2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
61 #include "tests/btrfs-tests.h"
65 #define CREATE_TRACE_POINTS
66 #include <trace/events/btrfs.h>
68 static const struct super_operations btrfs_super_ops;
69 static struct file_system_type btrfs_fs_type;
71 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
73 const char *btrfs_decode_error(int errno)
75 char *errstr = "unknown";
79 errstr = "IO failure";
82 errstr = "Out of memory";
85 errstr = "Readonly filesystem";
88 errstr = "Object already exists";
91 errstr = "No space left";
94 errstr = "No such entry";
101 /* btrfs handle error by forcing the filesystem readonly */
102 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
104 struct super_block *sb = fs_info->sb;
106 if (sb->s_flags & MS_RDONLY)
109 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
110 sb->s_flags |= MS_RDONLY;
111 btrfs_info(fs_info, "forced readonly");
113 * Note that a running device replace operation is not
114 * canceled here although there is no way to update
115 * the progress. It would add the risk of a deadlock,
116 * therefore the canceling is omitted. The only penalty
117 * is that some I/O remains active until the procedure
118 * completes. The next time when the filesystem is
119 * mounted writeable again, the device replace
120 * operation continues.
126 * __btrfs_handle_fs_error decodes expected errors from the caller and
127 * invokes the approciate error response.
130 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
131 unsigned int line, int errno, const char *fmt, ...)
133 struct super_block *sb = fs_info->sb;
139 * Special case: if the error is EROFS, and we're already
140 * under MS_RDONLY, then it is safe here.
142 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
146 errstr = btrfs_decode_error(errno);
148 struct va_format vaf;
155 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
156 sb->s_id, function, line, errno, errstr, &vaf);
159 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
160 sb->s_id, function, line, errno, errstr);
165 * Today we only save the error info to memory. Long term we'll
166 * also send it down to the disk
168 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
170 /* Don't go through full error handling during mount */
171 if (sb->s_flags & MS_BORN)
172 btrfs_handle_error(fs_info);
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 struct super_block *sb = fs_info->sb;
206 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
207 struct va_format vaf;
210 const char *type = logtypes[4];
211 struct ratelimit_state *ratelimit = &printk_limits[4];
215 while ((kern_level = printk_get_level(fmt)) != 0) {
216 size_t size = printk_skip_level(fmt) - fmt;
218 if (kern_level >= '0' && kern_level <= '7') {
219 memcpy(lvl, fmt, size);
221 type = logtypes[kern_level - '0'];
222 ratelimit = &printk_limits[kern_level - '0'];
230 if (__ratelimit(ratelimit))
231 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &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 && (fs_info->mount_opt & BTRFS_MOUNT_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));
311 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
312 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
313 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
314 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
315 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
316 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
317 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
318 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
319 Opt_skip_balance, Opt_check_integrity,
320 Opt_check_integrity_including_extent_data,
321 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
322 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
323 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
324 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
325 Opt_nologreplay, Opt_norecovery,
326 #ifdef CONFIG_BTRFS_DEBUG
327 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
332 static const match_table_t tokens = {
333 {Opt_degraded, "degraded"},
334 {Opt_subvol, "subvol=%s"},
335 {Opt_subvolid, "subvolid=%s"},
336 {Opt_device, "device=%s"},
337 {Opt_nodatasum, "nodatasum"},
338 {Opt_datasum, "datasum"},
339 {Opt_nodatacow, "nodatacow"},
340 {Opt_datacow, "datacow"},
341 {Opt_nobarrier, "nobarrier"},
342 {Opt_barrier, "barrier"},
343 {Opt_max_inline, "max_inline=%s"},
344 {Opt_alloc_start, "alloc_start=%s"},
345 {Opt_thread_pool, "thread_pool=%d"},
346 {Opt_compress, "compress"},
347 {Opt_compress_type, "compress=%s"},
348 {Opt_compress_force, "compress-force"},
349 {Opt_compress_force_type, "compress-force=%s"},
351 {Opt_ssd_spread, "ssd_spread"},
352 {Opt_nossd, "nossd"},
354 {Opt_noacl, "noacl"},
355 {Opt_notreelog, "notreelog"},
356 {Opt_treelog, "treelog"},
357 {Opt_nologreplay, "nologreplay"},
358 {Opt_norecovery, "norecovery"},
359 {Opt_flushoncommit, "flushoncommit"},
360 {Opt_noflushoncommit, "noflushoncommit"},
361 {Opt_ratio, "metadata_ratio=%d"},
362 {Opt_discard, "discard"},
363 {Opt_nodiscard, "nodiscard"},
364 {Opt_space_cache, "space_cache"},
365 {Opt_space_cache_version, "space_cache=%s"},
366 {Opt_clear_cache, "clear_cache"},
367 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
368 {Opt_enospc_debug, "enospc_debug"},
369 {Opt_noenospc_debug, "noenospc_debug"},
370 {Opt_subvolrootid, "subvolrootid=%d"},
371 {Opt_defrag, "autodefrag"},
372 {Opt_nodefrag, "noautodefrag"},
373 {Opt_inode_cache, "inode_cache"},
374 {Opt_noinode_cache, "noinode_cache"},
375 {Opt_no_space_cache, "nospace_cache"},
376 {Opt_recovery, "recovery"}, /* deprecated */
377 {Opt_usebackuproot, "usebackuproot"},
378 {Opt_skip_balance, "skip_balance"},
379 {Opt_check_integrity, "check_int"},
380 {Opt_check_integrity_including_extent_data, "check_int_data"},
381 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_fatal_errors, "fatal_errors=%s"},
384 {Opt_commit_interval, "commit=%d"},
385 #ifdef CONFIG_BTRFS_DEBUG
386 {Opt_fragment_data, "fragment=data"},
387 {Opt_fragment_metadata, "fragment=metadata"},
388 {Opt_fragment_all, "fragment=all"},
394 * Regular mount options parser. Everything that is needed only when
395 * reading in a new superblock is parsed here.
396 * XXX JDM: This needs to be cleaned up for remount.
