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 static 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 static void save_error_info(struct btrfs_fs_info *fs_info)
103 * today we only save the error info into ram. Long term we'll
104 * also send it down to the disk
106 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
112 struct super_block *sb = fs_info->sb;
114 if (sb->s_flags & MS_RDONLY)
117 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
118 sb->s_flags |= MS_RDONLY;
119 btrfs_info(fs_info, "forced readonly");
121 * Note that a running device replace operation is not
122 * canceled here although there is no way to update
123 * the progress. It would add the risk of a deadlock,
124 * therefore the canceling is ommited. The only penalty
125 * is that some I/O remains active until the procedure
126 * completes. The next time when the filesystem is
127 * mounted writeable again, the device replace
128 * operation continues.
135 * __btrfs_std_error decodes expected errors from the caller and
136 * invokes the approciate error response.
139 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
140 unsigned int line, int errno, const char *fmt, ...)
142 struct super_block *sb = fs_info->sb;
146 * Special case: if the error is EROFS, and we're already
147 * under MS_RDONLY, then it is safe here.
149 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
152 errstr = btrfs_decode_error(errno);
154 struct va_format vaf;
162 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
163 sb->s_id, function, line, errno, errstr, &vaf);
166 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
167 sb->s_id, function, line, errno, errstr);
170 /* Don't go through full error handling during mount */
171 save_error_info(fs_info);
172 if (sb->s_flags & MS_BORN)
173 btrfs_handle_error(fs_info);
176 static const char * const logtypes[] = {
187 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
189 struct super_block *sb = fs_info->sb;
191 struct va_format vaf;
193 const char *type = logtypes[4];
198 kern_level = printk_get_level(fmt);
200 size_t size = printk_skip_level(fmt) - fmt;
201 memcpy(lvl, fmt, size);
204 type = logtypes[kern_level - '0'];
211 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
218 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
219 unsigned int line, int errno, const char *fmt, ...)
221 struct super_block *sb = fs_info->sb;
224 * Special case: if the error is EROFS, and we're already
225 * under MS_RDONLY, then it is safe here.
227 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
230 /* Don't go through full error handling during mount */
231 if (sb->s_flags & MS_BORN) {
232 save_error_info(fs_info);
233 btrfs_handle_error(fs_info);
239 * We only mark the transaction aborted and then set the file system read-only.
240 * This will prevent new transactions from starting or trying to join this
243 * This means that error recovery at the call site is limited to freeing
244 * any local memory allocations and passing the error code up without
245 * further cleanup. The transaction should complete as it normally would
246 * in the call path but will return -EIO.
248 * We'll complete the cleanup in btrfs_end_transaction and
249 * btrfs_commit_transaction.
252 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
253 struct btrfs_root *root, const char *function,
254 unsigned int line, int errno)
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->blocks_used && list_empty(&trans->new_bgs)) {
262 errstr = btrfs_decode_error(errno);
263 btrfs_warn(root->fs_info,
264 "%s:%d: Aborting unused transaction(%s).",
265 function, line, errstr);
268 ACCESS_ONCE(trans->transaction->aborted) = errno;
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&root->fs_info->transaction_wait);
271 wake_up(&root->fs_info->transaction_blocked_wait);
272 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
280 unsigned int line, int errno, const char *fmt, ...)
282 char *s_id = "<unknown>";
284 struct va_format vaf = { .fmt = fmt };
288 s_id = fs_info->sb->s_id;
293 errstr = btrfs_decode_error(errno);
294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
296 s_id, function, line, &vaf, errno, errstr);
298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
299 function, line, &vaf, errno, errstr);
301 /* Caller calls BUG() */
304 static void btrfs_put_super(struct super_block *sb)
306 close_ctree(btrfs_sb(sb)->tree_root);
310 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
311 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
312 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
313 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
314 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
315 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
316 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
317 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
318 Opt_check_integrity, Opt_check_integrity_including_extent_data,
319 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
320 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
321 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
322 Opt_datasum, Opt_treelog, Opt_noinode_cache,
326 static match_table_t tokens = {
327 {Opt_degraded, "degraded"},
328 {Opt_subvol, "subvol=%s"},
329 {Opt_subvolid, "subvolid=%s"},
330 {Opt_device, "device=%s"},
331 {Opt_nodatasum, "nodatasum"},
332 {Opt_datasum, "datasum"},
333 {Opt_nodatacow, "nodatacow"},
334 {Opt_datacow, "datacow"},
335 {Opt_nobarrier, "nobarrier"},
336 {Opt_barrier, "barrier"},
337 {Opt_max_inline, "max_inline=%s"},
338 {Opt_alloc_start, "alloc_start=%s"},
339 {Opt_thread_pool, "thread_pool=%d"},
340 {Opt_compress, "compress"},
341 {Opt_compress_type, "compress=%s"},
342 {Opt_compress_force, "compress-force"},
343 {Opt_compress_force_type, "compress-force=%s"},
345 {Opt_ssd_spread, "ssd_spread"},
346 {Opt_nossd, "nossd"},
348 {Opt_noacl, "noacl"},
349 {Opt_notreelog, "notreelog"},
350 {Opt_treelog, "treelog"},
351 {Opt_flushoncommit, "flushoncommit"},
352 {Opt_noflushoncommit, "noflushoncommit"},
353 {Opt_ratio, "metadata_ratio=%d"},
354 {Opt_discard, "discard"},
355 {Opt_nodiscard, "nodiscard"},
356 {Opt_space_cache, "space_cache"},
357 {Opt_clear_cache, "clear_cache"},
358 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
359 {Opt_enospc_debug, "enospc_debug"},
360 {Opt_noenospc_debug, "noenospc_debug"},
361 {Opt_subvolrootid, "subvolrootid=%d"},
362 {Opt_defrag, "autodefrag"},
363 {Opt_nodefrag, "noautodefrag"},
364 {Opt_inode_cache, "inode_cache"},
365 {Opt_noinode_cache, "noinode_cache"},
366 {Opt_no_space_cache, "nospace_cache"},
367 {Opt_recovery, "recovery"},
368 {Opt_skip_balance, "skip_balance"},
369 {Opt_check_integrity, "check_int"},
370 {Opt_check_integrity_including_extent_data, "check_int_data"},
371 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
372 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
373 {Opt_fatal_errors, "fatal_errors=%s"},
374 {Opt_commit_interval, "commit=%d"},
379 * Regular mount options parser. Everything that is needed only when
380 * reading in a new superblock is parsed here.
