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>
44 #include "delayed-inode.h"
47 #include "transaction.h"
48 #include "btrfs_inode.h"
50 #include "print-tree.h"
55 #include "compression.h"
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/btrfs.h>
60 static const struct super_operations btrfs_super_ops;
62 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
69 errstr = "IO failure";
72 errstr = "Out of memory";
75 errstr = "Readonly filesystem";
79 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
88 static void __save_error_info(struct btrfs_fs_info *fs_info)
91 * today we only save the error info into ram. Long term we'll
92 * also send it down to the disk
94 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
98 * We move write_super stuff at umount in order to avoid deadlock
99 * for umount hold all lock.
101 static void save_error_info(struct btrfs_fs_info *fs_info)
103 __save_error_info(fs_info);
106 /* btrfs handle error by forcing the filesystem readonly */
107 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
109 struct super_block *sb = fs_info->sb;
111 if (sb->s_flags & MS_RDONLY)
114 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
115 sb->s_flags |= MS_RDONLY;
116 printk(KERN_INFO "btrfs is forced readonly\n");
121 * __btrfs_std_error decodes expected errors from the caller and
122 * invokes the approciate error response.
124 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
125 unsigned int line, int errno)
127 struct super_block *sb = fs_info->sb;
132 * Special case: if the error is EROFS, and we're already
133 * under MS_RDONLY, then it is safe here.
135 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
138 errstr = btrfs_decode_error(fs_info, errno, nbuf);
139 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
140 sb->s_id, function, line, errstr);
141 save_error_info(fs_info);
143 btrfs_handle_error(fs_info);
146 static void btrfs_put_super(struct super_block *sb)
148 struct btrfs_root *root = btrfs_sb(sb);
151 ret = close_ctree(root);
152 sb->s_fs_info = NULL;
154 (void)ret; /* FIXME: need to fix VFS to return error? */
158 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
159 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
160 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
161 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
162 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
163 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
164 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
165 Opt_inode_cache, Opt_err,
168 static match_table_t tokens = {
169 {Opt_degraded, "degraded"},
170 {Opt_subvol, "subvol=%s"},
171 {Opt_subvolid, "subvolid=%d"},
172 {Opt_device, "device=%s"},
173 {Opt_nodatasum, "nodatasum"},
174 {Opt_nodatacow, "nodatacow"},
175 {Opt_nobarrier, "nobarrier"},
176 {Opt_max_inline, "max_inline=%s"},
177 {Opt_alloc_start, "alloc_start=%s"},
178 {Opt_thread_pool, "thread_pool=%d"},
179 {Opt_compress, "compress"},
180 {Opt_compress_type, "compress=%s"},
181 {Opt_compress_force, "compress-force"},
182 {Opt_compress_force_type, "compress-force=%s"},
184 {Opt_ssd_spread, "ssd_spread"},
185 {Opt_nossd, "nossd"},
186 {Opt_noacl, "noacl"},
187 {Opt_notreelog, "notreelog"},
188 {Opt_flushoncommit, "flushoncommit"},
189 {Opt_ratio, "metadata_ratio=%d"},
190 {Opt_discard, "discard"},
191 {Opt_space_cache, "space_cache"},
192 {Opt_clear_cache, "clear_cache"},
193 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
194 {Opt_enospc_debug, "enospc_debug"},
195 {Opt_subvolrootid, "subvolrootid=%d"},
196 {Opt_defrag, "autodefrag"},
197 {Opt_inode_cache, "inode_cache"},
202 * Regular mount options parser. Everything that is needed only when
203 * reading in a new superblock is parsed here.
205 int btrfs_parse_options(struct btrfs_root *root, char *options)
207 struct btrfs_fs_info *info = root->fs_info;
208 substring_t args[MAX_OPT_ARGS];
209 char *p, *num, *orig;
213 bool compress_force = false;
219 * strsep changes the string, duplicate it because parse_options
222 options = kstrdup(options, GFP_NOFS);
228 while ((p = strsep(&options, ",")) != NULL) {
233 token = match_token(p, tokens, args);
236 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
237 btrfs_set_opt(info->mount_opt, DEGRADED);
241 case Opt_subvolrootid:
244 * These are parsed by btrfs_parse_early_options
245 * and can be happily ignored here.
