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>
43 #include "delayed-inode.h"
46 #include "transaction.h"
47 #include "btrfs_inode.h"
49 #include "print-tree.h"
54 #include "compression.h"
56 #define CREATE_TRACE_POINTS
57 #include <trace/events/btrfs.h>
59 static const struct super_operations btrfs_super_ops;
61 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
68 errstr = "IO failure";
71 errstr = "Out of memory";
74 errstr = "Readonly filesystem";
78 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
87 static void __save_error_info(struct btrfs_fs_info *fs_info)
90 * today we only save the error info into ram. Long term we'll
91 * also send it down to the disk
93 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
97 * We move write_super stuff at umount in order to avoid deadlock
98 * for umount hold all lock.
100 static void save_error_info(struct btrfs_fs_info *fs_info)
102 __save_error_info(fs_info);
105 /* btrfs handle error by forcing the filesystem readonly */
106 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
108 struct super_block *sb = fs_info->sb;
110 if (sb->s_flags & MS_RDONLY)
113 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
114 sb->s_flags |= MS_RDONLY;
115 printk(KERN_INFO "btrfs is forced readonly\n");
120 * __btrfs_std_error decodes expected errors from the caller and
121 * invokes the approciate error response.
123 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
124 unsigned int line, int errno)
126 struct super_block *sb = fs_info->sb;
131 * Special case: if the error is EROFS, and we're already
132 * under MS_RDONLY, then it is safe here.
134 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
137 errstr = btrfs_decode_error(fs_info, errno, nbuf);
138 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
139 sb->s_id, function, line, errstr);
140 save_error_info(fs_info);
142 btrfs_handle_error(fs_info);
145 static void btrfs_put_super(struct super_block *sb)
147 struct btrfs_root *root = btrfs_sb(sb);
150 ret = close_ctree(root);
151 sb->s_fs_info = NULL;
153 (void)ret; /* FIXME: need to fix VFS to return error? */
157 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
158 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
159 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
160 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
161 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
162 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
163 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
164 Opt_inode_cache, Opt_err,
167 static match_table_t tokens = {
168 {Opt_degraded, "degraded"},
169 {Opt_subvol, "subvol=%s"},
170 {Opt_subvolid, "subvolid=%d"},
171 {Opt_device, "device=%s"},
172 {Opt_nodatasum, "nodatasum"},
173 {Opt_nodatacow, "nodatacow"},
174 {Opt_nobarrier, "nobarrier"},
175 {Opt_max_inline, "max_inline=%s"},
176 {Opt_alloc_start, "alloc_start=%s"},
177 {Opt_thread_pool, "thread_pool=%d"},
178 {Opt_compress, "compress"},
179 {Opt_compress_type, "compress=%s"},
180 {Opt_compress_force, "compress-force"},
181 {Opt_compress_force_type, "compress-force=%s"},
183 {Opt_ssd_spread, "ssd_spread"},
184 {Opt_nossd, "nossd"},
185 {Opt_noacl, "noacl"},
186 {Opt_notreelog, "notreelog"},
187 {Opt_flushoncommit, "flushoncommit"},
188 {Opt_ratio, "metadata_ratio=%d"},
189 {Opt_discard, "discard"},
190 {Opt_space_cache, "space_cache"},
191 {Opt_clear_cache, "clear_cache"},
192 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
193 {Opt_enospc_debug, "enospc_debug"},
194 {Opt_subvolrootid, "subvolrootid=%d"},
195 {Opt_defrag, "autodefrag"},
196 {Opt_inode_cache, "inode_cache"},
201 * Regular mount options parser. Everything that is needed only when
202 * reading in a new superblock is parsed here.
204 int btrfs_parse_options(struct btrfs_root *root, char *options)
206 struct btrfs_fs_info *info = root->fs_info;
207 substring_t args[MAX_OPT_ARGS];
208 char *p, *num, *orig;
212 bool compress_force = false;
218 * strsep changes the string, duplicate it because parse_options
221 options = kstrdup(options, GFP_NOFS);
227 while ((p = strsep(&options, ",")) != NULL) {
232 token = match_token(p, tokens, args);
235 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
236 btrfs_set_opt(info->mount_opt, DEGRADED);
240 case Opt_subvolrootid:
243 * These are parsed by btrfs_parse_early_options
244 * and can be happily ignored here.
