2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
61 #include "tests/btrfs-tests.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
72 const char *btrfs_decode_error(int errno)
74 char *errstr = "unknown";
78 errstr = "IO failure";
81 errstr = "Out of memory";
84 errstr = "Readonly filesystem";
87 errstr = "Object already exists";
90 errstr = "No space left";
93 errstr = "No such entry";
100 /* btrfs handle error by forcing the filesystem readonly */
101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
103 struct super_block *sb = fs_info->sb;
105 if (sb->s_flags & MS_RDONLY)
108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
109 sb->s_flags |= MS_RDONLY;
110 btrfs_info(fs_info, "forced readonly");
112 * Note that a running device replace operation is not
113 * canceled here although there is no way to update
114 * the progress. It would add the risk of a deadlock,
115 * therefore the canceling is omitted. The only penalty
116 * is that some I/O remains active until the procedure
117 * completes. The next time when the filesystem is
118 * mounted writeable again, the device replace
119 * operation continues.
125 * __btrfs_handle_fs_error decodes expected errors from the caller and
126 * invokes the approciate error response.
129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
130 unsigned int line, int errno, const char *fmt, ...)
132 struct super_block *sb = fs_info->sb;
138 * Special case: if the error is EROFS, and we're already
139 * under MS_RDONLY, then it is safe here.
141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
145 errstr = btrfs_decode_error(errno);
147 struct va_format vaf;
154 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
155 sb->s_id, function, line, errno, errstr, &vaf);
158 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
164 * Today we only save the error info to memory. Long term we'll
165 * also send it down to the disk
167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
175 static const char * const logtypes[] = {
188 * Use one ratelimit state per log level so that a flood of less important
189 * messages doesn't cause more important ones to be dropped.
191 static struct ratelimit_state printk_limits[] = {
192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
204 struct super_block *sb = fs_info->sb;
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1];
206 struct va_format vaf;
208 const char *type = NULL;
210 struct ratelimit_state *ratelimit;
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
229 ratelimit = &printk_limits[4];
235 if (__ratelimit(ratelimit))
236 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
243 * We only mark the transaction aborted and then set the file system read-only.
244 * This will prevent new transactions from starting or trying to join this
247 * This means that error recovery at the call site is limited to freeing
248 * any local memory allocations and passing the error code up without
249 * further cleanup. The transaction should complete as it normally would
250 * in the call path but will return -EIO.
252 * We'll complete the cleanup in btrfs_end_transaction and
253 * btrfs_commit_transaction.
256 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
257 const char *function,
258 unsigned int line, int errno)
260 struct btrfs_fs_info *fs_info = trans->fs_info;
262 trans->aborted = errno;
263 /* Nothing used. The other threads that have joined this
264 * transaction may be able to continue. */
265 if (!trans->dirty && list_empty(&trans->new_bgs)) {
268 errstr = btrfs_decode_error(errno);
270 "%s:%d: Aborting unused transaction(%s).",
271 function, line, errstr);
274 ACCESS_ONCE(trans->transaction->aborted) = errno;
275 /* Wake up anybody who may be waiting on this transaction */
276 wake_up(&fs_info->transaction_wait);
277 wake_up(&fs_info->transaction_blocked_wait);
278 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
281 * __btrfs_panic decodes unexpected, fatal errors from the caller,
282 * issues an alert, and either panics or BUGs, depending on mount options.
285 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
286 unsigned int line, int errno, const char *fmt, ...)
288 char *s_id = "<unknown>";
290 struct va_format vaf = { .fmt = fmt };
294 s_id = fs_info->sb->s_id;
299 errstr = btrfs_decode_error(errno);
300 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
301 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
302 s_id, function, line, &vaf, errno, errstr);
304 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
305 function, line, &vaf, errno, errstr);
307 /* Caller calls BUG() */
310 static void btrfs_put_super(struct super_block *sb)
312 close_ctree(btrfs_sb(sb)->tree_root);
316 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
317 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
318 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
319 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
320 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
321 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
322 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
323 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
324 Opt_skip_balance, Opt_check_integrity,
325 Opt_check_integrity_including_extent_data,
326 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
327 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
328 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
329 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
330 Opt_nologreplay, Opt_norecovery,
331 #ifdef CONFIG_BTRFS_DEBUG
332 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
337 static const match_table_t tokens = {
338 {Opt_degraded, "degraded"},
339 {Opt_subvol, "subvol=%s"},
340 {Opt_subvolid, "subvolid=%s"},
341 {Opt_device, "device=%s"},
342 {Opt_nodatasum, "nodatasum"},
343 {Opt_datasum, "datasum"},
344 {Opt_nodatacow, "nodatacow"},
345 {Opt_datacow, "datacow"},
346 {Opt_nobarrier, "nobarrier"},
347 {Opt_barrier, "barrier"},
348 {Opt_max_inline, "max_inline=%s"},
349 {Opt_alloc_start, "alloc_start=%s"},
350 {Opt_thread_pool, "thread_pool=%d"},
351 {Opt_compress, "compress"},
352 {Opt_compress_type, "compress=%s"},
353 {Opt_compress_force, "compress-force"},
354 {Opt_compress_force_type, "compress-force=%s"},
356 {Opt_ssd_spread, "ssd_spread"},
357 {Opt_nossd, "nossd"},
359 {Opt_noacl, "noacl"},
360 {Opt_notreelog, "notreelog"},
361 {Opt_treelog, "treelog"},
362 {Opt_nologreplay, "nologreplay"},
363 {Opt_norecovery, "norecovery"},
364 {Opt_flushoncommit, "flushoncommit"},
365 {Opt_noflushoncommit, "noflushoncommit"},
366 {Opt_ratio, "metadata_ratio=%d"},
367 {Opt_discard, "discard"},
368 {Opt_nodiscard, "nodiscard"},
369 {Opt_space_cache, "space_cache"},
370 {Opt_space_cache_version, "space_cache=%s"},
371 {Opt_clear_cache, "clear_cache"},
372 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
373 {Opt_enospc_debug, "enospc_debug"},
374 {Opt_noenospc_debug, "noenospc_debug"},
375 {Opt_subvolrootid, "subvolrootid=%d"},
376 {Opt_defrag, "autodefrag"},
377 {Opt_nodefrag, "noautodefrag"},
378 {Opt_inode_cache, "inode_cache"},
379 {Opt_noinode_cache, "noinode_cache"},
380 {Opt_no_space_cache, "nospace_cache"},
381 {Opt_recovery, "recovery"}, /* deprecated */
382 {Opt_usebackuproot, "usebackuproot"},
383 {Opt_skip_balance, "skip_balance"},
384 {Opt_check_integrity, "check_int"},
385 {Opt_check_integrity_including_extent_data, "check_int_data"},
386 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
387 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
388 {Opt_fatal_errors, "fatal_errors=%s"},
389 {Opt_commit_interval, "commit=%d"},
390 #ifdef CONFIG_BTRFS_DEBUG
391 {Opt_fragment_data, "fragment=data"},
392 {Opt_fragment_metadata, "fragment=metadata"},
393 {Opt_fragment_all, "fragment=all"},
399 * Regular mount options parser. Everything that is needed only when
400 * reading in a new superblock is parsed here.