249 printk(KERN_INFO "btrfs: setting nodatasum\n");
250 btrfs_set_opt(info->mount_opt, NODATASUM);
253 printk(KERN_INFO "btrfs: setting nodatacow\n");
254 btrfs_set_opt(info->mount_opt, NODATACOW);
255 btrfs_set_opt(info->mount_opt, NODATASUM);
257 case Opt_compress_force:
258 case Opt_compress_force_type:
259 compress_force = true;
261 case Opt_compress_type:
262 if (token == Opt_compress ||
263 token == Opt_compress_force ||
264 strcmp(args[0].from, "zlib") == 0) {
265 compress_type = "zlib";
266 info->compress_type = BTRFS_COMPRESS_ZLIB;
267 } else if (strcmp(args[0].from, "lzo") == 0) {
268 compress_type = "lzo";
269 info->compress_type = BTRFS_COMPRESS_LZO;
275 btrfs_set_opt(info->mount_opt, COMPRESS);
276 if (compress_force) {
277 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
278 pr_info("btrfs: force %s compression\n",
281 pr_info("btrfs: use %s compression\n",
285 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
286 btrfs_set_opt(info->mount_opt, SSD);
289 printk(KERN_INFO "btrfs: use spread ssd "
290 "allocation scheme\n");
291 btrfs_set_opt(info->mount_opt, SSD);
292 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
295 printk(KERN_INFO "btrfs: not using ssd allocation "
297 btrfs_set_opt(info->mount_opt, NOSSD);
298 btrfs_clear_opt(info->mount_opt, SSD);
299 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
302 printk(KERN_INFO "btrfs: turning off barriers\n");
303 btrfs_set_opt(info->mount_opt, NOBARRIER);
305 case Opt_thread_pool:
307 match_int(&args[0], &intarg);
309 info->thread_pool_size = intarg;
310 printk(KERN_INFO "btrfs: thread pool %d\n",
311 info->thread_pool_size);
315 num = match_strdup(&args[0]);
317 info->max_inline = memparse(num, NULL);
320 if (info->max_inline) {
321 info->max_inline = max_t(u64,
325 printk(KERN_INFO "btrfs: max_inline at %llu\n",
326 (unsigned long long)info->max_inline);
329 case Opt_alloc_start:
330 num = match_strdup(&args[0]);
332 info->alloc_start = memparse(num, NULL);
335 "btrfs: allocations start at %llu\n",
336 (unsigned long long)info->alloc_start);
340 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
343 printk(KERN_INFO "btrfs: disabling tree log\n");
344 btrfs_set_opt(info->mount_opt, NOTREELOG);
346 case Opt_flushoncommit:
347 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
348 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
352 match_int(&args[0], &intarg);
354 info->metadata_ratio = intarg;
355 printk(KERN_INFO "btrfs: metadata ratio %d\n",
356 info->metadata_ratio);
360 btrfs_set_opt(info->mount_opt, DISCARD);
362 case Opt_space_cache:
363 printk(KERN_INFO "btrfs: enabling disk space caching\n");
364 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
366 case Opt_inode_cache:
367 printk(KERN_INFO "btrfs: enabling inode map caching\n");
368 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
370 case Opt_clear_cache:
371 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
372 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
374 case Opt_user_subvol_rm_allowed:
375 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
377 case Opt_enospc_debug:
378 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
381 printk(KERN_INFO "btrfs: enabling auto defrag");
382 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
385 printk(KERN_INFO "btrfs: unrecognized mount option "
399 * Parse mount options that are required early in the mount process.
401 * All other options will be parsed on much later in the mount process and
402 * only when we need to allocate a new super block.