398 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
399 unsigned long new_flags)
401 substring_t args[MAX_OPT_ARGS];
402 char *p, *num, *orig = NULL;
407 bool compress_force = false;
408 enum btrfs_compression_type saved_compress_type;
409 bool saved_compress_force;
412 cache_gen = btrfs_super_cache_generation(info->super_copy);
413 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
414 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
416 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
419 * Even the options are empty, we still need to do extra check
426 * strsep changes the string, duplicate it because parse_options
429 options = kstrdup(options, GFP_KERNEL);
435 while ((p = strsep(&options, ",")) != NULL) {
440 token = match_token(p, tokens, args);
443 btrfs_info(info, "allowing degraded mounts");
444 btrfs_set_opt(info->mount_opt, DEGRADED);
448 case Opt_subvolrootid:
451 * These are parsed by btrfs_parse_early_options
452 * and can be happily ignored here.
456 btrfs_set_and_info(info, NODATASUM,
457 "setting nodatasum");
460 if (btrfs_test_opt(info, NODATASUM)) {
461 if (btrfs_test_opt(info, NODATACOW))
463 "setting datasum, datacow enabled");
465 btrfs_info(info, "setting datasum");
467 btrfs_clear_opt(info->mount_opt, NODATACOW);
468 btrfs_clear_opt(info->mount_opt, NODATASUM);
471 if (!btrfs_test_opt(info, NODATACOW)) {
472 if (!btrfs_test_opt(info, COMPRESS) ||
473 !btrfs_test_opt(info, FORCE_COMPRESS)) {
475 "setting nodatacow, compression disabled");
477 btrfs_info(info, "setting nodatacow");
480 btrfs_clear_opt(info->mount_opt, COMPRESS);
481 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
482 btrfs_set_opt(info->mount_opt, NODATACOW);
483 btrfs_set_opt(info->mount_opt, NODATASUM);
486 btrfs_clear_and_info(info, NODATACOW,
489 case Opt_compress_force:
490 case Opt_compress_force_type:
491 compress_force = true;
494 case Opt_compress_type:
495 saved_compress_type = btrfs_test_opt(info,
497 info->compress_type : BTRFS_COMPRESS_NONE;
498 saved_compress_force =
499 btrfs_test_opt(info, FORCE_COMPRESS);
500 if (token == Opt_compress ||
501 token == Opt_compress_force ||
502 strncmp(args[0].from, "zlib", 4) == 0) {
503 compress_type = "zlib";
504 info->compress_type = BTRFS_COMPRESS_ZLIB;
505 btrfs_set_opt(info->mount_opt, COMPRESS);
506 btrfs_clear_opt(info->mount_opt, NODATACOW);
507 btrfs_clear_opt(info->mount_opt, NODATASUM);
509 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
510 compress_type = "lzo";
511 info->compress_type = BTRFS_COMPRESS_LZO;
512 btrfs_set_opt(info->mount_opt, COMPRESS);
513 btrfs_clear_opt(info->mount_opt, NODATACOW);
514 btrfs_clear_opt(info->mount_opt, NODATASUM);
515 btrfs_set_fs_incompat(info, COMPRESS_LZO);
517 } else if (strncmp(args[0].from, "no", 2) == 0) {
518 compress_type = "no";
519 btrfs_clear_opt(info->mount_opt, COMPRESS);
520 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
521 compress_force = false;
528 if (compress_force) {
529 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
532 * If we remount from compress-force=xxx to
533 * compress=xxx, we need clear FORCE_COMPRESS
534 * flag, otherwise, there is no way for users
535 * to disable forcible compression separately.
537 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
539 if ((btrfs_test_opt(info, COMPRESS) &&
540 (info->compress_type != saved_compress_type ||
541 compress_force != saved_compress_force)) ||
542 (!btrfs_test_opt(info, COMPRESS) &&
544 btrfs_info(info, "%s %s compression",
545 (compress_force) ? "force" : "use",
548 compress_force = false;
551 btrfs_set_and_info(info, SSD,
552 "enabling ssd optimizations");
553 btrfs_clear_opt(info->mount_opt, NOSSD);
556 btrfs_set_and_info(info, SSD,
557 "enabling ssd optimizations");
558 btrfs_set_and_info(info, SSD_SPREAD,
559 "using spread ssd allocation scheme");
560 btrfs_clear_opt(info->mount_opt, NOSSD);
563 btrfs_set_opt(info->mount_opt, NOSSD);
564 btrfs_clear_and_info(info, SSD,
565 "not using ssd optimizations");
566 btrfs_clear_and_info(info, SSD_SPREAD,
567 "not using spread ssd allocation scheme");
570 btrfs_clear_and_info(info, NOBARRIER,
571 "turning on barriers");
574 btrfs_set_and_info(info, NOBARRIER,
575 "turning off barriers");
577 case Opt_thread_pool:
578 ret = match_int(&args[0], &intarg);
581 } else if (intarg > 0) {
582 info->thread_pool_size = intarg;
589 num = match_strdup(&args[0]);
591 info->max_inline = memparse(num, NULL);
594 if (info->max_inline) {
595 info->max_inline = min_t(u64,
599 btrfs_info(info, "max_inline at %llu",
606 case Opt_alloc_start:
608 "option alloc_start is obsolete, ignored");
611 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
612 info->sb->s_flags |= MS_POSIXACL;
615 btrfs_err(info, "support for ACL not compiled in!");
620 info->sb->s_flags &= ~MS_POSIXACL;
623 btrfs_set_and_info(info, NOTREELOG,
624 "disabling tree log");
627 btrfs_clear_and_info(info, NOTREELOG,
628 "enabling tree log");
631 case Opt_nologreplay:
632 btrfs_set_and_info(info, NOLOGREPLAY,
633 "disabling log replay at mount time");
635 case Opt_flushoncommit:
636 btrfs_set_and_info(info, FLUSHONCOMMIT,
637 "turning on flush-on-commit");
639 case Opt_noflushoncommit:
640 btrfs_clear_and_info(info, FLUSHONCOMMIT,
641 "turning off flush-on-commit");
644 ret = match_int(&args[0], &intarg);
647 } else if (intarg >= 0) {
648 info->metadata_ratio = intarg;
649 btrfs_info(info, "metadata ratio %d",
650 info->metadata_ratio);
657 btrfs_set_and_info(info, DISCARD,
658 "turning on discard");
661 btrfs_clear_and_info(info, DISCARD,
662 "turning off discard");
664 case Opt_space_cache:
665 case Opt_space_cache_version:
666 if (token == Opt_space_cache ||
667 strcmp(args[0].from, "v1") == 0) {
668 btrfs_clear_opt(info->mount_opt,
670 btrfs_set_and_info(info, SPACE_CACHE,