401 * XXX JDM: This needs to be cleaned up for remount.
403 int btrfs_parse_options(struct btrfs_root *root, char *options,
404 unsigned long new_flags)
406 struct btrfs_fs_info *info = root->fs_info;
407 substring_t args[MAX_OPT_ARGS];
408 char *p, *num, *orig = NULL;
413 bool compress_force = false;
414 enum btrfs_compression_type saved_compress_type;
415 bool saved_compress_force;
418 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
419 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
420 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
422 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
425 * Even the options are empty, we still need to do extra check
432 * strsep changes the string, duplicate it because parse_options
435 options = kstrdup(options, GFP_NOFS);
441 while ((p = strsep(&options, ",")) != NULL) {
446 token = match_token(p, tokens, args);
449 btrfs_info(root->fs_info, "allowing degraded mounts");
450 btrfs_set_opt(info->mount_opt, DEGRADED);
454 case Opt_subvolrootid:
457 * These are parsed by btrfs_parse_early_options
458 * and can be happily ignored here.
462 btrfs_set_and_info(info, NODATASUM,
463 "setting nodatasum");
466 if (btrfs_test_opt(info, NODATASUM)) {
467 if (btrfs_test_opt(info, NODATACOW))
468 btrfs_info(root->fs_info,
469 "setting datasum, datacow enabled");
471 btrfs_info(root->fs_info,
474 btrfs_clear_opt(info->mount_opt, NODATACOW);
475 btrfs_clear_opt(info->mount_opt, NODATASUM);
478 if (!btrfs_test_opt(info, NODATACOW)) {
479 if (!btrfs_test_opt(info, COMPRESS) ||
480 !btrfs_test_opt(info, FORCE_COMPRESS)) {
481 btrfs_info(root->fs_info,
482 "setting nodatacow, compression disabled");
484 btrfs_info(root->fs_info,
485 "setting nodatacow");
488 btrfs_clear_opt(info->mount_opt, COMPRESS);
489 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
490 btrfs_set_opt(info->mount_opt, NODATACOW);
491 btrfs_set_opt(info->mount_opt, NODATASUM);
494 btrfs_clear_and_info(info, NODATACOW,
497 case Opt_compress_force:
498 case Opt_compress_force_type:
499 compress_force = true;
502 case Opt_compress_type:
503 saved_compress_type = btrfs_test_opt(info,
505 info->compress_type : BTRFS_COMPRESS_NONE;
506 saved_compress_force =
507 btrfs_test_opt(info, FORCE_COMPRESS);
508 if (token == Opt_compress ||
509 token == Opt_compress_force ||
510 strcmp(args[0].from, "zlib") == 0) {
511 compress_type = "zlib";
512 info->compress_type = BTRFS_COMPRESS_ZLIB;
513 btrfs_set_opt(info->mount_opt, COMPRESS);
514 btrfs_clear_opt(info->mount_opt, NODATACOW);
515 btrfs_clear_opt(info->mount_opt, NODATASUM);
517 } else if (strcmp(args[0].from, "lzo") == 0) {
518 compress_type = "lzo";
519 info->compress_type = BTRFS_COMPRESS_LZO;
520 btrfs_set_opt(info->mount_opt, COMPRESS);
521 btrfs_clear_opt(info->mount_opt, NODATACOW);
522 btrfs_clear_opt(info->mount_opt, NODATASUM);
523 btrfs_set_fs_incompat(info, COMPRESS_LZO);
525 } else if (strncmp(args[0].from, "no", 2) == 0) {
526 compress_type = "no";
527 btrfs_clear_opt(info->mount_opt, COMPRESS);
528 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
529 compress_force = false;
536 if (compress_force) {
537 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
540 * If we remount from compress-force=xxx to
541 * compress=xxx, we need clear FORCE_COMPRESS
542 * flag, otherwise, there is no way for users
543 * to disable forcible compression separately.
545 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
547 if ((btrfs_test_opt(info, COMPRESS) &&
548 (info->compress_type != saved_compress_type ||
549 compress_force != saved_compress_force)) ||
550 (!btrfs_test_opt(info, COMPRESS) &&
552 btrfs_info(root->fs_info,
554 (compress_force) ? "force" : "use",
557 compress_force = false;
560 btrfs_set_and_info(info, SSD,
561 "use ssd allocation scheme");
564 btrfs_set_and_info(info, SSD_SPREAD,
565 "use spread ssd allocation scheme");
566 btrfs_set_opt(info->mount_opt, SSD);
569 btrfs_set_and_info(info, NOSSD,
570 "not using ssd allocation scheme");
571 btrfs_clear_opt(info->mount_opt, SSD);
574 btrfs_clear_and_info(info, NOBARRIER,
575 "turning on barriers");
578 btrfs_set_and_info(info, NOBARRIER,
579 "turning off barriers");
581 case Opt_thread_pool:
582 ret = match_int(&args[0], &intarg);
585 } else if (intarg > 0) {
586 info->thread_pool_size = intarg;
593 num = match_strdup(&args[0]);
595 info->max_inline = memparse(num, NULL);
598 if (info->max_inline) {
599 info->max_inline = min_t(u64,
603 btrfs_info(root->fs_info, "max_inline at %llu",
610 case Opt_alloc_start:
611 num = match_strdup(&args[0]);
613 mutex_lock(&info->chunk_mutex);
614 info->alloc_start = memparse(num, NULL);
615 mutex_unlock(&info->chunk_mutex);
617 btrfs_info(root->fs_info,
618 "allocations start at %llu",
626 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
627 root->fs_info->sb->s_flags |= MS_POSIXACL;
630 btrfs_err(root->fs_info,
631 "support for ACL not compiled in!");
636 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
639 btrfs_set_and_info(info, NOTREELOG,
640 "disabling tree log");
643 btrfs_clear_and_info(info, NOTREELOG,
644 "enabling tree log");
647 case Opt_nologreplay:
648 btrfs_set_and_info(info, NOLOGREPLAY,
649 "disabling log replay at mount time");
651 case Opt_flushoncommit:
652 btrfs_set_and_info(info, FLUSHONCOMMIT,
653 "turning on flush-on-commit");
655 case Opt_noflushoncommit:
656 btrfs_clear_and_info(info, FLUSHONCOMMIT,
657 "turning off flush-on-commit");
660 ret = match_int(&args[0], &intarg);
663 } else if (intarg >= 0) {
664 info->metadata_ratio = intarg;
665 btrfs_info(root->fs_info, "metadata ratio %d",
666 info->metadata_ratio);
673 btrfs_set_and_info(info, DISCARD,
674 "turning on discard");
677 btrfs_clear_and_info(info, DISCARD,
678 "turning off discard");
680 case Opt_space_cache:
681 case Opt_space_cache_version:
682 if (token == Opt_space_cache ||
683 strcmp(args[0].from, "v1") == 0) {
684 btrfs_clear_opt(root->fs_info->mount_opt,
686 btrfs_set_and_info(info, SPACE_CACHE,
687 "enabling disk space caching");
688 } else if (strcmp(args[0].from, "v2") == 0) {