248 printk(KERN_INFO "btrfs: setting nodatasum\n");
249 btrfs_set_opt(info->mount_opt, NODATASUM);
252 printk(KERN_INFO "btrfs: setting nodatacow\n");
253 btrfs_set_opt(info->mount_opt, NODATACOW);
254 btrfs_set_opt(info->mount_opt, NODATASUM);
256 case Opt_compress_force:
257 case Opt_compress_force_type:
258 compress_force = true;
260 case Opt_compress_type:
261 if (token == Opt_compress ||
262 token == Opt_compress_force ||
263 strcmp(args[0].from, "zlib") == 0) {
264 compress_type = "zlib";
265 info->compress_type = BTRFS_COMPRESS_ZLIB;
266 } else if (strcmp(args[0].from, "lzo") == 0) {
267 compress_type = "lzo";
268 info->compress_type = BTRFS_COMPRESS_LZO;
274 btrfs_set_opt(info->mount_opt, COMPRESS);
275 if (compress_force) {
276 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
277 pr_info("btrfs: force %s compression\n",
280 pr_info("btrfs: use %s compression\n",
284 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
285 btrfs_set_opt(info->mount_opt, SSD);
288 printk(KERN_INFO "btrfs: use spread ssd "
289 "allocation scheme\n");
290 btrfs_set_opt(info->mount_opt, SSD);
291 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
294 printk(KERN_INFO "btrfs: not using ssd allocation "
296 btrfs_set_opt(info->mount_opt, NOSSD);
297 btrfs_clear_opt(info->mount_opt, SSD);
298 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
301 printk(KERN_INFO "btrfs: turning off barriers\n");
302 btrfs_set_opt(info->mount_opt, NOBARRIER);
304 case Opt_thread_pool:
306 match_int(&args[0], &intarg);
308 info->thread_pool_size = intarg;
309 printk(KERN_INFO "btrfs: thread pool %d\n",
310 info->thread_pool_size);
314 num = match_strdup(&args[0]);
316 info->max_inline = memparse(num, NULL);
319 if (info->max_inline) {
320 info->max_inline = max_t(u64,
324 printk(KERN_INFO "btrfs: max_inline at %llu\n",
325 (unsigned long long)info->max_inline);
328 case Opt_alloc_start:
329 num = match_strdup(&args[0]);
331 info->alloc_start = memparse(num, NULL);
334 "btrfs: allocations start at %llu\n",
335 (unsigned long long)info->alloc_start);
339 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
342 printk(KERN_INFO "btrfs: disabling tree log\n");
343 btrfs_set_opt(info->mount_opt, NOTREELOG);
345 case Opt_flushoncommit:
346 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
347 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
351 match_int(&args[0], &intarg);
353 info->metadata_ratio = intarg;
354 printk(KERN_INFO "btrfs: metadata ratio %d\n",
355 info->metadata_ratio);
359 btrfs_set_opt(info->mount_opt, DISCARD);
361 case Opt_space_cache:
362 printk(KERN_INFO "btrfs: enabling disk space caching\n");
363 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
365 case Opt_inode_cache:
366 printk(KERN_INFO "btrfs: enabling inode map caching\n");
367 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
369 case Opt_clear_cache:
370 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
371 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
373 case Opt_user_subvol_rm_allowed:
374 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
376 case Opt_enospc_debug:
377 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
380 printk(KERN_INFO "btrfs: enabling auto defrag");
381 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
384 printk(KERN_INFO "btrfs: unrecognized mount option "
398 * Parse mount options that are required early in the mount process.
400 * All other options will be parsed on much later in the mount process and
401 * only when we need to allocate a new super block.
403 static int btrfs_parse_early_options(const char *options, fmode_t flags,
404 void *holder, char **subvol_name, u64 *subvol_objectid,
405 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
407 substring_t args[MAX_OPT_ARGS];
408 char *opts, *orig, *p;
416 * strsep changes the string, duplicate it because parse_options
419 opts = kstrdup(options, GFP_KERNEL);
424 while ((p = strsep(&opts, ",")) != NULL) {
429 token = match_token(p, tokens, args);
432 *subvol_name = match_strdup(&args[0]);
436 error = match_int(&args[0], &intarg);
438 /* we want the original fs_tree */
441 BTRFS_FS_TREE_OBJECTID;
443 *subvol_objectid = intarg;
446 case Opt_subvolrootid:
448 error = match_int(&args[0], &intarg);
450 /* we want the original fs_tree */
453 BTRFS_FS_TREE_OBJECTID;
455 *subvol_rootid = intarg;
459 error = btrfs_scan_one_device(match_strdup(&args[0]),
460 flags, holder, fs_devices);
473 * If no subvolume name is specified we use the default one. Allocate
474 * a copy of the string "." here so that code later in the
475 * mount path doesn't care if it's the default volume or another one.