381 * XXX JDM: This needs to be cleaned up for remount.
383 int btrfs_parse_options(struct btrfs_root *root, char *options)
385 struct btrfs_fs_info *info = root->fs_info;
386 substring_t args[MAX_OPT_ARGS];
387 char *p, *num, *orig = NULL;
392 bool compress_force = false;
394 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
396 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
402 * strsep changes the string, duplicate it because parse_options
405 options = kstrdup(options, GFP_NOFS);
411 while ((p = strsep(&options, ",")) != NULL) {
416 token = match_token(p, tokens, args);
419 btrfs_info(root->fs_info, "allowing degraded mounts");
420 btrfs_set_opt(info->mount_opt, DEGRADED);
424 case Opt_subvolrootid:
427 * These are parsed by btrfs_parse_early_options
428 * and can be happily ignored here.
432 btrfs_set_and_info(root, NODATASUM,
433 "setting nodatasum");
436 if (btrfs_test_opt(root, NODATASUM)) {
437 if (btrfs_test_opt(root, NODATACOW))
438 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
440 btrfs_info(root->fs_info, "setting datasum");
442 btrfs_clear_opt(info->mount_opt, NODATACOW);
443 btrfs_clear_opt(info->mount_opt, NODATASUM);
446 if (!btrfs_test_opt(root, NODATACOW)) {
447 if (!btrfs_test_opt(root, COMPRESS) ||
448 !btrfs_test_opt(root, FORCE_COMPRESS)) {
449 btrfs_info(root->fs_info,
450 "setting nodatacow, compression disabled");
452 btrfs_info(root->fs_info, "setting nodatacow");
455 btrfs_clear_opt(info->mount_opt, COMPRESS);
456 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
457 btrfs_set_opt(info->mount_opt, NODATACOW);
458 btrfs_set_opt(info->mount_opt, NODATASUM);
461 btrfs_clear_and_info(root, NODATACOW,
464 case Opt_compress_force:
465 case Opt_compress_force_type:
466 compress_force = true;
469 case Opt_compress_type:
470 if (token == Opt_compress ||
471 token == Opt_compress_force ||
472 strcmp(args[0].from, "zlib") == 0) {
473 compress_type = "zlib";
474 info->compress_type = BTRFS_COMPRESS_ZLIB;
475 btrfs_set_opt(info->mount_opt, COMPRESS);
476 btrfs_clear_opt(info->mount_opt, NODATACOW);
477 btrfs_clear_opt(info->mount_opt, NODATASUM);
478 } else if (strcmp(args[0].from, "lzo") == 0) {
479 compress_type = "lzo";
480 info->compress_type = BTRFS_COMPRESS_LZO;
481 btrfs_set_opt(info->mount_opt, COMPRESS);
482 btrfs_clear_opt(info->mount_opt, NODATACOW);
483 btrfs_clear_opt(info->mount_opt, NODATASUM);
484 btrfs_set_fs_incompat(info, COMPRESS_LZO);
485 } else if (strncmp(args[0].from, "no", 2) == 0) {
486 compress_type = "no";
487 btrfs_clear_opt(info->mount_opt, COMPRESS);
488 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
489 compress_force = false;
495 if (compress_force) {
496 btrfs_set_and_info(root, FORCE_COMPRESS,
497 "force %s compression",
500 if (!btrfs_test_opt(root, COMPRESS))
501 btrfs_info(root->fs_info,
502 "btrfs: use %s compression",
505 * If we remount from compress-force=xxx to
506 * compress=xxx, we need clear FORCE_COMPRESS
507 * flag, otherwise, there is no way for users
508 * to disable forcible compression separately.
510 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
514 btrfs_set_and_info(root, SSD,
515 "use ssd allocation scheme");
518 btrfs_set_and_info(root, SSD_SPREAD,
519 "use spread ssd allocation scheme");
520 btrfs_set_opt(info->mount_opt, SSD);
523 btrfs_set_and_info(root, NOSSD,
524 "not using ssd allocation scheme");
525 btrfs_clear_opt(info->mount_opt, SSD);
528 btrfs_clear_and_info(root, NOBARRIER,
529 "turning on barriers");
532 btrfs_set_and_info(root, NOBARRIER,
533 "turning off barriers");
535 case Opt_thread_pool:
536 ret = match_int(&args[0], &intarg);
539 } else if (intarg > 0) {
540 info->thread_pool_size = intarg;
547 num = match_strdup(&args[0]);
549 info->max_inline = memparse(num, NULL);
552 if (info->max_inline) {
553 info->max_inline = min_t(u64,
557 btrfs_info(root->fs_info, "max_inline at %llu",
564 case Opt_alloc_start:
565 num = match_strdup(&args[0]);
567 mutex_lock(&info->chunk_mutex);
568 info->alloc_start = memparse(num, NULL);
569 mutex_unlock(&info->chunk_mutex);
571 btrfs_info(root->fs_info, "allocations start at %llu",
579 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
580 root->fs_info->sb->s_flags |= MS_POSIXACL;
583 btrfs_err(root->fs_info,
584 "support for ACL not compiled in!");
589 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
592 btrfs_set_and_info(root, NOTREELOG,
593 "disabling tree log");
596 btrfs_clear_and_info(root, NOTREELOG,
597 "enabling tree log");
599 case Opt_flushoncommit:
600 btrfs_set_and_info(root, FLUSHONCOMMIT,
601 "turning on flush-on-commit");
603 case Opt_noflushoncommit:
604 btrfs_clear_and_info(root, FLUSHONCOMMIT,
605 "turning off flush-on-commit");
608 ret = match_int(&args[0], &intarg);
611 } else if (intarg >= 0) {
612 info->metadata_ratio = intarg;
613 btrfs_info(root->fs_info, "metadata ratio %d",
614 info->metadata_ratio);
621 btrfs_set_and_info(root, DISCARD,
622 "turning on discard");
625 btrfs_clear_and_info(root, DISCARD,
626 "turning off discard");
628 case Opt_space_cache:
629 btrfs_set_and_info(root, SPACE_CACHE,
