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_extent, Opt_max_inline, Opt_alloc_start,
69 Opt_nobarrier, Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool,
70 Opt_noacl, Opt_compress, Opt_compress_force, Opt_notreelog, Opt_ratio,
75 static match_table_t tokens = {
76 {Opt_degraded, "degraded"},
77 {Opt_subvol, "subvol=%s"},
78 {Opt_subvolid, "subvolid=%d"},
79 {Opt_device, "device=%s"},
80 {Opt_nodatasum, "nodatasum"},
81 {Opt_nodatacow, "nodatacow"},
82 {Opt_nobarrier, "nobarrier"},
83 {Opt_max_extent, "max_extent=%s"},
84 {Opt_max_inline, "max_inline=%s"},
85 {Opt_alloc_start, "alloc_start=%s"},
86 {Opt_thread_pool, "thread_pool=%d"},
87 {Opt_compress, "compress"},
88 {Opt_compress_force, "compress-force"},
90 {Opt_ssd_spread, "ssd_spread"},
93 {Opt_notreelog, "notreelog"},
94 {Opt_flushoncommit, "flushoncommit"},
95 {Opt_ratio, "metadata_ratio=%d"},
96 {Opt_discard, "discard"},
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_extent = memparse(num, NULL);
198 info->max_extent = max_t(u64,
199 info->max_extent, root->sectorsize);
200 printk(KERN_INFO "btrfs: max_extent at %llu\n",
201 (unsigned long long)info->max_extent);
205 num = match_strdup(&args[0]);
207 info->max_inline = memparse(num, NULL);
210 if (info->max_inline) {
211 info->max_inline = max_t(u64,
215 printk(KERN_INFO "btrfs: max_inline at %llu\n",
216 (unsigned long long)info->max_inline);
219 case Opt_alloc_start:
220 num = match_strdup(&args[0]);
222 info->alloc_start = memparse(num, NULL);
225 "btrfs: allocations start at %llu\n",
226 (unsigned long long)info->alloc_start);
230 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
233 printk(KERN_INFO "btrfs: disabling tree log\n");
234 btrfs_set_opt(info->mount_opt, NOTREELOG);
236 case Opt_flushoncommit:
237 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
238 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
242 match_int(&args[0], &intarg);
244 info->metadata_ratio = intarg;
245 printk(KERN_INFO "btrfs: metadata ratio %d\n",
246 info->metadata_ratio);
250 btrfs_set_opt(info->mount_opt, DISCARD);
253 printk(KERN_INFO "btrfs: unrecognized mount option "
267 * Parse mount options that are required early in the mount process.
269 * All other options will be parsed on much later in the mount process and
270 * only when we need to allocate a new super block.
272 static int btrfs_parse_early_options(const char *options, fmode_t flags,
273 void *holder, char **subvol_name, u64 *subvol_objectid,
274 struct btrfs_fs_devices **fs_devices)
276 substring_t args[MAX_OPT_ARGS];
285 * strsep changes the string, duplicate it because parse_options
288 opts = kstrdup(options, GFP_KERNEL);
292 while ((p = strsep(&opts, ",")) != NULL) {
297 token = match_token(p, tokens, args);
300 *subvol_name = match_strdup(&args[0]);
304 error = match_int(&args[0], &intarg);
306 /* we want the original fs_tree */
309 BTRFS_FS_TREE_OBJECTID;
311 *subvol_objectid = intarg;
315 error = btrfs_scan_one_device(match_strdup(&args[0]),
316 flags, holder, fs_devices);
329 * If no subvolume name is specified we use the default one. Allocate
330 * a copy of the string "." here so that code later in the
331 * mount path doesn't care if it's the default volume or another one.
334 *subvol_name = kstrdup(".", GFP_KERNEL);
341 static struct dentry *get_default_root(struct super_block *sb,
344 struct btrfs_root *root = sb->s_fs_info;
345 struct btrfs_root *new_root;
346 struct btrfs_dir_item *di;
347 struct btrfs_path *path;
348 struct btrfs_key location;
350 struct dentry *dentry;
355 * We have a specific subvol we want to mount, just setup location and
356 * go look up the root.
358 if (subvol_objectid) {
359 location.objectid = subvol_objectid;
360 location.type = BTRFS_ROOT_ITEM_KEY;
361 location.offset = (u64)-1;
365 path = btrfs_alloc_path();
367 return ERR_PTR(-ENOMEM);
368 path->leave_spinning = 1;
371 * Find the "default" dir item which points to the root item that we
372 * will mount by default if we haven't been given a specific subvolume
375 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
376 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
379 * Ok the default dir item isn't there. This is weird since
380 * it's always been there, but don't freak out, just try and
381 * mount to root most subvolume.
