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>
42 #include <linux/cleancache.h>
43 #include <linux/mnt_namespace.h>
45 #include "delayed-inode.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
51 #include "print-tree.h"
56 #include "compression.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/btrfs.h>
61 static const struct super_operations btrfs_super_ops;
62 static struct file_system_type btrfs_fs_type;
64 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
71 errstr = "IO failure";
74 errstr = "Out of memory";
77 errstr = "Readonly filesystem";
81 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
90 static void __save_error_info(struct btrfs_fs_info *fs_info)
93 * today we only save the error info into ram. Long term we'll
94 * also send it down to the disk
96 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
100 * We move write_super stuff at umount in order to avoid deadlock
101 * for umount hold all lock.
103 static void save_error_info(struct btrfs_fs_info *fs_info)
105 __save_error_info(fs_info);
108 /* btrfs handle error by forcing the filesystem readonly */
109 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
111 struct super_block *sb = fs_info->sb;
113 if (sb->s_flags & MS_RDONLY)
116 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
117 sb->s_flags |= MS_RDONLY;
118 printk(KERN_INFO "btrfs is forced readonly\n");
123 * __btrfs_std_error decodes expected errors from the caller and
124 * invokes the approciate error response.
126 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
127 unsigned int line, int errno)
129 struct super_block *sb = fs_info->sb;
134 * Special case: if the error is EROFS, and we're already
135 * under MS_RDONLY, then it is safe here.
137 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
140 errstr = btrfs_decode_error(fs_info, errno, nbuf);
141 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
142 sb->s_id, function, line, errstr);
143 save_error_info(fs_info);
145 btrfs_handle_error(fs_info);
148 static void btrfs_put_super(struct super_block *sb)
150 struct btrfs_root *root = btrfs_sb(sb);
153 ret = close_ctree(root);
154 sb->s_fs_info = NULL;
156 (void)ret; /* FIXME: need to fix VFS to return error? */
160 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
161 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
162 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
163 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
164 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
165 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
166 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
167 Opt_inode_cache, Opt_no_space_cache, Opt_err,
170 static match_table_t tokens = {
171 {Opt_degraded, "degraded"},
172 {Opt_subvol, "subvol=%s"},
173 {Opt_subvolid, "subvolid=%d"},
174 {Opt_device, "device=%s"},
175 {Opt_nodatasum, "nodatasum"},
176 {Opt_nodatacow, "nodatacow"},
177 {Opt_nobarrier, "nobarrier"},
178 {Opt_max_inline, "max_inline=%s"},
179 {Opt_alloc_start, "alloc_start=%s"},
180 {Opt_thread_pool, "thread_pool=%d"},
181 {Opt_compress, "compress"},
182 {Opt_compress_type, "compress=%s"},
183 {Opt_compress_force, "compress-force"},
184 {Opt_compress_force_type, "compress-force=%s"},
186 {Opt_ssd_spread, "ssd_spread"},
187 {Opt_nossd, "nossd"},
188 {Opt_noacl, "noacl"},
189 {Opt_notreelog, "notreelog"},
190 {Opt_flushoncommit, "flushoncommit"},
191 {Opt_ratio, "metadata_ratio=%d"},
192 {Opt_discard, "discard"},
193 {Opt_space_cache, "space_cache"},
194 {Opt_clear_cache, "clear_cache"},
195 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
196 {Opt_enospc_debug, "enospc_debug"},
197 {Opt_subvolrootid, "subvolrootid=%d"},
198 {Opt_defrag, "autodefrag"},
199 {Opt_inode_cache, "inode_cache"},
200 {Opt_no_space_cache, "no_space_cache"},
205 * Regular mount options parser. Everything that is needed only when
206 * reading in a new superblock is parsed here.
208 int btrfs_parse_options(struct btrfs_root *root, char *options)
210 struct btrfs_fs_info *info = root->fs_info;
211 substring_t args[MAX_OPT_ARGS];
212 char *p, *num, *orig = NULL;
217 bool compress_force = false;
219 cache_gen = btrfs_super_cache_generation(&root->fs_info->super_copy);
221 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
227 * strsep changes the string, duplicate it because parse_options
230 options = kstrdup(options, GFP_NOFS);
236 while ((p = strsep(&options, ",")) != NULL) {
241 token = match_token(p, tokens, args);
244 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
245 btrfs_set_opt(info->mount_opt, DEGRADED);
249 case Opt_subvolrootid:
252 * These are parsed by btrfs_parse_early_options
253 * and can be happily ignored here.