404 static int btrfs_parse_early_options(const char *options, fmode_t flags,
405 void *holder, char **subvol_name, u64 *subvol_objectid,
406 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
408 substring_t args[MAX_OPT_ARGS];
409 char *opts, *orig, *p;
417 * strsep changes the string, duplicate it because parse_options
420 opts = kstrdup(options, GFP_KERNEL);
425 while ((p = strsep(&opts, ",")) != NULL) {
430 token = match_token(p, tokens, args);
433 *subvol_name = match_strdup(&args[0]);
437 error = match_int(&args[0], &intarg);
439 /* we want the original fs_tree */
442 BTRFS_FS_TREE_OBJECTID;
444 *subvol_objectid = intarg;
447 case Opt_subvolrootid:
449 error = match_int(&args[0], &intarg);
451 /* we want the original fs_tree */
454 BTRFS_FS_TREE_OBJECTID;
456 *subvol_rootid = intarg;
460 error = btrfs_scan_one_device(match_strdup(&args[0]),
461 flags, holder, fs_devices);
474 * If no subvolume name is specified we use the default one. Allocate
475 * a copy of the string "." here so that code later in the
476 * mount path doesn't care if it's the default volume or another one.
479 *subvol_name = kstrdup(".", GFP_KERNEL);
486 static struct dentry *get_default_root(struct super_block *sb,
489 struct btrfs_root *root = sb->s_fs_info;
490 struct btrfs_root *new_root;
491 struct btrfs_dir_item *di;
492 struct btrfs_path *path;
493 struct btrfs_key location;
499 * We have a specific subvol we want to mount, just setup location and
500 * go look up the root.
502 if (subvol_objectid) {
503 location.objectid = subvol_objectid;
504 location.type = BTRFS_ROOT_ITEM_KEY;
505 location.offset = (u64)-1;
509 path = btrfs_alloc_path();
511 return ERR_PTR(-ENOMEM);
512 path->leave_spinning = 1;
515 * Find the "default" dir item which points to the root item that we
516 * will mount by default if we haven't been given a specific subvolume
519 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
520 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
522 btrfs_free_path(path);
527 * Ok the default dir item isn't there. This is weird since
528 * it's always been there, but don't freak out, just try and
529 * mount to root most subvolume.
531 btrfs_free_path(path);
532 dir_id = BTRFS_FIRST_FREE_OBJECTID;
533 new_root = root->fs_info->fs_root;
537 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
538 btrfs_free_path(path);
541 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
542 if (IS_ERR(new_root))
543 return ERR_CAST(new_root);
545 if (btrfs_root_refs(&new_root->root_item) == 0)
546 return ERR_PTR(-ENOENT);
548 dir_id = btrfs_root_dirid(&new_root->root_item);
550 location.objectid = dir_id;
551 location.type = BTRFS_INODE_ITEM_KEY;
554 inode = btrfs_iget(sb, &location, new_root, &new);
556 return ERR_CAST(inode);
559 * If we're just mounting the root most subvol put the inode and return
560 * a reference to the dentry. We will have already gotten a reference
561 * to the inode in btrfs_fill_super so we're good to go.