689 btrfs_clear_opt(root->fs_info->mount_opt,
691 btrfs_set_and_info(info,
693 "enabling free space tree");
699 case Opt_rescan_uuid_tree:
700 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
702 case Opt_no_space_cache:
703 if (btrfs_test_opt(info, SPACE_CACHE)) {
704 btrfs_clear_and_info(info,
706 "disabling disk space caching");
708 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
709 btrfs_clear_and_info(info,
711 "disabling free space tree");
714 case Opt_inode_cache:
715 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
716 "enabling inode map caching");
718 case Opt_noinode_cache:
719 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
720 "disabling inode map caching");
722 case Opt_clear_cache:
723 btrfs_set_and_info(info, CLEAR_CACHE,
724 "force clearing of disk cache");
726 case Opt_user_subvol_rm_allowed:
727 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
729 case Opt_enospc_debug:
730 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
732 case Opt_noenospc_debug:
733 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
736 btrfs_set_and_info(info, AUTO_DEFRAG,
737 "enabling auto defrag");
740 btrfs_clear_and_info(info, AUTO_DEFRAG,
741 "disabling auto defrag");
744 btrfs_warn(root->fs_info,
745 "'recovery' is deprecated, use 'usebackuproot' instead");
746 case Opt_usebackuproot:
747 btrfs_info(root->fs_info,
748 "trying to use backup root at mount time");
749 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
751 case Opt_skip_balance:
752 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
754 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
755 case Opt_check_integrity_including_extent_data:
756 btrfs_info(root->fs_info,
757 "enabling check integrity including extent data");
758 btrfs_set_opt(info->mount_opt,
759 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
760 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
762 case Opt_check_integrity:
763 btrfs_info(root->fs_info, "enabling check integrity");
764 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
766 case Opt_check_integrity_print_mask:
767 ret = match_int(&args[0], &intarg);
770 } else if (intarg >= 0) {
771 info->check_integrity_print_mask = intarg;
772 btrfs_info(root->fs_info,
773 "check_integrity_print_mask 0x%x",
774 info->check_integrity_print_mask);
781 case Opt_check_integrity_including_extent_data:
782 case Opt_check_integrity:
783 case Opt_check_integrity_print_mask:
784 btrfs_err(root->fs_info,
785 "support for check_integrity* not compiled in!");
789 case Opt_fatal_errors:
790 if (strcmp(args[0].from, "panic") == 0)
791 btrfs_set_opt(info->mount_opt,
792 PANIC_ON_FATAL_ERROR);
793 else if (strcmp(args[0].from, "bug") == 0)
794 btrfs_clear_opt(info->mount_opt,
795 PANIC_ON_FATAL_ERROR);
801 case Opt_commit_interval:
803 ret = match_int(&args[0], &intarg);
805 btrfs_err(root->fs_info,
806 "invalid commit interval");
812 btrfs_warn(root->fs_info,
813 "excessive commit interval %d",
816 info->commit_interval = intarg;
818 btrfs_info(root->fs_info,
819 "using default commit interval %ds",
820 BTRFS_DEFAULT_COMMIT_INTERVAL);
821 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
824 #ifdef CONFIG_BTRFS_DEBUG
825 case Opt_fragment_all:
826 btrfs_info(root->fs_info, "fragmenting all space");
827 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
828 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
830 case Opt_fragment_metadata:
831 btrfs_info(root->fs_info, "fragmenting metadata");
832 btrfs_set_opt(info->mount_opt,
835 case Opt_fragment_data:
836 btrfs_info(root->fs_info, "fragmenting data");
837 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
841 btrfs_info(root->fs_info,
842 "unrecognized mount option '%s'", p);
851 * Extra check for current option against current flag
853 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
854 btrfs_err(root->fs_info,
855 "nologreplay must be used with ro mount option");
859 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
860 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
861 !btrfs_test_opt(info, CLEAR_CACHE)) {
862 btrfs_err(root->fs_info, "cannot disable free space tree");
866 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
867 btrfs_info(root->fs_info, "disk space caching is enabled");
868 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
869 btrfs_info(root->fs_info, "using free space tree");
875 * Parse mount options that are required early in the mount process.
877 * All other options will be parsed on much later in the mount process and
878 * only when we need to allocate a new super block.
880 static int btrfs_parse_early_options(const char *options, fmode_t flags,
881 void *holder, char **subvol_name, u64 *subvol_objectid,
882 struct btrfs_fs_devices **fs_devices)
884 substring_t args[MAX_OPT_ARGS];
885 char *device_name, *opts, *orig, *p;
893 * strsep changes the string, duplicate it because parse_options
896 opts = kstrdup(options, GFP_KERNEL);
901 while ((p = strsep(&opts, ",")) != NULL) {
906 token = match_token(p, tokens, args);
910 *subvol_name = match_strdup(&args[0]);
917 num = match_strdup(&args[0]);
919 *subvol_objectid = memparse(num, NULL);
921 /* we want the original fs_tree */
922 if (!*subvol_objectid)
924 BTRFS_FS_TREE_OBJECTID;
930 case Opt_subvolrootid:
931 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
934 device_name = match_strdup(&args[0]);
939 error = btrfs_scan_one_device(device_name,
940 flags, holder, fs_devices);
955 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
958 struct btrfs_root *root = fs_info->tree_root;
959 struct btrfs_root *fs_root;
960 struct btrfs_root_ref *root_ref;
961 struct btrfs_inode_ref *inode_ref;
962 struct btrfs_key key;
963 struct btrfs_path *path = NULL;
964 char *name = NULL, *ptr;
969 path = btrfs_alloc_path();
974 path->leave_spinning = 1;
976 name = kmalloc(PATH_MAX, GFP_NOFS);
981 ptr = name + PATH_MAX - 1;
985 * Walk up the subvolume trees in the tree of tree roots by root
986 * backrefs until we hit the top-level subvolume.