671 "enabling disk space caching");
672 } else if (strcmp(args[0].from, "v2") == 0) {
673 btrfs_clear_opt(info->mount_opt,
675 btrfs_set_and_info(info, FREE_SPACE_TREE,
676 "enabling free space tree");
682 case Opt_rescan_uuid_tree:
683 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
685 case Opt_no_space_cache:
686 if (btrfs_test_opt(info, SPACE_CACHE)) {
687 btrfs_clear_and_info(info, SPACE_CACHE,
688 "disabling disk space caching");
690 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
691 btrfs_clear_and_info(info, FREE_SPACE_TREE,
692 "disabling free space tree");
695 case Opt_inode_cache:
696 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
697 "enabling inode map caching");
699 case Opt_noinode_cache:
700 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
701 "disabling inode map caching");
703 case Opt_clear_cache:
704 btrfs_set_and_info(info, CLEAR_CACHE,
705 "force clearing of disk cache");
707 case Opt_user_subvol_rm_allowed:
708 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
710 case Opt_enospc_debug:
711 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
713 case Opt_noenospc_debug:
714 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
717 btrfs_set_and_info(info, AUTO_DEFRAG,
718 "enabling auto defrag");
721 btrfs_clear_and_info(info, AUTO_DEFRAG,
722 "disabling auto defrag");
726 "'recovery' is deprecated, use 'usebackuproot' instead");
727 case Opt_usebackuproot:
729 "trying to use backup root at mount time");
730 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
732 case Opt_skip_balance:
733 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
735 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
736 case Opt_check_integrity_including_extent_data:
738 "enabling check integrity including extent data");
739 btrfs_set_opt(info->mount_opt,
740 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
741 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
743 case Opt_check_integrity:
744 btrfs_info(info, "enabling check integrity");
745 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
747 case Opt_check_integrity_print_mask:
748 ret = match_int(&args[0], &intarg);
751 } else if (intarg >= 0) {
752 info->check_integrity_print_mask = intarg;
754 "check_integrity_print_mask 0x%x",
755 info->check_integrity_print_mask);
762 case Opt_check_integrity_including_extent_data:
763 case Opt_check_integrity:
764 case Opt_check_integrity_print_mask:
766 "support for check_integrity* not compiled in!");
770 case Opt_fatal_errors:
771 if (strcmp(args[0].from, "panic") == 0)
772 btrfs_set_opt(info->mount_opt,
773 PANIC_ON_FATAL_ERROR);
774 else if (strcmp(args[0].from, "bug") == 0)
775 btrfs_clear_opt(info->mount_opt,
776 PANIC_ON_FATAL_ERROR);
782 case Opt_commit_interval:
784 ret = match_int(&args[0], &intarg);
786 btrfs_err(info, "invalid commit interval");
793 "excessive commit interval %d",
796 info->commit_interval = intarg;
799 "using default commit interval %ds",
800 BTRFS_DEFAULT_COMMIT_INTERVAL);
801 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
804 #ifdef CONFIG_BTRFS_DEBUG
805 case Opt_fragment_all:
806 btrfs_info(info, "fragmenting all space");
807 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
808 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
810 case Opt_fragment_metadata:
811 btrfs_info(info, "fragmenting metadata");
812 btrfs_set_opt(info->mount_opt,
815 case Opt_fragment_data:
816 btrfs_info(info, "fragmenting data");
817 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
821 btrfs_info(info, "unrecognized mount option '%s'", p);
830 * Extra check for current option against current flag
832 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
834 "nologreplay must be used with ro mount option");
838 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
839 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
840 !btrfs_test_opt(info, CLEAR_CACHE)) {
841 btrfs_err(info, "cannot disable free space tree");
845 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
846 btrfs_info(info, "disk space caching is enabled");
847 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
848 btrfs_info(info, "using free space tree");
854 * Parse mount options that are required early in the mount process.
856 * All other options will be parsed on much later in the mount process and
857 * only when we need to allocate a new super block.
859 static int btrfs_parse_early_options(const char *options, fmode_t flags,
860 void *holder, char **subvol_name, u64 *subvol_objectid,
861 struct btrfs_fs_devices **fs_devices)
863 substring_t args[MAX_OPT_ARGS];
864 char *device_name, *opts, *orig, *p;
872 * strsep changes the string, duplicate it because parse_options
875 opts = kstrdup(options, GFP_KERNEL);
880 while ((p = strsep(&opts, ",")) != NULL) {
885 token = match_token(p, tokens, args);
889 *subvol_name = match_strdup(&args[0]);
896 num = match_strdup(&args[0]);
898 *subvol_objectid = memparse(num, NULL);
900 /* we want the original fs_tree */
901 if (!*subvol_objectid)
903 BTRFS_FS_TREE_OBJECTID;
909 case Opt_subvolrootid:
910 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
913 device_name = match_strdup(&args[0]);
918 error = btrfs_scan_one_device(device_name,
919 flags, holder, fs_devices);
934 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
937 struct btrfs_root *root = fs_info->tree_root;
938 struct btrfs_root *fs_root;
939 struct btrfs_root_ref *root_ref;
940 struct btrfs_inode_ref *inode_ref;
941 struct btrfs_key key;
942 struct btrfs_path *path = NULL;
943 char *name = NULL, *ptr;
948 path = btrfs_alloc_path();
953 path->leave_spinning = 1;
955 name = kmalloc(PATH_MAX, GFP_KERNEL);
960 ptr = name + PATH_MAX - 1;
964 * Walk up the subvolume trees in the tree of tree roots by root
965 * backrefs until we hit the top-level subvolume.