257 printk(KERN_INFO "btrfs: setting nodatasum\n");
258 btrfs_set_opt(info->mount_opt, NODATASUM);
261 printk(KERN_INFO "btrfs: setting nodatacow\n");
262 btrfs_set_opt(info->mount_opt, NODATACOW);
263 btrfs_set_opt(info->mount_opt, NODATASUM);
265 case Opt_compress_force:
266 case Opt_compress_force_type:
267 compress_force = true;
269 case Opt_compress_type:
270 if (token == Opt_compress ||
271 token == Opt_compress_force ||
272 strcmp(args[0].from, "zlib") == 0) {
273 compress_type = "zlib";
274 info->compress_type = BTRFS_COMPRESS_ZLIB;
275 } else if (strcmp(args[0].from, "lzo") == 0) {
276 compress_type = "lzo";
277 info->compress_type = BTRFS_COMPRESS_LZO;
283 btrfs_set_opt(info->mount_opt, COMPRESS);
284 if (compress_force) {
285 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
286 pr_info("btrfs: force %s compression\n",
289 pr_info("btrfs: use %s compression\n",
293 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
294 btrfs_set_opt(info->mount_opt, SSD);
297 printk(KERN_INFO "btrfs: use spread ssd "
298 "allocation scheme\n");
299 btrfs_set_opt(info->mount_opt, SSD);
300 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
303 printk(KERN_INFO "btrfs: not using ssd allocation "
305 btrfs_set_opt(info->mount_opt, NOSSD);
306 btrfs_clear_opt(info->mount_opt, SSD);
307 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
310 printk(KERN_INFO "btrfs: turning off barriers\n");
311 btrfs_set_opt(info->mount_opt, NOBARRIER);
313 case Opt_thread_pool:
315 match_int(&args[0], &intarg);
317 info->thread_pool_size = intarg;
318 printk(KERN_INFO "btrfs: thread pool %d\n",
319 info->thread_pool_size);
323 num = match_strdup(&args[0]);
325 info->max_inline = memparse(num, NULL);
328 if (info->max_inline) {
329 info->max_inline = max_t(u64,
333 printk(KERN_INFO "btrfs: max_inline at %llu\n",
334 (unsigned long long)info->max_inline);
337 case Opt_alloc_start:
338 num = match_strdup(&args[0]);
340 info->alloc_start = memparse(num, NULL);
343 "btrfs: allocations start at %llu\n",
344 (unsigned long long)info->alloc_start);
348 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
351 printk(KERN_INFO "btrfs: disabling tree log\n");
352 btrfs_set_opt(info->mount_opt, NOTREELOG);
354 case Opt_flushoncommit:
355 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
356 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
360 match_int(&args[0], &intarg);
362 info->metadata_ratio = intarg;
363 printk(KERN_INFO "btrfs: metadata ratio %d\n",
364 info->metadata_ratio);
368 btrfs_set_opt(info->mount_opt, DISCARD);
370 case Opt_space_cache:
371 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
373 case Opt_no_space_cache:
374 printk(KERN_INFO "btrfs: disabling disk space caching\n");
375 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
377 case Opt_inode_cache:
378 printk(KERN_INFO "btrfs: enabling inode map caching\n");
379 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
381 case Opt_clear_cache:
382 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
383 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
385 case Opt_user_subvol_rm_allowed:
386 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
388 case Opt_enospc_debug:
389 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
392 printk(KERN_INFO "btrfs: enabling auto defrag");
393 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
396 printk(KERN_INFO "btrfs: unrecognized mount option "
405 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
406 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
412 * Parse mount options that are required early in the mount process.
414 * All other options will be parsed on much later in the mount process and
415 * only when we need to allocate a new super block.