988 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
989 key.objectid = subvol_objectid;
990 key.type = BTRFS_ROOT_BACKREF_KEY;
991 key.offset = (u64)-1;
993 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
996 } else if (ret > 0) {
997 ret = btrfs_previous_item(root, path, subvol_objectid,
998 BTRFS_ROOT_BACKREF_KEY);
1001 } else if (ret > 0) {
1007 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1008 subvol_objectid = key.offset;
1010 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1011 struct btrfs_root_ref);
1012 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1015 ret = -ENAMETOOLONG;
1018 read_extent_buffer(path->nodes[0], ptr + 1,
1019 (unsigned long)(root_ref + 1), len);
1021 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1022 btrfs_release_path(path);
1024 key.objectid = subvol_objectid;
1025 key.type = BTRFS_ROOT_ITEM_KEY;
1026 key.offset = (u64)-1;
1027 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1028 if (IS_ERR(fs_root)) {
1029 ret = PTR_ERR(fs_root);
1034 * Walk up the filesystem tree by inode refs until we hit the
1037 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1038 key.objectid = dirid;
1039 key.type = BTRFS_INODE_REF_KEY;
1040 key.offset = (u64)-1;
1042 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1045 } else if (ret > 0) {
1046 ret = btrfs_previous_item(fs_root, path, dirid,
1047 BTRFS_INODE_REF_KEY);
1050 } else if (ret > 0) {
1056 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1059 inode_ref = btrfs_item_ptr(path->nodes[0],
1061 struct btrfs_inode_ref);
1062 len = btrfs_inode_ref_name_len(path->nodes[0],
1066 ret = -ENAMETOOLONG;
1069 read_extent_buffer(path->nodes[0], ptr + 1,
1070 (unsigned long)(inode_ref + 1), len);
1072 btrfs_release_path(path);
1076 btrfs_free_path(path);
1077 if (ptr == name + PATH_MAX - 1) {
1081 memmove(name, ptr, name + PATH_MAX - ptr);
1086 btrfs_free_path(path);
1088 return ERR_PTR(ret);
1091 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1093 struct btrfs_root *root = fs_info->tree_root;
1094 struct btrfs_dir_item *di;
1095 struct btrfs_path *path;
1096 struct btrfs_key location;
1099 path = btrfs_alloc_path();
1102 path->leave_spinning = 1;
1105 * Find the "default" dir item which points to the root item that we
1106 * will mount by default if we haven't been given a specific subvolume
1109 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1110 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1112 btrfs_free_path(path);
1117 * Ok the default dir item isn't there. This is weird since
1118 * it's always been there, but don't freak out, just try and
1119 * mount the top-level subvolume.
1121 btrfs_free_path(path);
1122 *objectid = BTRFS_FS_TREE_OBJECTID;
1126 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1127 btrfs_free_path(path);
1128 *objectid = location.objectid;
1132 static int btrfs_fill_super(struct super_block *sb,
1133 struct btrfs_fs_devices *fs_devices,
1134 void *data, int silent)
1136 struct inode *inode;
1137 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1138 struct btrfs_key key;
1141 sb->s_maxbytes = MAX_LFS_FILESIZE;
1142 sb->s_magic = BTRFS_SUPER_MAGIC;
1143 sb->s_op = &btrfs_super_ops;
1144 sb->s_d_op = &btrfs_dentry_operations;
1145 sb->s_export_op = &btrfs_export_ops;
1146 sb->s_xattr = btrfs_xattr_handlers;
1147 sb->s_time_gran = 1;
1148 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1149 sb->s_flags |= MS_POSIXACL;
1151 sb->s_flags |= MS_I_VERSION;
1152 sb->s_iflags |= SB_I_CGROUPWB;
1153 err = open_ctree(sb, fs_devices, (char *)data);
1155 btrfs_err(fs_info, "open_ctree failed");
1159 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1160 key.type = BTRFS_INODE_ITEM_KEY;
1162 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1163 if (IS_ERR(inode)) {
1164 err = PTR_ERR(inode);
1168 sb->s_root = d_make_root(inode);
1174 save_mount_options(sb, data);
1175 cleancache_init_fs(sb);
1176 sb->s_flags |= MS_ACTIVE;
1180 close_ctree(fs_info->tree_root);
1184 int btrfs_sync_fs(struct super_block *sb, int wait)
1186 struct btrfs_trans_handle *trans;
1187 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1188 struct btrfs_root *root = fs_info->tree_root;
1190 trace_btrfs_sync_fs(fs_info, wait);
1193 filemap_flush(fs_info->btree_inode->i_mapping);
1197 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1199 trans = btrfs_attach_transaction_barrier(root);
1200 if (IS_ERR(trans)) {
1201 /* no transaction, don't bother */
1202 if (PTR_ERR(trans) == -ENOENT) {
1204 * Exit unless we have some pending changes
1205 * that need to go through commit
1207 if (fs_info->pending_changes == 0)
1210 * A non-blocking test if the fs is frozen. We must not
1211 * start a new transaction here otherwise a deadlock
1212 * happens. The pending operations are delayed to the
1213 * next commit after thawing.
