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_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"},
100 * Regular mount options parser. Everything that is needed only when
101 * reading in a new superblock is parsed here.
103 int btrfs_parse_options(struct btrfs_root *root, char *options)
105 struct btrfs_fs_info *info = root->fs_info;
106 substring_t args[MAX_OPT_ARGS];
107 char *p, *num, *orig;
115 * strsep changes the string, duplicate it because parse_options
118 options = kstrdup(options, GFP_NOFS);
124 while ((p = strsep(&options, ",")) != NULL) {
129 token = match_token(p, tokens, args);
132 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
133 btrfs_set_opt(info->mount_opt, DEGRADED);
139 * These are parsed by btrfs_parse_early_options
140 * and can be happily ignored here.
144 printk(KERN_INFO "btrfs: setting nodatasum\n");
145 btrfs_set_opt(info->mount_opt, NODATASUM);
148 printk(KERN_INFO "btrfs: setting nodatacow\n");
149 btrfs_set_opt(info->mount_opt, NODATACOW);
150 btrfs_set_opt(info->mount_opt, NODATASUM);
153 printk(KERN_INFO "btrfs: use compression\n");
154 btrfs_set_opt(info->mount_opt, COMPRESS);
156 case Opt_compress_force:
157 printk(KERN_INFO "btrfs: forcing compression\n");
158 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
159 btrfs_set_opt(info->mount_opt, COMPRESS);
162 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
163 btrfs_set_opt(info->mount_opt, SSD);
166 printk(KERN_INFO "btrfs: use spread ssd "
167 "allocation scheme\n");
168 btrfs_set_opt(info->mount_opt, SSD);
169 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
172 printk(KERN_INFO "btrfs: not using ssd allocation "
174 btrfs_set_opt(info->mount_opt, NOSSD);
175 btrfs_clear_opt(info->mount_opt, SSD);
176 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
179 printk(KERN_INFO "btrfs: turning off barriers\n");
180 btrfs_set_opt(info->mount_opt, NOBARRIER);
182 case Opt_thread_pool:
184 match_int(&args[0], &intarg);
186 info->thread_pool_size = intarg;
187 printk(KERN_INFO "btrfs: thread pool %d\n",
188 info->thread_pool_size);
192 num = match_strdup(&args[0]);
194 info->max_inline = memparse(num, NULL);
197 if (info->max_inline) {
198 info->max_inline = max_t(u64,
202 printk(KERN_INFO "btrfs: max_inline at %llu\n",
203 (unsigned long long)info->max_inline);
206 case Opt_alloc_start:
207 num = match_strdup(&args[0]);
209 info->alloc_start = memparse(num, NULL);
212 "btrfs: allocations start at %llu\n",
213 (unsigned long long)info->alloc_start);
217 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
220 printk(KERN_INFO "btrfs: disabling tree log\n");
221 btrfs_set_opt(info->mount_opt, NOTREELOG);
223 case Opt_flushoncommit:
224 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
225 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
229 match_int(&args[0], &intarg);
231 info->metadata_ratio = intarg;
232 printk(KERN_INFO "btrfs: metadata ratio %d\n",
233 info->metadata_ratio);
237 btrfs_set_opt(info->mount_opt, DISCARD);
239 case Opt_space_cache:
240 printk(KERN_INFO "btrfs: enabling disk space caching\n");
241 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
244 printk(KERN_INFO "btrfs: unrecognized mount option "
258 * Parse mount options that are required early in the mount process.
260 * All other options will be parsed on much later in the mount process and
261 * only when we need to allocate a new super block.
263 static int btrfs_parse_early_options(const char *options, fmode_t flags,
264 void *holder, char **subvol_name, u64 *subvol_objectid,
265 struct btrfs_fs_devices **fs_devices)
267 substring_t args[MAX_OPT_ARGS];
276 * strsep changes the string, duplicate it because parse_options
279 opts = kstrdup(options, GFP_KERNEL);
283 while ((p = strsep(&opts, ",")) != NULL) {
288 token = match_token(p, tokens, args);
291 *subvol_name = match_strdup(&args[0]);
295 error = match_int(&args[0], &intarg);
297 /* we want the original fs_tree */
300 BTRFS_FS_TREE_OBJECTID;
302 *subvol_objectid = intarg;
306 error = btrfs_scan_one_device(match_strdup(&args[0]),
307 flags, holder, fs_devices);
320 * If no subvolume name is specified we use the default one. Allocate
321 * a copy of the string "." here so that code later in the
322 * mount path doesn't care if it's the default volume or another one.