967 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
968 key.objectid = subvol_objectid;
969 key.type = BTRFS_ROOT_BACKREF_KEY;
970 key.offset = (u64)-1;
972 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
975 } else if (ret > 0) {
976 ret = btrfs_previous_item(root, path, subvol_objectid,
977 BTRFS_ROOT_BACKREF_KEY);
980 } else if (ret > 0) {
986 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
987 subvol_objectid = key.offset;
989 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
990 struct btrfs_root_ref);
991 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
997 read_extent_buffer(path->nodes[0], ptr + 1,
998 (unsigned long)(root_ref + 1), len);
1000 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1001 btrfs_release_path(path);
1003 key.objectid = subvol_objectid;
1004 key.type = BTRFS_ROOT_ITEM_KEY;
1005 key.offset = (u64)-1;
1006 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1007 if (IS_ERR(fs_root)) {
1008 ret = PTR_ERR(fs_root);
1013 * Walk up the filesystem tree by inode refs until we hit the
1016 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1017 key.objectid = dirid;
1018 key.type = BTRFS_INODE_REF_KEY;
1019 key.offset = (u64)-1;
1021 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1024 } else if (ret > 0) {
1025 ret = btrfs_previous_item(fs_root, path, dirid,
1026 BTRFS_INODE_REF_KEY);
1029 } else if (ret > 0) {
1035 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1038 inode_ref = btrfs_item_ptr(path->nodes[0],
1040 struct btrfs_inode_ref);
1041 len = btrfs_inode_ref_name_len(path->nodes[0],
1045 ret = -ENAMETOOLONG;
1048 read_extent_buffer(path->nodes[0], ptr + 1,
1049 (unsigned long)(inode_ref + 1), len);
1051 btrfs_release_path(path);
1055 btrfs_free_path(path);
1056 if (ptr == name + PATH_MAX - 1) {
1060 memmove(name, ptr, name + PATH_MAX - ptr);
1065 btrfs_free_path(path);
1067 return ERR_PTR(ret);
1070 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1072 struct btrfs_root *root = fs_info->tree_root;
1073 struct btrfs_dir_item *di;
1074 struct btrfs_path *path;
1075 struct btrfs_key location;
1078 path = btrfs_alloc_path();
1081 path->leave_spinning = 1;
1084 * Find the "default" dir item which points to the root item that we
1085 * will mount by default if we haven't been given a specific subvolume
1088 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1089 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1091 btrfs_free_path(path);
1096 * Ok the default dir item isn't there. This is weird since
1097 * it's always been there, but don't freak out, just try and
1098 * mount the top-level subvolume.
1100 btrfs_free_path(path);
1101 *objectid = BTRFS_FS_TREE_OBJECTID;
1105 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1106 btrfs_free_path(path);
1107 *objectid = location.objectid;
1111 static int btrfs_fill_super(struct super_block *sb,
1112 struct btrfs_fs_devices *fs_devices,
1115 struct inode *inode;
1116 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1117 struct btrfs_key key;
1120 sb->s_maxbytes = MAX_LFS_FILESIZE;
1121 sb->s_magic = BTRFS_SUPER_MAGIC;
1122 sb->s_op = &btrfs_super_ops;
1123 sb->s_d_op = &btrfs_dentry_operations;
1124 sb->s_export_op = &btrfs_export_ops;
1125 sb->s_xattr = btrfs_xattr_handlers;
1126 sb->s_time_gran = 1;
1127 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1128 sb->s_flags |= MS_POSIXACL;
1130 sb->s_flags |= MS_I_VERSION;
1131 sb->s_iflags |= SB_I_CGROUPWB;
1133 err = super_setup_bdi(sb);
1135 btrfs_err(fs_info, "super_setup_bdi failed");
1139 err = open_ctree(sb, fs_devices, (char *)data);
1141 btrfs_err(fs_info, "open_ctree failed");
1145 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1146 key.type = BTRFS_INODE_ITEM_KEY;
1148 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1149 if (IS_ERR(inode)) {
1150 err = PTR_ERR(inode);
1154 sb->s_root = d_make_root(inode);
1160 cleancache_init_fs(sb);
1161 sb->s_flags |= MS_ACTIVE;
1165 close_ctree(fs_info);
1169 int btrfs_sync_fs(struct super_block *sb, int wait)
1171 struct btrfs_trans_handle *trans;
1172 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1173 struct btrfs_root *root = fs_info->tree_root;
1175 trace_btrfs_sync_fs(fs_info, wait);
1178 filemap_flush(fs_info->btree_inode->i_mapping);
1182 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1184 trans = btrfs_attach_transaction_barrier(root);
1185 if (IS_ERR(trans)) {
1186 /* no transaction, don't bother */
1187 if (PTR_ERR(trans) == -ENOENT) {
1189 * Exit unless we have some pending changes
1190 * that need to go through commit
1192 if (fs_info->pending_changes == 0)
1195 * A non-blocking test if the fs is frozen. We must not
1196 * start a new transaction here otherwise a deadlock
1197 * happens. The pending operations are delayed to the
1198 * next commit after thawing.
1200 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1201 __sb_end_write(sb, SB_FREEZE_WRITE);
1204 trans = btrfs_start_transaction(root, 0);
1207 return PTR_ERR(trans);
1209 return btrfs_commit_transaction(trans);
1212 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1214 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1215 char *compress_type;
1217 if (btrfs_test_opt(info, DEGRADED))
1218 seq_puts(seq, ",degraded");
1219 if (btrfs_test_opt(info, NODATASUM))
1220 seq_puts(seq, ",nodatasum");
1221 if (btrfs_test_opt(info, NODATACOW))
1222 seq_puts(seq, ",nodatacow");
1223 if (btrfs_test_opt(info, NOBARRIER))
1224 seq_puts(seq, ",nobarrier");
1225 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1226 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1227 if (info->thread_pool_size != min_t(unsigned long,
1228 num_online_cpus() + 2, 8))
1229 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1230 if (btrfs_test_opt(info, COMPRESS)) {
1231 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1232 compress_type = "zlib";
1234 compress_type = "lzo";
1235 if (btrfs_test_opt(info, FORCE_COMPRESS))
1236 seq_printf(seq, ",compress-force=%s", compress_type);
1238 seq_printf(seq, ",compress=%s", compress_type);
1240 if (btrfs_test_opt(info, NOSSD))
1241 seq_puts(seq, ",nossd");
1242 if (btrfs_test_opt(info, SSD_SPREAD))
1243 seq_puts(seq, ",ssd_spread");
1244 else if (btrfs_test_opt(info, SSD))
1245 seq_puts(seq, ",ssd");
1246 if (btrfs_test_opt(info, NOTREELOG))
1247 seq_puts(seq, ",notreelog");
1248 if (btrfs_test_opt(info, NOLOGREPLAY))
1249 seq_puts(seq, ",nologreplay");
1250 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1251 seq_puts(seq, ",flushoncommit");
1252 if (btrfs_test_opt(info, DISCARD))
1253 seq_puts(seq, ",discard");
1254 if (!(info->sb->s_flags & MS_POSIXACL))
1255 seq_puts(seq, ",noacl");
1256 if (btrfs_test_opt(info, SPACE_CACHE))
1257 seq_puts(seq, ",space_cache");
1258 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1259 seq_puts(seq, ",space_cache=v2");
1261 seq_puts(seq, ",nospace_cache");
1262 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1263 seq_puts(seq, ",rescan_uuid_tree");
1264 if (btrfs_test_opt(info, CLEAR_CACHE))