1215 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1216 __sb_end_write(sb, SB_FREEZE_WRITE);
1219 trans = btrfs_start_transaction(root, 0);
1222 return PTR_ERR(trans);
1224 return btrfs_commit_transaction(trans, root);
1227 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1229 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1230 struct btrfs_root *root = info->tree_root;
1231 char *compress_type;
1233 if (btrfs_test_opt(info, DEGRADED))
1234 seq_puts(seq, ",degraded");
1235 if (btrfs_test_opt(info, NODATASUM))
1236 seq_puts(seq, ",nodatasum");
1237 if (btrfs_test_opt(info, NODATACOW))
1238 seq_puts(seq, ",nodatacow");
1239 if (btrfs_test_opt(info, NOBARRIER))
1240 seq_puts(seq, ",nobarrier");
1241 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1242 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1243 if (info->alloc_start != 0)
1244 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1245 if (info->thread_pool_size != min_t(unsigned long,
1246 num_online_cpus() + 2, 8))
1247 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1248 if (btrfs_test_opt(info, COMPRESS)) {
1249 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1250 compress_type = "zlib";
1252 compress_type = "lzo";
1253 if (btrfs_test_opt(info, FORCE_COMPRESS))
1254 seq_printf(seq, ",compress-force=%s", compress_type);
1256 seq_printf(seq, ",compress=%s", compress_type);
1258 if (btrfs_test_opt(info, NOSSD))
1259 seq_puts(seq, ",nossd");
1260 if (btrfs_test_opt(info, SSD_SPREAD))
1261 seq_puts(seq, ",ssd_spread");
1262 else if (btrfs_test_opt(info, SSD))
1263 seq_puts(seq, ",ssd");
1264 if (btrfs_test_opt(info, NOTREELOG))
1265 seq_puts(seq, ",notreelog");
1266 if (btrfs_test_opt(info, NOLOGREPLAY))
1267 seq_puts(seq, ",nologreplay");
1268 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1269 seq_puts(seq, ",flushoncommit");
1270 if (btrfs_test_opt(info, DISCARD))
1271 seq_puts(seq, ",discard");
1272 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1273 seq_puts(seq, ",noacl");
1274 if (btrfs_test_opt(info, SPACE_CACHE))
1275 seq_puts(seq, ",space_cache");
1276 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1277 seq_puts(seq, ",space_cache=v2");
1279 seq_puts(seq, ",nospace_cache");
1280 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1281 seq_puts(seq, ",rescan_uuid_tree");
1282 if (btrfs_test_opt(info, CLEAR_CACHE))
1283 seq_puts(seq, ",clear_cache");
1284 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1285 seq_puts(seq, ",user_subvol_rm_allowed");
1286 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1287 seq_puts(seq, ",enospc_debug");
1288 if (btrfs_test_opt(info, AUTO_DEFRAG))
1289 seq_puts(seq, ",autodefrag");
1290 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1291 seq_puts(seq, ",inode_cache");
1292 if (btrfs_test_opt(info, SKIP_BALANCE))
1293 seq_puts(seq, ",skip_balance");
1294 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1295 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1296 seq_puts(seq, ",check_int_data");
1297 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1298 seq_puts(seq, ",check_int");
1299 if (info->check_integrity_print_mask)
1300 seq_printf(seq, ",check_int_print_mask=%d",
1301 info->check_integrity_print_mask);
1303 if (info->metadata_ratio)
1304 seq_printf(seq, ",metadata_ratio=%d",
1305 info->metadata_ratio);
1306 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1307 seq_puts(seq, ",fatal_errors=panic");
1308 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1309 seq_printf(seq, ",commit=%d", info->commit_interval);
1310 #ifdef CONFIG_BTRFS_DEBUG
1311 if (btrfs_test_opt(info, FRAGMENT_DATA))
1312 seq_puts(seq, ",fragment=data");
1313 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1314 seq_puts(seq, ",fragment=metadata");
1316 seq_printf(seq, ",subvolid=%llu",
1317 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1318 seq_puts(seq, ",subvol=");
1319 seq_dentry(seq, dentry, " \t\n\\");
1323 static int btrfs_test_super(struct super_block *s, void *data)
1325 struct btrfs_fs_info *p = data;
1326 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1328 return fs_info->fs_devices == p->fs_devices;
1331 static int btrfs_set_super(struct super_block *s, void *data)
1333 int err = set_anon_super(s, data);
1335 s->s_fs_info = data;
1340 * subvolumes are identified by ino 256
1342 static inline int is_subvolume_inode(struct inode *inode)
1344 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1350 * This will add subvolid=0 to the argument string while removing any subvol=
1351 * and subvolid= arguments to make sure we get the top-level root for path
1352 * walking to the subvol we want.
1354 static char *setup_root_args(char *args)
1356 char *buf, *dst, *sep;
1359 return kstrdup("subvolid=0", GFP_NOFS);
1361 /* The worst case is that we add ",subvolid=0" to the end. */
1362 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1367 sep = strchrnul(args, ',');
1368 if (!strstarts(args, "subvol=") &&
1369 !strstarts(args, "subvolid=")) {
1370 memcpy(dst, args, sep - args);
1379 strcpy(dst, "subvolid=0");
1384 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1385 int flags, const char *device_name,
1388 struct dentry *root;
1389 struct vfsmount *mnt = NULL;
1393 newargs = setup_root_args(data);
1395 root = ERR_PTR(-ENOMEM);
1399 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1400 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1401 if (flags & MS_RDONLY) {
1402 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1403 device_name, newargs);
1405 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1406 device_name, newargs);
1408 root = ERR_CAST(mnt);
1413 down_write(&mnt->mnt_sb->s_umount);
1414 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1415 up_write(&mnt->mnt_sb->s_umount);
1417 root = ERR_PTR(ret);
1423 root = ERR_CAST(mnt);
1429 if (!subvol_objectid) {
1430 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1433 root = ERR_PTR(ret);
1437 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1439 if (IS_ERR(subvol_name)) {
1440 root = ERR_CAST(subvol_name);
1447 root = mount_subtree(mnt, subvol_name);
1448 /* mount_subtree() drops our reference on the vfsmount. */
1451 if (!IS_ERR(root)) {
1452 struct super_block *s = root->d_sb;
1453 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1454 struct inode *root_inode = d_inode(root);
1455 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1458 if (!is_subvolume_inode(root_inode)) {
1459 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1463 if (subvol_objectid && root_objectid != subvol_objectid) {
1465 * This will also catch a race condition where a
1466 * subvolume which was passed by ID is renamed and
1467 * another subvolume is renamed over the old location.
1470 "subvol '%s' does not match subvolid %llu",
1471 subvol_name, subvol_objectid);
1476 root = ERR_PTR(ret);
1477 deactivate_locked_super(s);
1488 static int parse_security_options(char *orig_opts,
1489 struct security_mnt_opts *sec_opts)
1491 char *secdata = NULL;
1494 secdata = alloc_secdata();
1497 ret = security_sb_copy_data(orig_opts, secdata);
1499 free_secdata(secdata);
1502 ret = security_sb_parse_opts_str(secdata, sec_opts);
1503 free_secdata(secdata);
1507 static int setup_security_options(struct btrfs_fs_info *fs_info,
1508 struct super_block *sb,
1509 struct security_mnt_opts *sec_opts)
1514 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1517 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1521 #ifdef CONFIG_SECURITY
1522 if (!fs_info->security_opts.num_mnt_opts) {
1523 /* first time security setup, copy sec_opts to fs_info */
1524 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1527 * Since SELinux (the only one supporting security_mnt_opts)
1528 * does NOT support changing context during remount/mount of
1529 * the same sb, this must be the same or part of the same
1530 * security options, just free it.
1532 security_free_mnt_opts(sec_opts);
1539 * Find a superblock for the given device / mount point.
1541 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1542 * for multiple device setup. Make sure to keep it in sync.