417 static int btrfs_parse_early_options(const char *options, fmode_t flags,
418 void *holder, char **subvol_name, u64 *subvol_objectid,
419 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
421 substring_t args[MAX_OPT_ARGS];
422 char *device_name, *opts, *orig, *p;
430 * strsep changes the string, duplicate it because parse_options
433 opts = kstrdup(options, GFP_KERNEL);
438 while ((p = strsep(&opts, ",")) != NULL) {
443 token = match_token(p, tokens, args);
446 *subvol_name = match_strdup(&args[0]);
450 error = match_int(&args[0], &intarg);
452 /* we want the original fs_tree */
455 BTRFS_FS_TREE_OBJECTID;
457 *subvol_objectid = intarg;
460 case Opt_subvolrootid:
462 error = match_int(&args[0], &intarg);
464 /* we want the original fs_tree */
467 BTRFS_FS_TREE_OBJECTID;
469 *subvol_rootid = intarg;
473 device_name = match_strdup(&args[0]);
478 error = btrfs_scan_one_device(device_name,
479 flags, holder, fs_devices);
494 static struct dentry *get_default_root(struct super_block *sb,
497 struct btrfs_root *root = sb->s_fs_info;
498 struct btrfs_root *new_root;
499 struct btrfs_dir_item *di;
500 struct btrfs_path *path;
501 struct btrfs_key location;
507 * We have a specific subvol we want to mount, just setup location and
508 * go look up the root.
510 if (subvol_objectid) {
511 location.objectid = subvol_objectid;
512 location.type = BTRFS_ROOT_ITEM_KEY;
513 location.offset = (u64)-1;
517 path = btrfs_alloc_path();
519 return ERR_PTR(-ENOMEM);
520 path->leave_spinning = 1;
523 * Find the "default" dir item which points to the root item that we
524 * will mount by default if we haven't been given a specific subvolume
527 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
528 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
530 btrfs_free_path(path);
535 * Ok the default dir item isn't there. This is weird since
536 * it's always been there, but don't freak out, just try and
537 * mount to root most subvolume.
539 btrfs_free_path(path);
540 dir_id = BTRFS_FIRST_FREE_OBJECTID;
541 new_root = root->fs_info->fs_root;
545 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
546 btrfs_free_path(path);
549 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
550 if (IS_ERR(new_root))
551 return ERR_CAST(new_root);
553 if (btrfs_root_refs(&new_root->root_item) == 0)
554 return ERR_PTR(-ENOENT);
556 dir_id = btrfs_root_dirid(&new_root->root_item);
558 location.objectid = dir_id;
559 location.type = BTRFS_INODE_ITEM_KEY;
562 inode = btrfs_iget(sb, &location, new_root, &new);
564 return ERR_CAST(inode);
567 * If we're just mounting the root most subvol put the inode and return
568 * a reference to the dentry. We will have already gotten a reference
569 * to the inode in btrfs_fill_super so we're good to go.
571 if (!new && sb->s_root->d_inode == inode) {
573 return dget(sb->s_root);
576 return d_obtain_alias(inode);
579 static int btrfs_fill_super(struct super_block *sb,
580 struct btrfs_fs_devices *fs_devices,
581 void *data, int silent)
584 struct dentry *root_dentry;
585 struct btrfs_root *tree_root;
586 struct btrfs_key key;
589 sb->s_maxbytes = MAX_LFS_FILESIZE;
590 sb->s_magic = BTRFS_SUPER_MAGIC;
591 sb->s_op = &btrfs_super_ops;
592 sb->s_d_op = &btrfs_dentry_operations;
593 sb->s_export_op = &btrfs_export_ops;
594 sb->s_xattr = btrfs_xattr_handlers;
596 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
597 sb->s_flags |= MS_POSIXACL;
600 tree_root = open_ctree(sb, fs_devices, (char *)data);
602 if (IS_ERR(tree_root)) {
603 printk("btrfs: open_ctree failed\n");
604 return PTR_ERR(tree_root);
606 sb->s_fs_info = tree_root;
608 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
609 key.type = BTRFS_INODE_ITEM_KEY;
611 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
613 err = PTR_ERR(inode);
617 root_dentry = d_alloc_root(inode);
624 sb->s_root = root_dentry;
626 save_mount_options(sb, data);
627 cleancache_init_fs(sb);
631 close_ctree(tree_root);
635 int btrfs_sync_fs(struct super_block *sb, int wait)
637 struct btrfs_trans_handle *trans;