325 *subvol_name = kstrdup(".", GFP_KERNEL);
332 static struct dentry *get_default_root(struct super_block *sb,
335 struct btrfs_root *root = sb->s_fs_info;
336 struct btrfs_root *new_root;
337 struct btrfs_dir_item *di;
338 struct btrfs_path *path;
339 struct btrfs_key location;
341 struct dentry *dentry;
346 * We have a specific subvol we want to mount, just setup location and
347 * go look up the root.
349 if (subvol_objectid) {
350 location.objectid = subvol_objectid;
351 location.type = BTRFS_ROOT_ITEM_KEY;
352 location.offset = (u64)-1;
356 path = btrfs_alloc_path();
358 return ERR_PTR(-ENOMEM);
359 path->leave_spinning = 1;
362 * Find the "default" dir item which points to the root item that we
363 * will mount by default if we haven't been given a specific subvolume
366 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
367 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
372 * Ok the default dir item isn't there. This is weird since
373 * it's always been there, but don't freak out, just try and
374 * mount to root most subvolume.
376 btrfs_free_path(path);
377 dir_id = BTRFS_FIRST_FREE_OBJECTID;
378 new_root = root->fs_info->fs_root;
382 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
383 btrfs_free_path(path);
386 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
387 if (IS_ERR(new_root))
388 return ERR_PTR(PTR_ERR(new_root));
390 if (btrfs_root_refs(&new_root->root_item) == 0)
391 return ERR_PTR(-ENOENT);
393 dir_id = btrfs_root_dirid(&new_root->root_item);
395 location.objectid = dir_id;
396 location.type = BTRFS_INODE_ITEM_KEY;
399 inode = btrfs_iget(sb, &location, new_root, &new);
401 return ERR_CAST(inode);
404 * If we're just mounting the root most subvol put the inode and return
405 * a reference to the dentry. We will have already gotten a reference
406 * to the inode in btrfs_fill_super so we're good to go.
408 if (!new && sb->s_root->d_inode == inode) {
410 return dget(sb->s_root);
414 const struct qstr name = { .name = "/", .len = 1 };
417 * New inode, we need to make the dentry a sibling of s_root so
418 * everything gets cleaned up properly on unmount.
420 dentry = d_alloc(sb->s_root, &name);
423 return ERR_PTR(-ENOMEM);
425 d_splice_alias(inode, dentry);
428 * We found the inode in cache, just find a dentry for it and
429 * put the reference to the inode we just got.
431 dentry = d_find_alias(inode);
438 static int btrfs_fill_super(struct super_block *sb,
439 struct btrfs_fs_devices *fs_devices,
440 void *data, int silent)
443 struct dentry *root_dentry;
444 struct btrfs_super_block *disk_super;
445 struct btrfs_root *tree_root;
446 struct btrfs_key key;
449 sb->s_maxbytes = MAX_LFS_FILESIZE;
450 sb->s_magic = BTRFS_SUPER_MAGIC;
451 sb->s_op = &btrfs_super_ops;
452 sb->s_export_op = &btrfs_export_ops;
453 sb->s_xattr = btrfs_xattr_handlers;
455 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
456 sb->s_flags |= MS_POSIXACL;
459 tree_root = open_ctree(sb, fs_devices, (char *)data);
461 if (IS_ERR(tree_root)) {
462 printk("btrfs: open_ctree failed\n");
463 return PTR_ERR(tree_root);
465 sb->s_fs_info = tree_root;
466 disk_super = &tree_root->fs_info->super_copy;
468 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
469 key.type = BTRFS_INODE_ITEM_KEY;
471 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
473 err = PTR_ERR(inode);
477 root_dentry = d_alloc_root(inode);
484 sb->s_root = root_dentry;
486 save_mount_options(sb, data);
490 close_ctree(tree_root);
494 int btrfs_sync_fs(struct super_block *sb, int wait)
496 struct btrfs_trans_handle *trans;
497 struct btrfs_root *root = btrfs_sb(sb);
501 filemap_flush(root->fs_info->btree_inode->i_mapping);
505 btrfs_start_delalloc_inodes(root, 0);
506 btrfs_wait_ordered_extents(root, 0, 0);
508 trans = btrfs_start_transaction(root, 0);
509 ret = btrfs_commit_transaction(trans, root);
513 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
515 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
516 struct btrfs_fs_info *info = root->fs_info;
518 if (btrfs_test_opt(root, DEGRADED))
519 seq_puts(seq, ",degraded");
