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);
65 (void)ret; /* FIXME: need to fix VFS to return error? */
69 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
70 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
71 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
72 Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
73 Opt_discard, Opt_space_cache, Opt_clear_cache, Opt_err,
76 static match_table_t tokens = {
77 {Opt_degraded, "degraded"},
78 {Opt_subvol, "subvol=%s"},
79 {Opt_subvolid, "subvolid=%d"},
80 {Opt_device, "device=%s"},
81 {Opt_nodatasum, "nodatasum"},
82 {Opt_nodatacow, "nodatacow"},
83 {Opt_nobarrier, "nobarrier"},
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"},
97 {Opt_space_cache, "space_cache"},
98 {Opt_clear_cache, "clear_cache"},
103 * Regular mount options parser. Everything that is needed only when
104 * reading in a new superblock is parsed here.
106 int btrfs_parse_options(struct btrfs_root *root, char *options)
108 struct btrfs_fs_info *info = root->fs_info;
109 substring_t args[MAX_OPT_ARGS];
110 char *p, *num, *orig;
118 * strsep changes the string, duplicate it because parse_options
121 options = kstrdup(options, GFP_NOFS);
127 while ((p = strsep(&options, ",")) != NULL) {
132 token = match_token(p, tokens, args);
135 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
136 btrfs_set_opt(info->mount_opt, DEGRADED);
142 * These are parsed by btrfs_parse_early_options
143 * and can be happily ignored here.
147 printk(KERN_INFO "btrfs: setting nodatasum\n");
148 btrfs_set_opt(info->mount_opt, NODATASUM);
151 printk(KERN_INFO "btrfs: setting nodatacow\n");
152 btrfs_set_opt(info->mount_opt, NODATACOW);
153 btrfs_set_opt(info->mount_opt, NODATASUM);
156 printk(KERN_INFO "btrfs: use compression\n");
157 btrfs_set_opt(info->mount_opt, COMPRESS);
159 case Opt_compress_force:
160 printk(KERN_INFO "btrfs: forcing compression\n");
161 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
162 btrfs_set_opt(info->mount_opt, COMPRESS);
165 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
166 btrfs_set_opt(info->mount_opt, SSD);
169 printk(KERN_INFO "btrfs: use spread ssd "
170 "allocation scheme\n");
171 btrfs_set_opt(info->mount_opt, SSD);
172 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
175 printk(KERN_INFO "btrfs: not using ssd allocation "
177 btrfs_set_opt(info->mount_opt, NOSSD);
178 btrfs_clear_opt(info->mount_opt, SSD);
179 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
182 printk(KERN_INFO "btrfs: turning off barriers\n");
183 btrfs_set_opt(info->mount_opt, NOBARRIER);
185 case Opt_thread_pool:
187 match_int(&args[0], &intarg);
189 info->thread_pool_size = intarg;
190 printk(KERN_INFO "btrfs: thread pool %d\n",
191 info->thread_pool_size);
195 num = match_strdup(&args[0]);
197 info->max_inline = memparse(num, NULL);
200 if (info->max_inline) {
201 info->max_inline = max_t(u64,
205 printk(KERN_INFO "btrfs: max_inline at %llu\n",
206 (unsigned long long)info->max_inline);
209 case Opt_alloc_start:
210 num = match_strdup(&args[0]);
212 info->alloc_start = memparse(num, NULL);
215 "btrfs: allocations start at %llu\n",
216 (unsigned long long)info->alloc_start);
220 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
223 printk(KERN_INFO "btrfs: disabling tree log\n");
224 btrfs_set_opt(info->mount_opt, NOTREELOG);
226 case Opt_flushoncommit:
227 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
228 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
232 match_int(&args[0], &intarg);
234 info->metadata_ratio = intarg;
235 printk(KERN_INFO "btrfs: metadata ratio %d\n",
236 info->metadata_ratio);
240 btrfs_set_opt(info->mount_opt, DISCARD);
242 case Opt_space_cache:
243 printk(KERN_INFO "btrfs: enabling disk space caching\n");
244 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
245 case Opt_clear_cache:
246 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
247 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
250 printk(KERN_INFO "btrfs: unrecognized mount option "
264 * Parse mount options that are required early in the mount process.