638 struct btrfs_root *root = btrfs_sb(sb);
641 trace_btrfs_sync_fs(wait);
644 filemap_flush(root->fs_info->btree_inode->i_mapping);
648 btrfs_start_delalloc_inodes(root, 0);
649 btrfs_wait_ordered_extents(root, 0, 0);
651 trans = btrfs_start_transaction(root, 0);
653 return PTR_ERR(trans);
654 ret = btrfs_commit_transaction(trans, root);
658 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
660 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
661 struct btrfs_fs_info *info = root->fs_info;
664 if (btrfs_test_opt(root, DEGRADED))
665 seq_puts(seq, ",degraded");
666 if (btrfs_test_opt(root, NODATASUM))
667 seq_puts(seq, ",nodatasum");
668 if (btrfs_test_opt(root, NODATACOW))
669 seq_puts(seq, ",nodatacow");
670 if (btrfs_test_opt(root, NOBARRIER))
671 seq_puts(seq, ",nobarrier");
672 if (info->max_inline != 8192 * 1024)
673 seq_printf(seq, ",max_inline=%llu",
674 (unsigned long long)info->max_inline);
675 if (info->alloc_start != 0)
676 seq_printf(seq, ",alloc_start=%llu",
677 (unsigned long long)info->alloc_start);
678 if (info->thread_pool_size != min_t(unsigned long,
679 num_online_cpus() + 2, 8))
680 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
681 if (btrfs_test_opt(root, COMPRESS)) {
682 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
683 compress_type = "zlib";
685 compress_type = "lzo";
686 if (btrfs_test_opt(root, FORCE_COMPRESS))
687 seq_printf(seq, ",compress-force=%s", compress_type);
689 seq_printf(seq, ",compress=%s", compress_type);
691 if (btrfs_test_opt(root, NOSSD))
692 seq_puts(seq, ",nossd");
693 if (btrfs_test_opt(root, SSD_SPREAD))
694 seq_puts(seq, ",ssd_spread");
695 else if (btrfs_test_opt(root, SSD))
696 seq_puts(seq, ",ssd");
697 if (btrfs_test_opt(root, NOTREELOG))
698 seq_puts(seq, ",notreelog");
699 if (btrfs_test_opt(root, FLUSHONCOMMIT))
700 seq_puts(seq, ",flushoncommit");
701 if (btrfs_test_opt(root, DISCARD))
702 seq_puts(seq, ",discard");
703 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
704 seq_puts(seq, ",noacl");
705 if (btrfs_test_opt(root, SPACE_CACHE))
706 seq_puts(seq, ",space_cache");
708 seq_puts(seq, ",no_space_cache");
709 if (btrfs_test_opt(root, CLEAR_CACHE))
710 seq_puts(seq, ",clear_cache");
711 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
712 seq_puts(seq, ",user_subvol_rm_allowed");
713 if (btrfs_test_opt(root, ENOSPC_DEBUG))
714 seq_puts(seq, ",enospc_debug");
715 if (btrfs_test_opt(root, AUTO_DEFRAG))
716 seq_puts(seq, ",autodefrag");
717 if (btrfs_test_opt(root, INODE_MAP_CACHE))
718 seq_puts(seq, ",inode_cache");
722 static int btrfs_test_super(struct super_block *s, void *data)
724 struct btrfs_root *test_root = data;
725 struct btrfs_root *root = btrfs_sb(s);
728 * If this super block is going away, return false as it
729 * can't match as an existing super block.
731 if (!atomic_read(&s->s_active))
733 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
736 static int btrfs_set_super(struct super_block *s, void *data)
740 return set_anon_super(s, data);
744 * subvolumes are identified by ino 256
746 static inline int is_subvolume_inode(struct inode *inode)
748 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
754 * This will strip out the subvol=%s argument for an argument string and add
755 * subvolid=0 to make sure we get the actual tree root for path walking to the
758 static char *setup_root_args(char *args)
761 unsigned len = strlen(args) + 2;
766 * We need the same args as before, but minus
774 * which is a difference of 2 characters, so we allocate strlen(args) +
777 ret = kzalloc(len * sizeof(char), GFP_NOFS);
780 pos = strstr(args, "subvol=");
782 /* This shouldn't happen, but just in case.. */
789 * The subvol=<> arg is not at the front of the string, copy everybody
790 * up to that into ret.
795 copied += strlen(args);
799 strncpy(ret + copied, "subvolid=0", len - copied);
801 /* Length of subvolid=0 */
805 * If there is no , after the subvol= option then we know there's no
806 * other options and we can just return.
