1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/blkdev.h>
7 #include <linux/module.h>
9 #include <linux/pagemap.h>
10 #include <linux/highmem.h>
11 #include <linux/time.h>
12 #include <linux/init.h>
13 #include <linux/seq_file.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/writeback.h>
18 #include <linux/statfs.h>
19 #include <linux/compat.h>
20 #include <linux/parser.h>
21 #include <linux/ctype.h>
22 #include <linux/namei.h>
23 #include <linux/miscdevice.h>
24 #include <linux/magic.h>
25 #include <linux/slab.h>
26 #include <linux/cleancache.h>
27 #include <linux/ratelimit.h>
28 #include <linux/crc32c.h>
29 #include <linux/btrfs.h>
30 #include "delayed-inode.h"
33 #include "transaction.h"
34 #include "btrfs_inode.h"
35 #include "print-tree.h"
40 #include "compression.h"
41 #include "rcu-string.h"
42 #include "dev-replace.h"
43 #include "free-space-cache.h"
45 #include "tests/btrfs-tests.h"
48 #define CREATE_TRACE_POINTS
49 #include <trace/events/btrfs.h>
51 static const struct super_operations btrfs_super_ops;
54 * Types for mounting the default subvolume and a subvolume explicitly
55 * requested by subvol=/path. That way the callchain is straightforward and we
56 * don't have to play tricks with the mount options and recursive calls to
59 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
61 static struct file_system_type btrfs_fs_type;
62 static struct file_system_type btrfs_root_fs_type;
64 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
66 const char *btrfs_decode_error(int errno)
68 char *errstr = "unknown";
72 errstr = "IO failure";
75 errstr = "Out of memory";
78 errstr = "Readonly filesystem";
81 errstr = "Object already exists";
84 errstr = "No space left";
87 errstr = "No such entry";
95 * __btrfs_handle_fs_error decodes expected errors from the caller and
96 * invokes the appropriate error response.
99 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
100 unsigned int line, int errno, const char *fmt, ...)
102 struct super_block *sb = fs_info->sb;
108 * Special case: if the error is EROFS, and we're already
109 * under SB_RDONLY, then it is safe here.
111 if (errno == -EROFS && sb_rdonly(sb))
115 errstr = btrfs_decode_error(errno);
117 struct va_format vaf;
124 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
125 sb->s_id, function, line, errno, errstr, &vaf);
128 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
129 sb->s_id, function, line, errno, errstr);
134 * Today we only save the error info to memory. Long term we'll
135 * also send it down to the disk
137 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
139 /* Don't go through full error handling during mount */
140 if (!(sb->s_flags & SB_BORN))
146 /* btrfs handle error by forcing the filesystem readonly */
147 sb->s_flags |= SB_RDONLY;
148 btrfs_info(fs_info, "forced readonly");
150 * Note that a running device replace operation is not canceled here
151 * although there is no way to update the progress. It would add the
152 * risk of a deadlock, therefore the canceling is omitted. The only
153 * penalty is that some I/O remains active until the procedure
154 * completes. The next time when the filesystem is mounted writable
155 * again, the device replace operation continues.
160 static const char * const logtypes[] = {
173 * Use one ratelimit state per log level so that a flood of less important
174 * messages doesn't cause more important ones to be dropped.
176 static struct ratelimit_state printk_limits[] = {
177 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
178 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
179 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
180 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
181 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
182 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
183 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
184 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
187 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
189 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
190 struct va_format vaf;
193 const char *type = logtypes[4];
194 struct ratelimit_state *ratelimit = &printk_limits[4];
198 while ((kern_level = printk_get_level(fmt)) != 0) {
199 size_t size = printk_skip_level(fmt) - fmt;
201 if (kern_level >= '0' && kern_level <= '7') {
202 memcpy(lvl, fmt, size);
204 type = logtypes[kern_level - '0'];
205 ratelimit = &printk_limits[kern_level - '0'];
213 if (__ratelimit(ratelimit))
214 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
215 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
222 * We only mark the transaction aborted and then set the file system read-only.
223 * This will prevent new transactions from starting or trying to join this
226 * This means that error recovery at the call site is limited to freeing
227 * any local memory allocations and passing the error code up without
228 * further cleanup. The transaction should complete as it normally would
229 * in the call path but will return -EIO.
231 * We'll complete the cleanup in btrfs_end_transaction and
232 * btrfs_commit_transaction.
235 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
236 const char *function,
237 unsigned int line, int errno)
239 struct btrfs_fs_info *fs_info = trans->fs_info;
241 trans->aborted = errno;
242 /* Nothing used. The other threads that have joined this
243 * transaction may be able to continue. */
244 if (!trans->dirty && list_empty(&trans->new_bgs)) {
247 errstr = btrfs_decode_error(errno);
249 "%s:%d: Aborting unused transaction(%s).",
250 function, line, errstr);
253 WRITE_ONCE(trans->transaction->aborted, errno);
254 /* Wake up anybody who may be waiting on this transaction */
255 wake_up(&fs_info->transaction_wait);
256 wake_up(&fs_info->transaction_blocked_wait);
257 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
260 * __btrfs_panic decodes unexpected, fatal errors from the caller,
261 * issues an alert, and either panics or BUGs, depending on mount options.
264 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
265 unsigned int line, int errno, const char *fmt, ...)