520 if (btrfs_test_opt(root, NODATASUM))
521 seq_puts(seq, ",nodatasum");
522 if (btrfs_test_opt(root, NODATACOW))
523 seq_puts(seq, ",nodatacow");
524 if (btrfs_test_opt(root, NOBARRIER))
525 seq_puts(seq, ",nobarrier");
526 if (info->max_inline != 8192 * 1024)
527 seq_printf(seq, ",max_inline=%llu",
528 (unsigned long long)info->max_inline);
529 if (info->alloc_start != 0)
530 seq_printf(seq, ",alloc_start=%llu",
531 (unsigned long long)info->alloc_start);
532 if (info->thread_pool_size != min_t(unsigned long,
533 num_online_cpus() + 2, 8))
534 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
535 if (btrfs_test_opt(root, COMPRESS))
536 seq_puts(seq, ",compress");
537 if (btrfs_test_opt(root, NOSSD))
538 seq_puts(seq, ",nossd");
539 if (btrfs_test_opt(root, SSD_SPREAD))
540 seq_puts(seq, ",ssd_spread");
541 else if (btrfs_test_opt(root, SSD))
542 seq_puts(seq, ",ssd");
543 if (btrfs_test_opt(root, NOTREELOG))
544 seq_puts(seq, ",notreelog");
545 if (btrfs_test_opt(root, FLUSHONCOMMIT))
546 seq_puts(seq, ",flushoncommit");
547 if (btrfs_test_opt(root, DISCARD))
548 seq_puts(seq, ",discard");
549 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
550 seq_puts(seq, ",noacl");
554 static int btrfs_test_super(struct super_block *s, void *data)
556 struct btrfs_fs_devices *test_fs_devices = data;
557 struct btrfs_root *root = btrfs_sb(s);
559 return root->fs_info->fs_devices == test_fs_devices;
563 * Find a superblock for the given device / mount point.
565 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
566 * for multiple device setup. Make sure to keep it in sync.
568 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
569 const char *dev_name, void *data, struct vfsmount *mnt)
571 struct block_device *bdev = NULL;
572 struct super_block *s;
574 struct btrfs_fs_devices *fs_devices = NULL;
575 fmode_t mode = FMODE_READ;
576 char *subvol_name = NULL;
577 u64 subvol_objectid = 0;
581 if (!(flags & MS_RDONLY))
584 error = btrfs_parse_early_options(data, mode, fs_type,
585 &subvol_name, &subvol_objectid,
590 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
592 goto error_free_subvol_name;
594 error = btrfs_open_devices(fs_devices, mode, fs_type);
596 goto error_free_subvol_name;
598 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
600 goto error_close_devices;
603 bdev = fs_devices->latest_bdev;
604 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
609 if ((flags ^ s->s_flags) & MS_RDONLY) {
610 deactivate_locked_super(s);
612 goto error_close_devices;
616 btrfs_close_devices(fs_devices);
618 char b[BDEVNAME_SIZE];
621 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
622 error = btrfs_fill_super(s, fs_devices, data,
623 flags & MS_SILENT ? 1 : 0);
625 deactivate_locked_super(s);
626 goto error_free_subvol_name;
629 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
630 s->s_flags |= MS_ACTIVE;
633 root = get_default_root(s, subvol_objectid);
635 error = PTR_ERR(root);
636 deactivate_locked_super(s);
639 /* if they gave us a subvolume name bind mount into that */
640 if (strcmp(subvol_name, ".")) {
641 struct dentry *new_root;
642 mutex_lock(&root->d_inode->i_mutex);
643 new_root = lookup_one_len(subvol_name, root,
644 strlen(subvol_name));
645 mutex_unlock(&root->d_inode->i_mutex);
647 if (IS_ERR(new_root)) {
648 deactivate_locked_super(s);
649 error = PTR_ERR(new_root);
651 goto error_close_devices;
653 if (!new_root->d_inode) {
656 deactivate_locked_super(s);
658 goto error_close_devices;
665 mnt->mnt_root = root;
673 btrfs_close_devices(fs_devices);
674 error_free_subvol_name:
680 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
682 struct btrfs_root *root = btrfs_sb(sb);
685 ret = btrfs_parse_options(root, data);
689 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
692 if (*flags & MS_RDONLY) {
693 sb->s_flags |= MS_RDONLY;
695 ret = btrfs_commit_super(root);
698 if (root->fs_info->fs_devices->rw_devices == 0)
701 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
704 ret = btrfs_cleanup_fs_roots(root->fs_info);
707 /* recover relocation */