808 pos = strchr(pos, ',');
812 /* Copy the rest of the arguments into our buffer */
813 strncpy(ret + copied, pos, len - copied);
814 copied += strlen(pos);
819 static struct dentry *mount_subvol(const char *subvol_name, int flags,
820 const char *device_name, char *data)
822 struct super_block *s;
824 struct vfsmount *mnt;
825 struct mnt_namespace *ns_private;
830 newargs = setup_root_args(data);
832 return ERR_PTR(-ENOMEM);
833 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
837 return ERR_CAST(mnt);
839 ns_private = create_mnt_ns(mnt);
840 if (IS_ERR(ns_private)) {
842 return ERR_CAST(ns_private);
846 * This will trigger the automount of the subvol so we can just
847 * drop the mnt we have here and return the dentry that we
850 error = vfs_path_lookup(mnt->mnt_root, mnt, subvol_name,
851 LOOKUP_FOLLOW, &path);
852 put_mnt_ns(ns_private);
854 return ERR_PTR(error);
856 if (!is_subvolume_inode(path.dentry->d_inode)) {
860 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
862 return ERR_PTR(-EINVAL);
865 /* Get a ref to the sb and the dentry we found and return it */
866 s = path.mnt->mnt_sb;
867 atomic_inc(&s->s_active);
868 root = dget(path.dentry);
870 down_write(&s->s_umount);
876 * Find a superblock for the given device / mount point.
878 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
879 * for multiple device setup. Make sure to keep it in sync.
881 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
882 const char *device_name, void *data)
884 struct block_device *bdev = NULL;
885 struct super_block *s;
887 struct btrfs_fs_devices *fs_devices = NULL;
888 struct btrfs_root *tree_root = NULL;
889 struct btrfs_fs_info *fs_info = NULL;
890 fmode_t mode = FMODE_READ;
891 char *subvol_name = NULL;
892 u64 subvol_objectid = 0;
893 u64 subvol_rootid = 0;
896 if (!(flags & MS_RDONLY))
899 error = btrfs_parse_early_options(data, mode, fs_type,
900 &subvol_name, &subvol_objectid,
901 &subvol_rootid, &fs_devices);
903 return ERR_PTR(error);
906 root = mount_subvol(subvol_name, flags, device_name, data);
911 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
913 return ERR_PTR(error);
915 error = btrfs_open_devices(fs_devices, mode, fs_type);
917 return ERR_PTR(error);
919 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
921 goto error_close_devices;
925 * Setup a dummy root and fs_info for test/set super. This is because
926 * we don't actually fill this stuff out until open_ctree, but we need
927 * it for searching for existing supers, so this lets us do that and
928 * then open_ctree will properly initialize everything later.
930 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
931 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
932 if (!fs_info || !tree_root) {
934 goto error_close_devices;
936 fs_info->tree_root = tree_root;
937 fs_info->fs_devices = fs_devices;
938 tree_root->fs_info = fs_info;
940 bdev = fs_devices->latest_bdev;
941 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
944 goto error_close_devices;
948 if ((flags ^ s->s_flags) & MS_RDONLY) {
949 deactivate_locked_super(s);
950 return ERR_PTR(-EBUSY);
953 btrfs_close_devices(fs_devices);
957 char b[BDEVNAME_SIZE];
959 s->s_flags = flags | MS_NOSEC;
960 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
961 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
962 error = btrfs_fill_super(s, fs_devices, data,
963 flags & MS_SILENT ? 1 : 0);
965 deactivate_locked_super(s);
966 return ERR_PTR(error);
969 s->s_flags |= MS_ACTIVE;
972 root = get_default_root(s, subvol_objectid);
974 deactivate_locked_super(s);
981 btrfs_close_devices(fs_devices);
984 return ERR_PTR(error);
987 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
989 struct btrfs_root *root = btrfs_sb(sb);
992 ret = btrfs_parse_options(root, data);
996 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
999 if (*flags & MS_RDONLY) {
1000 sb->s_flags |= MS_RDONLY;
1002 ret = btrfs_commit_super(root);
1005 if (root->fs_info->fs_devices->rw_devices == 0)
1008 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
1011 ret = btrfs_cleanup_fs_roots(root->fs_info);
1014 /* recover relocation */
1015 ret = btrfs_recover_relocation(root);