478 *subvol_name = kstrdup(".", GFP_KERNEL);
485 static struct dentry *get_default_root(struct super_block *sb,
488 struct btrfs_root *root = sb->s_fs_info;
489 struct btrfs_root *new_root;
490 struct btrfs_dir_item *di;
491 struct btrfs_path *path;
492 struct btrfs_key location;
494 struct dentry *dentry;
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);
569 const struct qstr name = { .name = "/", .len = 1 };
572 * New inode, we need to make the dentry a sibling of s_root so
573 * everything gets cleaned up properly on unmount.
575 dentry = d_alloc(sb->s_root, &name);
578 return ERR_PTR(-ENOMEM);
580 d_splice_alias(inode, dentry);
583 * We found the inode in cache, just find a dentry for it and
584 * put the reference to the inode we just got.
586 dentry = d_find_alias(inode);
593 static int btrfs_fill_super(struct super_block *sb,
594 struct btrfs_fs_devices *fs_devices,
595 void *data, int silent)
598 struct dentry *root_dentry;
599 struct btrfs_root *tree_root;
600 struct btrfs_key key;
603 sb->s_maxbytes = MAX_LFS_FILESIZE;
604 sb->s_magic = BTRFS_SUPER_MAGIC;
605 sb->s_op = &btrfs_super_ops;
606 sb->s_d_op = &btrfs_dentry_operations;
607 sb->s_export_op = &btrfs_export_ops;
608 sb->s_xattr = btrfs_xattr_handlers;
610 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
611 sb->s_flags |= MS_POSIXACL;
614 tree_root = open_ctree(sb, fs_devices, (char *)data);
616 if (IS_ERR(tree_root)) {
617 printk("btrfs: open_ctree failed\n");
618 return PTR_ERR(tree_root);
620 sb->s_fs_info = tree_root;
622 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
623 key.type = BTRFS_INODE_ITEM_KEY;
625 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
627 err = PTR_ERR(inode);
631 root_dentry = d_alloc_root(inode);
638 sb->s_root = root_dentry;
640 save_mount_options(sb, data);
644 close_ctree(tree_root);
648 int btrfs_sync_fs(struct super_block *sb, int wait)
650 struct btrfs_trans_handle *trans;
651 struct btrfs_root *root = btrfs_sb(sb);
654 trace_btrfs_sync_fs(wait);
657 filemap_flush(root->fs_info->btree_inode->i_mapping);
661 btrfs_start_delalloc_inodes(root, 0);
662 btrfs_wait_ordered_extents(root, 0, 0);
664 trans = btrfs_start_transaction(root, 0);
666 return PTR_ERR(trans);
667 ret = btrfs_commit_transaction(trans, root);
671 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
673 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
674 struct btrfs_fs_info *info = root->fs_info;
677 if (btrfs_test_opt(root, DEGRADED))
678 seq_puts(seq, ",degraded");
679 if (btrfs_test_opt(root, NODATASUM))
680 seq_puts(seq, ",nodatasum");
681 if (btrfs_test_opt(root, NODATACOW))
682 seq_puts(seq, ",nodatacow");
683 if (btrfs_test_opt(root, NOBARRIER))
684 seq_puts(seq, ",nobarrier");
685 if (info->max_inline != 8192 * 1024)
686 seq_printf(seq, ",max_inline=%llu",
687 (unsigned long long)info->max_inline);
688 if (info->alloc_start != 0)
689 seq_printf(seq, ",alloc_start=%llu",
690 (unsigned long long)info->alloc_start);
691 if (info->thread_pool_size != min_t(unsigned long,
692 num_online_cpus() + 2, 8))
693 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
694 if (btrfs_test_opt(root, COMPRESS)) {
695 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
696 compress_type = "zlib";
698 compress_type = "lzo";
699 if (btrfs_test_opt(root, FORCE_COMPRESS))
700 seq_printf(seq, ",compress-force=%s", compress_type);
702 seq_printf(seq, ",compress=%s", compress_type);
704 if (btrfs_test_opt(root, NOSSD))
705 seq_puts(seq, ",nossd");
706 if (btrfs_test_opt(root, SSD_SPREAD))
707 seq_puts(seq, ",ssd_spread");
708 else if (btrfs_test_opt(root, SSD))
709 seq_puts(seq, ",ssd");
710 if (btrfs_test_opt(root, NOTREELOG))
711 seq_puts(seq, ",notreelog");
712 if (btrfs_test_opt(root, FLUSHONCOMMIT))
713 seq_puts(seq, ",flushoncommit");
714 if (btrfs_test_opt(root, DISCARD))
715 seq_puts(seq, ",discard");
716 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
717 seq_puts(seq, ",noacl");
718 if (btrfs_test_opt(root, SPACE_CACHE))
719 seq_puts(seq, ",space_cache");
720 if (btrfs_test_opt(root, CLEAR_CACHE))
721 seq_puts(seq, ",clear_cache");
722 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
723 seq_puts(seq, ",user_subvol_rm_allowed");
727 static int btrfs_test_super(struct super_block *s, void *data)
729 struct btrfs_root *test_root = data;
730 struct btrfs_root *root = btrfs_sb(s);
733 * If this super block is going away, return false as it
734 * can't match as an existing super block.