1265 seq_puts(seq, ",clear_cache");
1266 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1267 seq_puts(seq, ",user_subvol_rm_allowed");
1268 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1269 seq_puts(seq, ",enospc_debug");
1270 if (btrfs_test_opt(info, AUTO_DEFRAG))
1271 seq_puts(seq, ",autodefrag");
1272 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1273 seq_puts(seq, ",inode_cache");
1274 if (btrfs_test_opt(info, SKIP_BALANCE))
1275 seq_puts(seq, ",skip_balance");
1276 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1277 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1278 seq_puts(seq, ",check_int_data");
1279 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1280 seq_puts(seq, ",check_int");
1281 if (info->check_integrity_print_mask)
1282 seq_printf(seq, ",check_int_print_mask=%d",
1283 info->check_integrity_print_mask);
1285 if (info->metadata_ratio)
1286 seq_printf(seq, ",metadata_ratio=%d",
1287 info->metadata_ratio);
1288 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1289 seq_puts(seq, ",fatal_errors=panic");
1290 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1291 seq_printf(seq, ",commit=%d", info->commit_interval);
1292 #ifdef CONFIG_BTRFS_DEBUG
1293 if (btrfs_test_opt(info, FRAGMENT_DATA))
1294 seq_puts(seq, ",fragment=data");
1295 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1296 seq_puts(seq, ",fragment=metadata");
1298 seq_printf(seq, ",subvolid=%llu",
1299 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1300 seq_puts(seq, ",subvol=");
1301 seq_dentry(seq, dentry, " \t\n\\");
1305 static int btrfs_test_super(struct super_block *s, void *data)
1307 struct btrfs_fs_info *p = data;
1308 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1310 return fs_info->fs_devices == p->fs_devices;
1313 static int btrfs_set_super(struct super_block *s, void *data)
1315 int err = set_anon_super(s, data);
1317 s->s_fs_info = data;
1322 * subvolumes are identified by ino 256
1324 static inline int is_subvolume_inode(struct inode *inode)
1326 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1332 * This will add subvolid=0 to the argument string while removing any subvol=
1333 * and subvolid= arguments to make sure we get the top-level root for path
1334 * walking to the subvol we want.
1336 static char *setup_root_args(char *args)
1338 char *buf, *dst, *sep;
1341 return kstrdup("subvolid=0", GFP_KERNEL);
1343 /* The worst case is that we add ",subvolid=0" to the end. */
1344 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1,
1350 sep = strchrnul(args, ',');
1351 if (!strstarts(args, "subvol=") &&
1352 !strstarts(args, "subvolid=")) {
1353 memcpy(dst, args, sep - args);
1362 strcpy(dst, "subvolid=0");
1367 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1368 int flags, const char *device_name,
1371 struct dentry *root;
1372 struct vfsmount *mnt = NULL;
1376 newargs = setup_root_args(data);
1378 root = ERR_PTR(-ENOMEM);
1382 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1383 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1384 if (flags & MS_RDONLY) {
1385 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1386 device_name, newargs);
1388 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1389 device_name, newargs);
1391 root = ERR_CAST(mnt);
1396 down_write(&mnt->mnt_sb->s_umount);
1397 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1398 up_write(&mnt->mnt_sb->s_umount);
1400 root = ERR_PTR(ret);
1406 root = ERR_CAST(mnt);
1412 if (!subvol_objectid) {
1413 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1416 root = ERR_PTR(ret);
1420 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1422 if (IS_ERR(subvol_name)) {
1423 root = ERR_CAST(subvol_name);
1430 root = mount_subtree(mnt, subvol_name);
1431 /* mount_subtree() drops our reference on the vfsmount. */
1434 if (!IS_ERR(root)) {
1435 struct super_block *s = root->d_sb;
1436 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1437 struct inode *root_inode = d_inode(root);
1438 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1441 if (!is_subvolume_inode(root_inode)) {
1442 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1446 if (subvol_objectid && root_objectid != subvol_objectid) {
1448 * This will also catch a race condition where a
1449 * subvolume which was passed by ID is renamed and
1450 * another subvolume is renamed over the old location.
1453 "subvol '%s' does not match subvolid %llu",
1454 subvol_name, subvol_objectid);
1459 root = ERR_PTR(ret);
1460 deactivate_locked_super(s);
1471 static int parse_security_options(char *orig_opts,
1472 struct security_mnt_opts *sec_opts)
1474 char *secdata = NULL;
1477 secdata = alloc_secdata();
1480 ret = security_sb_copy_data(orig_opts, secdata);
1482 free_secdata(secdata);
1485 ret = security_sb_parse_opts_str(secdata, sec_opts);
1486 free_secdata(secdata);
1490 static int setup_security_options(struct btrfs_fs_info *fs_info,
1491 struct super_block *sb,
1492 struct security_mnt_opts *sec_opts)
1497 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1500 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1504 #ifdef CONFIG_SECURITY
1505 if (!fs_info->security_opts.num_mnt_opts) {
1506 /* first time security setup, copy sec_opts to fs_info */
1507 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1510 * Since SELinux (the only one supporting security_mnt_opts)
1511 * does NOT support changing context during remount/mount of
1512 * the same sb, this must be the same or part of the same
1513 * security options, just free it.
1515 security_free_mnt_opts(sec_opts);
1522 * Find a superblock for the given device / mount point.
1524 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1525 * for multiple device setup. Make sure to keep it in sync.
1527 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1528 const char *device_name, void *data)
1530 struct block_device *bdev = NULL;
1531 struct super_block *s;
1532 struct btrfs_fs_devices *fs_devices = NULL;
1533 struct btrfs_fs_info *fs_info = NULL;
1534 struct security_mnt_opts new_sec_opts;
1535 fmode_t mode = FMODE_READ;
1536 char *subvol_name = NULL;
1537 u64 subvol_objectid = 0;
1540 if (!(flags & MS_RDONLY))
1541 mode |= FMODE_WRITE;
1543 error = btrfs_parse_early_options(data, mode, fs_type,
1544 &subvol_name, &subvol_objectid,
1548 return ERR_PTR(error);
1551 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1552 /* mount_subvol() will free subvol_name. */
1553 return mount_subvol(subvol_name, subvol_objectid, flags,
1557 security_init_mnt_opts(&new_sec_opts);
1559 error = parse_security_options(data, &new_sec_opts);
1561 return ERR_PTR(error);
1564 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1566 goto error_sec_opts;
1569 * Setup a dummy root and fs_info for test/set super. This is because
1570 * we don't actually fill this stuff out until open_ctree, but we need
1571 * it for searching for existing supers, so this lets us do that and
1572 * then open_ctree will properly initialize everything later.
