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>
45 #include "transaction.h"
46 #include "btrfs_inode.h"
48 #include "print-tree.h"
53 #include "compression.h"
55 static const struct super_operations btrfs_super_ops;
57 static void btrfs_put_super(struct super_block *sb)
59 struct btrfs_root *root = btrfs_sb(sb);
62 ret = close_ctree(root);
67 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
68 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
69 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
70 Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
71 Opt_discard, Opt_space_cache, Opt_clear_cache, Opt_err,
74 static match_table_t tokens = {
75 {Opt_degraded, "degraded"},
76 {Opt_subvol, "subvol=%s"},
77 {Opt_subvolid, "subvolid=%d"},
78 {Opt_device, "device=%s"},
79 {Opt_nodatasum, "nodatasum"},
80 {Opt_nodatacow, "nodatacow"},
81 {Opt_nobarrier, "nobarrier"},
82 {Opt_max_inline, "max_inline=%s"},
83 {Opt_alloc_start, "alloc_start=%s"},
84 {Opt_thread_pool, "thread_pool=%d"},
85 {Opt_compress, "compress"},
86 {Opt_compress_force, "compress-force"},
88 {Opt_ssd_spread, "ssd_spread"},
91 {Opt_notreelog, "notreelog"},
92 {Opt_flushoncommit, "flushoncommit"},
93 {Opt_ratio, "metadata_ratio=%d"},
94 {Opt_discard, "discard"},
95 {Opt_space_cache, "space_cache"},
96 {Opt_clear_cache, "clear_cache"},
101 * Regular mount options parser. Everything that is needed only when
102 * reading in a new superblock is parsed here.
104 int btrfs_parse_options(struct btrfs_root *root, char *options)
106 struct btrfs_fs_info *info = root->fs_info;
107 substring_t args[MAX_OPT_ARGS];
108 char *p, *num, *orig;
116 * strsep changes the string, duplicate it because parse_options
119 options = kstrdup(options, GFP_NOFS);
125 while ((p = strsep(&options, ",")) != NULL) {
130 token = match_token(p, tokens, args);
133 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
134 btrfs_set_opt(info->mount_opt, DEGRADED);
140 * These are parsed by btrfs_parse_early_options
141 * and can be happily ignored here.
145 printk(KERN_INFO "btrfs: setting nodatasum\n");
146 btrfs_set_opt(info->mount_opt, NODATASUM);
149 printk(KERN_INFO "btrfs: setting nodatacow\n");
150 btrfs_set_opt(info->mount_opt, NODATACOW);
151 btrfs_set_opt(info->mount_opt, NODATASUM);
154 printk(KERN_INFO "btrfs: use compression\n");
155 btrfs_set_opt(info->mount_opt, COMPRESS);
157 case Opt_compress_force:
158 printk(KERN_INFO "btrfs: forcing compression\n");
159 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
160 btrfs_set_opt(info->mount_opt, COMPRESS);
163 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
164 btrfs_set_opt(info->mount_opt, SSD);
167 printk(KERN_INFO "btrfs: use spread ssd "
168 "allocation scheme\n");
169 btrfs_set_opt(info->mount_opt, SSD);
170 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
173 printk(KERN_INFO "btrfs: not using ssd allocation "
175 btrfs_set_opt(info->mount_opt, NOSSD);
176 btrfs_clear_opt(info->mount_opt, SSD);
177 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
180 printk(KERN_INFO "btrfs: turning off barriers\n");
181 btrfs_set_opt(info->mount_opt, NOBARRIER);
183 case Opt_thread_pool:
185 match_int(&args[0], &intarg);
187 info->thread_pool_size = intarg;
188 printk(KERN_INFO "btrfs: thread pool %d\n",
189 info->thread_pool_size);
193 num = match_strdup(&args[0]);
195 info->max_inline = memparse(num, NULL);
198 if (info->max_inline) {
199 info->max_inline = max_t(u64,
203 printk(KERN_INFO "btrfs: max_inline at %llu\n",
204 (unsigned long long)info->max_inline);
207 case Opt_alloc_start:
208 num = match_strdup(&args[0]);
210 info->alloc_start = memparse(num, NULL);
213 "btrfs: allocations start at %llu\n",
214 (unsigned long long)info->alloc_start);
218 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
221 printk(KERN_INFO "btrfs: disabling tree log\n");
222 btrfs_set_opt(info->mount_opt, NOTREELOG);
224 case Opt_flushoncommit:
225 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
226 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
230 match_int(&args[0], &intarg);
232 info->metadata_ratio = intarg;
233 printk(KERN_INFO "btrfs: metadata ratio %d\n",
234 info->metadata_ratio);
238 btrfs_set_opt(info->mount_opt, DISCARD);
240 case Opt_space_cache:
241 printk(KERN_INFO "btrfs: enabling disk space caching\n");
242 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
243 case Opt_clear_cache:
244 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
245 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
248 printk(KERN_INFO "btrfs: unrecognized mount option "
262 * Parse mount options that are required early in the mount process.