383 btrfs_free_path(path);
384 dir_id = BTRFS_FIRST_FREE_OBJECTID;
385 new_root = root->fs_info->fs_root;
389 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
390 btrfs_free_path(path);
393 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
394 if (IS_ERR(new_root))
395 return ERR_PTR(PTR_ERR(new_root));
397 if (btrfs_root_refs(&new_root->root_item) == 0)
398 return ERR_PTR(-ENOENT);
400 dir_id = btrfs_root_dirid(&new_root->root_item);
402 location.objectid = dir_id;
403 location.type = BTRFS_INODE_ITEM_KEY;
406 inode = btrfs_iget(sb, &location, new_root, &new);
408 return ERR_PTR(-ENOMEM);
411 * If we're just mounting the root most subvol put the inode and return
412 * a reference to the dentry. We will have already gotten a reference
413 * to the inode in btrfs_fill_super so we're good to go.
415 if (!new && sb->s_root->d_inode == inode) {
417 return dget(sb->s_root);
421 const struct qstr name = { .name = "/", .len = 1 };
424 * New inode, we need to make the dentry a sibling of s_root so
425 * everything gets cleaned up properly on unmount.
427 dentry = d_alloc(sb->s_root, &name);
430 return ERR_PTR(-ENOMEM);
432 d_splice_alias(inode, dentry);
435 * We found the inode in cache, just find a dentry for it and
436 * put the reference to the inode we just got.
438 dentry = d_find_alias(inode);
445 static int btrfs_fill_super(struct super_block *sb,
446 struct btrfs_fs_devices *fs_devices,
447 void *data, int silent)
450 struct dentry *root_dentry;
451 struct btrfs_super_block *disk_super;
452 struct btrfs_root *tree_root;
453 struct btrfs_key key;
456 sb->s_maxbytes = MAX_LFS_FILESIZE;
457 sb->s_magic = BTRFS_SUPER_MAGIC;
458 sb->s_op = &btrfs_super_ops;
459 sb->s_export_op = &btrfs_export_ops;
460 sb->s_xattr = btrfs_xattr_handlers;
462 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
463 sb->s_flags |= MS_POSIXACL;
466 tree_root = open_ctree(sb, fs_devices, (char *)data);
468 if (IS_ERR(tree_root)) {
469 printk("btrfs: open_ctree failed\n");
470 return PTR_ERR(tree_root);
472 sb->s_fs_info = tree_root;
473 disk_super = &tree_root->fs_info->super_copy;
475 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
476 key.type = BTRFS_INODE_ITEM_KEY;
478 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
480 err = PTR_ERR(inode);
484 root_dentry = d_alloc_root(inode);
491 sb->s_root = root_dentry;
493 save_mount_options(sb, data);
497 close_ctree(tree_root);
501 int btrfs_sync_fs(struct super_block *sb, int wait)
503 struct btrfs_trans_handle *trans;
504 struct btrfs_root *root = btrfs_sb(sb);
508 filemap_flush(root->fs_info->btree_inode->i_mapping);
512 btrfs_start_delalloc_inodes(root, 0);
513 btrfs_wait_ordered_extents(root, 0, 0);
515 trans = btrfs_start_transaction(root, 1);
516 ret = btrfs_commit_transaction(trans, root);
520 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
522 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
523 struct btrfs_fs_info *info = root->fs_info;
525 if (btrfs_test_opt(root, DEGRADED))
526 seq_puts(seq, ",degraded");
527 if (btrfs_test_opt(root, NODATASUM))
528 seq_puts(seq, ",nodatasum");
529 if (btrfs_test_opt(root, NODATACOW))
530 seq_puts(seq, ",nodatacow");
531 if (btrfs_test_opt(root, NOBARRIER))
532 seq_puts(seq, ",nobarrier");
533 if (info->max_extent != (u64)-1)
534 seq_printf(seq, ",max_extent=%llu",
535 (unsigned long long)info->max_extent);
536 if (info->max_inline != 8192 * 1024)
537 seq_printf(seq, ",max_inline=%llu",
538 (unsigned long long)info->max_inline);
539 if (info->alloc_start != 0)
540 seq_printf(seq, ",alloc_start=%llu",
541 (unsigned long long)info->alloc_start);
542 if (info->thread_pool_size != min_t(unsigned long,
543 num_online_cpus() + 2, 8))
544 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
545 if (btrfs_test_opt(root, COMPRESS))
546 seq_puts(seq, ",compress");
547 if (btrfs_test_opt(root, NOSSD))
548 seq_puts(seq, ",nossd");
549 if (btrfs_test_opt(root, SSD_SPREAD))
550 seq_puts(seq, ",ssd_spread");
551 else if (btrfs_test_opt(root, SSD))
552 seq_puts(seq, ",ssd");
553 if (btrfs_test_opt(root, NOTREELOG))
554 seq_puts(seq, ",notreelog");
555 if (btrfs_test_opt(root, FLUSHONCOMMIT))
556 seq_puts(seq, ",flushoncommit");
557 if (btrfs_test_opt(root, DISCARD))
558 seq_puts(seq, ",discard");
559 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
560 seq_puts(seq, ",noacl");
564 static int btrfs_test_super(struct super_block *s, void *data)
566 struct btrfs_fs_devices *test_fs_devices = data;
567 struct btrfs_root *root = btrfs_sb(s);
569 return root->fs_info->fs_devices == test_fs_devices;
573 * Find a superblock for the given device / mount point.
