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/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/swap.h>
34 #include <linux/writeback.h>
35 #include <linux/statfs.h>
36 #include <linux/compat.h>
37 #include <linux/parser.h>
38 #include <linux/ctype.h>
39 #include <linux/namei.h>
40 #include <linux/miscdevice.h>
41 #include <linux/magic.h>
45 #include "transaction.h"
46 #include "btrfs_inode.h"
48 #include "print-tree.h"
53 #include "compression.h"
55 static 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_device, Opt_nodatasum, Opt_nodatacow,
68 Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
69 Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl,
70 Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err,
73 static match_table_t tokens = {
74 {Opt_degraded, "degraded"},
75 {Opt_subvol, "subvol=%s"},
76 {Opt_device, "device=%s"},
77 {Opt_nodatasum, "nodatasum"},
78 {Opt_nodatacow, "nodatacow"},
79 {Opt_nobarrier, "nobarrier"},
80 {Opt_max_extent, "max_extent=%s"},
81 {Opt_max_inline, "max_inline=%s"},
82 {Opt_alloc_start, "alloc_start=%s"},
83 {Opt_thread_pool, "thread_pool=%d"},
84 {Opt_compress, "compress"},
86 {Opt_ssd_spread, "ssd_spread"},
89 {Opt_notreelog, "notreelog"},
90 {Opt_flushoncommit, "flushoncommit"},
91 {Opt_ratio, "metadata_ratio=%d"},
95 u64 btrfs_parse_size(char *str)
102 res = simple_strtoul(str, &end, 10);
106 last = tolower(last);
121 * Regular mount options parser. Everything that is needed only when
122 * reading in a new superblock is parsed here.
124 int btrfs_parse_options(struct btrfs_root *root, char *options)
126 struct btrfs_fs_info *info = root->fs_info;
127 substring_t args[MAX_OPT_ARGS];
135 * strsep changes the string, duplicate it because parse_options
138 options = kstrdup(options, GFP_NOFS);
143 while ((p = strsep(&options, ",")) != NULL) {
148 token = match_token(p, tokens, args);
151 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
152 btrfs_set_opt(info->mount_opt, DEGRADED);
157 * These are parsed by btrfs_parse_early_options
158 * and can be happily ignored here.
162 printk(KERN_INFO "btrfs: setting nodatacsum\n");
163 btrfs_set_opt(info->mount_opt, NODATASUM);
166 printk(KERN_INFO "btrfs: setting nodatacow\n");
167 btrfs_set_opt(info->mount_opt, NODATACOW);
168 btrfs_set_opt(info->mount_opt, NODATASUM);
171 printk(KERN_INFO "btrfs: use compression\n");
172 btrfs_set_opt(info->mount_opt, COMPRESS);
175 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
176 btrfs_set_opt(info->mount_opt, SSD);
179 printk(KERN_INFO "btrfs: use spread ssd "
180 "allocation scheme\n");
181 btrfs_set_opt(info->mount_opt, SSD);
182 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
185 printk(KERN_INFO "btrfs: not using ssd allocation "
187 btrfs_clear_opt(info->mount_opt, SSD);
188 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
191 printk(KERN_INFO "btrfs: turning off barriers\n");
192 btrfs_set_opt(info->mount_opt, NOBARRIER);
194 case Opt_thread_pool:
196 match_int(&args[0], &intarg);
198 info->thread_pool_size = intarg;
199 printk(KERN_INFO "btrfs: thread pool %d\n",
200 info->thread_pool_size);
204 num = match_strdup(&args[0]);
206 info->max_extent = btrfs_parse_size(num);
209 info->max_extent = max_t(u64,
210 info->max_extent, root->sectorsize);
211 printk(KERN_INFO "btrfs: max_extent at %llu\n",
212 (unsigned long long)info->max_extent);
216 num = match_strdup(&args[0]);
218 info->max_inline = btrfs_parse_size(num);
221 if (info->max_inline) {
222 info->max_inline = max_t(u64,
226 printk(KERN_INFO "btrfs: max_inline at %llu\n",
227 (unsigned long long)info->max_inline);
230 case Opt_alloc_start:
231 num = match_strdup(&args[0]);
233 info->alloc_start = btrfs_parse_size(num);
236 "btrfs: allocations start at %llu\n",
237 (unsigned long long)info->alloc_start);
241 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
244 printk(KERN_INFO "btrfs: disabling tree log\n");
245 btrfs_set_opt(info->mount_opt, NOTREELOG);
247 case Opt_flushoncommit:
248 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
249 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
253 match_int(&args[0], &intarg);
255 info->metadata_ratio = intarg;
256 printk(KERN_INFO "btrfs: metadata ratio %d\n",
257 info->metadata_ratio);
269 * Parse mount options that are required early in the mount process.
