u64 total_free_meta = 0;
2095 int bits = dentry->d_sb->s_blocksize_bits;
2096 __be32 *fsid = (__be32 *)fs_info->fsid;
2097 unsigned factor = 1;
2098 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
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, &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
2175 if (!mixed && total_free_meta - thresh < block_rsv->size)
2178 buf->f_type = BTRFS_SUPER_MAGIC;
2179 buf->f_bsize = dentry->d_sb->s_blocksize;
2180 buf->f_namelen = BTRFS_NAME_LEN;
2182 /* We treat it as constant endianness (it doesn't matter _which_)
2183 because we want the fsid to come out the same whether mounted
2184 on a big-endian or little-endian host */
2185 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2186 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2187 /* Mask in the root object ID too, to disambiguate subvols */
2188 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2189 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2194 static void btrfs_kill_super(struct super_block *sb)
2196 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2197 kill_anon_super(sb);
2198 free_fs_info(fs_info);
2201 static struct file_system_type btrfs_fs_type = {
2202 .owner = THIS_MODULE,
2204 .mount = btrfs_mount,
2205 .kill_sb = btrfs_kill_super,
2206 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2209 static struct file_system_type btrfs_root_fs_type = {
2210 .owner = THIS_MODULE,
2212 .mount = btrfs_mount_root,
2213 .kill_sb = btrfs_kill_super,
2214 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2217 MODULE_ALIAS_FS("btrfs");
2219 static int btrfs_control_open(struct inode *inode, struct file *file)
2222 * The control file's private_data is used to hold the
2223 * transaction when it is started and is used to keep
2224 * track of whether a transaction is already in progress.
2226 file->private_data = NULL;
2231 * used by btrfsctl to scan devices when no FS is mounted
2233 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2236 struct btrfs_ioctl_vol_args *vol;
2237 struct btrfs_fs_devices *fs_devices;
2240 if (!capable(CAP_SYS_ADMIN))
2243 vol = memdup_user((void __user *)arg, sizeof(*vol));
2245 return PTR_ERR(vol);
2248 case BTRFS_IOC_SCAN_DEV:
2249 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2250 &btrfs_root_fs_type, &fs_devices);
2252 case BTRFS_IOC_DEVICES_READY:
2253 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2254 &btrfs_root_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_fs_info *fs_info = btrfs_sb(sb);
2272 struct btrfs_root *root = fs_info->tree_root;
2274 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2276 * We don't need a barrier here, we'll wait for any transaction that
2277 * could be in progress on other threads (and do delayed iputs that
2278 * we want to avoid on a frozen filesystem), or do the commit
2281 trans = btrfs_attach_transaction_barrier(root);
2282 if (IS_ERR(trans)) {
2283 /* no transaction, don't bother */
2284 if (PTR_ERR(trans) == -ENOENT)
2286 return PTR_ERR(trans);
2288 return btrfs_commit_transaction(trans);
2291 static int btrfs_unfreeze(struct super_block *sb)
2293 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2295 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2299 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2301 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2302 struct btrfs_fs_devices *cur_devices;
2303 struct btrfs_device *dev, *first_dev = NULL;
2304 struct list_head *head;
2305 struct rcu_string *name;
2307 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2308 cur_devices = fs_info->fs_devices;
2309 while (cur_devices) {
2310 head = &cur_devices->devices;
2311 list_for_each_entry(dev, head, dev_list) {
2312 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2316 if (!first_dev || dev->devid < first_dev->devid)
2319 cur_devices = cur_devices->seed;
2324 name = rcu_dereference(first_dev->name);
2325 seq_escape(m, name->str, " \t\n\\");
2330 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2334 static const struct super_operations btrfs_super_ops = {
2335 .drop_inode = btrfs_drop_inode,
2336 .evict_inode = btrfs_evict_inode,
2337 .put_super = btrfs_put_super,
2338 .sync_fs = btrfs_sync_fs,
2339 .show_options = btrfs_show_options,
2340 .show_devname = btrfs_show_devname,
2341 .write_inode = btrfs_write_inode,
2342 .alloc_inode = btrfs_alloc_inode,
2343 .destroy_inode = btrfs_destroy_inode,
2344 .statfs = btrfs_statfs,
2345 .remount_fs = btrfs_remount,
2346 .freeze_fs = btrfs_freeze,
2347 .unfreeze_fs = btrfs_unfreeze,
2350 static const struct file_operations btrfs_ctl_fops = {
2351 .open = btrfs_control_open,
2352 .unlocked_ioctl = btrfs_control_ioctl,
2353 .compat_ioctl = btrfs_control_ioctl,
2354 .owner = THIS_MODULE,
2355 .llseek = noop_llseek,
2358 static struct miscdevice btrfs_misc = {
2359 .minor = BTRFS_MINOR,
2360 .name = "btrfs-control",
2361 .fops = &btrfs_ctl_fops
2364 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2365 MODULE_ALIAS("devname:btrfs-control");
2367 static int __init btrfs_interface_init(void)
2369 return misc_register(&btrfs_misc);
2372 static __cold void btrfs_interface_exit(void)
2374 misc_deregister(&btrfs_misc);
2377 static void __init btrfs_print_mod_info(void)
2379 pr_info("Btrfs loaded, crc32c=%s"
2380 #ifdef CONFIG_BTRFS_DEBUG
2383 #ifdef CONFIG_BTRFS_ASSERT
2386 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2387 ", integrity-checker=on"
2389 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2396 static int __init init_btrfs_fs(void)
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();
2489 static void __exit exit_btrfs_fs(void)
2491 btrfs_destroy_cachep();
2492 btrfs_delayed_ref_exit();
2493 btrfs_auto_defrag_exit();
2494 btrfs_delayed_inode_exit();
2495 btrfs_prelim_ref_exit();
2496 ordered_data_exit();
2499 btrfs_interface_exit();
2500 btrfs_end_io_wq_exit();
2501 unregister_filesystem(&btrfs_fs_type);
2503 btrfs_cleanup_fs_uuids();
2504 btrfs_exit_compress();
2507 late_initcall(init_btrfs_fs);
2508 module_exit(exit_btrfs_fs)
2510 MODULE_LICENSE("GPL");