736 if (!atomic_read(&s->s_active))
738 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
741 static int btrfs_set_super(struct super_block *s, void *data)
745 return set_anon_super(s, data);
750 * Find a superblock for the given device / mount point.
752 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
753 * for multiple device setup. Make sure to keep it in sync.
755 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
756 const char *device_name, void *data)
758 struct block_device *bdev = NULL;
759 struct super_block *s;
761 struct btrfs_fs_devices *fs_devices = NULL;
762 struct btrfs_root *tree_root = NULL;
763 struct btrfs_fs_info *fs_info = NULL;
764 fmode_t mode = FMODE_READ;
765 char *subvol_name = NULL;
766 u64 subvol_objectid = 0;
767 u64 subvol_rootid = 0;
770 if (!(flags & MS_RDONLY))
773 error = btrfs_parse_early_options(data, mode, fs_type,
774 &subvol_name, &subvol_objectid,
775 &subvol_rootid, &fs_devices);
777 return ERR_PTR(error);
779 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
781 goto error_free_subvol_name;
783 error = btrfs_open_devices(fs_devices, mode, fs_type);
785 goto error_free_subvol_name;
787 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
789 goto error_close_devices;
793 * Setup a dummy root and fs_info for test/set super. This is because
794 * we don't actually fill this stuff out until open_ctree, but we need
795 * it for searching for existing supers, so this lets us do that and
796 * then open_ctree will properly initialize everything later.
798 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
799 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
800 if (!fs_info || !tree_root) {
802 goto error_close_devices;
804 fs_info->tree_root = tree_root;
805 fs_info->fs_devices = fs_devices;
806 tree_root->fs_info = fs_info;
808 bdev = fs_devices->latest_bdev;
809 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
814 if ((flags ^ s->s_flags) & MS_RDONLY) {
815 deactivate_locked_super(s);
817 goto error_close_devices;
820 btrfs_close_devices(fs_devices);
824 char b[BDEVNAME_SIZE];
827 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
828 error = btrfs_fill_super(s, fs_devices, data,
829 flags & MS_SILENT ? 1 : 0);
831 deactivate_locked_super(s);
832 goto error_free_subvol_name;
835 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
836 s->s_flags |= MS_ACTIVE;
839 /* if they gave us a subvolume name bind mount into that */
840 if (strcmp(subvol_name, ".")) {
841 struct dentry *new_root;
843 root = get_default_root(s, subvol_rootid);
845 error = PTR_ERR(root);
846 deactivate_locked_super(s);
847 goto error_free_subvol_name;
850 mutex_lock(&root->d_inode->i_mutex);
851 new_root = lookup_one_len(subvol_name, root,
852 strlen(subvol_name));
853 mutex_unlock(&root->d_inode->i_mutex);
855 if (IS_ERR(new_root)) {
857 deactivate_locked_super(s);
858 error = PTR_ERR(new_root);
859 goto error_free_subvol_name;
861 if (!new_root->d_inode) {
864 deactivate_locked_super(s);
866 goto error_free_subvol_name;
871 root = get_default_root(s, subvol_objectid);
873 error = PTR_ERR(root);
874 deactivate_locked_super(s);
875 goto error_free_subvol_name;
885 btrfs_close_devices(fs_devices);
888 error_free_subvol_name:
890 return ERR_PTR(error);
893 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
895 struct btrfs_root *root = btrfs_sb(sb);
898 ret = btrfs_parse_options(root, data);
902 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
905 if (*flags & MS_RDONLY) {
906 sb->s_flags |= MS_RDONLY;
908 ret = btrfs_commit_super(root);
911 if (root->fs_info->fs_devices->rw_devices == 0)
914 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
917 ret = btrfs_cleanup_fs_roots(root->fs_info);
920 /* recover relocation */
921 ret = btrfs_recover_relocation(root);
924 sb->s_flags &= ~MS_RDONLY;
930 /* Used to sort the devices by max_avail(descending sort) */
931 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
932 const void *dev_info2)