271 * All other options will be parsed on much later in the mount process and
272 * only when we need to allocate a new super block.
274 static int btrfs_parse_early_options(const char *options, fmode_t flags,
275 void *holder, char **subvol_name,
276 struct btrfs_fs_devices **fs_devices)
278 substring_t args[MAX_OPT_ARGS];
286 * strsep changes the string, duplicate it because parse_options
289 opts = kstrdup(options, GFP_KERNEL);
293 while ((p = strsep(&opts, ",")) != NULL) {
298 token = match_token(p, tokens, args);
301 *subvol_name = match_strdup(&args[0]);
304 error = btrfs_scan_one_device(match_strdup(&args[0]),
305 flags, holder, fs_devices);
318 * If no subvolume name is specified we use the default one. Allocate
319 * a copy of the string "." here so that code later in the
320 * mount path doesn't care if it's the default volume or another one.
323 *subvol_name = kstrdup(".", GFP_KERNEL);
330 static int btrfs_fill_super(struct super_block *sb,
331 struct btrfs_fs_devices *fs_devices,
332 void *data, int silent)
335 struct dentry *root_dentry;
336 struct btrfs_super_block *disk_super;
337 struct btrfs_root *tree_root;
338 struct btrfs_key key;
341 sb->s_maxbytes = MAX_LFS_FILESIZE;
342 sb->s_magic = BTRFS_SUPER_MAGIC;
343 sb->s_op = &btrfs_super_ops;
344 sb->s_export_op = &btrfs_export_ops;
345 sb->s_xattr = btrfs_xattr_handlers;
347 sb->s_flags |= MS_POSIXACL;
349 tree_root = open_ctree(sb, fs_devices, (char *)data);
351 if (IS_ERR(tree_root)) {
352 printk("btrfs: open_ctree failed\n");
353 return PTR_ERR(tree_root);
355 sb->s_fs_info = tree_root;
356 disk_super = &tree_root->fs_info->super_copy;
358 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
359 key.type = BTRFS_INODE_ITEM_KEY;
361 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root);
363 err = PTR_ERR(inode);
367 root_dentry = d_alloc_root(inode);
374 /* this does the super kobj at the same time */
375 err = btrfs_sysfs_add_super(tree_root->fs_info);
380 sb->s_root = root_dentry;
382 save_mount_options(sb, data);
386 close_ctree(tree_root);
390 int btrfs_sync_fs(struct super_block *sb, int wait)
392 struct btrfs_trans_handle *trans;
393 struct btrfs_root *root = btrfs_sb(sb);
396 if (sb->s_flags & MS_RDONLY)
401 filemap_flush(root->fs_info->btree_inode->i_mapping);
405 btrfs_start_delalloc_inodes(root);
406 btrfs_wait_ordered_extents(root, 0);
408 trans = btrfs_start_transaction(root, 1);
409 ret = btrfs_commit_transaction(trans, root);
414 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
416 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
417 struct btrfs_fs_info *info = root->fs_info;
419 if (btrfs_test_opt(root, DEGRADED))
420 seq_puts(seq, ",degraded");
421 if (btrfs_test_opt(root, NODATASUM))
422 seq_puts(seq, ",nodatasum");
423 if (btrfs_test_opt(root, NODATACOW))
424 seq_puts(seq, ",nodatacow");
425 if (btrfs_test_opt(root, NOBARRIER))
426 seq_puts(seq, ",nobarrier");
427 if (info->max_extent != (u64)-1)
428 seq_printf(seq, ",max_extent=%llu",
429 (unsigned long long)info->max_extent);
430 if (info->max_inline != 8192 * 1024)
431 seq_printf(seq, ",max_inline=%llu",
432 (unsigned long long)info->max_inline);
433 if (info->alloc_start != 0)
434 seq_printf(seq, ",alloc_start=%llu",
435 (unsigned long long)info->alloc_start);
436 if (info->thread_pool_size != min_t(unsigned long,
437 num_online_cpus() + 2, 8))
438 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
439 if (btrfs_test_opt(root, COMPRESS))