1544 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1545 const char *device_name, void *data)
1547 struct block_device *bdev = NULL;
1548 struct super_block *s;
1549 struct btrfs_fs_devices *fs_devices = NULL;
1550 struct btrfs_fs_info *fs_info = NULL;
1551 struct security_mnt_opts new_sec_opts;
1552 fmode_t mode = FMODE_READ;
1553 char *subvol_name = NULL;
1554 u64 subvol_objectid = 0;
1557 if (!(flags & MS_RDONLY))
1558 mode |= FMODE_WRITE;
1560 error = btrfs_parse_early_options(data, mode, fs_type,
1561 &subvol_name, &subvol_objectid,
1565 return ERR_PTR(error);
1568 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1569 /* mount_subvol() will free subvol_name. */
1570 return mount_subvol(subvol_name, subvol_objectid, flags,
1574 security_init_mnt_opts(&new_sec_opts);
1576 error = parse_security_options(data, &new_sec_opts);
1578 return ERR_PTR(error);
1581 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1583 goto error_sec_opts;
1586 * Setup a dummy root and fs_info for test/set super. This is because
1587 * we don't actually fill this stuff out until open_ctree, but we need
1588 * it for searching for existing supers, so this lets us do that and
1589 * then open_ctree will properly initialize everything later.
1591 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1594 goto error_sec_opts;
1597 fs_info->fs_devices = fs_devices;
1599 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1600 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1601 security_init_mnt_opts(&fs_info->security_opts);
1602 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1607 error = btrfs_open_devices(fs_devices, mode, fs_type);
1611 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1613 goto error_close_devices;
1616 bdev = fs_devices->latest_bdev;
1617 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1621 goto error_close_devices;
1625 btrfs_close_devices(fs_devices);
1626 free_fs_info(fs_info);
1627 if ((flags ^ s->s_flags) & MS_RDONLY)
1630 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1631 btrfs_sb(s)->bdev_holder = fs_type;
1632 error = btrfs_fill_super(s, fs_devices, data,
1633 flags & MS_SILENT ? 1 : 0);
1636 deactivate_locked_super(s);
1637 goto error_sec_opts;
1640 fs_info = btrfs_sb(s);
1641 error = setup_security_options(fs_info, s, &new_sec_opts);
1643 deactivate_locked_super(s);
1644 goto error_sec_opts;
1647 return dget(s->s_root);
1649 error_close_devices:
1650 btrfs_close_devices(fs_devices);
1652 free_fs_info(fs_info);
1654 security_free_mnt_opts(&new_sec_opts);
1655 return ERR_PTR(error);
1658 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1659 int new_pool_size, int 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 & MS_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 & MS_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) ||
1713 (fs_info->sb->s_flags & MS_RDONLY))) {
1714 btrfs_cleanup_defrag_inodes(fs_info);
1717 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1720 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1722 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1723 struct btrfs_root *root = fs_info->tree_root;
1724 unsigned old_flags = sb->s_flags;
1725 unsigned long old_opts = fs_info->mount_opt;
1726 unsigned long old_compress_type = fs_info->compress_type;
1727 u64 old_max_inline = fs_info->max_inline;
1728 u64 old_alloc_start = fs_info->alloc_start;
1729 int old_thread_pool_size = fs_info->thread_pool_size;
1730 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1733 sync_filesystem(sb);
1734 btrfs_remount_prepare(fs_info);
1737 struct security_mnt_opts new_sec_opts;
1739 security_init_mnt_opts(&new_sec_opts);
1740 ret = parse_security_options(data, &new_sec_opts);
1743 ret = setup_security_options(fs_info, sb,
1746 security_free_mnt_opts(&new_sec_opts);
1751 ret = btrfs_parse_options(root, data, *flags);
1757 btrfs_remount_begin(fs_info, old_opts, *flags);
1758 btrfs_resize_thread_pool(fs_info,
1759 fs_info->thread_pool_size, old_thread_pool_size);
1761 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1764 if (*flags & MS_RDONLY) {
1766 * this also happens on 'umount -rf' or on shutdown, when
1767 * the filesystem is busy.
1769 cancel_work_sync(&fs_info->async_reclaim_work);
1771 /* wait for the uuid_scan task to finish */
1772 down(&fs_info->uuid_tree_rescan_sem);
1773 /* avoid complains from lockdep et al. */
1774 up(&fs_info->uuid_tree_rescan_sem);
1776 sb->s_flags |= MS_RDONLY;
1779 * Setting MS_RDONLY will put the cleaner thread to
1780 * sleep at the next loop if it's already active.
1781 * If it's already asleep, we'll leave unused block
1782 * groups on disk until we're mounted read-write again
1783 * unless we clean them up here.
1785 btrfs_delete_unused_bgs(fs_info);
1787 btrfs_dev_replace_suspend_for_unmount(fs_info);
1788 btrfs_scrub_cancel(fs_info);
1789 btrfs_pause_balance(fs_info);
1791 ret = btrfs_commit_super(root);
1795 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1797 "Remounting read-write after error is not allowed");
1801 if (fs_info->fs_devices->rw_devices == 0) {
1806 if (fs_info->fs_devices->missing_devices >
1807 fs_info->num_tolerated_disk_barrier_failures &&
1808 !(*flags & MS_RDONLY)) {
1810 "too many missing devices, writeable remount is not allowed");
1815 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1820 ret = btrfs_cleanup_fs_roots(fs_info);
1824 /* recover relocation */
1825 mutex_lock(&fs_info->cleaner_mutex);
1826 ret = btrfs_recover_relocation(root);
1827 mutex_unlock(&fs_info->cleaner_mutex);
1831 ret = btrfs_resume_balance_async(fs_info);
1835 ret = btrfs_resume_dev_replace_async(fs_info);
1837 btrfs_warn(fs_info, "failed to resume dev_replace");
1841 if (!fs_info->uuid_root) {
1842 btrfs_info(fs_info, "creating UUID tree");
1843 ret = btrfs_create_uuid_tree(fs_info);
1846 "failed to create the UUID tree %d",
1851 sb->s_flags &= ~MS_RDONLY;
1853 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1856 wake_up_process(fs_info->transaction_kthread);
1857 btrfs_remount_cleanup(fs_info, old_opts);
1861 /* We've hit an error - don't reset MS_RDONLY */
1862 if (sb->s_flags & MS_RDONLY)
1863 old_flags |= MS_RDONLY;
1864 sb->s_flags = old_flags;
1865 fs_info->mount_opt = old_opts;
1866 fs_info->compress_type = old_compress_type;
1867 fs_info->max_inline = old_max_inline;
1868 mutex_lock(&fs_info->chunk_mutex);
1869 fs_info->alloc_start = old_alloc_start;
1870 mutex_unlock(&fs_info->chunk_mutex);
1871 btrfs_resize_thread_pool(fs_info,
1872 old_thread_pool_size, fs_info->thread_pool_size);
1873 fs_info->metadata_ratio = old_metadata_ratio;
1874 btrfs_remount_cleanup(fs_info, old_opts);
1878 /* Used to sort the devices by max_avail(descending sort) */
1879 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1880 const void *dev_info2)
1882 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1883 ((struct btrfs_device_info *)dev_info2)->max_avail)
1885 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1886 ((struct btrfs_device_info *)dev_info2)->max_avail)
1893 * sort the devices by max_avail, in which max free extent size of each device
1894 * is stored.(Descending Sort)
1896 static inline void btrfs_descending_sort_devices(
1897 struct btrfs_device_info *devices,
1900 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1901 btrfs_cmp_device_free_bytes, NULL);
1905 * The helper to calc the free space on the devices that can be used to store
1908 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1910 struct btrfs_fs_info *fs_info = root->fs_info;
1911 struct btrfs_device_info *devices_info;
1912 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1913 struct btrfs_device *device;
1918 u64 min_stripe_size;
1919 int min_stripes = 1, num_stripes = 1;
1920 int i = 0, nr_devices;
1924 * We aren't under the device list lock, so this is racy-ish, but good
1925 * enough for our purposes.