1018 sb->s_flags &= ~MS_RDONLY;
1024 /* Used to sort the devices by max_avail(descending sort) */
1025 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1026 const void *dev_info2)
1028 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1029 ((struct btrfs_device_info *)dev_info2)->max_avail)
1031 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1032 ((struct btrfs_device_info *)dev_info2)->max_avail)
1039 * sort the devices by max_avail, in which max free extent size of each device
1040 * is stored.(Descending Sort)
1042 static inline void btrfs_descending_sort_devices(
1043 struct btrfs_device_info *devices,
1046 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1047 btrfs_cmp_device_free_bytes, NULL);
1051 * The helper to calc the free space on the devices that can be used to store
1054 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1056 struct btrfs_fs_info *fs_info = root->fs_info;
1057 struct btrfs_device_info *devices_info;
1058 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1059 struct btrfs_device *device;
1064 u64 min_stripe_size;
1065 int min_stripes = 1;
1066 int i = 0, nr_devices;
1069 nr_devices = fs_info->fs_devices->rw_devices;
1070 BUG_ON(!nr_devices);
1072 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1077 /* calc min stripe number for data space alloction */
1078 type = btrfs_get_alloc_profile(root, 1);
1079 if (type & BTRFS_BLOCK_GROUP_RAID0)
1081 else if (type & BTRFS_BLOCK_GROUP_RAID1)
1083 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1086 if (type & BTRFS_BLOCK_GROUP_DUP)
1087 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1089 min_stripe_size = BTRFS_STRIPE_LEN;
1091 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
1092 if (!device->in_fs_metadata)
1095 avail_space = device->total_bytes - device->bytes_used;
1097 /* align with stripe_len */
1098 do_div(avail_space, BTRFS_STRIPE_LEN);
1099 avail_space *= BTRFS_STRIPE_LEN;
1102 * In order to avoid overwritting the superblock on the drive,
1103 * btrfs starts at an offset of at least 1MB when doing chunk
1106 skip_space = 1024 * 1024;
1108 /* user can set the offset in fs_info->alloc_start. */
1109 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1110 device->total_bytes)
1111 skip_space = max(fs_info->alloc_start, skip_space);
1114 * btrfs can not use the free space in [0, skip_space - 1],
1115 * we must subtract it from the total. In order to implement
1116 * it, we account the used space in this range first.
1118 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1121 kfree(devices_info);
1125 /* calc the free space in [0, skip_space - 1] */
1126 skip_space -= used_space;
1129 * we can use the free space in [0, skip_space - 1], subtract
1130 * it from the total.
1132 if (avail_space && avail_space >= skip_space)
1133 avail_space -= skip_space;
1137 if (avail_space < min_stripe_size)
1140 devices_info[i].dev = device;
1141 devices_info[i].max_avail = avail_space;
1148 btrfs_descending_sort_devices(devices_info, nr_devices);
1152 while (nr_devices >= min_stripes) {
1153 if (devices_info[i].max_avail >= min_stripe_size) {
1157 avail_space += devices_info[i].max_avail * min_stripes;
1158 alloc_size = devices_info[i].max_avail;
1159 for (j = i + 1 - min_stripes; j <= i; j++)
1160 devices_info[j].max_avail -= alloc_size;
1166 kfree(devices_info);
1167 *free_bytes = avail_space;
1171 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1173 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1174 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
1175 struct list_head *head = &root->fs_info->space_info;
1176 struct btrfs_space_info *found;
1178 u64 total_free_data = 0;
1179 int bits = dentry->d_sb->s_blocksize_bits;
1180 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1183 /* holding chunk_muext to avoid allocating new chunks */
1184 mutex_lock(&root->fs_info->chunk_mutex);
1186 list_for_each_entry_rcu(found, head, list) {
1187 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1188 total_free_data += found->disk_total - found->disk_used;
1190 btrfs_account_ro_block_groups_free_space(found);
1193 total_used += found->disk_used;