264 * All other options will be parsed on much later in the mount process and
265 * only when we need to allocate a new super block.
267 static int btrfs_parse_early_options(const char *options, fmode_t flags,
268 void *holder, char **subvol_name, u64 *subvol_objectid,
269 struct btrfs_fs_devices **fs_devices)
271 substring_t args[MAX_OPT_ARGS];
280 * strsep changes the string, duplicate it because parse_options
283 opts = kstrdup(options, GFP_KERNEL);
287 while ((p = strsep(&opts, ",")) != NULL) {
292 token = match_token(p, tokens, args);
295 *subvol_name = match_strdup(&args[0]);
299 error = match_int(&args[0], &intarg);
301 /* we want the original fs_tree */
304 BTRFS_FS_TREE_OBJECTID;
306 *subvol_objectid = intarg;
310 error = btrfs_scan_one_device(match_strdup(&args[0]),
311 flags, holder, fs_devices);
324 * If no subvolume name is specified we use the default one. Allocate
325 * a copy of the string "." here so that code later in the
326 * mount path doesn't care if it's the default volume or another one.
329 *subvol_name = kstrdup(".", GFP_KERNEL);
336 static struct dentry *get_default_root(struct super_block *sb,
339 struct btrfs_root *root = sb->s_fs_info;
340 struct btrfs_root *new_root;
341 struct btrfs_dir_item *di;
342 struct btrfs_path *path;
343 struct btrfs_key location;
345 struct dentry *dentry;
350 * We have a specific subvol we want to mount, just setup location and
351 * go look up the root.
353 if (subvol_objectid) {
354 location.objectid = subvol_objectid;
355 location.type = BTRFS_ROOT_ITEM_KEY;
356 location.offset = (u64)-1;
360 path = btrfs_alloc_path();
362 return ERR_PTR(-ENOMEM);
363 path->leave_spinning = 1;
366 * Find the "default" dir item which points to the root item that we
367 * will mount by default if we haven't been given a specific subvolume
370 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
371 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
376 * Ok the default dir item isn't there. This is weird since
377 * it's always been there, but don't freak out, just try and
378 * mount to root most subvolume.
380 btrfs_free_path(path);
381 dir_id = BTRFS_FIRST_FREE_OBJECTID;
382 new_root = root->fs_info->fs_root;
386 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
387 btrfs_free_path(path);
390 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
391 if (IS_ERR(new_root))
392 return ERR_CAST(new_root);
394 if (btrfs_root_refs(&new_root->root_item) == 0)
395 return ERR_PTR(-ENOENT);
397 dir_id = btrfs_root_dirid(&new_root->root_item);
399 location.objectid = dir_id;
400 location.type = BTRFS_INODE_ITEM_KEY;
403 inode = btrfs_iget(sb, &location, new_root, &new);
405 return ERR_CAST(inode);
408 * If we're just mounting the root most subvol put the inode and return
409 * a reference to the dentry. We will have already gotten a reference
410 * to the inode in btrfs_fill_super so we're good to go.
