xref: /linux/fs/btrfs/super.c (revision a1087ef6abedf0bfd60e5e3fddf33192cb2c1325)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
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.
7  *
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.
12  *
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.
17  */
18 
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
22 #include <linux/fs.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 "compat.h"
43 #include "ctree.h"
44 #include "disk-io.h"
45 #include "transaction.h"
46 #include "btrfs_inode.h"
47 #include "ioctl.h"
48 #include "print-tree.h"
49 #include "xattr.h"
50 #include "volumes.h"
51 #include "version.h"
52 #include "export.h"
53 #include "compression.h"
54 
55 static const struct super_operations btrfs_super_ops;
56 
57 static void btrfs_put_super(struct super_block *sb)
58 {
59 	struct btrfs_root *root = btrfs_sb(sb);
60 	int ret;
61 
62 	ret = close_ctree(root);
63 	sb->s_fs_info = NULL;
64 
65 	(void)ret; /* FIXME: need to fix VFS to return error? */
66 }
67 
68 enum {
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,
74 	Opt_user_subvol_rm_allowed,
75 };
76 
77 static match_table_t tokens = {
78 	{Opt_degraded, "degraded"},
79 	{Opt_subvol, "subvol=%s"},
80 	{Opt_subvolid, "subvolid=%d"},
81 	{Opt_device, "device=%s"},
82 	{Opt_nodatasum, "nodatasum"},
83 	{Opt_nodatacow, "nodatacow"},
84 	{Opt_nobarrier, "nobarrier"},
85 	{Opt_max_inline, "max_inline=%s"},
86 	{Opt_alloc_start, "alloc_start=%s"},
87 	{Opt_thread_pool, "thread_pool=%d"},
88 	{Opt_compress, "compress"},
89 	{Opt_compress_force, "compress-force"},
90 	{Opt_ssd, "ssd"},
91 	{Opt_ssd_spread, "ssd_spread"},
92 	{Opt_nossd, "nossd"},
93 	{Opt_noacl, "noacl"},
94 	{Opt_notreelog, "notreelog"},
95 	{Opt_flushoncommit, "flushoncommit"},
96 	{Opt_ratio, "metadata_ratio=%d"},
97 	{Opt_discard, "discard"},
98 	{Opt_space_cache, "space_cache"},
99 	{Opt_clear_cache, "clear_cache"},
100 	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
101 	{Opt_err, NULL},
102 };
103 
104 /*
105  * Regular mount options parser.  Everything that is needed only when
106  * reading in a new superblock is parsed here.
107  */
108 int btrfs_parse_options(struct btrfs_root *root, char *options)
109 {
110 	struct btrfs_fs_info *info = root->fs_info;
111 	substring_t args[MAX_OPT_ARGS];
112 	char *p, *num, *orig;
113 	int intarg;
114 	int ret = 0;
115 
116 	if (!options)
117 		return 0;
118 
119 	/*
120 	 * strsep changes the string, duplicate it because parse_options
121 	 * gets called twice
122 	 */
123 	options = kstrdup(options, GFP_NOFS);
124 	if (!options)
125 		return -ENOMEM;
126 
127 	orig = options;
128 
129 	while ((p = strsep(&options, ",")) != NULL) {
130 		int token;
131 		if (!*p)
132 			continue;
133 
134 		token = match_token(p, tokens, args);
135 		switch (token) {
136 		case Opt_degraded:
137 			printk(KERN_INFO "btrfs: allowing degraded mounts\n");
138 			btrfs_set_opt(info->mount_opt, DEGRADED);
139 			break;
140 		case Opt_subvol:
141 		case Opt_subvolid:
142 		case Opt_device:
143 			/*
144 			 * These are parsed by btrfs_parse_early_options
145 			 * and can be happily ignored here.