934 if (((struct btrfs_device_info *)dev_info1)->max_avail >
935 ((struct btrfs_device_info *)dev_info2)->max_avail)
937 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
938 ((struct btrfs_device_info *)dev_info2)->max_avail)
945 * sort the devices by max_avail, in which max free extent size of each device
946 * is stored.(Descending Sort)
948 static inline void btrfs_descending_sort_devices(
949 struct btrfs_device_info *devices,
952 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
953 btrfs_cmp_device_free_bytes, NULL);
957 * The helper to calc the free space on the devices that can be used to store
960 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
962 struct btrfs_fs_info *fs_info = root->fs_info;
963 struct btrfs_device_info *devices_info;
964 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
965 struct btrfs_device *device;
972 int i = 0, nr_devices;
975 nr_devices = fs_info->fs_devices->rw_devices;
978 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
983 /* calc min stripe number for data space alloction */
984 type = btrfs_get_alloc_profile(root, 1);
985 if (type & BTRFS_BLOCK_GROUP_RAID0)
987 else if (type & BTRFS_BLOCK_GROUP_RAID1)
989 else if (type & BTRFS_BLOCK_GROUP_RAID10)
992 if (type & BTRFS_BLOCK_GROUP_DUP)
993 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
995 min_stripe_size = BTRFS_STRIPE_LEN;
997 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
998 if (!device->in_fs_metadata)
1001 avail_space = device->total_bytes - device->bytes_used;
1003 /* align with stripe_len */
1004 do_div(avail_space, BTRFS_STRIPE_LEN);
1005 avail_space *= BTRFS_STRIPE_LEN;
1008 * In order to avoid overwritting the superblock on the drive,
1009 * btrfs starts at an offset of at least 1MB when doing chunk
1012 skip_space = 1024 * 1024;
1014 /* user can set the offset in fs_info->alloc_start. */
1015 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1016 device->total_bytes)
1017 skip_space = max(fs_info->alloc_start, skip_space);
1020 * btrfs can not use the free space in [0, skip_space - 1],
1021 * we must subtract it from the total. In order to implement
1022 * it, we account the used space in this range first.
1024 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1027 kfree(devices_info);
1031 /* calc the free space in [0, skip_space - 1] */
1032 skip_space -= used_space;
1035 * we can use the free space in [0, skip_space - 1], subtract
1036 * it from the total.
1038 if (avail_space && avail_space >= skip_space)
1039 avail_space -= skip_space;
1043 if (avail_space < min_stripe_size)
1046 devices_info[i].dev = device;
1047 devices_info[i].max_avail = avail_space;
1054 btrfs_descending_sort_devices(devices_info, nr_devices);
1058 while (nr_devices >= min_stripes) {
1059 if (devices_info[i].max_avail >= min_stripe_size) {
1063 avail_space += devices_info[i].max_avail * min_stripes;
1064 alloc_size = devices_info[i].max_avail;
1065 for (j = i + 1 - min_stripes; j <= i; j++)
1066 devices_info[j].max_avail -= alloc_size;
1072 kfree(devices_info);
1073 *free_bytes = avail_space;
1077 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1079 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1080 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
1081 struct list_head *head = &root->fs_info->space_info;
1082 struct btrfs_space_info *found;
1084 u64 total_free_data = 0;
1085 int bits = dentry->d_sb->s_blocksize_bits;
1086 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1089 /* holding chunk_muext to avoid allocating new chunks */
1090 mutex_lock(&root->fs_info->chunk_mutex);
1092 list_for_each_entry_rcu(found, head, list) {
1093 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1094 total_free_data += found->disk_total - found->disk_used;
1096 btrfs_account_ro_block_groups_free_space(found);
1099 total_used += found->disk_used;
1103 buf->f_namelen = BTRFS_NAME_LEN;
1104 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1105 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1106 buf->f_bsize = dentry->d_sb->s_blocksize;