267 char *s_id = "<unknown>";
269 struct va_format vaf = { .fmt = fmt };
273 s_id = fs_info->sb->s_id;
278 errstr = btrfs_decode_error(errno);
279 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
280 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
281 s_id, function, line, &vaf, errno, errstr);
283 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
284 function, line, &vaf, errno, errstr);
286 /* Caller calls BUG() */
289 static void btrfs_put_super(struct super_block *sb)
291 close_ctree(btrfs_sb(sb));
300 Opt_compress_force_type,
305 Opt_flushoncommit, Opt_noflushoncommit,
306 Opt_inode_cache, Opt_noinode_cache,
308 Opt_barrier, Opt_nobarrier,
309 Opt_datacow, Opt_nodatacow,
310 Opt_datasum, Opt_nodatasum,
311 Opt_defrag, Opt_nodefrag,
312 Opt_discard, Opt_nodiscard,
316 Opt_rescan_uuid_tree,
318 Opt_space_cache, Opt_no_space_cache,
319 Opt_space_cache_version,
321 Opt_ssd_spread, Opt_nossd_spread,
326 Opt_treelog, Opt_notreelog,
328 Opt_user_subvol_rm_allowed,
330 /* Deprecated options */
335 /* Debugging options */
337 Opt_check_integrity_including_extent_data,
338 Opt_check_integrity_print_mask,
339 Opt_enospc_debug, Opt_noenospc_debug,
340 #ifdef CONFIG_BTRFS_DEBUG
341 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
343 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
349 static const match_table_t tokens = {
351 {Opt_noacl, "noacl"},
352 {Opt_clear_cache, "clear_cache"},
353 {Opt_commit_interval, "commit=%u"},
354 {Opt_compress, "compress"},
355 {Opt_compress_type, "compress=%s"},
356 {Opt_compress_force, "compress-force"},
357 {Opt_compress_force_type, "compress-force=%s"},
358 {Opt_degraded, "degraded"},
359 {Opt_device, "device=%s"},
360 {Opt_fatal_errors, "fatal_errors=%s"},
361 {Opt_flushoncommit, "flushoncommit"},
362 {Opt_noflushoncommit, "noflushoncommit"},
363 {Opt_inode_cache, "inode_cache"},
364 {Opt_noinode_cache, "noinode_cache"},
365 {Opt_max_inline, "max_inline=%s"},
366 {Opt_barrier, "barrier"},
367 {Opt_nobarrier, "nobarrier"},
368 {Opt_datacow, "datacow"},
369 {Opt_nodatacow, "nodatacow"},
370 {Opt_datasum, "datasum"},
371 {Opt_nodatasum, "nodatasum"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_discard, "discard"},
375 {Opt_nodiscard, "nodiscard"},
376 {Opt_nologreplay, "nologreplay"},
377 {Opt_norecovery, "norecovery"},
378 {Opt_ratio, "metadata_ratio=%u"},
379 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
380 {Opt_skip_balance, "skip_balance"},
381 {Opt_space_cache, "space_cache"},
382 {Opt_no_space_cache, "nospace_cache"},
383 {Opt_space_cache_version, "space_cache=%s"},
385 {Opt_nossd, "nossd"},
386 {Opt_ssd_spread, "ssd_spread"},
387 {Opt_nossd_spread, "nossd_spread"},
388 {Opt_subvol, "subvol=%s"},
389 {Opt_subvol_empty, "subvol="},
390 {Opt_subvolid, "subvolid=%s"},
391 {Opt_thread_pool, "thread_pool=%u"},
392 {Opt_treelog, "treelog"},
393 {Opt_notreelog, "notreelog"},
394 {Opt_usebackuproot, "usebackuproot"},
395 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
397 /* Deprecated options */
398 {Opt_alloc_start, "alloc_start=%s"},
399 {Opt_recovery, "recovery"},
400 {Opt_subvolrootid, "subvolrootid=%d"},
402 /* Debugging options */
403 {Opt_check_integrity, "check_int"},
404 {Opt_check_integrity_including_extent_data, "check_int_data"},
405 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
406 {Opt_enospc_debug, "enospc_debug"},
407 {Opt_noenospc_debug, "noenospc_debug"},
408 #ifdef CONFIG_BTRFS_DEBUG
409 {Opt_fragment_data, "fragment=data"},
410 {Opt_fragment_metadata, "fragment=metadata"},
411 {Opt_fragment_all, "fragment=all"},
413 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
414 {Opt_ref_verify, "ref_verify"},
420 * Regular mount options parser. Everything that is needed only when
421 * reading in a new superblock is parsed here.
422 * XXX JDM: This needs to be cleaned up for remount.
424 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
425 unsigned long new_flags)
427 substring_t args[MAX_OPT_ARGS];
433 bool compress_force = false;
434 enum btrfs_compression_type saved_compress_type;
435 bool saved_compress_force;
438 cache_gen = btrfs_super_cache_generation(info->super_copy);
439 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
440 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
442 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
445 * Even the options are empty, we still need to do extra check
451 while ((p = strsep(&options, ",")) != NULL) {
456 token = match_token(p, tokens, args);
459 btrfs_info(info, "allowing degraded mounts");
460 btrfs_set_opt(info->mount_opt, DEGRADED);
463 case Opt_subvol_empty:
465 case Opt_subvolrootid:
468 * These are parsed by btrfs_parse_subvol_options or
469 * btrfs_parse_device_options and can be ignored here.
473 btrfs_set_and_info(info, NODATASUM,
474 "setting nodatasum");
477 if (btrfs_test_opt(info, NODATASUM)) {
478 if (btrfs_test_opt(info, NODATACOW))
480 "setting datasum, datacow enabled");
482 btrfs_info(info, "setting datasum");
484 btrfs_clear_opt(info->mount_opt, NODATACOW);
485 btrfs_clear_opt(info->mount_opt, NODATASUM);
488 if (!btrfs_test_opt(info, NODATACOW)) {
489 if (!btrfs_test_opt(info, COMPRESS) ||
490 !btrfs_test_opt(info, FORCE_COMPRESS)) {
492 "setting nodatacow, compression disabled");
494 btrfs_info(info, "setting nodatacow");
497 btrfs_clear_opt(info->mount_opt, COMPRESS);
498 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
499 btrfs_set_opt(info->mount_opt, NODATACOW);
500 btrfs_set_opt(info->mount_opt, NODATASUM);
503 btrfs_clear_and_info(info, NODATACOW,
506 case Opt_compress_force:
507 case Opt_compress_force_type:
508 compress_force = true;
511 case Opt_compress_type:
512 saved_compress_type = btrfs_test_opt(info,
514 info->compress_type : BTRFS_COMPRESS_NONE;
515 saved_compress_force =
516 btrfs_test_opt(info, FORCE_COMPRESS);
517 if (token == Opt_compress ||
518 token == Opt_compress_force ||
519 strncmp(args[0].from, "zlib", 4) == 0) {
520 compress_type = "zlib";
522 info->compress_type = BTRFS_COMPRESS_ZLIB;
523 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
525 * args[0] contains uninitialized data since
526 * for these tokens we don't expect any
529 if (token != Opt_compress &&
530 token != Opt_compress_force)
531 info->compress_level =
532 btrfs_compress_str2level(
535 btrfs_set_opt(info->mount_opt, COMPRESS);
536 btrfs_clear_opt(info->mount_opt, NODATACOW);
537 btrfs_clear_opt(info->mount_opt, NODATASUM);
539 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
540 compress_type = "lzo";
541 info->compress_type = BTRFS_COMPRESS_LZO;
542 btrfs_set_opt(info->mount_opt, COMPRESS);
543 btrfs_clear_opt(info->mount_opt, NODATACOW);
544 btrfs_clear_opt(info->mount_opt, NODATASUM);
545 btrfs_set_fs_incompat(info, COMPRESS_LZO);
547 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
548 compress_type = "zstd";
549 info->compress_type = BTRFS_COMPRESS_ZSTD;
550 info->compress_level =
551 btrfs_compress_str2level(
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_ZSTD);
559 } else if (strncmp(args[0].from, "no", 2) == 0) {
560 compress_type = "no";
561 btrfs_clear_opt(info->mount_opt, COMPRESS);
562 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
563 compress_force = false;
570 if (compress_force) {
571 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
574 * If we remount from compress-force=xxx to
575 * compress=xxx, we need clear FORCE_COMPRESS
576 * flag, otherwise, there is no way for users
577 * to disable forcible compression separately.