440 seq_puts(seq, ",compress");
441 if (btrfs_test_opt(root, SSD_SPREAD))
442 seq_puts(seq, ",ssd_spread");
443 else if (btrfs_test_opt(root, SSD))
444 seq_puts(seq, ",ssd");
445 if (btrfs_test_opt(root, NOTREELOG))
446 seq_puts(seq, ",notreelog");
447 if (btrfs_test_opt(root, FLUSHONCOMMIT))
448 seq_puts(seq, ",flushoncommit");
449 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
450 seq_puts(seq, ",noacl");
454 static void btrfs_write_super(struct super_block *sb)
459 static int btrfs_test_super(struct super_block *s, void *data)
461 struct btrfs_fs_devices *test_fs_devices = data;
462 struct btrfs_root *root = btrfs_sb(s);
464 return root->fs_info->fs_devices == test_fs_devices;
468 * Find a superblock for the given device / mount point.
470 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
471 * for multiple device setup. Make sure to keep it in sync.
473 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
474 const char *dev_name, void *data, struct vfsmount *mnt)
476 char *subvol_name = NULL;
477 struct block_device *bdev = NULL;
478 struct super_block *s;
480 struct btrfs_fs_devices *fs_devices = NULL;
481 fmode_t mode = FMODE_READ;
484 if (!(flags & MS_RDONLY))
487 error = btrfs_parse_early_options(data, mode, fs_type,
488 &subvol_name, &fs_devices);
492 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
494 goto error_free_subvol_name;
496 error = btrfs_open_devices(fs_devices, mode, fs_type);
498 goto error_free_subvol_name;
500 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
502 goto error_close_devices;
505 bdev = fs_devices->latest_bdev;
506 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
511 if ((flags ^ s->s_flags) & MS_RDONLY) {
512 deactivate_locked_super(s);
514 goto error_close_devices;
517 btrfs_close_devices(fs_devices);
519 char b[BDEVNAME_SIZE];
522 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
523 error = btrfs_fill_super(s, fs_devices, data,
524 flags & MS_SILENT ? 1 : 0);
526 deactivate_locked_super(s);
527 goto error_free_subvol_name;
530 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
531 s->s_flags |= MS_ACTIVE;
534 if (!strcmp(subvol_name, "."))
535 root = dget(s->s_root);
537 mutex_lock(&s->s_root->d_inode->i_mutex);
538 root = lookup_one_len(subvol_name, s->s_root,
539 strlen(subvol_name));
540 mutex_unlock(&s->s_root->d_inode->i_mutex);
543 deactivate_locked_super(s);
544 error = PTR_ERR(root);
545 goto error_free_subvol_name;
547 if (!root->d_inode) {
549 deactivate_locked_super(s);
551 goto error_free_subvol_name;
556 mnt->mnt_root = root;
564 btrfs_close_devices(fs_devices);
565 error_free_subvol_name:
570 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
572 struct btrfs_root *root = btrfs_sb(sb);
575 ret = btrfs_parse_options(root, data);
579 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
582 if (*flags & MS_RDONLY) {
583 sb->s_flags |= MS_RDONLY;
585 ret = btrfs_commit_super(root);
588 if (root->fs_info->fs_devices->rw_devices == 0)
591 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
594 /* recover relocation */
595 ret = btrfs_recover_relocation(root);
598 ret = btrfs_cleanup_fs_roots(root->fs_info);
601 sb->s_flags &= ~MS_RDONLY;
607 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
609 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
610 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
611 int bits = dentry->d_sb->s_blocksize_bits;
612 __be32 *fsid = (__be32 *)root->fs_info->fsid;
614 buf->f_namelen = BTRFS_NAME_LEN;