575 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
576 * for multiple device setup. Make sure to keep it in sync.
578 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
579 const char *dev_name, void *data, struct vfsmount *mnt)
581 struct block_device *bdev = NULL;
582 struct super_block *s;
584 struct btrfs_fs_devices *fs_devices = NULL;
585 fmode_t mode = FMODE_READ;
586 char *subvol_name = NULL;
587 u64 subvol_objectid = 0;
591 if (!(flags & MS_RDONLY))
594 error = btrfs_parse_early_options(data, mode, fs_type,
595 &subvol_name, &subvol_objectid,
600 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
602 goto error_free_subvol_name;
604 error = btrfs_open_devices(fs_devices, mode, fs_type);
606 goto error_free_subvol_name;
608 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
610 goto error_close_devices;
613 bdev = fs_devices->latest_bdev;
614 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
619 if ((flags ^ s->s_flags) & MS_RDONLY) {
620 deactivate_locked_super(s);
622 goto error_close_devices;
626 btrfs_close_devices(fs_devices);
628 char b[BDEVNAME_SIZE];
631 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
632 error = btrfs_fill_super(s, fs_devices, data,
633 flags & MS_SILENT ? 1 : 0);
635 deactivate_locked_super(s);
636 goto error_free_subvol_name;
639 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
640 s->s_flags |= MS_ACTIVE;
643 root = get_default_root(s, subvol_objectid);
645 error = PTR_ERR(root);
646 deactivate_locked_super(s);
649 /* if they gave us a subvolume name bind mount into that */
650 if (strcmp(subvol_name, ".")) {
651 struct dentry *new_root;
652 mutex_lock(&root->d_inode->i_mutex);
653 new_root = lookup_one_len(subvol_name, root,
654 strlen(subvol_name));
655 mutex_unlock(&root->d_inode->i_mutex);
657 if (IS_ERR(new_root)) {
658 deactivate_locked_super(s);
659 error = PTR_ERR(new_root);
661 goto error_close_devices;
663 if (!new_root->d_inode) {
666 deactivate_locked_super(s);
668 goto error_close_devices;
675 mnt->mnt_root = root;
683 btrfs_close_devices(fs_devices);
684 error_free_subvol_name:
690 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
692 struct btrfs_root *root = btrfs_sb(sb);
695 ret = btrfs_parse_options(root, data);
699 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
702 if (*flags & MS_RDONLY) {
703 sb->s_flags |= MS_RDONLY;
705 ret = btrfs_commit_super(root);
708 if (root->fs_info->fs_devices->rw_devices == 0)
711 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
714 /* recover relocation */
715 ret = btrfs_recover_relocation(root);
718 ret = btrfs_cleanup_fs_roots(root->fs_info);
721 sb->s_flags &= ~MS_RDONLY;
727 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
729 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
730 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
731 struct list_head *head = &root->fs_info->space_info;
732 struct btrfs_space_info *found;
735 int bits = dentry->d_sb->s_blocksize_bits;
736 __be32 *fsid = (__be32 *)root->fs_info->fsid;
739 list_for_each_entry_rcu(found, head, list) {
740 if (found->flags & (BTRFS_BLOCK_GROUP_DUP|
741 BTRFS_BLOCK_GROUP_RAID10|
742 BTRFS_BLOCK_GROUP_RAID1)) {
743 total_used += found->bytes_used;
744 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
745 data_used += found->bytes_used;
747 data_used += found->total_bytes;
750 total_used += found->bytes_used;
751 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
752 data_used += found->bytes_used;
754 data_used += found->total_bytes;