1197 buf->f_namelen = BTRFS_NAME_LEN;
1198 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1199 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1200 buf->f_bsize = dentry->d_sb->s_blocksize;
1201 buf->f_type = BTRFS_SUPER_MAGIC;
1202 buf->f_bavail = total_free_data;
1203 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1205 mutex_unlock(&root->fs_info->chunk_mutex);
1208 buf->f_bavail += total_free_data;
1209 buf->f_bavail = buf->f_bavail >> bits;
1210 mutex_unlock(&root->fs_info->chunk_mutex);
1212 /* We treat it as constant endianness (it doesn't matter _which_)
1213 because we want the fsid to come out the same whether mounted
1214 on a big-endian or little-endian host */
1215 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1216 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1217 /* Mask in the root object ID too, to disambiguate subvols */
1218 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1219 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1224 static struct file_system_type btrfs_fs_type = {
1225 .owner = THIS_MODULE,
1227 .mount = btrfs_mount,
1228 .kill_sb = kill_anon_super,
1229 .fs_flags = FS_REQUIRES_DEV,
1233 * used by btrfsctl to scan devices when no FS is mounted
1235 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1238 struct btrfs_ioctl_vol_args *vol;
1239 struct btrfs_fs_devices *fs_devices;
1242 if (!capable(CAP_SYS_ADMIN))
1245 vol = memdup_user((void __user *)arg, sizeof(*vol));
1247 return PTR_ERR(vol);
1250 case BTRFS_IOC_SCAN_DEV:
1251 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1252 &btrfs_fs_type, &fs_devices);
1260 static int btrfs_freeze(struct super_block *sb)
1262 struct btrfs_root *root = btrfs_sb(sb);
1263 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1264 mutex_lock(&root->fs_info->cleaner_mutex);
1268 static int btrfs_unfreeze(struct super_block *sb)
1270 struct btrfs_root *root = btrfs_sb(sb);
1271 mutex_unlock(&root->fs_info->cleaner_mutex);
1272 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1276 static const struct super_operations btrfs_super_ops = {
1277 .drop_inode = btrfs_drop_inode,
1278 .evict_inode = btrfs_evict_inode,
1279 .put_super = btrfs_put_super,
1280 .sync_fs = btrfs_sync_fs,
1281 .show_options = btrfs_show_options,
1282 .write_inode = btrfs_write_inode,
1283 .dirty_inode = btrfs_dirty_inode,
1284 .alloc_inode = btrfs_alloc_inode,
1285 .destroy_inode = btrfs_destroy_inode,
1286 .statfs = btrfs_statfs,
1287 .remount_fs = btrfs_remount,
1288 .freeze_fs = btrfs_freeze,
1289 .unfreeze_fs = btrfs_unfreeze,
1292 static const struct file_operations btrfs_ctl_fops = {
1293 .unlocked_ioctl = btrfs_control_ioctl,
1294 .compat_ioctl = btrfs_control_ioctl,
1295 .owner = THIS_MODULE,
1296 .llseek = noop_llseek,
1299 static struct miscdevice btrfs_misc = {
1300 .minor = BTRFS_MINOR,
1301 .name = "btrfs-control",
1302 .fops = &btrfs_ctl_fops
1305 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1306 MODULE_ALIAS("devname:btrfs-control");
1308 static int btrfs_interface_init(void)
1310 return misc_register(&btrfs_misc);
1313 static void btrfs_interface_exit(void)
1315 if (misc_deregister(&btrfs_misc) < 0)
1316 printk(KERN_INFO "misc_deregister failed for control device");
1319 static int __init init_btrfs_fs(void)
1323 err = btrfs_init_sysfs();
1327 err = btrfs_init_compress();
1331 err = btrfs_init_cachep();
1335 err = extent_io_init();
1339 err = extent_map_init();
1341 goto free_extent_io;
1343 err = btrfs_delayed_inode_init();
1345 goto free_extent_map;
1347 err = btrfs_interface_init();
1349 goto free_delayed_inode;
1351 err = register_filesystem(&btrfs_fs_type);
1353 goto unregister_ioctl;
1355 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1359 btrfs_interface_exit();
1361 btrfs_delayed_inode_exit();
1367 btrfs_destroy_cachep();
1369 btrfs_exit_compress();
1375 static void __exit exit_btrfs_fs(void)
1377 btrfs_destroy_cachep();
1378 btrfs_delayed_inode_exit();
1381 btrfs_interface_exit();
1382 unregister_filesystem(&btrfs_fs_type);
1384 btrfs_cleanup_fs_uuids();
1385 btrfs_exit_compress();
1388 module_init(init_btrfs_fs)
1389 module_exit(exit_btrfs_fs)
1391 MODULE_LICENSE("GPL");