615 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
616 buf->f_bfree = buf->f_blocks -
617 (btrfs_super_bytes_used(disk_super) >> bits);
618 buf->f_bavail = buf->f_bfree;
619 buf->f_bsize = dentry->d_sb->s_blocksize;
620 buf->f_type = BTRFS_SUPER_MAGIC;
622 /* We treat it as constant endianness (it doesn't matter _which_)
623 because we want the fsid to come out the same whether mounted
624 on a big-endian or little-endian host */
625 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
626 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
627 /* Mask in the root object ID too, to disambiguate subvols */
628 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
629 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
634 static struct file_system_type btrfs_fs_type = {
635 .owner = THIS_MODULE,
637 .get_sb = btrfs_get_sb,
638 .kill_sb = kill_anon_super,
639 .fs_flags = FS_REQUIRES_DEV,
643 * used by btrfsctl to scan devices when no FS is mounted
645 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
648 struct btrfs_ioctl_vol_args *vol;
649 struct btrfs_fs_devices *fs_devices;
652 if (!capable(CAP_SYS_ADMIN))
655 vol = memdup_user((void __user *)arg, sizeof(*vol));
660 case BTRFS_IOC_SCAN_DEV:
661 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
662 &btrfs_fs_type, &fs_devices);
670 static int btrfs_freeze(struct super_block *sb)
672 struct btrfs_root *root = btrfs_sb(sb);
673 mutex_lock(&root->fs_info->transaction_kthread_mutex);
674 mutex_lock(&root->fs_info->cleaner_mutex);
678 static int btrfs_unfreeze(struct super_block *sb)
680 struct btrfs_root *root = btrfs_sb(sb);
681 mutex_unlock(&root->fs_info->cleaner_mutex);
682 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
686 static struct super_operations btrfs_super_ops = {
687 .delete_inode = btrfs_delete_inode,
688 .put_super = btrfs_put_super,
689 .write_super = btrfs_write_super,
690 .sync_fs = btrfs_sync_fs,
691 .show_options = btrfs_show_options,
692 .write_inode = btrfs_write_inode,
693 .dirty_inode = btrfs_dirty_inode,
694 .alloc_inode = btrfs_alloc_inode,
695 .destroy_inode = btrfs_destroy_inode,
696 .statfs = btrfs_statfs,
697 .remount_fs = btrfs_remount,
698 .freeze_fs = btrfs_freeze,
699 .unfreeze_fs = btrfs_unfreeze,
702 static const struct file_operations btrfs_ctl_fops = {
703 .unlocked_ioctl = btrfs_control_ioctl,
704 .compat_ioctl = btrfs_control_ioctl,
705 .owner = THIS_MODULE,
708 static struct miscdevice btrfs_misc = {
709 .minor = MISC_DYNAMIC_MINOR,
710 .name = "btrfs-control",
711 .fops = &btrfs_ctl_fops
714 static int btrfs_interface_init(void)
716 return misc_register(&btrfs_misc);
719 static void btrfs_interface_exit(void)
721 if (misc_deregister(&btrfs_misc) < 0)
722 printk(KERN_INFO "misc_deregister failed for control device");
725 static int __init init_btrfs_fs(void)
729 err = btrfs_init_sysfs();
733 err = btrfs_init_cachep();
737 err = extent_io_init();
741 err = extent_map_init();
745 err = btrfs_interface_init();
747 goto free_extent_map;
749 err = register_filesystem(&btrfs_fs_type);
751 goto unregister_ioctl;
753 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
757 btrfs_interface_exit();
763 btrfs_destroy_cachep();
769 static void __exit exit_btrfs_fs(void)
771 btrfs_destroy_cachep();
774 btrfs_interface_exit();
775 unregister_filesystem(&btrfs_fs_type);
777 btrfs_cleanup_fs_uuids();
781 module_init(init_btrfs_fs)
782 module_exit(exit_btrfs_fs)
784 MODULE_LICENSE("GPL");