1574 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1577 goto error_sec_opts;
1580 fs_info->fs_devices = fs_devices;
1582 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1583 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1584 security_init_mnt_opts(&fs_info->security_opts);
1585 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1590 error = btrfs_open_devices(fs_devices, mode, fs_type);
1594 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1596 goto error_close_devices;
1599 bdev = fs_devices->latest_bdev;
1600 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1604 goto error_close_devices;
1608 btrfs_close_devices(fs_devices);
1609 free_fs_info(fs_info);
1610 if ((flags ^ s->s_flags) & MS_RDONLY)
1613 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1614 btrfs_sb(s)->bdev_holder = fs_type;
1615 error = btrfs_fill_super(s, fs_devices, data);
1618 deactivate_locked_super(s);
1619 goto error_sec_opts;
1622 fs_info = btrfs_sb(s);
1623 error = setup_security_options(fs_info, s, &new_sec_opts);
1625 deactivate_locked_super(s);
1626 goto error_sec_opts;
1629 return dget(s->s_root);
1631 error_close_devices:
1632 btrfs_close_devices(fs_devices);
1634 free_fs_info(fs_info);
1636 security_free_mnt_opts(&new_sec_opts);
1637 return ERR_PTR(error);
1640 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1641 int new_pool_size, int old_pool_size)
1643 if (new_pool_size == old_pool_size)
1646 fs_info->thread_pool_size = new_pool_size;
1648 btrfs_info(fs_info, "resize thread pool %d -> %d",
1649 old_pool_size, new_pool_size);
1651 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1652 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1653 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1654 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1655 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1656 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1657 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1659 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1660 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1661 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1662 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1663 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1667 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1669 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1672 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1673 unsigned long old_opts, int flags)
1675 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1676 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1677 (flags & MS_RDONLY))) {
1678 /* wait for any defraggers to finish */
1679 wait_event(fs_info->transaction_wait,
1680 (atomic_read(&fs_info->defrag_running) == 0));
1681 if (flags & MS_RDONLY)
1682 sync_filesystem(fs_info->sb);
1686 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1687 unsigned long old_opts)
1690 * We need to cleanup all defragable inodes if the autodefragment is
1691 * close or the filesystem is read only.
1693 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1694 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1695 (fs_info->sb->s_flags & MS_RDONLY))) {
1696 btrfs_cleanup_defrag_inodes(fs_info);
1699 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1702 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1704 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1705 struct btrfs_root *root = fs_info->tree_root;
1706 unsigned old_flags = sb->s_flags;
1707 unsigned long old_opts = fs_info->mount_opt;
1708 unsigned long old_compress_type = fs_info->compress_type;
1709 u64 old_max_inline = fs_info->max_inline;
1710 int old_thread_pool_size = fs_info->thread_pool_size;
1711 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1714 sync_filesystem(sb);
1715 btrfs_remount_prepare(fs_info);
1718 struct security_mnt_opts new_sec_opts;
1720 security_init_mnt_opts(&new_sec_opts);
1721 ret = parse_security_options(data, &new_sec_opts);
1724 ret = setup_security_options(fs_info, sb,
1727 security_free_mnt_opts(&new_sec_opts);
1732 ret = btrfs_parse_options(fs_info, data, *flags);
1738 btrfs_remount_begin(fs_info, old_opts, *flags);
1739 btrfs_resize_thread_pool(fs_info,
1740 fs_info->thread_pool_size, old_thread_pool_size);
1742 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1745 if (*flags & MS_RDONLY) {
1747 * this also happens on 'umount -rf' or on shutdown, when
1748 * the filesystem is busy.
1750 cancel_work_sync(&fs_info->async_reclaim_work);
1752 /* wait for the uuid_scan task to finish */
1753 down(&fs_info->uuid_tree_rescan_sem);
1754 /* avoid complains from lockdep et al. */
1755 up(&fs_info->uuid_tree_rescan_sem);
1757 sb->s_flags |= MS_RDONLY;
1760 * Setting MS_RDONLY will put the cleaner thread to
1761 * sleep at the next loop if it's already active.
1762 * If it's already asleep, we'll leave unused block
1763 * groups on disk until we're mounted read-write again
1764 * unless we clean them up here.
1766 btrfs_delete_unused_bgs(fs_info);
1768 btrfs_dev_replace_suspend_for_unmount(fs_info);
1769 btrfs_scrub_cancel(fs_info);
1770 btrfs_pause_balance(fs_info);
1772 ret = btrfs_commit_super(fs_info);
1776 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1778 "Remounting read-write after error is not allowed");
1782 if (fs_info->fs_devices->rw_devices == 0) {
1787 if (!btrfs_check_rw_degradable(fs_info)) {
1789 "too many missing devices, writeable remount is not allowed");
1794 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1799 ret = btrfs_cleanup_fs_roots(fs_info);
1803 /* recover relocation */
1804 mutex_lock(&fs_info->cleaner_mutex);
1805 ret = btrfs_recover_relocation(root);
1806 mutex_unlock(&fs_info->cleaner_mutex);
1810 ret = btrfs_resume_balance_async(fs_info);
1814 ret = btrfs_resume_dev_replace_async(fs_info);
1816 btrfs_warn(fs_info, "failed to resume dev_replace");
1820 btrfs_qgroup_rescan_resume(fs_info);
1822 if (!fs_info->uuid_root) {
1823 btrfs_info(fs_info, "creating UUID tree");
1824 ret = btrfs_create_uuid_tree(fs_info);
1827 "failed to create the UUID tree %d",
1832 sb->s_flags &= ~MS_RDONLY;
1834 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1837 wake_up_process(fs_info->transaction_kthread);
1838 btrfs_remount_cleanup(fs_info, old_opts);
1842 /* We've hit an error - don't reset MS_RDONLY */
1843 if (sb->s_flags & MS_RDONLY)
1844 old_flags |= MS_RDONLY;
1845 sb->s_flags = old_flags;
1846 fs_info->mount_opt = old_opts;
1847 fs_info->compress_type = old_compress_type;
1848 fs_info->max_inline = old_max_inline;
1849 btrfs_resize_thread_pool(fs_info,
1850 old_thread_pool_size, fs_info->thread_pool_size);
1851 fs_info->metadata_ratio = old_metadata_ratio;
1852 btrfs_remount_cleanup(fs_info, old_opts);
1856 /* Used to sort the devices by max_avail(descending sort) */
1857 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1858 const void *dev_info2)
1860 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1861 ((struct btrfs_device_info *)dev_info2)->max_avail)
1863 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1864 ((struct btrfs_device_info *)dev_info2)->max_avail)
1871 * sort the devices by max_avail, in which max free extent size of each device
1872 * is stored.(Descending Sort)
1874 static inline void btrfs_descending_sort_devices(
1875 struct btrfs_device_info *devices,
1878 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1879 btrfs_cmp_device_free_bytes, NULL);
1883 * The helper to calc the free space on the devices that can be used to store
1886 static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1889 struct btrfs_device_info *devices_info;
1890 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1891 struct btrfs_device *device;
1895 u64 min_stripe_size;
1896 int min_stripes = 1, num_stripes = 1;
1897 int i = 0, nr_devices;
1900 * We aren't under the device list lock, so this is racy-ish, but good
1901 * enough for our purposes.