630 "enabling disk space caching");
632 case Opt_rescan_uuid_tree:
633 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
635 case Opt_no_space_cache:
636 btrfs_clear_and_info(root, SPACE_CACHE,
637 "disabling disk space caching");
639 case Opt_inode_cache:
640 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
641 "enabling inode map caching");
643 case Opt_noinode_cache:
644 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
645 "disabling inode map caching");
647 case Opt_clear_cache:
648 btrfs_set_and_info(root, CLEAR_CACHE,
649 "force clearing of disk cache");
651 case Opt_user_subvol_rm_allowed:
652 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
654 case Opt_enospc_debug:
655 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
657 case Opt_noenospc_debug:
658 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
661 btrfs_set_and_info(root, AUTO_DEFRAG,
662 "enabling auto defrag");
665 btrfs_clear_and_info(root, AUTO_DEFRAG,
666 "disabling auto defrag");
669 btrfs_info(root->fs_info, "enabling auto recovery");
670 btrfs_set_opt(info->mount_opt, RECOVERY);
672 case Opt_skip_balance:
673 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
675 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
676 case Opt_check_integrity_including_extent_data:
677 btrfs_info(root->fs_info,
678 "enabling check integrity including extent data");
679 btrfs_set_opt(info->mount_opt,
680 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
681 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
683 case Opt_check_integrity:
684 btrfs_info(root->fs_info, "enabling check integrity");
685 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
687 case Opt_check_integrity_print_mask:
688 ret = match_int(&args[0], &intarg);
691 } else if (intarg >= 0) {
692 info->check_integrity_print_mask = intarg;
693 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
694 info->check_integrity_print_mask);
701 case Opt_check_integrity_including_extent_data:
702 case Opt_check_integrity:
703 case Opt_check_integrity_print_mask:
704 btrfs_err(root->fs_info,
705 "support for check_integrity* not compiled in!");
709 case Opt_fatal_errors:
710 if (strcmp(args[0].from, "panic") == 0)
711 btrfs_set_opt(info->mount_opt,
712 PANIC_ON_FATAL_ERROR);
713 else if (strcmp(args[0].from, "bug") == 0)
714 btrfs_clear_opt(info->mount_opt,
715 PANIC_ON_FATAL_ERROR);
721 case Opt_commit_interval:
723 ret = match_int(&args[0], &intarg);
725 btrfs_err(root->fs_info, "invalid commit interval");
731 btrfs_warn(root->fs_info, "excessive commit interval %d",
734 info->commit_interval = intarg;
736 btrfs_info(root->fs_info, "using default commit interval %ds",
737 BTRFS_DEFAULT_COMMIT_INTERVAL);
738 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
742 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
750 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
751 btrfs_info(root->fs_info, "disk space caching is enabled");
757 * Parse mount options that are required early in the mount process.
759 * All other options will be parsed on much later in the mount process and
760 * only when we need to allocate a new super block.
762 static int btrfs_parse_early_options(const char *options, fmode_t flags,
763 void *holder, char **subvol_name, u64 *subvol_objectid,
764 struct btrfs_fs_devices **fs_devices)
766 substring_t args[MAX_OPT_ARGS];
767 char *device_name, *opts, *orig, *p;
775 * strsep changes the string, duplicate it because parse_options
778 opts = kstrdup(options, GFP_KERNEL);
783 while ((p = strsep(&opts, ",")) != NULL) {
788 token = match_token(p, tokens, args);
792 *subvol_name = match_strdup(&args[0]);
799 num = match_strdup(&args[0]);
801 *subvol_objectid = memparse(num, NULL);
803 /* we want the original fs_tree */
804 if (!*subvol_objectid)
806 BTRFS_FS_TREE_OBJECTID;
812 case Opt_subvolrootid:
814 "BTRFS: 'subvolrootid' mount option is deprecated and has "
818 device_name = match_strdup(&args[0]);
823 error = btrfs_scan_one_device(device_name,
824 flags, holder, fs_devices);
839 static struct dentry *get_default_root(struct super_block *sb,
842 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
843 struct btrfs_root *root = fs_info->tree_root;
844 struct btrfs_root *new_root;
845 struct btrfs_dir_item *di;
846 struct btrfs_path *path;
847 struct btrfs_key location;
853 * We have a specific subvol we want to mount, just setup location and
854 * go look up the root.
856 if (subvol_objectid) {
857 location.objectid = subvol_objectid;
858 location.type = BTRFS_ROOT_ITEM_KEY;
859 location.offset = (u64)-1;
863 path = btrfs_alloc_path();
865 return ERR_PTR(-ENOMEM);
866 path->leave_spinning = 1;
869 * Find the "default" dir item which points to the root item that we
870 * will mount by default if we haven't been given a specific subvolume
873 dir_id = btrfs_super_root_dir(fs_info->super_copy);
874 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
876 btrfs_free_path(path);
881 * Ok the default dir item isn't there. This is weird since
882 * it's always been there, but don't freak out, just try and
883 * mount to root most subvolume.