1927 nr_devices = fs_info->fs_devices->open_devices;
1930 nr_devices = fs_info->fs_devices->open_devices;
1938 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1943 /* calc min stripe number for data space allocation */
1944 type = btrfs_get_alloc_profile(root, 1);
1945 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1947 num_stripes = nr_devices;
1948 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1951 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1956 if (type & BTRFS_BLOCK_GROUP_DUP)
1957 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1959 min_stripe_size = BTRFS_STRIPE_LEN;
1961 if (fs_info->alloc_start)
1962 mutex_lock(&fs_devices->device_list_mutex);
1964 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1965 if (!device->in_fs_metadata || !device->bdev ||
1966 device->is_tgtdev_for_dev_replace)
1969 if (i >= nr_devices)
1972 avail_space = device->total_bytes - device->bytes_used;
1974 /* align with stripe_len */
1975 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1976 avail_space *= BTRFS_STRIPE_LEN;
1979 * In order to avoid overwriting the superblock on the drive,
1980 * btrfs starts at an offset of at least 1MB when doing chunk
1985 /* user can set the offset in fs_info->alloc_start. */
1986 if (fs_info->alloc_start &&
1987 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1988 device->total_bytes) {
1990 skip_space = max(fs_info->alloc_start, skip_space);
1993 * btrfs can not use the free space in
1994 * [0, skip_space - 1], we must subtract it from the
1995 * total. In order to implement it, we account the used
1996 * space in this range first.
1998 ret = btrfs_account_dev_extents_size(device, 0,
2002 kfree(devices_info);
2003 mutex_unlock(&fs_devices->device_list_mutex);
2009 /* calc the free space in [0, skip_space - 1] */
2010 skip_space -= used_space;
2014 * we can use the free space in [0, skip_space - 1], subtract
2015 * it from the total.
2017 if (avail_space && avail_space >= skip_space)
2018 avail_space -= skip_space;
2022 if (avail_space < min_stripe_size)
2025 devices_info[i].dev = device;
2026 devices_info[i].max_avail = avail_space;
2031 if (fs_info->alloc_start)
2032 mutex_unlock(&fs_devices->device_list_mutex);
2036 btrfs_descending_sort_devices(devices_info, nr_devices);
2040 while (nr_devices >= min_stripes) {
2041 if (num_stripes > nr_devices)
2042 num_stripes = nr_devices;
2044 if (devices_info[i].max_avail >= min_stripe_size) {
2048 avail_space += devices_info[i].max_avail * num_stripes;
2049 alloc_size = devices_info[i].max_avail;
2050 for (j = i + 1 - num_stripes; j <= i; j++)
2051 devices_info[j].max_avail -= alloc_size;
2057 kfree(devices_info);
2058 *free_bytes = avail_space;
2063 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2065 * If there's a redundant raid level at DATA block groups, use the respective
2066 * multiplier to scale the sizes.
2068 * Unused device space usage is based on simulating the chunk allocator
2069 * algorithm that respects the device sizes, order of allocations and the
2070 * 'alloc_start' value, this is a close approximation of the actual use but
2071 * there are other factors that may change the result (like a new metadata
2074 * If metadata is exhausted, f_bavail will be 0.
2076 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2078 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2079 struct btrfs_super_block *disk_super = fs_info->super_copy;
2080 struct list_head *head = &fs_info->space_info;
2081 struct btrfs_space_info *found;
2083 u64 total_free_data = 0;
2084 u64 total_free_meta = 0;
2085 int bits = dentry->d_sb->s_blocksize_bits;
2086 __be32 *fsid = (__be32 *)fs_info->fsid;
2087 unsigned factor = 1;
2088 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2094 * holding chunk_mutex to avoid allocating new chunks, holding
2095 * device_list_mutex to avoid the device being removed
2098 list_for_each_entry_rcu(found, head, list) {
2099 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2102 total_free_data += found->disk_total - found->disk_used;
2104 btrfs_account_ro_block_groups_free_space(found);
2106 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2107 if (!list_empty(&found->block_groups[i])) {
2109 case BTRFS_RAID_DUP:
2110 case BTRFS_RAID_RAID1:
2111 case BTRFS_RAID_RAID10:
2119 * Metadata in mixed block goup profiles are accounted in data
2121 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2122 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2125 total_free_meta += found->disk_total -
2129 total_used += found->disk_used;
2134 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2135 buf->f_blocks >>= bits;
2136 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2138 /* Account global block reserve as used, it's in logical size already */
2139 spin_lock(&block_rsv->lock);
2140 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2141 if (buf->f_bfree >= block_rsv->size >> bits)
2142 buf->f_bfree -= block_rsv->size >> bits;
2145 spin_unlock(&block_rsv->lock);
2147 buf->f_bavail = div_u64(total_free_data, factor);
2148 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2151 buf->f_bavail += div_u64(total_free_data, factor);
2152 buf->f_bavail = buf->f_bavail >> bits;
2155 * We calculate the remaining metadata space minus global reserve. If
2156 * this is (supposedly) smaller than zero, there's no space. But this
2157 * does not hold in practice, the exhausted state happens where's still
2158 * some positive delta. So we apply some guesswork and compare the
2159 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2161 * We probably cannot calculate the exact threshold value because this
2162 * depends on the internal reservations requested by various
2163 * operations, so some operations that consume a few metadata will
2164 * succeed even if the Avail is zero. But this is better than the other
2167 thresh = 4 * 1024 * 1024;
2169 if (!mixed && total_free_meta - thresh < block_rsv->size)
2172 buf->f_type = BTRFS_SUPER_MAGIC;
2173 buf->f_bsize = dentry->d_sb->s_blocksize;
2174 buf->f_namelen = BTRFS_NAME_LEN;
2176 /* We treat it as constant endianness (it doesn't matter _which_)
2177 because we want the fsid to come out the same whether mounted
2178 on a big-endian or little-endian host */
2179 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2180 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2181 /* Mask in the root object ID too, to disambiguate subvols */
2182 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2183 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2188 static void btrfs_kill_super(struct super_block *sb)
2190 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2191 kill_anon_super(sb);
2192 free_fs_info(fs_info);
2195 static struct file_system_type btrfs_fs_type = {
2196 .owner = THIS_MODULE,
2198 .mount = btrfs_mount,
2199 .kill_sb = btrfs_kill_super,
2200 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2202 MODULE_ALIAS_FS("btrfs");
2204 static int btrfs_control_open(struct inode *inode, struct file *file)
2207 * The control file's private_data is used to hold the
2208 * transaction when it is started and is used to keep
2209 * track of whether a transaction is already in progress.