412 if (!new && sb->s_root->d_inode == inode) {
414 return dget(sb->s_root);
418 const struct qstr name = { .name = "/", .len = 1 };
421 * New inode, we need to make the dentry a sibling of s_root so
422 * everything gets cleaned up properly on unmount.
424 dentry = d_alloc(sb->s_root, &name);
427 return ERR_PTR(-ENOMEM);
429 d_splice_alias(inode, dentry);
432 * We found the inode in cache, just find a dentry for it and
433 * put the reference to the inode we just got.
435 dentry = d_find_alias(inode);
442 static int btrfs_fill_super(struct super_block *sb,
443 struct btrfs_fs_devices *fs_devices,
444 void *data, int silent)
447 struct dentry *root_dentry;
448 struct btrfs_super_block *disk_super;
449 struct btrfs_root *tree_root;
450 struct btrfs_key key;
453 sb->s_maxbytes = MAX_LFS_FILESIZE;
454 sb->s_magic = BTRFS_SUPER_MAGIC;
455 sb->s_op = &btrfs_super_ops;
456 sb->s_export_op = &btrfs_export_ops;
457 sb->s_xattr = btrfs_xattr_handlers;
459 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
460 sb->s_flags |= MS_POSIXACL;
463 tree_root = open_ctree(sb, fs_devices, (char *)data);
465 if (IS_ERR(tree_root)) {
466 printk("btrfs: open_ctree failed\n");
467 return PTR_ERR(tree_root);
469 sb->s_fs_info = tree_root;
470 disk_super = &tree_root->fs_info->super_copy;
472 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
473 key.type = BTRFS_INODE_ITEM_KEY;
475 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
477 err = PTR_ERR(inode);
481 root_dentry = d_alloc_root(inode);
488 sb->s_root = root_dentry;
490 save_mount_options(sb, data);
494 close_ctree(tree_root);
498 int btrfs_sync_fs(struct super_block *sb, int wait)
500 struct btrfs_trans_handle *trans;
501 struct btrfs_root *root = btrfs_sb(sb);
505 filemap_flush(root->fs_info->btree_inode->i_mapping);
509 btrfs_start_delalloc_inodes(root, 0);
510 btrfs_wait_ordered_extents(root, 0, 0);
512 trans = btrfs_start_transaction(root, 0);
513 ret = btrfs_commit_transaction(trans, root);
517 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
519 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
520 struct btrfs_fs_info *info = root->fs_info;
522 if (btrfs_test_opt(root, DEGRADED))
523 seq_puts(seq, ",degraded");
524 if (btrfs_test_opt(root, NODATASUM))
525 seq_puts(seq, ",nodatasum");
526 if (btrfs_test_opt(root, NODATACOW))
527 seq_puts(seq, ",nodatacow");
528 if (btrfs_test_opt(root, NOBARRIER))
529 seq_puts(seq, ",nobarrier");
530 if (info->max_inline != 8192 * 1024)
531 seq_printf(seq, ",max_inline=%llu",
532 (unsigned long long)info->max_inline);
533 if (info->alloc_start != 0)
534 seq_printf(seq, ",alloc_start=%llu",
535 (unsigned long long)info->alloc_start);
536 if (info->thread_pool_size != min_t(unsigned long,
537 num_online_cpus() + 2, 8))
538 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
539 if (btrfs_test_opt(root, COMPRESS))
540 seq_puts(seq, ",compress");
541 if (btrfs_test_opt(root, NOSSD))
542 seq_puts(seq, ",nossd");
543 if (btrfs_test_opt(root, SSD_SPREAD))
544 seq_puts(seq, ",ssd_spread");
545 else if (btrfs_test_opt(root, SSD))
546 seq_puts(seq, ",ssd");
547 if (btrfs_test_opt(root, NOTREELOG))
548 seq_puts(seq, ",notreelog");
549 if (btrfs_test_opt(root, FLUSHONCOMMIT))
550 seq_puts(seq, ",flushoncommit");
551 if (btrfs_test_opt(root, DISCARD))
552 seq_puts(seq, ",discard");
553 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
554 seq_puts(seq, ",noacl");
558 static int btrfs_test_super(struct super_block *s, void *data)
560 struct btrfs_fs_devices *test_fs_devices = data;
561 struct btrfs_root *root = btrfs_sb(s);
563 return root->fs_info->fs_devices == test_fs_devices;
567 * Find a superblock for the given device / mount point.