579 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
581 if ((btrfs_test_opt(info, COMPRESS) &&
582 (info->compress_type != saved_compress_type ||
583 compress_force != saved_compress_force)) ||
584 (!btrfs_test_opt(info, COMPRESS) &&
586 btrfs_info(info, "%s %s compression, level %d",
587 (compress_force) ? "force" : "use",
588 compress_type, info->compress_level);
590 compress_force = false;
593 btrfs_set_and_info(info, SSD,
594 "enabling ssd optimizations");
595 btrfs_clear_opt(info->mount_opt, NOSSD);
598 btrfs_set_and_info(info, SSD,
599 "enabling ssd optimizations");
600 btrfs_set_and_info(info, SSD_SPREAD,
601 "using spread ssd allocation scheme");
602 btrfs_clear_opt(info->mount_opt, NOSSD);
605 btrfs_set_opt(info->mount_opt, NOSSD);
606 btrfs_clear_and_info(info, SSD,
607 "not using ssd optimizations");
609 case Opt_nossd_spread:
610 btrfs_clear_and_info(info, SSD_SPREAD,
611 "not using spread ssd allocation scheme");
614 btrfs_clear_and_info(info, NOBARRIER,
615 "turning on barriers");
618 btrfs_set_and_info(info, NOBARRIER,
619 "turning off barriers");
621 case Opt_thread_pool:
622 ret = match_int(&args[0], &intarg);
625 } else if (intarg == 0) {
629 info->thread_pool_size = intarg;
632 num = match_strdup(&args[0]);
634 info->max_inline = memparse(num, NULL);
637 if (info->max_inline) {
638 info->max_inline = min_t(u64,
642 btrfs_info(info, "max_inline at %llu",
649 case Opt_alloc_start:
651 "option alloc_start is obsolete, ignored");
654 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
655 info->sb->s_flags |= SB_POSIXACL;
658 btrfs_err(info, "support for ACL not compiled in!");
663 info->sb->s_flags &= ~SB_POSIXACL;
666 btrfs_set_and_info(info, NOTREELOG,
667 "disabling tree log");
670 btrfs_clear_and_info(info, NOTREELOG,
671 "enabling tree log");
674 case Opt_nologreplay:
675 btrfs_set_and_info(info, NOLOGREPLAY,
676 "disabling log replay at mount time");
678 case Opt_flushoncommit:
679 btrfs_set_and_info(info, FLUSHONCOMMIT,
680 "turning on flush-on-commit");
682 case Opt_noflushoncommit:
683 btrfs_clear_and_info(info, FLUSHONCOMMIT,
684 "turning off flush-on-commit");
687 ret = match_int(&args[0], &intarg);
690 info->metadata_ratio = intarg;
691 btrfs_info(info, "metadata ratio %u",
692 info->metadata_ratio);
695 btrfs_set_and_info(info, DISCARD,
696 "turning on discard");
699 btrfs_clear_and_info(info, DISCARD,
700 "turning off discard");
702 case Opt_space_cache:
703 case Opt_space_cache_version:
704 if (token == Opt_space_cache ||
705 strcmp(args[0].from, "v1") == 0) {
706 btrfs_clear_opt(info->mount_opt,
708 btrfs_set_and_info(info, SPACE_CACHE,
709 "enabling disk space caching");
710 } else if (strcmp(args[0].from, "v2") == 0) {
711 btrfs_clear_opt(info->mount_opt,
713 btrfs_set_and_info(info, FREE_SPACE_TREE,
714 "enabling free space tree");
720 case Opt_rescan_uuid_tree:
721 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
723 case Opt_no_space_cache:
724 if (btrfs_test_opt(info, SPACE_CACHE)) {
725 btrfs_clear_and_info(info, SPACE_CACHE,
726 "disabling disk space caching");
728 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
729 btrfs_clear_and_info(info, FREE_SPACE_TREE,
730 "disabling free space tree");
733 case Opt_inode_cache:
734 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
735 "enabling inode map caching");
737 case Opt_noinode_cache:
738 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
739 "disabling inode map caching");
741 case Opt_clear_cache:
742 btrfs_set_and_info(info, CLEAR_CACHE,
743 "force clearing of disk cache");
745 case Opt_user_subvol_rm_allowed:
746 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
748 case Opt_enospc_debug:
749 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
751 case Opt_noenospc_debug:
752 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
755 btrfs_set_and_info(info, AUTO_DEFRAG,
756 "enabling auto defrag");
759 btrfs_clear_and_info(info, AUTO_DEFRAG,
760 "disabling auto defrag");
764 "'recovery' is deprecated, use 'usebackuproot' instead");
766 case Opt_usebackuproot:
768 "trying to use backup root at mount time");
769 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
771 case Opt_skip_balance:
772 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
774 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
775 case Opt_check_integrity_including_extent_data:
777 "enabling check integrity including extent data");
778 btrfs_set_opt(info->mount_opt,
779 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
780 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
782 case Opt_check_integrity:
783 btrfs_info(info, "enabling check integrity");
784 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
786 case Opt_check_integrity_print_mask:
787 ret = match_int(&args[0], &intarg);
790 info->check_integrity_print_mask = intarg;
791 btrfs_info(info, "check_integrity_print_mask 0x%x",
792 info->check_integrity_print_mask);
795 case Opt_check_integrity_including_extent_data:
796 case Opt_check_integrity:
797 case Opt_check_integrity_print_mask:
799 "support for check_integrity* not compiled in!");
803 case Opt_fatal_errors:
804 if (strcmp(args[0].from, "panic") == 0)
805 btrfs_set_opt(info->mount_opt,
806 PANIC_ON_FATAL_ERROR);
807 else if (strcmp(args[0].from, "bug") == 0)
808 btrfs_clear_opt(info->mount_opt,
809 PANIC_ON_FATAL_ERROR);
815 case Opt_commit_interval:
817 ret = match_int(&args[0], &intarg);
822 "using default commit interval %us",
823 BTRFS_DEFAULT_COMMIT_INTERVAL);
824 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
825 } else if (intarg > 300) {
826 btrfs_warn(info, "excessive commit interval %d",
829 info->commit_interval = intarg;
831 #ifdef CONFIG_BTRFS_DEBUG
832 case Opt_fragment_all:
833 btrfs_info(info, "fragmenting all space");
834 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
835 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
837 case Opt_fragment_metadata:
838 btrfs_info(info, "fragmenting metadata");
839 btrfs_set_opt(info->mount_opt,
842 case Opt_fragment_data:
843 btrfs_info(info, "fragmenting data");
844 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
847 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
849 btrfs_info(info, "doing ref verification");
850 btrfs_set_opt(info->mount_opt, REF_VERIFY);
854 btrfs_info(info, "unrecognized mount option '%s'", p);
863 * Extra check for current option against current flag
865 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
867 "nologreplay must be used with ro mount option");
871 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
872 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
873 !btrfs_test_opt(info, CLEAR_CACHE)) {
874 btrfs_err(info, "cannot disable free space tree");
878 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
879 btrfs_info(info, "disk space caching is enabled");
880 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
881 btrfs_info(info, "using free space tree");
886 * Parse mount options that are required early in the mount process.
888 * All other options will be parsed on much later in the mount process and
889 * only when we need to allocate a new super block.