266 * All other options will be parsed on much later in the mount process and
267 * only when we need to allocate a new super block.
269 static int btrfs_parse_early_options(const char *options, fmode_t flags,
270 void *holder, char **subvol_name, u64 *subvol_objectid,
271 struct btrfs_fs_devices **fs_devices)
273 substring_t args[MAX_OPT_ARGS];
282 * strsep changes the string, duplicate it because parse_options
285 opts = kstrdup(options, GFP_KERNEL);
289 while ((p = strsep(&opts, ",")) != NULL) {
294 token = match_token(p, tokens, args);
297 *subvol_name = match_strdup(&args[0]);
301 error = match_int(&args[0], &intarg);
303 /* we want the original fs_tree */
306 BTRFS_FS_TREE_OBJECTID;
308 *subvol_objectid = intarg;
312 error = btrfs_scan_one_device(match_strdup(&args[0]),
313 flags, holder, fs_devices);
326 * If no subvolume name is specified we use the default one. Allocate
327 * a copy of the string "." here so that code later in the
328 * mount path doesn't care if it's the default volume or another one.
331 *subvol_name = kstrdup(".", GFP_KERNEL);
338 static struct dentry *get_default_root(struct super_block *sb,
341 struct btrfs_root *root = sb->s_fs_info;
342 struct btrfs_root *new_root;
343 struct btrfs_dir_item *di;
344 struct btrfs_path *path;
345 struct btrfs_key location;
347 struct dentry *dentry;
352 * We have a specific subvol we want to mount, just setup location and
353 * go look up the root.
355 if (subvol_objectid) {
356 location.objectid = subvol_objectid;
357 location.type = BTRFS_ROOT_ITEM_KEY;
358 location.offset = (u64)-1;
362 path = btrfs_alloc_path();
364 return ERR_PTR(-ENOMEM);
365 path->leave_spinning = 1;
368 * Find the "default" dir item which points to the root item that we
369 * will mount by default if we haven't been given a specific subvolume
372 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
373 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
378 * Ok the default dir item isn't there. This is weird since
379 * it's always been there, but don't freak out, just try and
380 * mount to root most subvolume.
382 btrfs_free_path(path);
383 dir_id = BTRFS_FIRST_FREE_OBJECTID;
384 new_root = root->fs_info->fs_root;
388 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
389 btrfs_free_path(path);
392 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
393 if (IS_ERR(new_root))
394 return ERR_CAST(new_root);
396 if (btrfs_root_refs(&new_root->root_item) == 0)
397 return ERR_PTR(-ENOENT);
399 dir_id = btrfs_root_dirid(&new_root->root_item);
401 location.objectid = dir_id;
402 location.type = BTRFS_INODE_ITEM_KEY;
405 inode = btrfs_iget(sb, &location, new_root, &new);
407 return ERR_CAST(inode);
410 * If we're just mounting the root most subvol put the inode and return
411 * a reference to the dentry. We will have already gotten a reference
412 * to the inode in btrfs_fill_super so we're good to go.
414 if (!new && sb->s_root->d_inode == inode) {
416 return dget(sb->s_root);
420 const struct qstr name = { .name = "/", .len = 1 };
423 * New inode, we need to make the dentry a sibling of s_root so
424 * everything gets cleaned up properly on unmount.
426 dentry = d_alloc(sb->s_root, &name);
429 return ERR_PTR(-ENOMEM);
431 d_splice_alias(inode, dentry);
434 * We found the inode in cache, just find a dentry for it and
435 * put the reference to the inode we just got.