569 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
570 * for multiple device setup. Make sure to keep it in sync.
572 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
573 const char *dev_name, void *data, struct vfsmount *mnt)
575 struct block_device *bdev = NULL;
576 struct super_block *s;
578 struct btrfs_fs_devices *fs_devices = NULL;
579 fmode_t mode = FMODE_READ;
580 char *subvol_name = NULL;
581 u64 subvol_objectid = 0;
585 if (!(flags & MS_RDONLY))
588 error = btrfs_parse_early_options(data, mode, fs_type,
589 &subvol_name, &subvol_objectid,
594 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
596 goto error_free_subvol_name;
598 error = btrfs_open_devices(fs_devices, mode, fs_type);
600 goto error_free_subvol_name;
602 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
604 goto error_close_devices;
607 bdev = fs_devices->latest_bdev;
608 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
613 if ((flags ^ s->s_flags) & MS_RDONLY) {
614 deactivate_locked_super(s);
616 goto error_close_devices;
620 btrfs_close_devices(fs_devices);
622 char b[BDEVNAME_SIZE];
625 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
626 error = btrfs_fill_super(s, fs_devices, data,
627 flags & MS_SILENT ? 1 : 0);
629 deactivate_locked_super(s);
630 goto error_free_subvol_name;
633 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
634 s->s_flags |= MS_ACTIVE;
637 root = get_default_root(s, subvol_objectid);
639 error = PTR_ERR(root);
640 deactivate_locked_super(s);
641 goto error_free_subvol_name;
643 /* if they gave us a subvolume name bind mount into that */
644 if (strcmp(subvol_name, ".")) {
645 struct dentry *new_root;
646 mutex_lock(&root->d_inode->i_mutex);
647 new_root = lookup_one_len(subvol_name, root,
648 strlen(subvol_name));
649 mutex_unlock(&root->d_inode->i_mutex);
651 if (IS_ERR(new_root)) {
652 deactivate_locked_super(s);
653 error = PTR_ERR(new_root);
655 goto error_free_subvol_name;
657 if (!new_root->d_inode) {
660 deactivate_locked_super(s);
662 goto error_free_subvol_name;
669 mnt->mnt_root = root;
677 btrfs_close_devices(fs_devices);
678 error_free_subvol_name:
683 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
685 struct btrfs_root *root = btrfs_sb(sb);
688 ret = btrfs_parse_options(root, data);
692 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
695 if (*flags & MS_RDONLY) {
696 sb->s_flags |= MS_RDONLY;
698 ret = btrfs_commit_super(root);
701 if (root->fs_info->fs_devices->rw_devices == 0)
704 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
707 ret = btrfs_cleanup_fs_roots(root->fs_info);
710 /* recover relocation */
711 ret = btrfs_recover_relocation(root);
714 sb->s_flags &= ~MS_RDONLY;
720 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
722 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
723 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
724 struct list_head *head = &root->fs_info->space_info;
725 struct btrfs_space_info *found;
727 u64 total_used_data = 0;
728 int bits = dentry->d_sb->s_blocksize_bits;
729 __be32 *fsid = (__be32 *)root->fs_info->fsid;
732 list_for_each_entry_rcu(found, head, list) {
733 if (found->flags & (BTRFS_BLOCK_GROUP_METADATA |
734 BTRFS_BLOCK_GROUP_SYSTEM))
735 total_used_data += found->disk_total;
737 total_used_data += found->disk_used;
738 total_used += found->disk_used;
742 buf->f_namelen = BTRFS_NAME_LEN;
743 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