891 static int btrfs_parse_device_options(const char *options, fmode_t flags,
894 substring_t args[MAX_OPT_ARGS];
895 char *device_name, *opts, *orig, *p;
896 struct btrfs_device *device = NULL;
899 lockdep_assert_held(&uuid_mutex);
905 * strsep changes the string, duplicate it because btrfs_parse_options
908 opts = kstrdup(options, GFP_KERNEL);
913 while ((p = strsep(&opts, ",")) != NULL) {
919 token = match_token(p, tokens, args);
920 if (token == Opt_device) {
921 device_name = match_strdup(&args[0]);
926 device = btrfs_scan_one_device(device_name, flags,
929 if (IS_ERR(device)) {
930 error = PTR_ERR(device);
942 * Parse mount options that are related to subvolume id
944 * The value is later passed to mount_subvol()
946 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
947 u64 *subvol_objectid)
949 substring_t args[MAX_OPT_ARGS];
950 char *opts, *orig, *p;
958 * strsep changes the string, duplicate it because
959 * btrfs_parse_device_options gets called later
961 opts = kstrdup(options, GFP_KERNEL);
966 while ((p = strsep(&opts, ",")) != NULL) {
971 token = match_token(p, tokens, args);
975 *subvol_name = match_strdup(&args[0]);
982 error = match_u64(&args[0], &subvolid);
986 /* we want the original fs_tree */
988 subvolid = BTRFS_FS_TREE_OBJECTID;
990 *subvol_objectid = subvolid;
992 case Opt_subvolrootid:
993 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
1005 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1006 u64 subvol_objectid)
1008 struct btrfs_root *root = fs_info->tree_root;
1009 struct btrfs_root *fs_root;
1010 struct btrfs_root_ref *root_ref;
1011 struct btrfs_inode_ref *inode_ref;
1012 struct btrfs_key key;
1013 struct btrfs_path *path = NULL;
1014 char *name = NULL, *ptr;
1019 path = btrfs_alloc_path();
1024 path->leave_spinning = 1;
1026 name = kmalloc(PATH_MAX, GFP_KERNEL);
1031 ptr = name + PATH_MAX - 1;
1035 * Walk up the subvolume trees in the tree of tree roots by root
1036 * backrefs until we hit the top-level subvolume.
1038 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1039 key.objectid = subvol_objectid;
1040 key.type = BTRFS_ROOT_BACKREF_KEY;
1041 key.offset = (u64)-1;
1043 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1046 } else if (ret > 0) {
1047 ret = btrfs_previous_item(root, path, subvol_objectid,
1048 BTRFS_ROOT_BACKREF_KEY);
1051 } else if (ret > 0) {
1057 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1058 subvol_objectid = key.offset;
1060 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1061 struct btrfs_root_ref);
1062 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1065 ret = -ENAMETOOLONG;
1068 read_extent_buffer(path->nodes[0], ptr + 1,
1069 (unsigned long)(root_ref + 1), len);
1071 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1072 btrfs_release_path(path);
1074 key.objectid = subvol_objectid;
1075 key.type = BTRFS_ROOT_ITEM_KEY;
1076 key.offset = (u64)-1;
1077 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1078 if (IS_ERR(fs_root)) {
1079 ret = PTR_ERR(fs_root);
1084 * Walk up the filesystem tree by inode refs until we hit the
1087 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1088 key.objectid = dirid;
1089 key.type = BTRFS_INODE_REF_KEY;
1090 key.offset = (u64)-1;
1092 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1095 } else if (ret > 0) {
1096 ret = btrfs_previous_item(fs_root, path, dirid,
1097 BTRFS_INODE_REF_KEY);
1100 } else if (ret > 0) {
1106 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1109 inode_ref = btrfs_item_ptr(path->nodes[0],
1111 struct btrfs_inode_ref);
1112 len = btrfs_inode_ref_name_len(path->nodes[0],
1116 ret = -ENAMETOOLONG;
1119 read_extent_buffer(path->nodes[0], ptr + 1,
1120 (unsigned long)(inode_ref + 1), len);
1122 btrfs_release_path(path);
1126 btrfs_free_path(path);
1127 if (ptr == name + PATH_MAX - 1) {
1131 memmove(name, ptr, name + PATH_MAX - ptr);
1136 btrfs_free_path(path);
1138 return ERR_PTR(ret);
1141 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1143 struct btrfs_root *root = fs_info->tree_root;
1144 struct btrfs_dir_item *di;
1145 struct btrfs_path *path;
1146 struct btrfs_key location;
1149 path = btrfs_alloc_path();
1152 path->leave_spinning = 1;
1155 * Find the "default" dir item which points to the root item that we
1156 * will mount by default if we haven't been given a specific subvolume
1159 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1160 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1162 btrfs_free_path(path);
1167 * Ok the default dir item isn't there. This is weird since
1168 * it's always been there, but don't freak out, just try and
1169 * mount the top-level subvolume.
1171 btrfs_free_path(path);
1172 *objectid = BTRFS_FS_TREE_OBJECTID;
1176 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1177 btrfs_free_path(path);
1178 *objectid = location.objectid;
1182 static int btrfs_fill_super(struct super_block *sb,
1183 struct btrfs_fs_devices *fs_devices,
1186 struct inode *inode;
1187 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1188 struct btrfs_key key;
1191 sb->s_maxbytes = MAX_LFS_FILESIZE;
1192 sb->s_magic = BTRFS_SUPER_MAGIC;
1193 sb->s_op = &btrfs_super_ops;
1194 sb->s_d_op = &btrfs_dentry_operations;
1195 sb->s_export_op = &btrfs_export_ops;
1196 sb->s_xattr = btrfs_xattr_handlers;
1197 sb->s_time_gran = 1;
1198 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1199 sb->s_flags |= SB_POSIXACL;
1201 sb->s_flags |= SB_I_VERSION;
1202 sb->s_iflags |= SB_I_CGROUPWB;
1204 err = super_setup_bdi(sb);
1206 btrfs_err(fs_info, "super_setup_bdi failed");
1210 err = open_ctree(sb, fs_devices, (char *)data);
1212 btrfs_err(fs_info, "open_ctree failed");
1216 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1217 key.type = BTRFS_INODE_ITEM_KEY;
1219 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1220 if (IS_ERR(inode)) {
1221 err = PTR_ERR(inode);
1225 sb->s_root = d_make_root(inode);
1231 cleancache_init_fs(sb);
1232 sb->s_flags |= SB_ACTIVE;
1236 close_ctree(fs_info);
1240 int btrfs_sync_fs(struct super_block *sb, int wait)
1242 struct btrfs_trans_handle *trans;
1243 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1244 struct btrfs_root *root = fs_info->tree_root;
1246 trace_btrfs_sync_fs(fs_info, wait);
1249 filemap_flush(fs_info->btree_inode->i_mapping);
1253 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1255 trans = btrfs_attach_transaction_barrier(root);
1256 if (IS_ERR(trans)) {
1257 /* no transaction, don't bother */
1258 if (PTR_ERR(trans) == -ENOENT) {
1260 * Exit unless we have some pending changes
1261 * that need to go through commit
1263 if (fs_info->pending_changes == 0)
1266 * A non-blocking test if the fs is frozen. We must not
1267 * start a new transaction here otherwise a deadlock
1268 * happens. The pending operations are delayed to the
1269 * next commit after thawing.