1903 nr_devices = fs_info->fs_devices->open_devices;
1906 nr_devices = fs_info->fs_devices->open_devices;
1914 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1919 /* calc min stripe number for data space allocation */
1920 type = btrfs_data_alloc_profile(fs_info);
1921 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1923 num_stripes = nr_devices;
1924 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1927 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1932 if (type & BTRFS_BLOCK_GROUP_DUP)
1933 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1935 min_stripe_size = BTRFS_STRIPE_LEN;
1938 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1939 if (!device->in_fs_metadata || !device->bdev ||
1940 device->is_tgtdev_for_dev_replace)
1943 if (i >= nr_devices)
1946 avail_space = device->total_bytes - device->bytes_used;
1948 /* align with stripe_len */
1949 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1950 avail_space *= BTRFS_STRIPE_LEN;
1953 * In order to avoid overwriting the superblock on the drive,
1954 * btrfs starts at an offset of at least 1MB when doing chunk
1960 * we can use the free space in [0, skip_space - 1], subtract
1961 * it from the total.
1963 if (avail_space && avail_space >= skip_space)
1964 avail_space -= skip_space;
1968 if (avail_space < min_stripe_size)
1971 devices_info[i].dev = device;
1972 devices_info[i].max_avail = avail_space;
1980 btrfs_descending_sort_devices(devices_info, nr_devices);
1984 while (nr_devices >= min_stripes) {
1985 if (num_stripes > nr_devices)
1986 num_stripes = nr_devices;
1988 if (devices_info[i].max_avail >= min_stripe_size) {
1992 avail_space += devices_info[i].max_avail * num_stripes;
1993 alloc_size = devices_info[i].max_avail;
1994 for (j = i + 1 - num_stripes; j <= i; j++)
1995 devices_info[j].max_avail -= alloc_size;
2001 kfree(devices_info);
2002 *free_bytes = avail_space;
2007 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2009 * If there's a redundant raid level at DATA block groups, use the respective
2010 * multiplier to scale the sizes.
2012 * Unused device space usage is based on simulating the chunk allocator
2013 * algorithm that respects the device sizes and order of allocations. This is
2014 * a close approximation of the actual use but there are other factors that may
2015 * change the result (like a new metadata chunk).
2017 * If metadata is exhausted, f_bavail will be 0.
2019 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2021 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2022 struct btrfs_super_block *disk_super = fs_info->super_copy;
2023 struct list_head *head = &fs_info->space_info;
2024 struct btrfs_space_info *found;
2026 u64 total_free_data = 0;
2027 u64 total_free_meta = 0;
2028 int bits = dentry->d_sb->s_blocksize_bits;
2029 __be32 *fsid = (__be32 *)fs_info->fsid;
2030 unsigned factor = 1;
2031 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2037 list_for_each_entry_rcu(found, head, list) {
2038 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2041 total_free_data += found->disk_total - found->disk_used;
2043 btrfs_account_ro_block_groups_free_space(found);
2045 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2046 if (!list_empty(&found->block_groups[i])) {
2048 case BTRFS_RAID_DUP:
2049 case BTRFS_RAID_RAID1:
2050 case BTRFS_RAID_RAID10:
2058 * Metadata in mixed block goup profiles are accounted in data
2060 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2061 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2064 total_free_meta += found->disk_total -
2068 total_used += found->disk_used;
2073 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2074 buf->f_blocks >>= bits;
2075 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2077 /* Account global block reserve as used, it's in logical size already */
2078 spin_lock(&block_rsv->lock);
2079 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2080 if (buf->f_bfree >= block_rsv->size >> bits)
2081 buf->f_bfree -= block_rsv->size >> bits;
2084 spin_unlock(&block_rsv->lock);
2086 buf->f_bavail = div_u64(total_free_data, factor);
2087 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2090 buf->f_bavail += div_u64(total_free_data, factor);
2091 buf->f_bavail = buf->f_bavail >> bits;
2094 * We calculate the remaining metadata space minus global reserve. If
2095 * this is (supposedly) smaller than zero, there's no space. But this
2096 * does not hold in practice, the exhausted state happens where's still
2097 * some positive delta. So we apply some guesswork and compare the
2098 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2100 * We probably cannot calculate the exact threshold value because this
2101 * depends on the internal reservations requested by various
2102 * operations, so some operations that consume a few metadata will
2103 * succeed even if the Avail is zero. But this is better than the other
2106 thresh = 4 * 1024 * 1024;
2108 if (!mixed && total_free_meta - thresh < block_rsv->size)
2111 buf->f_type = BTRFS_SUPER_MAGIC;
2112 buf->f_bsize = dentry->d_sb->s_blocksize;
2113 buf->f_namelen = BTRFS_NAME_LEN;
2115 /* We treat it as constant endianness (it doesn't matter _which_)
2116 because we want the fsid to come out the same whether mounted
2117 on a big-endian or little-endian host */
2118 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2119 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2120 /* Mask in the root object ID too, to disambiguate subvols */
2121 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2122 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2127 static void btrfs_kill_super(struct super_block *sb)
2129 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2130 kill_anon_super(sb);
2131 free_fs_info(fs_info);
2134 static struct file_system_type btrfs_fs_type = {
2135 .owner = THIS_MODULE,
2137 .mount = btrfs_mount,
2138 .kill_sb = btrfs_kill_super,
2139 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2141 MODULE_ALIAS_FS("btrfs");
2143 static int btrfs_control_open(struct inode *inode, struct file *file)
2146 * The control file's private_data is used to hold the
2147 * transaction when it is started and is used to keep
2148 * track of whether a transaction is already in progress.