758 buf->f_namelen = BTRFS_NAME_LEN;
759 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
760 buf->f_bfree = buf->f_blocks - (total_used >> bits);
761 buf->f_bavail = buf->f_blocks - (data_used >> bits);
762 buf->f_bsize = dentry->d_sb->s_blocksize;
763 buf->f_type = BTRFS_SUPER_MAGIC;
765 /* We treat it as constant endianness (it doesn't matter _which_)
766 because we want the fsid to come out the same whether mounted
767 on a big-endian or little-endian host */
768 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
769 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
770 /* Mask in the root object ID too, to disambiguate subvols */
771 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
772 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
777 static struct file_system_type btrfs_fs_type = {
778 .owner = THIS_MODULE,
780 .get_sb = btrfs_get_sb,
781 .kill_sb = kill_anon_super,
782 .fs_flags = FS_REQUIRES_DEV,
786 * used by btrfsctl to scan devices when no FS is mounted
788 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
791 struct btrfs_ioctl_vol_args *vol;
792 struct btrfs_fs_devices *fs_devices;
795 if (!capable(CAP_SYS_ADMIN))
798 vol = memdup_user((void __user *)arg, sizeof(*vol));
803 case BTRFS_IOC_SCAN_DEV:
804 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
805 &btrfs_fs_type, &fs_devices);
813 static int btrfs_freeze(struct super_block *sb)
815 struct btrfs_root *root = btrfs_sb(sb);
816 mutex_lock(&root->fs_info->transaction_kthread_mutex);
817 mutex_lock(&root->fs_info->cleaner_mutex);
821 static int btrfs_unfreeze(struct super_block *sb)
823 struct btrfs_root *root = btrfs_sb(sb);
824 mutex_unlock(&root->fs_info->cleaner_mutex);
825 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
829 static const struct super_operations btrfs_super_ops = {
830 .drop_inode = btrfs_drop_inode,
831 .delete_inode = btrfs_delete_inode,
832 .put_super = btrfs_put_super,
833 .sync_fs = btrfs_sync_fs,
834 .show_options = btrfs_show_options,
835 .write_inode = btrfs_write_inode,
836 .dirty_inode = btrfs_dirty_inode,
837 .alloc_inode = btrfs_alloc_inode,
838 .destroy_inode = btrfs_destroy_inode,
839 .statfs = btrfs_statfs,
840 .remount_fs = btrfs_remount,
841 .freeze_fs = btrfs_freeze,
842 .unfreeze_fs = btrfs_unfreeze,
845 static const struct file_operations btrfs_ctl_fops = {
846 .unlocked_ioctl = btrfs_control_ioctl,
847 .compat_ioctl = btrfs_control_ioctl,
848 .owner = THIS_MODULE,
851 static struct miscdevice btrfs_misc = {
852 .minor = MISC_DYNAMIC_MINOR,
853 .name = "btrfs-control",
854 .fops = &btrfs_ctl_fops
857 static int btrfs_interface_init(void)
859 return misc_register(&btrfs_misc);
862 static void btrfs_interface_exit(void)
864 if (misc_deregister(&btrfs_misc) < 0)
865 printk(KERN_INFO "misc_deregister failed for control device");
868 static int __init init_btrfs_fs(void)
872 err = btrfs_init_sysfs();
876 err = btrfs_init_cachep();
880 err = extent_io_init();
884 err = extent_map_init();
888 err = btrfs_interface_init();
890 goto free_extent_map;
892 err = register_filesystem(&btrfs_fs_type);
894 goto unregister_ioctl;
896 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
900 btrfs_interface_exit();
906 btrfs_destroy_cachep();
912 static void __exit exit_btrfs_fs(void)
914 btrfs_destroy_cachep();
917 btrfs_interface_exit();
918 unregister_filesystem(&btrfs_fs_type);
920 btrfs_cleanup_fs_uuids();
924 module_init(init_btrfs_fs)
925 module_exit(exit_btrfs_fs)
927 MODULE_LICENSE("GPL");