563 if (!new && sb->s_root->d_inode == inode) {
565 return dget(sb->s_root);
568 return d_obtain_alias(inode);
571 static int btrfs_fill_super(struct super_block *sb,
572 struct btrfs_fs_devices *fs_devices,
573 void *data, int silent)
576 struct dentry *root_dentry;
577 struct btrfs_root *tree_root;
578 struct btrfs_key key;
581 sb->s_maxbytes = MAX_LFS_FILESIZE;
582 sb->s_magic = BTRFS_SUPER_MAGIC;
583 sb->s_op = &btrfs_super_ops;
584 sb->s_d_op = &btrfs_dentry_operations;
585 sb->s_export_op = &btrfs_export_ops;
586 sb->s_xattr = btrfs_xattr_handlers;
588 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
589 sb->s_flags |= MS_POSIXACL;
592 tree_root = open_ctree(sb, fs_devices, (char *)data);
594 if (IS_ERR(tree_root)) {
595 printk("btrfs: open_ctree failed\n");
596 return PTR_ERR(tree_root);
598 sb->s_fs_info = tree_root;
600 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
601 key.type = BTRFS_INODE_ITEM_KEY;
603 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
605 err = PTR_ERR(inode);
609 root_dentry = d_alloc_root(inode);
616 sb->s_root = root_dentry;
618 save_mount_options(sb, data);
619 cleancache_init_fs(sb);
623 close_ctree(tree_root);
627 int btrfs_sync_fs(struct super_block *sb, int wait)
629 struct btrfs_trans_handle *trans;
630 struct btrfs_root *root = btrfs_sb(sb);
633 trace_btrfs_sync_fs(wait);
636 filemap_flush(root->fs_info->btree_inode->i_mapping);
640 btrfs_start_delalloc_inodes(root, 0);
641 btrfs_wait_ordered_extents(root, 0, 0);
643 trans = btrfs_start_transaction(root, 0);
645 return PTR_ERR(trans);
646 ret = btrfs_commit_transaction(trans, root);
650 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
652 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
653 struct btrfs_fs_info *info = root->fs_info;
656 if (btrfs_test_opt(root, DEGRADED))
657 seq_puts(seq, ",degraded");
658 if (btrfs_test_opt(root, NODATASUM))
659 seq_puts(seq, ",nodatasum");
660 if (btrfs_test_opt(root, NODATACOW))
661 seq_puts(seq, ",nodatacow");
662 if (btrfs_test_opt(root, NOBARRIER))
663 seq_puts(seq, ",nobarrier");
664 if (info->max_inline != 8192 * 1024)
665 seq_printf(seq, ",max_inline=%llu",
666 (unsigned long long)info->max_inline);
667 if (info->alloc_start != 0)
668 seq_printf(seq, ",alloc_start=%llu",
669 (unsigned long long)info->alloc_start);
670 if (info->thread_pool_size != min_t(unsigned long,
671 num_online_cpus() + 2, 8))
672 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
673 if (btrfs_test_opt(root, COMPRESS)) {
674 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
675 compress_type = "zlib";
677 compress_type = "lzo";
678 if (btrfs_test_opt(root, FORCE_COMPRESS))
679 seq_printf(seq, ",compress-force=%s", compress_type);
681 seq_printf(seq, ",compress=%s", compress_type);
683 if (btrfs_test_opt(root, NOSSD))
684 seq_puts(seq, ",nossd");
685 if (btrfs_test_opt(root, SSD_SPREAD))
686 seq_puts(seq, ",ssd_spread");
687 else if (btrfs_test_opt(root, SSD))
688 seq_puts(seq, ",ssd");
689 if (btrfs_test_opt(root, NOTREELOG))
690 seq_puts(seq, ",notreelog");
691 if (btrfs_test_opt(root, FLUSHONCOMMIT))
692 seq_puts(seq, ",flushoncommit");
693 if (btrfs_test_opt(root, DISCARD))
694 seq_puts(seq, ",discard");
695 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
696 seq_puts(seq, ",noacl");
697 if (btrfs_test_opt(root, SPACE_CACHE))
698 seq_puts(seq, ",space_cache");
699 if (btrfs_test_opt(root, CLEAR_CACHE))
700 seq_puts(seq, ",clear_cache");
701 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
702 seq_puts(seq, ",user_subvol_rm_allowed");
703 if (btrfs_test_opt(root, ENOSPC_DEBUG))
704 seq_puts(seq, ",enospc_debug");
705 if (btrfs_test_opt(root, AUTO_DEFRAG))
706 seq_puts(seq, ",autodefrag");
707 if (btrfs_test_opt(root, INODE_MAP_CACHE))
708 seq_puts(seq, ",inode_cache");
712 static int btrfs_test_super(struct super_block *s, void *data)
714 struct btrfs_root *test_root = data;
715 struct btrfs_root *root = btrfs_sb(s);
718 * If this super block is going away, return false as it
719 * can't match as an existing super block.
721 if (!atomic_read(&s->s_active))
723 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
726 static int btrfs_set_super(struct super_block *s, void *data)
730 return set_anon_super(s, data);
735 * Find a superblock for the given device / mount point.
737 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
738 * for multiple device setup. Make sure to keep it in sync.