885 btrfs_free_path(path);
886 dir_id = BTRFS_FIRST_FREE_OBJECTID;
887 new_root = fs_info->fs_root;
891 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
892 btrfs_free_path(path);
895 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
896 if (IS_ERR(new_root))
897 return ERR_CAST(new_root);
899 if (!(sb->s_flags & MS_RDONLY)) {
901 down_read(&fs_info->cleanup_work_sem);
902 ret = btrfs_orphan_cleanup(new_root);
903 up_read(&fs_info->cleanup_work_sem);
908 dir_id = btrfs_root_dirid(&new_root->root_item);
910 location.objectid = dir_id;
911 location.type = BTRFS_INODE_ITEM_KEY;
914 inode = btrfs_iget(sb, &location, new_root, &new);
916 return ERR_CAST(inode);
919 * If we're just mounting the root most subvol put the inode and return
920 * a reference to the dentry. We will have already gotten a reference
921 * to the inode in btrfs_fill_super so we're good to go.
923 if (!new && d_inode(sb->s_root) == inode) {
925 return dget(sb->s_root);
928 return d_obtain_root(inode);
931 static int btrfs_fill_super(struct super_block *sb,
932 struct btrfs_fs_devices *fs_devices,
933 void *data, int silent)
936 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
937 struct btrfs_key key;
940 sb->s_maxbytes = MAX_LFS_FILESIZE;
941 sb->s_magic = BTRFS_SUPER_MAGIC;
942 sb->s_op = &btrfs_super_ops;
943 sb->s_d_op = &btrfs_dentry_operations;
944 sb->s_export_op = &btrfs_export_ops;
945 sb->s_xattr = btrfs_xattr_handlers;
947 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
948 sb->s_flags |= MS_POSIXACL;
950 sb->s_flags |= MS_I_VERSION;
951 err = open_ctree(sb, fs_devices, (char *)data);
953 printk(KERN_ERR "BTRFS: open_ctree failed\n");
957 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
958 key.type = BTRFS_INODE_ITEM_KEY;
960 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
962 err = PTR_ERR(inode);
966 sb->s_root = d_make_root(inode);
972 save_mount_options(sb, data);
973 cleancache_init_fs(sb);
974 sb->s_flags |= MS_ACTIVE;
978 close_ctree(fs_info->tree_root);
982 int btrfs_sync_fs(struct super_block *sb, int wait)
984 struct btrfs_trans_handle *trans;
985 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
986 struct btrfs_root *root = fs_info->tree_root;
988 trace_btrfs_sync_fs(wait);
991 filemap_flush(fs_info->btree_inode->i_mapping);
995 btrfs_wait_ordered_roots(fs_info, -1);
997 trans = btrfs_attach_transaction_barrier(root);
999 /* no transaction, don't bother */
1000 if (PTR_ERR(trans) == -ENOENT) {
1002 * Exit unless we have some pending changes
1003 * that need to go through commit
1005 if (fs_info->pending_changes == 0)
1008 * A non-blocking test if the fs is frozen. We must not
1009 * start a new transaction here otherwise a deadlock
1010 * happens. The pending operations are delayed to the
1011 * next commit after thawing.
1013 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1014 __sb_end_write(sb, SB_FREEZE_WRITE);
1017 trans = btrfs_start_transaction(root, 0);
1020 return PTR_ERR(trans);
1022 return btrfs_commit_transaction(trans, root);
1025 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1027 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1028 struct btrfs_root *root = info->tree_root;
1029 char *compress_type;
1031 if (btrfs_test_opt(root, DEGRADED))
1032 seq_puts(seq, ",degraded");
1033 if (btrfs_test_opt(root, NODATASUM))
1034 seq_puts(seq, ",nodatasum");
1035 if (btrfs_test_opt(root, NODATACOW))
1036 seq_puts(seq, ",nodatacow");
1037 if (btrfs_test_opt(root, NOBARRIER))
1038 seq_puts(seq, ",nobarrier");
1039 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1040 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1041 if (info->alloc_start != 0)
1042 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1043 if (info->thread_pool_size != min_t(unsigned long,
1044 num_online_cpus() + 2, 8))
1045 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1046 if (btrfs_test_opt(root, COMPRESS)) {
1047 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1048 compress_type = "zlib";
1050 compress_type = "lzo";
1051 if (btrfs_test_opt(root, FORCE_COMPRESS))
1052 seq_printf(seq, ",compress-force=%s", compress_type);
1054 seq_printf(seq, ",compress=%s", compress_type);
1056 if (btrfs_test_opt(root, NOSSD))
1057 seq_puts(seq, ",nossd");
1058 if (btrfs_test_opt(root, SSD_SPREAD))
1059 seq_puts(seq, ",ssd_spread");
1060 else if (btrfs_test_opt(root, SSD))
1061 seq_puts(seq, ",ssd");
1062 if (btrfs_test_opt(root, NOTREELOG))
1063 seq_puts(seq, ",notreelog");
1064 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1065 seq_puts(seq, ",flushoncommit");
1066 if (btrfs_test_opt(root, DISCARD))
1067 seq_puts(seq, ",discard");
1068 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1069 seq_puts(seq, ",noacl");
1070 if (btrfs_test_opt(root, SPACE_CACHE))
1071 seq_puts(seq, ",space_cache");
1073 seq_puts(seq, ",nospace_cache");
1074 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1075 seq_puts(seq, ",rescan_uuid_tree");
1076 if (btrfs_test_opt(root, CLEAR_CACHE))
1077 seq_puts(seq, ",clear_cache");
1078 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1079 seq_puts(seq, ",user_subvol_rm_allowed");
1080 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1081 seq_puts(seq, ",enospc_debug");
1082 if (btrfs_test_opt(root, AUTO_DEFRAG))
1083 seq_puts(seq, ",autodefrag");
1084 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1085 seq_puts(seq, ",inode_cache");
1086 if (btrfs_test_opt(root, SKIP_BALANCE))
1087 seq_puts(seq, ",skip_balance");
1088 if (btrfs_test_opt(root, RECOVERY))
1089 seq_puts(seq, ",recovery");
1090 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1091 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1092 seq_puts(seq, ",check_int_data");
1093 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1094 seq_puts(seq, ",check_int");
1095 if (info->check_integrity_print_mask)
1096 seq_printf(seq, ",check_int_print_mask=%d",
1097 info->check_integrity_print_mask);
1099 if (info->metadata_ratio)
1100 seq_printf(seq, ",metadata_ratio=%d",
1101 info->metadata_ratio);
1102 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1103 seq_puts(seq, ",fatal_errors=panic");
1104 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1105 seq_printf(seq, ",commit=%d", info->commit_interval);
1109 static int btrfs_test_super(struct super_block *s, void *data)
1111 struct btrfs_fs_info *p = data;
1112 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1114 return fs_info->fs_devices == p->fs_devices;
1117 static int btrfs_set_super(struct super_block *s, void *data)
1119 int err = set_anon_super(s, data);
1121 s->s_fs_info = data;
1126 * subvolumes are identified by ino 256
1128 static inline int is_subvolume_inode(struct inode *inode)
1130 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1136 * This will add subvolid=0 to the argument string while removing any subvol=
1137 * and subvolid= arguments to make sure we get the top-level root for path
1138 * walking to the subvol we want.