2150 file->private_data = NULL;
2155 * used by btrfsctl to scan devices when no FS is mounted
2157 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2160 struct btrfs_ioctl_vol_args *vol;
2161 struct btrfs_fs_devices *fs_devices;
2164 if (!capable(CAP_SYS_ADMIN))
2167 vol = memdup_user((void __user *)arg, sizeof(*vol));
2169 return PTR_ERR(vol);
2172 case BTRFS_IOC_SCAN_DEV:
2173 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2174 &btrfs_fs_type, &fs_devices);
2176 case BTRFS_IOC_DEVICES_READY:
2177 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2178 &btrfs_fs_type, &fs_devices);
2181 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2183 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2184 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2192 static int btrfs_freeze(struct super_block *sb)
2194 struct btrfs_trans_handle *trans;
2195 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2196 struct btrfs_root *root = fs_info->tree_root;
2198 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2200 * We don't need a barrier here, we'll wait for any transaction that
2201 * could be in progress on other threads (and do delayed iputs that
2202 * we want to avoid on a frozen filesystem), or do the commit
2205 trans = btrfs_attach_transaction_barrier(root);
2206 if (IS_ERR(trans)) {
2207 /* no transaction, don't bother */
2208 if (PTR_ERR(trans) == -ENOENT)
2210 return PTR_ERR(trans);
2212 return btrfs_commit_transaction(trans);
2215 static int btrfs_unfreeze(struct super_block *sb)
2217 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2219 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2223 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2225 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2226 struct btrfs_fs_devices *cur_devices;
2227 struct btrfs_device *dev, *first_dev = NULL;
2228 struct list_head *head;
2229 struct rcu_string *name;
2231 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2232 cur_devices = fs_info->fs_devices;
2233 while (cur_devices) {
2234 head = &cur_devices->devices;
2235 list_for_each_entry(dev, head, dev_list) {
2240 if (!first_dev || dev->devid < first_dev->devid)
2243 cur_devices = cur_devices->seed;
2248 name = rcu_dereference(first_dev->name);
2249 seq_escape(m, name->str, " \t\n\\");
2254 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2258 static const struct super_operations btrfs_super_ops = {
2259 .drop_inode = btrfs_drop_inode,
2260 .evict_inode = btrfs_evict_inode,
2261 .put_super = btrfs_put_super,
2262 .sync_fs = btrfs_sync_fs,
2263 .show_options = btrfs_show_options,
2264 .show_devname = btrfs_show_devname,
2265 .write_inode = btrfs_write_inode,
2266 .alloc_inode = btrfs_alloc_inode,
2267 .destroy_inode = btrfs_destroy_inode,
2268 .statfs = btrfs_statfs,
2269 .remount_fs = btrfs_remount,
2270 .freeze_fs = btrfs_freeze,
2271 .unfreeze_fs = btrfs_unfreeze,
2274 static const struct file_operations btrfs_ctl_fops = {
2275 .open = btrfs_control_open,
2276 .unlocked_ioctl = btrfs_control_ioctl,
2277 .compat_ioctl = btrfs_control_ioctl,
2278 .owner = THIS_MODULE,
2279 .llseek = noop_llseek,
2282 static struct miscdevice btrfs_misc = {
2283 .minor = BTRFS_MINOR,
2284 .name = "btrfs-control",
2285 .fops = &btrfs_ctl_fops
2288 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2289 MODULE_ALIAS("devname:btrfs-control");
2291 static int btrfs_interface_init(void)
2293 return misc_register(&btrfs_misc);
2296 static void btrfs_interface_exit(void)
2298 misc_deregister(&btrfs_misc);
2301 static void btrfs_print_mod_info(void)
2303 pr_info("Btrfs loaded, crc32c=%s"
2304 #ifdef CONFIG_BTRFS_DEBUG
2307 #ifdef CONFIG_BTRFS_ASSERT
2310 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2311 ", integrity-checker=on"
2314 btrfs_crc32c_impl());
2317 static int __init init_btrfs_fs(void)
2321 err = btrfs_hash_init();
2327 err = btrfs_init_sysfs();
2331 btrfs_init_compress();
2333 err = btrfs_init_cachep();
2337 err = extent_io_init();
2341 err = extent_map_init();
2343 goto free_extent_io;
2345 err = ordered_data_init();
2347 goto free_extent_map;
2349 err = btrfs_delayed_inode_init();
2351 goto free_ordered_data;
2353 err = btrfs_auto_defrag_init();
2355 goto free_delayed_inode;
2357 err = btrfs_delayed_ref_init();
2359 goto free_auto_defrag;
2361 err = btrfs_prelim_ref_init();
2363 goto free_delayed_ref;
2365 err = btrfs_end_io_wq_init();
2367 goto free_prelim_ref;
2369 err = btrfs_interface_init();
2371 goto free_end_io_wq;
2373 btrfs_init_lockdep();
2375 btrfs_print_mod_info();
2377 err = btrfs_run_sanity_tests();
2379 goto unregister_ioctl;
2381 err = register_filesystem(&btrfs_fs_type);
2383 goto unregister_ioctl;
2388 btrfs_interface_exit();
2390 btrfs_end_io_wq_exit();
2392 btrfs_prelim_ref_exit();
2394 btrfs_delayed_ref_exit();
2396 btrfs_auto_defrag_exit();
2398 btrfs_delayed_inode_exit();
2400 ordered_data_exit();
2406 btrfs_destroy_cachep();
2408 btrfs_exit_compress();
2415 static void __exit exit_btrfs_fs(void)
2417 btrfs_destroy_cachep();
2418 btrfs_delayed_ref_exit();
2419 btrfs_auto_defrag_exit();
2420 btrfs_delayed_inode_exit();
2421 btrfs_prelim_ref_exit();
2422 ordered_data_exit();
2425 btrfs_interface_exit();
2426 btrfs_end_io_wq_exit();
2427 unregister_filesystem(&btrfs_fs_type);
2429 btrfs_cleanup_fs_uuids();
2430 btrfs_exit_compress();
2434 late_initcall(init_btrfs_fs);
2435 module_exit(exit_btrfs_fs)
2437 MODULE_LICENSE("GPL");