1271 if (sb_start_write_trylock(sb))
1275 trans = btrfs_start_transaction(root, 0);
1278 return PTR_ERR(trans);
1280 return btrfs_commit_transaction(trans);
1283 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1285 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1286 const char *compress_type;
1288 if (btrfs_test_opt(info, DEGRADED))
1289 seq_puts(seq, ",degraded");
1290 if (btrfs_test_opt(info, NODATASUM))
1291 seq_puts(seq, ",nodatasum");
1292 if (btrfs_test_opt(info, NODATACOW))
1293 seq_puts(seq, ",nodatacow");
1294 if (btrfs_test_opt(info, NOBARRIER))
1295 seq_puts(seq, ",nobarrier");
1296 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1297 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1298 if (info->thread_pool_size != min_t(unsigned long,
1299 num_online_cpus() + 2, 8))
1300 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1301 if (btrfs_test_opt(info, COMPRESS)) {
1302 compress_type = btrfs_compress_type2str(info->compress_type);
1303 if (btrfs_test_opt(info, FORCE_COMPRESS))
1304 seq_printf(seq, ",compress-force=%s", compress_type);
1306 seq_printf(seq, ",compress=%s", compress_type);
1307 if (info->compress_level)
1308 seq_printf(seq, ":%d", info->compress_level);
1310 if (btrfs_test_opt(info, NOSSD))
1311 seq_puts(seq, ",nossd");
1312 if (btrfs_test_opt(info, SSD_SPREAD))
1313 seq_puts(seq, ",ssd_spread");
1314 else if (btrfs_test_opt(info, SSD))
1315 seq_puts(seq, ",ssd");
1316 if (btrfs_test_opt(info, NOTREELOG))
1317 seq_puts(seq, ",notreelog");
1318 if (btrfs_test_opt(info, NOLOGREPLAY))
1319 seq_puts(seq, ",nologreplay");
1320 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1321 seq_puts(seq, ",flushoncommit");
1322 if (btrfs_test_opt(info, DISCARD))
1323 seq_puts(seq, ",discard");
1324 if (!(info->sb->s_flags & SB_POSIXACL))
1325 seq_puts(seq, ",noacl");
1326 if (btrfs_test_opt(info, SPACE_CACHE))
1327 seq_puts(seq, ",space_cache");
1328 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1329 seq_puts(seq, ",space_cache=v2");
1331 seq_puts(seq, ",nospace_cache");
1332 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1333 seq_puts(seq, ",rescan_uuid_tree");
1334 if (btrfs_test_opt(info, CLEAR_CACHE))
1335 seq_puts(seq, ",clear_cache");
1336 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1337 seq_puts(seq, ",user_subvol_rm_allowed");
1338 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1339 seq_puts(seq, ",enospc_debug");
1340 if (btrfs_test_opt(info, AUTO_DEFRAG))
1341 seq_puts(seq, ",autodefrag");
1342 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1343 seq_puts(seq, ",inode_cache");
1344 if (btrfs_test_opt(info, SKIP_BALANCE))
1345 seq_puts(seq, ",skip_balance");
1346 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1347 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1348 seq_puts(seq, ",check_int_data");
1349 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1350 seq_puts(seq, ",check_int");
1351 if (info->check_integrity_print_mask)
1352 seq_printf(seq, ",check_int_print_mask=%d",
1353 info->check_integrity_print_mask);
1355 if (info->metadata_ratio)
1356 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1357 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1358 seq_puts(seq, ",fatal_errors=panic");
1359 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1360 seq_printf(seq, ",commit=%u", info->commit_interval);
1361 #ifdef CONFIG_BTRFS_DEBUG
1362 if (btrfs_test_opt(info, FRAGMENT_DATA))
1363 seq_puts(seq, ",fragment=data");
1364 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1365 seq_puts(seq, ",fragment=metadata");
1367 if (btrfs_test_opt(info, REF_VERIFY))
1368 seq_puts(seq, ",ref_verify");
1369 seq_printf(seq, ",subvolid=%llu",
1370 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1371 seq_puts(seq, ",subvol=");
1372 seq_dentry(seq, dentry, " \t\n\\");
1376 static int btrfs_test_super(struct super_block *s, void *data)
1378 struct btrfs_fs_info *p = data;
1379 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1381 return fs_info->fs_devices == p->fs_devices;
1384 static int btrfs_set_super(struct super_block *s, void *data)
1386 int err = set_anon_super(s, data);
1388 s->s_fs_info = data;
1393 * subvolumes are identified by ino 256
1395 static inline int is_subvolume_inode(struct inode *inode)
1397 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1402 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1403 const char *device_name, struct vfsmount *mnt)
1405 struct dentry *root;
1409 if (!subvol_objectid) {
1410 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1413 root = ERR_PTR(ret);
1417 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1419 if (IS_ERR(subvol_name)) {
1420 root = ERR_CAST(subvol_name);
1427 root = mount_subtree(mnt, subvol_name);
1428 /* mount_subtree() drops our reference on the vfsmount. */
1431 if (!IS_ERR(root)) {
1432 struct super_block *s = root->d_sb;
1433 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1434 struct inode *root_inode = d_inode(root);
1435 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1438 if (!is_subvolume_inode(root_inode)) {
1439 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1443 if (subvol_objectid && root_objectid != subvol_objectid) {
1445 * This will also catch a race condition where a
1446 * subvolume which was passed by ID is renamed and
1447 * another subvolume is renamed over the old location.
1450 "subvol '%s' does not match subvolid %llu",
1451 subvol_name, subvol_objectid);
1456 root = ERR_PTR(ret);
1457 deactivate_locked_super(s);
1468 * Find a superblock for the given device / mount point.
1470 * Note: This is based on mount_bdev from fs/super.c with a few additions
1471 * for multiple device setup. Make sure to keep it in sync.
1473 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1474 int flags, const char *device_name, void *data)
1476 struct block_device *bdev = NULL;
1477 struct super_block *s;
1478 struct btrfs_device *device = NULL;
1479 struct btrfs_fs_devices *fs_devices = NULL;
1480 struct btrfs_fs_info *fs_info = NULL;
1481 void *new_sec_opts = NULL;
1482 fmode_t mode = FMODE_READ;
1485 if (!(flags & SB_RDONLY))
1486 mode |= FMODE_WRITE;
1489 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1491 return ERR_PTR(error);
1495 * Setup a dummy root and fs_info for test/set super. This is because
1496 * we don't actually fill this stuff out until open_ctree, but we need
1497 * it for searching for existing supers, so this lets us do that and
1498 * then open_ctree will properly initialize everything later.