1140 static char *setup_root_args(char *args)
1142 char *buf, *dst, *sep;
1145 return kstrdup("subvolid=0", GFP_NOFS);
1147 /* The worst case is that we add ",subvolid=0" to the end. */
1148 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1153 sep = strchrnul(args, ',');
1154 if (!strstarts(args, "subvol=") &&
1155 !strstarts(args, "subvolid=")) {
1156 memcpy(dst, args, sep - args);
1165 strcpy(dst, "subvolid=0");
1170 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1171 const char *device_name, char *data)
1173 struct dentry *root;
1174 struct vfsmount *mnt;
1177 newargs = setup_root_args(data);
1179 return ERR_PTR(-ENOMEM);
1180 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1183 if (PTR_RET(mnt) == -EBUSY) {
1184 if (flags & MS_RDONLY) {
1185 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1189 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1193 return ERR_CAST(mnt);
1196 down_write(&mnt->mnt_sb->s_umount);
1197 r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1198 up_write(&mnt->mnt_sb->s_umount);
1200 /* FIXME: release vfsmount mnt ??*/
1210 return ERR_CAST(mnt);
1212 root = mount_subtree(mnt, subvol_name);
1214 if (!IS_ERR(root) && !is_subvolume_inode(d_inode(root))) {
1215 struct super_block *s = root->d_sb;
1217 root = ERR_PTR(-EINVAL);
1218 deactivate_locked_super(s);
1219 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1226 static int parse_security_options(char *orig_opts,
1227 struct security_mnt_opts *sec_opts)
1229 char *secdata = NULL;
1232 secdata = alloc_secdata();
1235 ret = security_sb_copy_data(orig_opts, secdata);
1237 free_secdata(secdata);
1240 ret = security_sb_parse_opts_str(secdata, sec_opts);
1241 free_secdata(secdata);
1245 static int setup_security_options(struct btrfs_fs_info *fs_info,
1246 struct super_block *sb,
1247 struct security_mnt_opts *sec_opts)
1252 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1255 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1259 #ifdef CONFIG_SECURITY
1260 if (!fs_info->security_opts.num_mnt_opts) {
1261 /* first time security setup, copy sec_opts to fs_info */
1262 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1265 * Since SELinux(the only one supports security_mnt_opts) does
1266 * NOT support changing context during remount/mount same sb,
1267 * This must be the same or part of the same security options,
1270 security_free_mnt_opts(sec_opts);
1277 * Find a superblock for the given device / mount point.
1279 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1280 * for multiple device setup. Make sure to keep it in sync.
1282 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1283 const char *device_name, void *data)
1285 struct block_device *bdev = NULL;
1286 struct super_block *s;
1287 struct dentry *root;
1288 struct btrfs_fs_devices *fs_devices = NULL;
1289 struct btrfs_fs_info *fs_info = NULL;
1290 struct security_mnt_opts new_sec_opts;
1291 fmode_t mode = FMODE_READ;
1292 char *subvol_name = NULL;
1293 u64 subvol_objectid = 0;
1296 if (!(flags & MS_RDONLY))
1297 mode |= FMODE_WRITE;
1299 error = btrfs_parse_early_options(data, mode, fs_type,
1300 &subvol_name, &subvol_objectid,
1304 return ERR_PTR(error);
1308 root = mount_subvol(subvol_name, flags, device_name, data);
1313 security_init_mnt_opts(&new_sec_opts);
1315 error = parse_security_options(data, &new_sec_opts);
1317 return ERR_PTR(error);
1320 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1322 goto error_sec_opts;
1325 * Setup a dummy root and fs_info for test/set super. This is because
1326 * we don't actually fill this stuff out until open_ctree, but we need
1327 * it for searching for existing supers, so this lets us do that and
1328 * then open_ctree will properly initialize everything later.
1330 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1333 goto error_sec_opts;
1336 fs_info->fs_devices = fs_devices;
1338 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1339 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1340 security_init_mnt_opts(&fs_info->security_opts);
1341 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1346 error = btrfs_open_devices(fs_devices, mode, fs_type);
1350 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1352 goto error_close_devices;
1355 bdev = fs_devices->latest_bdev;
1356 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1360 goto error_close_devices;
1364 btrfs_close_devices(fs_devices);
1365 free_fs_info(fs_info);
1366 if ((flags ^ s->s_flags) & MS_RDONLY)
1369 char b[BDEVNAME_SIZE];
1371 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1372 btrfs_sb(s)->bdev_holder = fs_type;
1373 error = btrfs_fill_super(s, fs_devices, data,
1374 flags & MS_SILENT ? 1 : 0);
1377 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1379 deactivate_locked_super(s);
1380 error = PTR_ERR(root);
1381 goto error_sec_opts;
1384 fs_info = btrfs_sb(s);
1385 error = setup_security_options(fs_info, s, &new_sec_opts);
1388 deactivate_locked_super(s);
1389 goto error_sec_opts;
1394 error_close_devices:
1395 btrfs_close_devices(fs_devices);
1397 free_fs_info(fs_info);
1399 security_free_mnt_opts(&new_sec_opts);
1400 return ERR_PTR(error);
1403 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1404 int new_pool_size, int old_pool_size)
1406 if (new_pool_size == old_pool_size)
1409 fs_info->thread_pool_size = new_pool_size;
1411 btrfs_info(fs_info, "resize thread pool %d -> %d",
1412 old_pool_size, new_pool_size);
1414 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1415 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1416 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1417 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1418 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1419 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1420 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1422 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1423 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1424 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1425 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1426 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1430 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1432 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1435 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1436 unsigned long old_opts, int flags)
1438 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1439 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1440 (flags & MS_RDONLY))) {
1441 /* wait for any defraggers to finish */
1442 wait_event(fs_info->transaction_wait,
1443 (atomic_read(&fs_info->defrag_running) == 0));
1444 if (flags & MS_RDONLY)
1445 sync_filesystem(fs_info->sb);
1449 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1450 unsigned long old_opts)
1453 * We need cleanup all defragable inodes if the autodefragment is
1454 * close or the fs is R/O.