437 dentry = d_find_alias(inode);
444 static int btrfs_fill_super(struct super_block *sb,
445 struct btrfs_fs_devices *fs_devices,
446 void *data, int silent)
449 struct dentry *root_dentry;
450 struct btrfs_root *tree_root;
451 struct btrfs_key key;
454 sb->s_maxbytes = MAX_LFS_FILESIZE;
455 sb->s_magic = BTRFS_SUPER_MAGIC;
456 sb->s_op = &btrfs_super_ops;
457 sb->s_export_op = &btrfs_export_ops;
458 sb->s_xattr = btrfs_xattr_handlers;
460 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
461 sb->s_flags |= MS_POSIXACL;
464 tree_root = open_ctree(sb, fs_devices, (char *)data);
466 if (IS_ERR(tree_root)) {
467 printk("btrfs: open_ctree failed\n");
468 return PTR_ERR(tree_root);
470 sb->s_fs_info = tree_root;
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;
584 if (!(flags & MS_RDONLY))
587 error = btrfs_parse_early_options(data, mode, fs_type,
588 &subvol_name, &subvol_objectid,
593 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
595 goto error_free_subvol_name;
597 error = btrfs_open_devices(fs_devices, mode, fs_type);
599 goto error_free_subvol_name;
601 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
603 goto error_close_devices;
606 bdev = fs_devices->latest_bdev;
607 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
612 if ((flags ^ s->s_flags) & MS_RDONLY) {
613 deactivate_locked_super(s);
615 goto error_close_devices;
618 btrfs_close_devices(fs_devices);
620 char b[BDEVNAME_SIZE];
623 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
624 error = btrfs_fill_super(s, fs_devices, data,
625 flags & MS_SILENT ? 1 : 0);
627 deactivate_locked_super(s);
628 goto error_free_subvol_name;
631 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
632 s->s_flags |= MS_ACTIVE;
635 root = get_default_root(s, subvol_objectid);
637 error = PTR_ERR(root);
638 deactivate_locked_super(s);
639 goto error_free_subvol_name;
641 /* if they gave us a subvolume name bind mount into that */
642 if (strcmp(subvol_name, ".")) {
643 struct dentry *new_root;
644 mutex_lock(&root->d_inode->i_mutex);
645 new_root = lookup_one_len(subvol_name, root,
646 strlen(subvol_name));
647 mutex_unlock(&root->d_inode->i_mutex);
649 if (IS_ERR(new_root)) {
650 deactivate_locked_super(s);
651 error = PTR_ERR(new_root);
653 goto error_free_subvol_name;
655 if (!new_root->d_inode) {
658 deactivate_locked_super(s);
660 goto error_free_subvol_name;
667 mnt->mnt_root = root;
675 btrfs_close_devices(fs_devices);
676 error_free_subvol_name:
681 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
683 struct btrfs_root *root = btrfs_sb(sb);
686 ret = btrfs_parse_options(root, data);
690 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
693 if (*flags & MS_RDONLY) {
694 sb->s_flags |= MS_RDONLY;
696 ret = btrfs_commit_super(root);
699 if (root->fs_info->fs_devices->rw_devices == 0)
702 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
705 ret = btrfs_cleanup_fs_roots(root->fs_info);
708 /* recover relocation */
709 ret = btrfs_recover_relocation(root);
712 sb->s_flags &= ~MS_RDONLY;
718 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
720 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
721 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
722 struct list_head *head = &root->fs_info->space_info;
723 struct btrfs_space_info *found;
725 u64 total_used_data = 0;
726 int bits = dentry->d_sb->s_blocksize_bits;
727 __be32 *fsid = (__be32 *)root->fs_info->fsid;
730 list_for_each_entry_rcu(found, head, list) {
731 if (found->flags & (BTRFS_BLOCK_GROUP_METADATA |
732 BTRFS_BLOCK_GROUP_SYSTEM))
733 total_used_data += found->disk_total;
735 total_used_data += found->disk_used;