1500 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1503 goto error_sec_opts;
1506 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1507 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1508 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1513 mutex_lock(&uuid_mutex);
1514 error = btrfs_parse_device_options(data, mode, fs_type);
1516 mutex_unlock(&uuid_mutex);
1520 device = btrfs_scan_one_device(device_name, mode, fs_type);
1521 if (IS_ERR(device)) {
1522 mutex_unlock(&uuid_mutex);
1523 error = PTR_ERR(device);
1527 fs_devices = device->fs_devices;
1528 fs_info->fs_devices = fs_devices;
1530 error = btrfs_open_devices(fs_devices, mode, fs_type);
1531 mutex_unlock(&uuid_mutex);
1535 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1537 goto error_close_devices;
1540 bdev = fs_devices->latest_bdev;
1541 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1545 goto error_close_devices;
1549 btrfs_close_devices(fs_devices);
1550 free_fs_info(fs_info);
1551 if ((flags ^ s->s_flags) & SB_RDONLY)
1554 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1555 btrfs_sb(s)->bdev_holder = fs_type;
1556 error = btrfs_fill_super(s, fs_devices, data);
1559 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1560 security_free_mnt_opts(&new_sec_opts);
1562 deactivate_locked_super(s);
1563 return ERR_PTR(error);
1566 return dget(s->s_root);
1568 error_close_devices:
1569 btrfs_close_devices(fs_devices);
1571 free_fs_info(fs_info);
1573 security_free_mnt_opts(&new_sec_opts);
1574 return ERR_PTR(error);
1578 * Mount function which is called by VFS layer.
1580 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1581 * which needs vfsmount* of device's root (/). This means device's root has to
1582 * be mounted internally in any case.
1585 * 1. Parse subvol id related options for later use in mount_subvol().
1587 * 2. Mount device's root (/) by calling vfs_kern_mount().
1589 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1590 * first place. In order to avoid calling btrfs_mount() again, we use
1591 * different file_system_type which is not registered to VFS by
1592 * register_filesystem() (btrfs_root_fs_type). As a result,
1593 * btrfs_mount_root() is called. The return value will be used by
1594 * mount_subtree() in mount_subvol().
1596 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1597 * "btrfs subvolume set-default", mount_subvol() is called always.
1599 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1600 const char *device_name, void *data)
1602 struct vfsmount *mnt_root;
1603 struct dentry *root;
1604 fmode_t mode = FMODE_READ;
1605 char *subvol_name = NULL;
1606 u64 subvol_objectid = 0;
1609 if (!(flags & SB_RDONLY))
1610 mode |= FMODE_WRITE;
1612 error = btrfs_parse_subvol_options(data, &subvol_name,
1616 return ERR_PTR(error);
1619 /* mount device's root (/) */
1620 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1621 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1622 if (flags & SB_RDONLY) {
1623 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1624 flags & ~SB_RDONLY, device_name, data);
1626 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1627 flags | SB_RDONLY, device_name, data);
1628 if (IS_ERR(mnt_root)) {
1629 root = ERR_CAST(mnt_root);
1634 down_write(&mnt_root->mnt_sb->s_umount);
1635 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1636 up_write(&mnt_root->mnt_sb->s_umount);
1638 root = ERR_PTR(error);
1645 if (IS_ERR(mnt_root)) {
1646 root = ERR_CAST(mnt_root);
1651 /* mount_subvol() will free subvol_name and mnt_root */
1652 root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1658 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1659 u32 new_pool_size, u32 old_pool_size)
1661 if (new_pool_size == old_pool_size)
1664 fs_info->thread_pool_size = new_pool_size;
1666 btrfs_info(fs_info, "resize thread pool %d -> %d",
1667 old_pool_size, new_pool_size);
1669 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1670 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1671 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1672 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1673 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1674 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1675 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1677 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1678 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1679 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1680 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1681 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1685 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1687 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1690 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1691 unsigned long old_opts, int flags)
1693 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1694 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1695 (flags & SB_RDONLY))) {
1696 /* wait for any defraggers to finish */
1697 wait_event(fs_info->transaction_wait,
1698 (atomic_read(&fs_info->defrag_running) == 0));
1699 if (flags & SB_RDONLY)
1700 sync_filesystem(fs_info->sb);
1704 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1705 unsigned long old_opts)
1708 * We need to cleanup all defragable inodes if the autodefragment is
1709 * close or the filesystem is read only.
1711 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1712 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1713 btrfs_cleanup_defrag_inodes(fs_info);
1716 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1719 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1721 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1722 struct btrfs_root *root = fs_info->tree_root;
1723 unsigned old_flags = sb->s_flags;
1724 unsigned long old_opts = fs_info->mount_opt;
1725 unsigned long old_compress_type = fs_info->compress_type;
1726 u64 old_max_inline = fs_info->max_inline;
1727 u32 old_thread_pool_size = fs_info->thread_pool_size;
1728 u32 old_metadata_ratio = fs_info->metadata_ratio;
1731 sync_filesystem(sb);
1732 btrfs_remount_prepare(fs_info);
1735 void *new_sec_opts = NULL;
1737 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1739 ret = security_sb_remount(sb, new_sec_opts);
1740 security_free_mnt_opts(&new_sec_opts);
1745 ret = btrfs_parse_options(fs_info, data, *flags);
1749 btrfs_remount_begin(fs_info, old_opts, *flags);
1750 btrfs_resize_thread_pool(fs_info,
1751 fs_info->thread_pool_size, old_thread_pool_size);
1753 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1756 if (*flags & SB_RDONLY) {
1758 * this also happens on 'umount -rf' or on shutdown, when
1759 * the filesystem is busy.
1761 cancel_work_sync(&fs_info->async_reclaim_work);
1763 /* wait for the uuid_scan task to finish */
1764 down(&fs_info->uuid_tree_rescan_sem);
1765 /* avoid complains from lockdep et al. */
1766 up(&fs_info->uuid_tree_rescan_sem);
1768 sb->s_flags |= SB_RDONLY;
1771 * Setting SB_RDONLY will put the cleaner thread to
1772 * sleep at the next loop if it's already active.
1773 * If it's already asleep, we'll leave unused block
1774 * groups on disk until we're mounted read-write again
1775 * unless we clean them up here.