2211 file->private_data = NULL;
2216 * used by btrfsctl to scan devices when no FS is mounted
2218 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2221 struct btrfs_ioctl_vol_args *vol;
2222 struct btrfs_fs_devices *fs_devices;
2225 if (!capable(CAP_SYS_ADMIN))
2228 vol = memdup_user((void __user *)arg, sizeof(*vol));
2230 return PTR_ERR(vol);
2233 case BTRFS_IOC_SCAN_DEV:
2234 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2235 &btrfs_fs_type, &fs_devices);
2237 case BTRFS_IOC_DEVICES_READY:
2238 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2239 &btrfs_fs_type, &fs_devices);
2242 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2244 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2245 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2253 static int btrfs_freeze(struct super_block *sb)
2255 struct btrfs_trans_handle *trans;
2256 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2258 root->fs_info->fs_frozen = 1;
2260 * We don't need a barrier here, we'll wait for any transaction that
2261 * could be in progress on other threads (and do delayed iputs that
2262 * we want to avoid on a frozen filesystem), or do the commit
2265 trans = btrfs_attach_transaction_barrier(root);
2266 if (IS_ERR(trans)) {
2267 /* no transaction, don't bother */
2268 if (PTR_ERR(trans) == -ENOENT)
2270 return PTR_ERR(trans);
2272 return btrfs_commit_transaction(trans, root);
2275 static int btrfs_unfreeze(struct super_block *sb)
2277 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2279 root->fs_info->fs_frozen = 0;
2283 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2285 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2286 struct btrfs_fs_devices *cur_devices;
2287 struct btrfs_device *dev, *first_dev = NULL;
2288 struct list_head *head;
2289 struct rcu_string *name;
2291 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2292 cur_devices = fs_info->fs_devices;
2293 while (cur_devices) {
2294 head = &cur_devices->devices;
2295 list_for_each_entry(dev, head, dev_list) {
2300 if (!first_dev || dev->devid < first_dev->devid)
2303 cur_devices = cur_devices->seed;
2308 name = rcu_dereference(first_dev->name);
2309 seq_escape(m, name->str, " \t\n\\");
2314 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2318 static const struct super_operations btrfs_super_ops = {
2319 .drop_inode = btrfs_drop_inode,
2320 .evict_inode = btrfs_evict_inode,
2321 .put_super = btrfs_put_super,
2322 .sync_fs = btrfs_sync_fs,
2323 .show_options = btrfs_show_options,
2324 .show_devname = btrfs_show_devname,
2325 .write_inode = btrfs_write_inode,
2326 .alloc_inode = btrfs_alloc_inode,
2327 .destroy_inode = btrfs_destroy_inode,
2328 .statfs = btrfs_statfs,
2329 .remount_fs = btrfs_remount,
2330 .freeze_fs = btrfs_freeze,
2331 .unfreeze_fs = btrfs_unfreeze,
2334 static const struct file_operations btrfs_ctl_fops = {
2335 .open = btrfs_control_open,
2336 .unlocked_ioctl = btrfs_control_ioctl,
2337 .compat_ioctl = btrfs_control_ioctl,
2338 .owner = THIS_MODULE,
2339 .llseek = noop_llseek,
2342 static struct miscdevice btrfs_misc = {
2343 .minor = BTRFS_MINOR,
2344 .name = "btrfs-control",
2345 .fops = &btrfs_ctl_fops
2348 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2349 MODULE_ALIAS("devname:btrfs-control");
2351 static int btrfs_interface_init(void)
2353 return misc_register(&btrfs_misc);
2356 static void btrfs_interface_exit(void)
2358 misc_deregister(&btrfs_misc);
2361 static void btrfs_print_mod_info(void)
2363 pr_info("Btrfs loaded, crc32c=%s"
2364 #ifdef CONFIG_BTRFS_DEBUG
2367 #ifdef CONFIG_BTRFS_ASSERT
2370 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2371 ", integrity-checker=on"
2374 btrfs_crc32c_impl());
2377 static int __init init_btrfs_fs(void)
2381 err = btrfs_hash_init();
2387 err = btrfs_init_sysfs();
2391 btrfs_init_compress();
2393 err = btrfs_init_cachep();
2397 err = extent_io_init();
2401 err = extent_map_init();
2403 goto free_extent_io;
2405 err = ordered_data_init();
2407 goto free_extent_map;
2409 err = btrfs_delayed_inode_init();
2411 goto free_ordered_data;
2413 err = btrfs_auto_defrag_init();
2415 goto free_delayed_inode;
2417 err = btrfs_delayed_ref_init();
2419 goto free_auto_defrag;
2421 err = btrfs_prelim_ref_init();
2423 goto free_delayed_ref;
2425 err = btrfs_end_io_wq_init();
2427 goto free_prelim_ref;
2429 err = btrfs_interface_init();
2431 goto free_end_io_wq;
2433 btrfs_init_lockdep();
2435 btrfs_print_mod_info();
2437 err = btrfs_run_sanity_tests();
2439 goto unregister_ioctl;
2441 err = register_filesystem(&btrfs_fs_type);
2443 goto unregister_ioctl;
2448 btrfs_interface_exit();
2450 btrfs_end_io_wq_exit();
2452 btrfs_prelim_ref_exit();
2454 btrfs_delayed_ref_exit();
2456 btrfs_auto_defrag_exit();
2458 btrfs_delayed_inode_exit();
2460 ordered_data_exit();
2466 btrfs_destroy_cachep();
2468 btrfs_exit_compress();
2475 static void __exit exit_btrfs_fs(void)
2477 btrfs_destroy_cachep();
2478 btrfs_delayed_ref_exit();
2479 btrfs_auto_defrag_exit();
2480 btrfs_delayed_inode_exit();
2481 btrfs_prelim_ref_exit();
2482 ordered_data_exit();
2485 btrfs_interface_exit();
2486 btrfs_end_io_wq_exit();
2487 unregister_filesystem(&btrfs_fs_type);
2489 btrfs_cleanup_fs_uuids();
2490 btrfs_exit_compress();
2494 late_initcall(init_btrfs_fs);
2495 module_exit(exit_btrfs_fs)
2497 MODULE_LICENSE("GPL");