1456 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1457 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1458 (fs_info->sb->s_flags & MS_RDONLY))) {
1459 btrfs_cleanup_defrag_inodes(fs_info);
1462 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1465 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1467 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1468 struct btrfs_root *root = fs_info->tree_root;
1469 unsigned old_flags = sb->s_flags;
1470 unsigned long old_opts = fs_info->mount_opt;
1471 unsigned long old_compress_type = fs_info->compress_type;
1472 u64 old_max_inline = fs_info->max_inline;
1473 u64 old_alloc_start = fs_info->alloc_start;
1474 int old_thread_pool_size = fs_info->thread_pool_size;
1475 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1478 sync_filesystem(sb);
1479 btrfs_remount_prepare(fs_info);
1482 struct security_mnt_opts new_sec_opts;
1484 security_init_mnt_opts(&new_sec_opts);
1485 ret = parse_security_options(data, &new_sec_opts);
1488 ret = setup_security_options(fs_info, sb,
1491 security_free_mnt_opts(&new_sec_opts);
1496 ret = btrfs_parse_options(root, data);
1502 btrfs_remount_begin(fs_info, old_opts, *flags);
1503 btrfs_resize_thread_pool(fs_info,
1504 fs_info->thread_pool_size, old_thread_pool_size);
1506 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1509 if (*flags & MS_RDONLY) {
1511 * this also happens on 'umount -rf' or on shutdown, when
1512 * the filesystem is busy.
1514 cancel_work_sync(&fs_info->async_reclaim_work);
1516 /* wait for the uuid_scan task to finish */
1517 down(&fs_info->uuid_tree_rescan_sem);
1518 /* avoid complains from lockdep et al. */
1519 up(&fs_info->uuid_tree_rescan_sem);
1521 sb->s_flags |= MS_RDONLY;
1523 btrfs_dev_replace_suspend_for_unmount(fs_info);
1524 btrfs_scrub_cancel(fs_info);
1525 btrfs_pause_balance(fs_info);
1527 ret = btrfs_commit_super(root);
1531 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1533 "Remounting read-write after error is not allowed");
1537 if (fs_info->fs_devices->rw_devices == 0) {
1542 if (fs_info->fs_devices->missing_devices >
1543 fs_info->num_tolerated_disk_barrier_failures &&
1544 !(*flags & MS_RDONLY)) {
1546 "too many missing devices, writeable remount is not allowed");
1551 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1556 ret = btrfs_cleanup_fs_roots(fs_info);
1560 /* recover relocation */
1561 mutex_lock(&fs_info->cleaner_mutex);
1562 ret = btrfs_recover_relocation(root);
1563 mutex_unlock(&fs_info->cleaner_mutex);
1567 ret = btrfs_resume_balance_async(fs_info);
1571 ret = btrfs_resume_dev_replace_async(fs_info);
1573 btrfs_warn(fs_info, "failed to resume dev_replace");
1577 if (!fs_info->uuid_root) {
1578 btrfs_info(fs_info, "creating UUID tree");
1579 ret = btrfs_create_uuid_tree(fs_info);
1581 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1585 sb->s_flags &= ~MS_RDONLY;
1588 wake_up_process(fs_info->transaction_kthread);
1589 btrfs_remount_cleanup(fs_info, old_opts);
1593 /* We've hit an error - don't reset MS_RDONLY */
1594 if (sb->s_flags & MS_RDONLY)
1595 old_flags |= MS_RDONLY;
1596 sb->s_flags = old_flags;
1597 fs_info->mount_opt = old_opts;
1598 fs_info->compress_type = old_compress_type;
1599 fs_info->max_inline = old_max_inline;
1600 mutex_lock(&fs_info->chunk_mutex);
1601 fs_info->alloc_start = old_alloc_start;
1602 mutex_unlock(&fs_info->chunk_mutex);
1603 btrfs_resize_thread_pool(fs_info,
1604 old_thread_pool_size, fs_info->thread_pool_size);
1605 fs_info->metadata_ratio = old_metadata_ratio;
1606 btrfs_remount_cleanup(fs_info, old_opts);
1610 /* Used to sort the devices by max_avail(descending sort) */
1611 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1612 const void *dev_info2)
1614 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1615 ((struct btrfs_device_info *)dev_info2)->max_avail)
1617 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1618 ((struct btrfs_device_info *)dev_info2)->max_avail)
1625 * sort the devices by max_avail, in which max free extent size of each device
1626 * is stored.(Descending Sort)
1628 static inline void btrfs_descending_sort_devices(
1629 struct btrfs_device_info *devices,
1632 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1633 btrfs_cmp_device_free_bytes, NULL);
1637 * The helper to calc the free space on the devices that can be used to store
1640 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1642 struct btrfs_fs_info *fs_info = root->fs_info;
1643 struct btrfs_device_info *devices_info;
1644 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1645 struct btrfs_device *device;
1650 u64 min_stripe_size;
1651 int min_stripes = 1, num_stripes = 1;
1652 int i = 0, nr_devices;
1656 * We aren't under the device list lock, so this is racey-ish, but good
1657 * enough for our purposes.