1777 btrfs_delete_unused_bgs(fs_info);
1779 btrfs_dev_replace_suspend_for_unmount(fs_info);
1780 btrfs_scrub_cancel(fs_info);
1781 btrfs_pause_balance(fs_info);
1783 ret = btrfs_commit_super(fs_info);
1787 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1789 "Remounting read-write after error is not allowed");
1793 if (fs_info->fs_devices->rw_devices == 0) {
1798 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1800 "too many missing devices, writable remount is not allowed");
1805 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1810 ret = btrfs_cleanup_fs_roots(fs_info);
1814 /* recover relocation */
1815 mutex_lock(&fs_info->cleaner_mutex);
1816 ret = btrfs_recover_relocation(root);
1817 mutex_unlock(&fs_info->cleaner_mutex);
1821 ret = btrfs_resume_balance_async(fs_info);
1825 ret = btrfs_resume_dev_replace_async(fs_info);
1827 btrfs_warn(fs_info, "failed to resume dev_replace");
1831 btrfs_qgroup_rescan_resume(fs_info);
1833 if (!fs_info->uuid_root) {
1834 btrfs_info(fs_info, "creating UUID tree");
1835 ret = btrfs_create_uuid_tree(fs_info);
1838 "failed to create the UUID tree %d",
1843 sb->s_flags &= ~SB_RDONLY;
1845 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1848 wake_up_process(fs_info->transaction_kthread);
1849 btrfs_remount_cleanup(fs_info, old_opts);
1853 /* We've hit an error - don't reset SB_RDONLY */
1855 old_flags |= SB_RDONLY;
1856 sb->s_flags = old_flags;
1857 fs_info->mount_opt = old_opts;
1858 fs_info->compress_type = old_compress_type;
1859 fs_info->max_inline = old_max_inline;
1860 btrfs_resize_thread_pool(fs_info,
1861 old_thread_pool_size, fs_info->thread_pool_size);
1862 fs_info->metadata_ratio = old_metadata_ratio;
1863 btrfs_remount_cleanup(fs_info, old_opts);
1867 /* Used to sort the devices by max_avail(descending sort) */
1868 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
1869 const void *dev_info2)
1871 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1872 ((struct btrfs_device_info *)dev_info2)->max_avail)
1874 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1875 ((struct btrfs_device_info *)dev_info2)->max_avail)
1882 * sort the devices by max_avail, in which max free extent size of each device
1883 * is stored.(Descending Sort)
1885 static inline void btrfs_descending_sort_devices(
1886 struct btrfs_device_info *devices,
1889 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1890 btrfs_cmp_device_free_bytes, NULL);
1894 * The helper to calc the free space on the devices that can be used to store
1897 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1900 struct btrfs_device_info *devices_info;
1901 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1902 struct btrfs_device *device;
1906 u64 min_stripe_size;
1907 int min_stripes = 1, num_stripes = 1;
1908 int i = 0, nr_devices;
1911 * We aren't under the device list lock, so this is racy-ish, but good
1912 * enough for our purposes.
1914 nr_devices = fs_info->fs_devices->open_devices;
1917 nr_devices = fs_info->fs_devices->open_devices;
1925 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1930 /* calc min stripe number for data space allocation */
1931 type = btrfs_data_alloc_profile(fs_info);
1932 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1934 num_stripes = nr_devices;
1935 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1938 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1943 if (type & BTRFS_BLOCK_GROUP_DUP)
1944 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1946 min_stripe_size = BTRFS_STRIPE_LEN;
1949 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1950 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1951 &device->dev_state) ||
1953 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1956 if (i >= nr_devices)
1959 avail_space = device->total_bytes - device->bytes_used;
1961 /* align with stripe_len */
1962 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1963 avail_space *= BTRFS_STRIPE_LEN;
1966 * In order to avoid overwriting the superblock on the drive,
1967 * btrfs starts at an offset of at least 1MB when doing chunk
1973 * we can use the free space in [0, skip_space - 1], subtract
1974 * it from the total.
1976 if (avail_space && avail_space >= skip_space)
1977 avail_space -= skip_space;
1981 if (avail_space < min_stripe_size)
1984 devices_info[i].dev = device;
1985 devices_info[i].max_avail = avail_space;
1993 btrfs_descending_sort_devices(devices_info, nr_devices);
1997 while (nr_devices >= min_stripes) {
1998 if (num_stripes > nr_devices)
1999 num_stripes = nr_devices;
2001 if (devices_info[i].max_avail >= min_stripe_size) {
2005 avail_space += devices_info[i].max_avail * num_stripes;
2006 alloc_size = devices_info[i].max_avail;
2007 for (j = i + 1 - num_stripes; j <= i; j++)
2008 devices_info[j].max_avail -= alloc_size;
2014 kfree(devices_info);
2015 *free_bytes = avail_space;
2020 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2022 * If there's a redundant raid level at DATA block groups, use the respective
2023 * multiplier to scale the sizes.
2025 * Unused device space usage is based on simulating the chunk allocator
2026 * algorithm that respects the device sizes and order of allocations. This is
2027 * a close approximation of the actual use but there are other factors that may
2028 * change the result (like a new metadata chunk).
2030 * If metadata is exhausted, f_bavail will be 0.
2032 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2034 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2035 struct btrfs_super_block *disk_super = fs_info->super_copy;
2036 struct list_head *head = &fs_info->space_info;
2037 struct btrfs_space_info *found;
2039 u64 total_free_data = 0;
2040 u64 total_free_meta = 0;
2041 int bits = dentry->d_sb->s_blocksize_bits;
2042 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2043 unsigned factor = 1;
2044 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2050 list_for_each_entry_rcu(found, head, list) {
2051 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2054 total_free_data += found->disk_total - found->disk_used;
2056 btrfs_account_ro_block_groups_free_space(found);
2058 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2059 if (!list_empty(&found->block_groups[i]))
2060 factor = btrfs_bg_type_to_factor(
2061 btrfs_raid_array[i].bg_flag);
2066 * Metadata in mixed block goup profiles are accounted in data
2068 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2069 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2072 total_free_meta += found->disk_total -
2076 total_used += found->disk_used;
2081 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2082 buf->f_blocks >>= bits;
2083 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2085 /* Account global block reserve as used, it's in logical size already */
2086 spin_lock(&block_rsv->lock);
2087 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2088 if (buf->f_bfree >= block_rsv->size >> bits)
2089 buf->f_bfree -= block_rsv->size >> bits;
2092 spin_unlock(&block_rsv->lock);
2094 buf->f_bavail = div_u64(total_free_data, factor);
2095 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2098 buf->f_bavail += div_u64(total_free_data, factor);
2099 buf->f_bavail = buf->f_bavail >> bits;
2102 * We calculate the remaining metadata space minus global reserve. If
2103 * this is (supposedly) smaller than zero, there's no space. But this
2104 * does not hold in practice, the exhausted state happens where's still
2105 * some positive delta. So we apply some guesswork and compare the
2106 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2108 * We probably cannot calculate the exact threshold value because this
2109 * depends on the internal reservations requested by various
2110 * operations, so some operations that consume a few metadata will
2111 * succeed even if the Avail is zero. But this is better than the other
2116 if (!mixed && total_free_meta - thresh < block_rsv->size)
2119 buf->f_type = BTRFS_SUPER_MAGIC;
2120 buf->f_bsize = dentry->d_sb->s_blocksize;
2121 buf->f_namelen = BTRFS_NAME_LEN;
2123 /* We treat it as constant endianness (it doesn't matter _which_)
2124 because we want the fsid to come out the same whether mounted
2125 on a big-endian or little-endian host */
2126 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2127 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2128 /* Mask in the root object ID too, to disambiguate subvols */
2129 buf->f_fsid.val[0] ^=
2130 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2131 buf->f_fsid.val[1] ^=
2132 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2137 static void btrfs_kill_super(struct super_block *sb)
2139 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2140 kill_anon_super(sb);
2141 free_fs_info(fs_info);
2144 static struct file_system_type btrfs_fs_type = {
2145 .owner = THIS_MODULE,
2147 .mount = btrfs_mount,
2148 .kill_sb = btrfs_kill_super,
2149 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2152 static struct file_system_type btrfs_root_fs_type = {
2153 .owner = THIS_MODULE,
2155 .mount = btrfs_mount_root,
2156 .kill_sb = btrfs_kill_super,
2157 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2160 MODULE_ALIAS_FS("btrfs");
2162 static int btrfs_control_open(struct inode *inode, struct file *file)
2165 * The control file's private_data is used to hold the
2166 * transaction when it is started and is used to keep
2167 * track of whether a transaction is already in progress.