740 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
741 const char *device_name, void *data)
743 struct block_device *bdev = NULL;
744 struct super_block *s;
746 struct btrfs_fs_devices *fs_devices = NULL;
747 struct btrfs_root *tree_root = NULL;
748 struct btrfs_fs_info *fs_info = NULL;
749 fmode_t mode = FMODE_READ;
750 char *subvol_name = NULL;
751 u64 subvol_objectid = 0;
752 u64 subvol_rootid = 0;
755 if (!(flags & MS_RDONLY))
758 error = btrfs_parse_early_options(data, mode, fs_type,
759 &subvol_name, &subvol_objectid,
760 &subvol_rootid, &fs_devices);
762 return ERR_PTR(error);
764 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
766 goto error_free_subvol_name;
768 error = btrfs_open_devices(fs_devices, mode, fs_type);
770 goto error_free_subvol_name;
772 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
774 goto error_close_devices;
778 * Setup a dummy root and fs_info for test/set super. This is because
779 * we don't actually fill this stuff out until open_ctree, but we need
780 * it for searching for existing supers, so this lets us do that and
781 * then open_ctree will properly initialize everything later.
783 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
784 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
785 if (!fs_info || !tree_root) {
787 goto error_close_devices;
789 fs_info->tree_root = tree_root;
790 fs_info->fs_devices = fs_devices;
791 tree_root->fs_info = fs_info;
793 bdev = fs_devices->latest_bdev;
794 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
799 if ((flags ^ s->s_flags) & MS_RDONLY) {
800 deactivate_locked_super(s);
802 goto error_close_devices;
805 btrfs_close_devices(fs_devices);
809 char b[BDEVNAME_SIZE];
811 s->s_flags = flags | MS_NOSEC;
812 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
813 error = btrfs_fill_super(s, fs_devices, data,
814 flags & MS_SILENT ? 1 : 0);
816 deactivate_locked_super(s);
817 goto error_free_subvol_name;
820 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
821 s->s_flags |= MS_ACTIVE;
824 /* if they gave us a subvolume name bind mount into that */
825 if (strcmp(subvol_name, ".")) {
826 struct dentry *new_root;
828 root = get_default_root(s, subvol_rootid);
830 error = PTR_ERR(root);
831 deactivate_locked_super(s);
832 goto error_free_subvol_name;
835 mutex_lock(&root->d_inode->i_mutex);
836 new_root = lookup_one_len(subvol_name, root,
837 strlen(subvol_name));
838 mutex_unlock(&root->d_inode->i_mutex);
840 if (IS_ERR(new_root)) {
842 deactivate_locked_super(s);
843 error = PTR_ERR(new_root);
844 goto error_free_subvol_name;
846 if (!new_root->d_inode) {
849 deactivate_locked_super(s);
851 goto error_free_subvol_name;
856 root = get_default_root(s, subvol_objectid);
858 error = PTR_ERR(root);
859 deactivate_locked_super(s);
860 goto error_free_subvol_name;
870 btrfs_close_devices(fs_devices);
873 error_free_subvol_name:
875 return ERR_PTR(error);
878 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
880 struct btrfs_root *root = btrfs_sb(sb);
883 ret = btrfs_parse_options(root, data);
887 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
890 if (*flags & MS_RDONLY) {
891 sb->s_flags |= MS_RDONLY;
893 ret = btrfs_commit_super(root);
896 if (root->fs_info->fs_devices->rw_devices == 0)
899 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
902 ret = btrfs_cleanup_fs_roots(root->fs_info);
905 /* recover relocation */
906 ret = btrfs_recover_relocation(root);
909 sb->s_flags &= ~MS_RDONLY;
915 /* Used to sort the devices by max_avail(descending sort) */
916 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
917 const void *dev_info2)
919 if (((struct btrfs_device_info *)dev_info1)->max_avail >
920 ((struct btrfs_device_info *)dev_info2)->max_avail)
922 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
923 ((struct btrfs_device_info *)dev_info2)->max_avail)