708 ret = btrfs_recover_relocation(root);
711 sb->s_flags &= ~MS_RDONLY;
717 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
719 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
720 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
721 struct list_head *head = &root->fs_info->space_info;
722 struct btrfs_space_info *found;
724 int bits = dentry->d_sb->s_blocksize_bits;
725 __be32 *fsid = (__be32 *)root->fs_info->fsid;
728 list_for_each_entry_rcu(found, head, list)
729 total_used += found->disk_used;
732 buf->f_namelen = BTRFS_NAME_LEN;
733 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
734 buf->f_bfree = buf->f_blocks - (total_used >> bits);
735 buf->f_bavail = buf->f_bfree;
736 buf->f_bsize = dentry->d_sb->s_blocksize;
737 buf->f_type = BTRFS_SUPER_MAGIC;
739 /* We treat it as constant endianness (it doesn't matter _which_)
740 because we want the fsid to come out the same whether mounted
741 on a big-endian or little-endian host */
742 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
743 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
744 /* Mask in the root object ID too, to disambiguate subvols */
745 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
746 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
751 static struct file_system_type btrfs_fs_type = {
752 .owner = THIS_MODULE,
754 .get_sb = btrfs_get_sb,
755 .kill_sb = kill_anon_super,
756 .fs_flags = FS_REQUIRES_DEV,
760 * used by btrfsctl to scan devices when no FS is mounted
762 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
765 struct btrfs_ioctl_vol_args *vol;
766 struct btrfs_fs_devices *fs_devices;
769 if (!capable(CAP_SYS_ADMIN))
772 vol = memdup_user((void __user *)arg, sizeof(*vol));
777 case BTRFS_IOC_SCAN_DEV:
778 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
779 &btrfs_fs_type, &fs_devices);
787 static int btrfs_freeze(struct super_block *sb)
789 struct btrfs_root *root = btrfs_sb(sb);
790 mutex_lock(&root->fs_info->transaction_kthread_mutex);
791 mutex_lock(&root->fs_info->cleaner_mutex);
795 static int btrfs_unfreeze(struct super_block *sb)
797 struct btrfs_root *root = btrfs_sb(sb);
798 mutex_unlock(&root->fs_info->cleaner_mutex);
799 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
803 static const struct super_operations btrfs_super_ops = {
804 .drop_inode = btrfs_drop_inode,
805 .evict_inode = btrfs_evict_inode,
806 .put_super = btrfs_put_super,
807 .sync_fs = btrfs_sync_fs,
808 .show_options = btrfs_show_options,
809 .write_inode = btrfs_write_inode,
810 .dirty_inode = btrfs_dirty_inode,
811 .alloc_inode = btrfs_alloc_inode,
812 .destroy_inode = btrfs_destroy_inode,
813 .statfs = btrfs_statfs,
814 .remount_fs = btrfs_remount,
815 .freeze_fs = btrfs_freeze,
816 .unfreeze_fs = btrfs_unfreeze,
819 static const struct file_operations btrfs_ctl_fops = {
820 .unlocked_ioctl = btrfs_control_ioctl,
821 .compat_ioctl = btrfs_control_ioctl,
822 .owner = THIS_MODULE,
825 static struct miscdevice btrfs_misc = {
826 .minor = BTRFS_MINOR,
827 .name = "btrfs-control",
828 .fops = &btrfs_ctl_fops
831 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
832 MODULE_ALIAS("devname:btrfs-control");
834 static int btrfs_interface_init(void)
836 return misc_register(&btrfs_misc);
839 static void btrfs_interface_exit(void)
841 if (misc_deregister(&btrfs_misc) < 0)
842 printk(KERN_INFO "misc_deregister failed for control device");
845 static int __init init_btrfs_fs(void)
849 err = btrfs_init_sysfs();
853 err = btrfs_init_cachep();
857 err = extent_io_init();
861 err = extent_map_init();
865 err = btrfs_interface_init();
867 goto free_extent_map;
869 err = register_filesystem(&btrfs_fs_type);
871 goto unregister_ioctl;
873 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
877 btrfs_interface_exit();
883 btrfs_destroy_cachep();
889 static void __exit exit_btrfs_fs(void)
891 btrfs_destroy_cachep();
894 btrfs_interface_exit();
895 unregister_filesystem(&btrfs_fs_type);
897 btrfs_cleanup_fs_uuids();
901 module_init(init_btrfs_fs)
902 module_exit(exit_btrfs_fs)
904 MODULE_LICENSE("GPL");