2169 file->private_data = NULL;
2174 * used by btrfsctl to scan devices when no FS is mounted
2176 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2179 struct btrfs_ioctl_vol_args *vol;
2180 struct btrfs_device *device = NULL;
2183 if (!capable(CAP_SYS_ADMIN))
2186 vol = memdup_user((void __user *)arg, sizeof(*vol));
2188 return PTR_ERR(vol);
2189 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2192 case BTRFS_IOC_SCAN_DEV:
2193 mutex_lock(&uuid_mutex);
2194 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2195 &btrfs_root_fs_type);
2196 ret = PTR_ERR_OR_ZERO(device);
2197 mutex_unlock(&uuid_mutex);
2199 case BTRFS_IOC_FORGET_DEV:
2200 ret = btrfs_forget_devices(vol->name);
2202 case BTRFS_IOC_DEVICES_READY:
2203 mutex_lock(&uuid_mutex);
2204 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2205 &btrfs_root_fs_type);
2206 if (IS_ERR(device)) {
2207 mutex_unlock(&uuid_mutex);
2208 ret = PTR_ERR(device);
2211 ret = !(device->fs_devices->num_devices ==
2212 device->fs_devices->total_devices);
2213 mutex_unlock(&uuid_mutex);
2215 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2216 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2224 static int btrfs_freeze(struct super_block *sb)
2226 struct btrfs_trans_handle *trans;
2227 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2228 struct btrfs_root *root = fs_info->tree_root;
2230 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2232 * We don't need a barrier here, we'll wait for any transaction that
2233 * could be in progress on other threads (and do delayed iputs that
2234 * we want to avoid on a frozen filesystem), or do the commit
2237 trans = btrfs_attach_transaction_barrier(root);
2238 if (IS_ERR(trans)) {
2239 /* no transaction, don't bother */
2240 if (PTR_ERR(trans) == -ENOENT)
2242 return PTR_ERR(trans);
2244 return btrfs_commit_transaction(trans);
2247 static int btrfs_unfreeze(struct super_block *sb)
2249 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2251 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2255 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2257 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2258 struct btrfs_fs_devices *cur_devices;
2259 struct btrfs_device *dev, *first_dev = NULL;
2260 struct list_head *head;
2263 * Lightweight locking of the devices. We should not need
2264 * device_list_mutex here as we only read the device data and the list
2265 * is protected by RCU. Even if a device is deleted during the list
2266 * traversals, we'll get valid data, the freeing callback will wait at
2267 * least until the rcu_read_unlock.
2270 cur_devices = fs_info->fs_devices;
2271 while (cur_devices) {
2272 head = &cur_devices->devices;
2273 list_for_each_entry_rcu(dev, head, dev_list) {
2274 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2278 if (!first_dev || dev->devid < first_dev->devid)
2281 cur_devices = cur_devices->seed;
2285 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2292 static const struct super_operations btrfs_super_ops = {
2293 .drop_inode = btrfs_drop_inode,
2294 .evict_inode = btrfs_evict_inode,
2295 .put_super = btrfs_put_super,
2296 .sync_fs = btrfs_sync_fs,
2297 .show_options = btrfs_show_options,
2298 .show_devname = btrfs_show_devname,
2299 .alloc_inode = btrfs_alloc_inode,
2300 .destroy_inode = btrfs_destroy_inode,
2301 .statfs = btrfs_statfs,
2302 .remount_fs = btrfs_remount,
2303 .freeze_fs = btrfs_freeze,
2304 .unfreeze_fs = btrfs_unfreeze,
2307 static const struct file_operations btrfs_ctl_fops = {
2308 .open = btrfs_control_open,
2309 .unlocked_ioctl = btrfs_control_ioctl,
2310 .compat_ioctl = btrfs_control_ioctl,
2311 .owner = THIS_MODULE,
2312 .llseek = noop_llseek,
2315 static struct miscdevice btrfs_misc = {
2316 .minor = BTRFS_MINOR,
2317 .name = "btrfs-control",
2318 .fops = &btrfs_ctl_fops
2321 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2322 MODULE_ALIAS("devname:btrfs-control");
2324 static int __init btrfs_interface_init(void)
2326 return misc_register(&btrfs_misc);
2329 static __cold void btrfs_interface_exit(void)
2331 misc_deregister(&btrfs_misc);
2334 static void __init btrfs_print_mod_info(void)
2336 static const char options[] = ""
2337 #ifdef CONFIG_BTRFS_DEBUG
2340 #ifdef CONFIG_BTRFS_ASSERT
2343 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2344 ", integrity-checker=on"
2346 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2350 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2353 static int __init init_btrfs_fs(void)
2359 err = btrfs_init_sysfs();
2363 btrfs_init_compress();
2365 err = btrfs_init_cachep();
2369 err = extent_io_init();
2373 err = extent_map_init();
2375 goto free_extent_io;
2377 err = ordered_data_init();
2379 goto free_extent_map;
2381 err = btrfs_delayed_inode_init();
2383 goto free_ordered_data;
2385 err = btrfs_auto_defrag_init();
2387 goto free_delayed_inode;
2389 err = btrfs_delayed_ref_init();
2391 goto free_auto_defrag;
2393 err = btrfs_prelim_ref_init();
2395 goto free_delayed_ref;
2397 err = btrfs_end_io_wq_init();
2399 goto free_prelim_ref;
2401 err = btrfs_interface_init();
2403 goto free_end_io_wq;
2405 btrfs_init_lockdep();
2407 btrfs_print_mod_info();
2409 err = btrfs_run_sanity_tests();
2411 goto unregister_ioctl;
2413 err = register_filesystem(&btrfs_fs_type);
2415 goto unregister_ioctl;
2420 btrfs_interface_exit();
2422 btrfs_end_io_wq_exit();
2424 btrfs_prelim_ref_exit();
2426 btrfs_delayed_ref_exit();
2428 btrfs_auto_defrag_exit();
2430 btrfs_delayed_inode_exit();
2432 ordered_data_exit();
2438 btrfs_destroy_cachep();
2440 btrfs_exit_compress();
2446 static void __exit exit_btrfs_fs(void)
2448 btrfs_destroy_cachep();
2449 btrfs_delayed_ref_exit();
2450 btrfs_auto_defrag_exit();
2451 btrfs_delayed_inode_exit();
2452 btrfs_prelim_ref_exit();
2453 ordered_data_exit();
2456 btrfs_interface_exit();
2457 btrfs_end_io_wq_exit();
2458 unregister_filesystem(&btrfs_fs_type);
2460 btrfs_cleanup_fs_uuids();
2461 btrfs_exit_compress();
2464 late_initcall(init_btrfs_fs);
2465 module_exit(exit_btrfs_fs)
2467 MODULE_LICENSE("GPL");