736 total_used += found->disk_used;
740 buf->f_namelen = BTRFS_NAME_LEN;
741 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
742 buf->f_bfree = buf->f_blocks - (total_used >> bits);
743 buf->f_bavail = buf->f_blocks - (total_used_data >> bits);
744 buf->f_bsize = dentry->d_sb->s_blocksize;
745 buf->f_type = BTRFS_SUPER_MAGIC;
747 /* We treat it as constant endianness (it doesn't matter _which_)
748 because we want the fsid to come out the same whether mounted
749 on a big-endian or little-endian host */
750 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
751 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
752 /* Mask in the root object ID too, to disambiguate subvols */
753 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
754 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
759 static struct file_system_type btrfs_fs_type = {
760 .owner = THIS_MODULE,
762 .get_sb = btrfs_get_sb,
763 .kill_sb = kill_anon_super,
764 .fs_flags = FS_REQUIRES_DEV,
768 * used by btrfsctl to scan devices when no FS is mounted
770 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
773 struct btrfs_ioctl_vol_args *vol;
774 struct btrfs_fs_devices *fs_devices;
777 if (!capable(CAP_SYS_ADMIN))
780 vol = memdup_user((void __user *)arg, sizeof(*vol));
785 case BTRFS_IOC_SCAN_DEV:
786 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
787 &btrfs_fs_type, &fs_devices);
795 static int btrfs_freeze(struct super_block *sb)
797 struct btrfs_root *root = btrfs_sb(sb);
798 mutex_lock(&root->fs_info->transaction_kthread_mutex);
799 mutex_lock(&root->fs_info->cleaner_mutex);
803 static int btrfs_unfreeze(struct super_block *sb)
805 struct btrfs_root *root = btrfs_sb(sb);
806 mutex_unlock(&root->fs_info->cleaner_mutex);
807 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
811 static const struct super_operations btrfs_super_ops = {
812 .drop_inode = btrfs_drop_inode,
813 .evict_inode = btrfs_evict_inode,
814 .put_super = btrfs_put_super,
815 .sync_fs = btrfs_sync_fs,
816 .show_options = btrfs_show_options,
817 .write_inode = btrfs_write_inode,
818 .dirty_inode = btrfs_dirty_inode,
819 .alloc_inode = btrfs_alloc_inode,
820 .destroy_inode = btrfs_destroy_inode,
821 .statfs = btrfs_statfs,
822 .remount_fs = btrfs_remount,
823 .freeze_fs = btrfs_freeze,
824 .unfreeze_fs = btrfs_unfreeze,
827 static const struct file_operations btrfs_ctl_fops = {
828 .unlocked_ioctl = btrfs_control_ioctl,
829 .compat_ioctl = btrfs_control_ioctl,
830 .owner = THIS_MODULE,
833 static struct miscdevice btrfs_misc = {
834 .minor = BTRFS_MINOR,
835 .name = "btrfs-control",
836 .fops = &btrfs_ctl_fops
839 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
840 MODULE_ALIAS("devname:btrfs-control");
842 static int btrfs_interface_init(void)
844 return misc_register(&btrfs_misc);
847 static void btrfs_interface_exit(void)
849 if (misc_deregister(&btrfs_misc) < 0)
850 printk(KERN_INFO "misc_deregister failed for control device");
853 static int __init init_btrfs_fs(void)
857 err = btrfs_init_sysfs();
861 err = btrfs_init_cachep();
865 err = extent_io_init();
869 err = extent_map_init();
873 err = btrfs_interface_init();
875 goto free_extent_map;
877 err = register_filesystem(&btrfs_fs_type);
879 goto unregister_ioctl;
881 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
885 btrfs_interface_exit();
891 btrfs_destroy_cachep();
897 static void __exit exit_btrfs_fs(void)
899 btrfs_destroy_cachep();
902 btrfs_interface_exit();
903 unregister_filesystem(&btrfs_fs_type);
905 btrfs_cleanup_fs_uuids();
909 module_init(init_btrfs_fs)
910 module_exit(exit_btrfs_fs)
912 MODULE_LICENSE("GPL");