744 buf->f_bfree = buf->f_blocks - (total_used >> bits);
745 buf->f_bavail = buf->f_blocks - (total_used_data >> bits);
746 buf->f_bsize = dentry->d_sb->s_blocksize;
747 buf->f_type = BTRFS_SUPER_MAGIC;
749 /* We treat it as constant endianness (it doesn't matter _which_)
750 because we want the fsid to come out the same whether mounted
751 on a big-endian or little-endian host */
752 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
753 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
754 /* Mask in the root object ID too, to disambiguate subvols */
755 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
756 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
761 static struct file_system_type btrfs_fs_type = {
762 .owner = THIS_MODULE,
764 .get_sb = btrfs_get_sb,
765 .kill_sb = kill_anon_super,
766 .fs_flags = FS_REQUIRES_DEV,
770 * used by btrfsctl to scan devices when no FS is mounted
772 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
775 struct btrfs_ioctl_vol_args *vol;
776 struct btrfs_fs_devices *fs_devices;
779 if (!capable(CAP_SYS_ADMIN))
782 vol = memdup_user((void __user *)arg, sizeof(*vol));
787 case BTRFS_IOC_SCAN_DEV:
788 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
789 &btrfs_fs_type, &fs_devices);
797 static int btrfs_freeze(struct super_block *sb)
799 struct btrfs_root *root = btrfs_sb(sb);
800 mutex_lock(&root->fs_info->transaction_kthread_mutex);
801 mutex_lock(&root->fs_info->cleaner_mutex);
805 static int btrfs_unfreeze(struct super_block *sb)
807 struct btrfs_root *root = btrfs_sb(sb);
808 mutex_unlock(&root->fs_info->cleaner_mutex);
809 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
813 static const struct super_operations btrfs_super_ops = {
814 .drop_inode = btrfs_drop_inode,
815 .evict_inode = btrfs_evict_inode,
816 .put_super = btrfs_put_super,
817 .sync_fs = btrfs_sync_fs,
818 .show_options = btrfs_show_options,
819 .write_inode = btrfs_write_inode,
820 .dirty_inode = btrfs_dirty_inode,
821 .alloc_inode = btrfs_alloc_inode,
822 .destroy_inode = btrfs_destroy_inode,
823 .statfs = btrfs_statfs,
824 .remount_fs = btrfs_remount,
825 .freeze_fs = btrfs_freeze,
826 .unfreeze_fs = btrfs_unfreeze,
829 static const struct file_operations btrfs_ctl_fops = {
830 .unlocked_ioctl = btrfs_control_ioctl,
831 .compat_ioctl = btrfs_control_ioctl,
832 .owner = THIS_MODULE,
835 static struct miscdevice btrfs_misc = {
836 .minor = BTRFS_MINOR,
837 .name = "btrfs-control",
838 .fops = &btrfs_ctl_fops
841 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
842 MODULE_ALIAS("devname:btrfs-control");
844 static int btrfs_interface_init(void)
846 return misc_register(&btrfs_misc);
849 static void btrfs_interface_exit(void)
851 if (misc_deregister(&btrfs_misc) < 0)
852 printk(KERN_INFO "misc_deregister failed for control device");
855 static int __init init_btrfs_fs(void)
859 err = btrfs_init_sysfs();
863 err = btrfs_init_cachep();
867 err = extent_io_init();
871 err = extent_map_init();
875 err = btrfs_interface_init();
877 goto free_extent_map;
879 err = register_filesystem(&btrfs_fs_type);
881 goto unregister_ioctl;
883 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
887 btrfs_interface_exit();
893 btrfs_destroy_cachep();
899 static void __exit exit_btrfs_fs(void)
901 btrfs_destroy_cachep();
904 btrfs_interface_exit();
905 unregister_filesystem(&btrfs_fs_type);
907 btrfs_cleanup_fs_uuids();
911 module_init(init_btrfs_fs)
912 module_exit(exit_btrfs_fs)
914 MODULE_LICENSE("GPL");