1107 buf->f_type = BTRFS_SUPER_MAGIC;
1108 buf->f_bavail = total_free_data;
1109 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1111 mutex_unlock(&root->fs_info->chunk_mutex);
1114 buf->f_bavail += total_free_data;
1115 buf->f_bavail = buf->f_bavail >> bits;
1116 mutex_unlock(&root->fs_info->chunk_mutex);
1118 /* We treat it as constant endianness (it doesn't matter _which_)
1119 because we want the fsid to come out the same whether mounted
1120 on a big-endian or little-endian host */
1121 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1122 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1123 /* Mask in the root object ID too, to disambiguate subvols */
1124 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1125 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1130 static struct file_system_type btrfs_fs_type = {
1131 .owner = THIS_MODULE,
1133 .mount = btrfs_mount,
1134 .kill_sb = kill_anon_super,
1135 .fs_flags = FS_REQUIRES_DEV,
1139 * used by btrfsctl to scan devices when no FS is mounted
1141 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1144 struct btrfs_ioctl_vol_args *vol;
1145 struct btrfs_fs_devices *fs_devices;
1148 if (!capable(CAP_SYS_ADMIN))
1151 vol = memdup_user((void __user *)arg, sizeof(*vol));
1153 return PTR_ERR(vol);
1156 case BTRFS_IOC_SCAN_DEV:
1157 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1158 &btrfs_fs_type, &fs_devices);
1166 static int btrfs_freeze(struct super_block *sb)
1168 struct btrfs_root *root = btrfs_sb(sb);
1169 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1170 mutex_lock(&root->fs_info->cleaner_mutex);
1174 static int btrfs_unfreeze(struct super_block *sb)
1176 struct btrfs_root *root = btrfs_sb(sb);
1177 mutex_unlock(&root->fs_info->cleaner_mutex);
1178 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1182 static const struct super_operations btrfs_super_ops = {
1183 .drop_inode = btrfs_drop_inode,
1184 .evict_inode = btrfs_evict_inode,
1185 .put_super = btrfs_put_super,
1186 .sync_fs = btrfs_sync_fs,
1187 .show_options = btrfs_show_options,
1188 .write_inode = btrfs_write_inode,
1189 .dirty_inode = btrfs_dirty_inode,
1190 .alloc_inode = btrfs_alloc_inode,
1191 .destroy_inode = btrfs_destroy_inode,
1192 .statfs = btrfs_statfs,
1193 .remount_fs = btrfs_remount,
1194 .freeze_fs = btrfs_freeze,
1195 .unfreeze_fs = btrfs_unfreeze,
1198 static const struct file_operations btrfs_ctl_fops = {
1199 .unlocked_ioctl = btrfs_control_ioctl,
1200 .compat_ioctl = btrfs_control_ioctl,
1201 .owner = THIS_MODULE,
1202 .llseek = noop_llseek,
1205 static struct miscdevice btrfs_misc = {
1206 .minor = BTRFS_MINOR,
1207 .name = "btrfs-control",
1208 .fops = &btrfs_ctl_fops
1211 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1212 MODULE_ALIAS("devname:btrfs-control");
1214 static int btrfs_interface_init(void)
1216 return misc_register(&btrfs_misc);
1219 static void btrfs_interface_exit(void)
1221 if (misc_deregister(&btrfs_misc) < 0)
1222 printk(KERN_INFO "misc_deregister failed for control device");
1225 static int __init init_btrfs_fs(void)
1229 err = btrfs_init_sysfs();
1233 err = btrfs_init_compress();
1237 err = btrfs_init_cachep();
1241 err = extent_io_init();
1245 err = extent_map_init();
1247 goto free_extent_io;
1249 err = btrfs_delayed_inode_init();
1251 goto free_extent_map;
1253 err = btrfs_interface_init();
1255 goto free_delayed_inode;
1257 err = register_filesystem(&btrfs_fs_type);
1259 goto unregister_ioctl;
1261 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1265 btrfs_interface_exit();
1267 btrfs_delayed_inode_exit();
1273 btrfs_destroy_cachep();
1275 btrfs_exit_compress();
1281 static void __exit exit_btrfs_fs(void)
1283 btrfs_destroy_cachep();
1284 btrfs_delayed_inode_exit();
1287 btrfs_interface_exit();
1288 unregister_filesystem(&btrfs_fs_type);
1290 btrfs_cleanup_fs_uuids();
1291 btrfs_exit_compress();
1294 module_init(init_btrfs_fs)
1295 module_exit(exit_btrfs_fs)
1297 MODULE_LICENSE("GPL");