930 * sort the devices by max_avail, in which max free extent size of each device
931 * is stored.(Descending Sort)
933 static inline void btrfs_descending_sort_devices(
934 struct btrfs_device_info *devices,
937 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
938 btrfs_cmp_device_free_bytes, NULL);
942 * The helper to calc the free space on the devices that can be used to store
945 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
947 struct btrfs_fs_info *fs_info = root->fs_info;
948 struct btrfs_device_info *devices_info;
949 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
950 struct btrfs_device *device;
957 int i = 0, nr_devices;
960 nr_devices = fs_info->fs_devices->rw_devices;
963 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
968 /* calc min stripe number for data space alloction */
969 type = btrfs_get_alloc_profile(root, 1);
970 if (type & BTRFS_BLOCK_GROUP_RAID0)
972 else if (type & BTRFS_BLOCK_GROUP_RAID1)
974 else if (type & BTRFS_BLOCK_GROUP_RAID10)
977 if (type & BTRFS_BLOCK_GROUP_DUP)
978 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
980 min_stripe_size = BTRFS_STRIPE_LEN;
982 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
983 if (!device->in_fs_metadata)
986 avail_space = device->total_bytes - device->bytes_used;
988 /* align with stripe_len */
989 do_div(avail_space, BTRFS_STRIPE_LEN);
990 avail_space *= BTRFS_STRIPE_LEN;
993 * In order to avoid overwritting the superblock on the drive,
994 * btrfs starts at an offset of at least 1MB when doing chunk
997 skip_space = 1024 * 1024;
999 /* user can set the offset in fs_info->alloc_start. */
1000 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1001 device->total_bytes)
1002 skip_space = max(fs_info->alloc_start, skip_space);
1005 * btrfs can not use the free space in [0, skip_space - 1],
1006 * we must subtract it from the total. In order to implement
1007 * it, we account the used space in this range first.
1009 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1012 kfree(devices_info);
1016 /* calc the free space in [0, skip_space - 1] */
1017 skip_space -= used_space;
1020 * we can use the free space in [0, skip_space - 1], subtract
1021 * it from the total.
1023 if (avail_space && avail_space >= skip_space)
1024 avail_space -= skip_space;
1028 if (avail_space < min_stripe_size)
1031 devices_info[i].dev = device;
1032 devices_info[i].max_avail = avail_space;
1039 btrfs_descending_sort_devices(devices_info, nr_devices);
1043 while (nr_devices >= min_stripes) {
1044 if (devices_info[i].max_avail >= min_stripe_size) {
1048 avail_space += devices_info[i].max_avail * min_stripes;
1049 alloc_size = devices_info[i].max_avail;
1050 for (j = i + 1 - min_stripes; j <= i; j++)
1051 devices_info[j].max_avail -= alloc_size;
1057 kfree(devices_info);
1058 *free_bytes = avail_space;
1062 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1064 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1065 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
1066 struct list_head *head = &root->fs_info->space_info;
1067 struct btrfs_space_info *found;
1069 u64 total_free_data = 0;
1070 int bits = dentry->d_sb->s_blocksize_bits;
1071 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1074 /* holding chunk_muext to avoid allocating new chunks */
1075 mutex_lock(&root->fs_info->chunk_mutex);
1077 list_for_each_entry_rcu(found, head, list) {
1078 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1079 total_free_data += found->disk_total - found->disk_used;
1081 btrfs_account_ro_block_groups_free_space(found);
1084 total_used += found->disk_used;
1088 buf->f_namelen = BTRFS_NAME_LEN;
1089 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1090 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1091 buf->f_bsize = dentry->d_sb->s_blocksize;
1092 buf->f_type = BTRFS_SUPER_MAGIC;
1093 buf->f_bavail = total_free_data;