1659 nr_devices = fs_info->fs_devices->open_devices;
1662 nr_devices = fs_info->fs_devices->open_devices;
1670 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1675 /* calc min stripe number for data space alloction */
1676 type = btrfs_get_alloc_profile(root, 1);
1677 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1679 num_stripes = nr_devices;
1680 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1683 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1688 if (type & BTRFS_BLOCK_GROUP_DUP)
1689 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1691 min_stripe_size = BTRFS_STRIPE_LEN;
1693 if (fs_info->alloc_start)
1694 mutex_lock(&fs_devices->device_list_mutex);
1696 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1697 if (!device->in_fs_metadata || !device->bdev ||
1698 device->is_tgtdev_for_dev_replace)
1701 if (i >= nr_devices)
1704 avail_space = device->total_bytes - device->bytes_used;
1706 /* align with stripe_len */
1707 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1708 avail_space *= BTRFS_STRIPE_LEN;
1711 * In order to avoid overwritting the superblock on the drive,
1712 * btrfs starts at an offset of at least 1MB when doing chunk
1715 skip_space = 1024 * 1024;
1717 /* user can set the offset in fs_info->alloc_start. */
1718 if (fs_info->alloc_start &&
1719 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1720 device->total_bytes) {
1722 skip_space = max(fs_info->alloc_start, skip_space);
1725 * btrfs can not use the free space in
1726 * [0, skip_space - 1], we must subtract it from the
1727 * total. In order to implement it, we account the used
1728 * space in this range first.
1730 ret = btrfs_account_dev_extents_size(device, 0,
1734 kfree(devices_info);
1735 mutex_unlock(&fs_devices->device_list_mutex);
1741 /* calc the free space in [0, skip_space - 1] */
1742 skip_space -= used_space;
1746 * we can use the free space in [0, skip_space - 1], subtract
1747 * it from the total.
1749 if (avail_space && avail_space >= skip_space)
1750 avail_space -= skip_space;
1754 if (avail_space < min_stripe_size)
1757 devices_info[i].dev = device;
1758 devices_info[i].max_avail = avail_space;
1763 if (fs_info->alloc_start)
1764 mutex_unlock(&fs_devices->device_list_mutex);
1768 btrfs_descending_sort_devices(devices_info, nr_devices);
1772 while (nr_devices >= min_stripes) {
1773 if (num_stripes > nr_devices)
1774 num_stripes = nr_devices;
1776 if (devices_info[i].max_avail >= min_stripe_size) {
1780 avail_space += devices_info[i].max_avail * num_stripes;
1781 alloc_size = devices_info[i].max_avail;
1782 for (j = i + 1 - num_stripes; j <= i; j++)
1783 devices_info[j].max_avail -= alloc_size;
1789 kfree(devices_info);
1790 *free_bytes = avail_space;
1795 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1797 * If there's a redundant raid level at DATA block groups, use the respective
1798 * multiplier to scale the sizes.
1800 * Unused device space usage is based on simulating the chunk allocator
1801 * algorithm that respects the device sizes, order of allocations and the
1802 * 'alloc_start' value, this is a close approximation of the actual use but
1803 * there are other factors that may change the result (like a new metadata
1806 * FIXME: not accurate for mixed block groups, total and free/used are ok,
1807 * available appears slightly larger.
1809 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1811 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1812 struct btrfs_super_block *disk_super = fs_info->super_copy;
1813 struct list_head *head = &fs_info->space_info;
1814 struct btrfs_space_info *found;
1816 u64 total_free_data = 0;
1817 int bits = dentry->d_sb->s_blocksize_bits;
1818 __be32 *fsid = (__be32 *)fs_info->fsid;
1819 unsigned factor = 1;
1820 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1824 * holding chunk_muext to avoid allocating new chunks, holding
1825 * device_list_mutex to avoid the device being removed
1828 list_for_each_entry_rcu(found, head, list) {
1829 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1832 total_free_data += found->disk_total - found->disk_used;
1834 btrfs_account_ro_block_groups_free_space(found);
1836 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1837 if (!list_empty(&found->block_groups[i])) {
1839 case BTRFS_RAID_DUP:
1840 case BTRFS_RAID_RAID1:
1841 case BTRFS_RAID_RAID10:
1848 total_used += found->disk_used;
1853 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1854 buf->f_blocks >>= bits;
1855 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1857 /* Account global block reserve as used, it's in logical size already */
1858 spin_lock(&block_rsv->lock);
1859 buf->f_bfree -= block_rsv->size >> bits;
1860 spin_unlock(&block_rsv->lock);
1862 buf->f_bavail = div_u64(total_free_data, factor);
1863 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1866 buf->f_bavail += div_u64(total_free_data, factor);
1867 buf->f_bavail = buf->f_bavail >> bits;
1869 buf->f_type = BTRFS_SUPER_MAGIC;
1870 buf->f_bsize = dentry->d_sb->s_blocksize;
1871 buf->f_namelen = BTRFS_NAME_LEN;
1873 /* We treat it as constant endianness (it doesn't matter _which_)
1874 because we want the fsid to come out the same whether mounted
1875 on a big-endian or little-endian host */
1876 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1877 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1878 /* Mask in the root object ID too, to disambiguate subvols */
1879 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
1880 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
1885 static void btrfs_kill_super(struct super_block *sb)
1887 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1888 kill_anon_super(sb);
1889 free_fs_info(fs_info);
1892 static struct file_system_type btrfs_fs_type = {
1893 .owner = THIS_MODULE,
1895 .mount = btrfs_mount,
1896 .kill_sb = btrfs_kill_super,
1897 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
1899 MODULE_ALIAS_FS("btrfs");
1901 static int btrfs_control_open(struct inode *inode, struct file *file)
1904 * The control file's private_data is used to hold the
1905 * transaction when it is started and is used to keep
1906 * track of whether a transaction is already in progress.