1094 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1096 mutex_unlock(&root->fs_info->chunk_mutex);
1099 buf->f_bavail += total_free_data;
1100 buf->f_bavail = buf->f_bavail >> bits;
1101 mutex_unlock(&root->fs_info->chunk_mutex);
1103 /* We treat it as constant endianness (it doesn't matter _which_)
1104 because we want the fsid to come out the same whether mounted
1105 on a big-endian or little-endian host */
1106 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1107 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1108 /* Mask in the root object ID too, to disambiguate subvols */
1109 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1110 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1115 static struct file_system_type btrfs_fs_type = {
1116 .owner = THIS_MODULE,
1118 .mount = btrfs_mount,
1119 .kill_sb = kill_anon_super,
1120 .fs_flags = FS_REQUIRES_DEV,
1124 * used by btrfsctl to scan devices when no FS is mounted
1126 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1129 struct btrfs_ioctl_vol_args *vol;
1130 struct btrfs_fs_devices *fs_devices;
1133 if (!capable(CAP_SYS_ADMIN))
1136 vol = memdup_user((void __user *)arg, sizeof(*vol));
1138 return PTR_ERR(vol);
1141 case BTRFS_IOC_SCAN_DEV:
1142 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1143 &btrfs_fs_type, &fs_devices);
1151 static int btrfs_freeze(struct super_block *sb)
1153 struct btrfs_root *root = btrfs_sb(sb);
1154 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1155 mutex_lock(&root->fs_info->cleaner_mutex);
1159 static int btrfs_unfreeze(struct super_block *sb)
1161 struct btrfs_root *root = btrfs_sb(sb);
1162 mutex_unlock(&root->fs_info->cleaner_mutex);
1163 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1167 static const struct super_operations btrfs_super_ops = {
1168 .drop_inode = btrfs_drop_inode,
1169 .evict_inode = btrfs_evict_inode,
1170 .put_super = btrfs_put_super,
1171 .sync_fs = btrfs_sync_fs,
1172 .show_options = btrfs_show_options,
1173 .write_inode = btrfs_write_inode,
1174 .dirty_inode = btrfs_dirty_inode,
1175 .alloc_inode = btrfs_alloc_inode,
1176 .destroy_inode = btrfs_destroy_inode,
1177 .statfs = btrfs_statfs,
1178 .remount_fs = btrfs_remount,
1179 .freeze_fs = btrfs_freeze,
1180 .unfreeze_fs = btrfs_unfreeze,
1183 static const struct file_operations btrfs_ctl_fops = {
1184 .unlocked_ioctl = btrfs_control_ioctl,
1185 .compat_ioctl = btrfs_control_ioctl,
1186 .owner = THIS_MODULE,
1187 .llseek = noop_llseek,
1190 static struct miscdevice btrfs_misc = {
1191 .minor = BTRFS_MINOR,
1192 .name = "btrfs-control",
1193 .fops = &btrfs_ctl_fops
1196 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1197 MODULE_ALIAS("devname:btrfs-control");
1199 static int btrfs_interface_init(void)
1201 return misc_register(&btrfs_misc);
1204 static void btrfs_interface_exit(void)
1206 if (misc_deregister(&btrfs_misc) < 0)
1207 printk(KERN_INFO "misc_deregister failed for control device");
1210 static int __init init_btrfs_fs(void)
1214 err = btrfs_init_sysfs();
1218 err = btrfs_init_compress();
1222 err = btrfs_init_cachep();
1226 err = extent_io_init();
1230 err = extent_map_init();
1232 goto free_extent_io;
1234 err = btrfs_delayed_inode_init();
1236 goto free_extent_map;
1238 err = btrfs_interface_init();
1240 goto free_delayed_inode;
1242 err = register_filesystem(&btrfs_fs_type);
1244 goto unregister_ioctl;
1246 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1250 btrfs_interface_exit();
1252 btrfs_delayed_inode_exit();
1258 btrfs_destroy_cachep();
1260 btrfs_exit_compress();
1266 static void __exit exit_btrfs_fs(void)
1268 btrfs_destroy_cachep();
1269 btrfs_delayed_inode_exit();
1272 btrfs_interface_exit();
1273 unregister_filesystem(&btrfs_fs_type);
1275 btrfs_cleanup_fs_uuids();
1276 btrfs_exit_compress();
1279 module_init(init_btrfs_fs)
1280 module_exit(exit_btrfs_fs)
1282 MODULE_LICENSE("GPL");