1908 file->private_data = NULL;
1913 * used by btrfsctl to scan devices when no FS is mounted
1915 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1918 struct btrfs_ioctl_vol_args *vol;
1919 struct btrfs_fs_devices *fs_devices;
1922 if (!capable(CAP_SYS_ADMIN))
1925 vol = memdup_user((void __user *)arg, sizeof(*vol));
1927 return PTR_ERR(vol);
1930 case BTRFS_IOC_SCAN_DEV:
1931 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1932 &btrfs_fs_type, &fs_devices);
1934 case BTRFS_IOC_DEVICES_READY:
1935 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1936 &btrfs_fs_type, &fs_devices);
1939 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1947 static int btrfs_freeze(struct super_block *sb)
1949 struct btrfs_trans_handle *trans;
1950 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1952 trans = btrfs_attach_transaction_barrier(root);
1953 if (IS_ERR(trans)) {
1954 /* no transaction, don't bother */
1955 if (PTR_ERR(trans) == -ENOENT)
1957 return PTR_ERR(trans);
1959 return btrfs_commit_transaction(trans, root);
1962 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1964 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1965 struct btrfs_fs_devices *cur_devices;
1966 struct btrfs_device *dev, *first_dev = NULL;
1967 struct list_head *head;
1968 struct rcu_string *name;
1970 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1971 cur_devices = fs_info->fs_devices;
1972 while (cur_devices) {
1973 head = &cur_devices->devices;
1974 list_for_each_entry(dev, head, dev_list) {
1979 if (!first_dev || dev->devid < first_dev->devid)
1982 cur_devices = cur_devices->seed;
1987 name = rcu_dereference(first_dev->name);
1988 seq_escape(m, name->str, " \t\n\\");
1993 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1997 static const struct super_operations btrfs_super_ops = {
1998 .drop_inode = btrfs_drop_inode,
1999 .evict_inode = btrfs_evict_inode,
2000 .put_super = btrfs_put_super,
2001 .sync_fs = btrfs_sync_fs,
2002 .show_options = btrfs_show_options,
2003 .show_devname = btrfs_show_devname,
2004 .write_inode = btrfs_write_inode,
2005 .alloc_inode = btrfs_alloc_inode,
2006 .destroy_inode = btrfs_destroy_inode,
2007 .statfs = btrfs_statfs,
2008 .remount_fs = btrfs_remount,
2009 .freeze_fs = btrfs_freeze,
2012 static const struct file_operations btrfs_ctl_fops = {
2013 .open = btrfs_control_open,
2014 .unlocked_ioctl = btrfs_control_ioctl,
2015 .compat_ioctl = btrfs_control_ioctl,
2016 .owner = THIS_MODULE,
2017 .llseek = noop_llseek,
2020 static struct miscdevice btrfs_misc = {
2021 .minor = BTRFS_MINOR,
2022 .name = "btrfs-control",
2023 .fops = &btrfs_ctl_fops
2026 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2027 MODULE_ALIAS("devname:btrfs-control");
2029 static int btrfs_interface_init(void)
2031 return misc_register(&btrfs_misc);
2034 static void btrfs_interface_exit(void)
2036 if (misc_deregister(&btrfs_misc) < 0)
2037 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
2040 static void btrfs_print_info(void)
2042 printk(KERN_INFO "Btrfs loaded"
2043 #ifdef CONFIG_BTRFS_DEBUG
2046 #ifdef CONFIG_BTRFS_ASSERT
2049 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2050 ", integrity-checker=on"
2055 static int btrfs_run_sanity_tests(void)
2059 ret = btrfs_init_test_fs();
2063 ret = btrfs_test_free_space_cache();
2066 ret = btrfs_test_extent_buffer_operations();
2069 ret = btrfs_test_extent_io();
2072 ret = btrfs_test_inodes();
2075 ret = btrfs_test_qgroups();
2077 btrfs_destroy_test_fs();
2081 static int __init init_btrfs_fs(void)
2085 err = btrfs_hash_init();
2091 err = btrfs_init_sysfs();
2095 btrfs_init_compress();
2097 err = btrfs_init_cachep();
2101 err = extent_io_init();
2105 err = extent_map_init();
2107 goto free_extent_io;
2109 err = ordered_data_init();
2111 goto free_extent_map;
2113 err = btrfs_delayed_inode_init();
2115 goto free_ordered_data;
2117 err = btrfs_auto_defrag_init();
2119 goto free_delayed_inode;
2121 err = btrfs_delayed_ref_init();
2123 goto free_auto_defrag;
2125 err = btrfs_prelim_ref_init();
2127 goto free_delayed_ref;
2129 err = btrfs_end_io_wq_init();
2131 goto free_prelim_ref;
2133 err = btrfs_interface_init();
2135 goto free_end_io_wq;
2137 btrfs_init_lockdep();
2141 err = btrfs_run_sanity_tests();
2143 goto unregister_ioctl;
2145 err = register_filesystem(&btrfs_fs_type);
2147 goto unregister_ioctl;
2152 btrfs_interface_exit();
2154 btrfs_end_io_wq_exit();
2156 btrfs_prelim_ref_exit();
2158 btrfs_delayed_ref_exit();
2160 btrfs_auto_defrag_exit();
2162 btrfs_delayed_inode_exit();
2164 ordered_data_exit();
2170 btrfs_destroy_cachep();
2172 btrfs_exit_compress();
2179 static void __exit exit_btrfs_fs(void)
2181 btrfs_destroy_cachep();
2182 btrfs_delayed_ref_exit();
2183 btrfs_auto_defrag_exit();
2184 btrfs_delayed_inode_exit();
2185 btrfs_prelim_ref_exit();
2186 ordered_data_exit();
2189 btrfs_interface_exit();
2190 btrfs_end_io_wq_exit();
2191 unregister_filesystem(&btrfs_fs_type);
2193 btrfs_cleanup_fs_uuids();
2194 btrfs_exit_compress();
2198 late_initcall(init_btrfs_fs);
2199 module_exit(exit_btrfs_fs)
2201 MODULE_LICENSE("GPL");