146 			 */
147 			break;
148 		case Opt_nodatasum:
149 			printk(KERN_INFO "btrfs: setting nodatasum\n");
150 			btrfs_set_opt(info->mount_opt, NODATASUM);
151 			break;
152 		case Opt_nodatacow:
153 			printk(KERN_INFO "btrfs: setting nodatacow\n");
154 			btrfs_set_opt(info->mount_opt, NODATACOW);
155 			btrfs_set_opt(info->mount_opt, NODATASUM);
156 			break;
157 		case Opt_compress:
158 			printk(KERN_INFO "btrfs: use compression\n");
159 			btrfs_set_opt(info->mount_opt, COMPRESS);
160 			break;
161 		case Opt_compress_force:
162 			printk(KERN_INFO "btrfs: forcing compression\n");
163 			btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
164 			btrfs_set_opt(info->mount_opt, COMPRESS);
165 			break;
166 		case Opt_ssd:
167 			printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
168 			btrfs_set_opt(info->mount_opt, SSD);
169 			break;
170 		case Opt_ssd_spread:
171 			printk(KERN_INFO "btrfs: use spread ssd "
172 			       "allocation scheme\n");
173 			btrfs_set_opt(info->mount_opt, SSD);
174 			btrfs_set_opt(info->mount_opt, SSD_SPREAD);
175 			break;
176 		case Opt_nossd:
177 			printk(KERN_INFO "btrfs: not using ssd allocation "
178 			       "scheme\n");
179 			btrfs_set_opt(info->mount_opt, NOSSD);
180 			btrfs_clear_opt(info->mount_opt, SSD);
181 			btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
182 			break;
183 		case Opt_nobarrier:
184 			printk(KERN_INFO "btrfs: turning off barriers\n");
185 			btrfs_set_opt(info->mount_opt, NOBARRIER);
186 			break;
187 		case Opt_thread_pool:
188 			intarg = 0;
189 			match_int(&args[0], &intarg);
190 			if (intarg) {
191 				info->thread_pool_size = intarg;
192 				printk(KERN_INFO "btrfs: thread pool %d\n",
193 				       info->thread_pool_size);
194 			}
195 			break;
196 		case Opt_max_inline:
197 			num = match_strdup(&args[0]);
198 			if (num) {
199 				info->max_inline = memparse(num, NULL);
200 				kfree(num);
201 
202 				if (info->max_inline) {
203 					info->max_inline = max_t(u64,
204 						info->max_inline,
205 						root->sectorsize);
206 				}
207 				printk(KERN_INFO "btrfs: max_inline at %llu\n",
208 					(unsigned long long)info->max_inline);
209 			}
210 			break;
211 		case Opt_alloc_start:
212 			num = match_strdup(&args[0]);
213 			if (num) {
214 				info->alloc_start = memparse(num, NULL);
215 				kfree(num);
216 				printk(KERN_INFO
217 					"btrfs: allocations start at %llu\n",
218 					(unsigned long long)info->alloc_start);
219 			}
220 			break;
221 		case Opt_noacl:
222 			root->fs_info->sb->s_flags &= ~MS_POSIXACL;
223 			break;
224 		case Opt_notreelog:
225 			printk(KERN_INFO "btrfs: disabling tree log\n");
226 			btrfs_set_opt(info->mount_opt, NOTREELOG);
227 			break;
228 		case Opt_flushoncommit:
229 			printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
230 			btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
231 			break;
232 		case Opt_ratio:
233 			intarg = 0;
234 			match_int(&args[0], &intarg);
235 			if (intarg) {
236 				info->metadata_ratio = intarg;
237 				printk(KERN_INFO "btrfs: metadata ratio %d\n",
238 				       info->metadata_ratio);
239 			}
240 			break;
241 		case Opt_discard:
242 			btrfs_set_opt(info->mount_opt, DISCARD);
243 			break;
244 		case Opt_space_cache:
245 			printk(KERN_INFO "btrfs: enabling disk space caching\n");
246 			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
247 		case Opt_clear_cache:
248 			printk(KERN_INFO "btrfs: force clearing of disk cache\n");
249 			btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
250 			break;
251 		case Opt_user_subvol_rm_allowed:
252 			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
253 			break;
254 		case Opt_err:
255 			printk(KERN_INFO "btrfs: unrecognized mount option "
256 			       "'%s'\n", p);
257 			ret = -EINVAL;
258 			goto out;
259 		default:
260 			break;
261 		}
262 	}
263 out:
264 	kfree(orig);
265 	return ret;
266 }
267 
268 /*
269  * Parse mount options that are required early in the mount process.
270  *
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.
273  */
274 static int btrfs_parse_early_options(const char *options, fmode_t flags,
275 		void *holder, char **subvol_name, u64 *subvol_objectid,
276 		struct btrfs_fs_devices **fs_devices)
277 {
278 	substring_t args[MAX_OPT_ARGS];
279 	char *opts, *p;
280 	int error = 0;
281 	int intarg;
282 
283 	if (!options)
284 		goto out;
285 
286 	/*
287 	 * strsep changes the string, duplicate it because parse_options
288 	 * gets called twice
289 	 */
290 	opts = kstrdup(options, GFP_KERNEL);
291 	if (!opts)
292 		return -ENOMEM;
293 
294 	while ((p = strsep(&opts, ",")) != NULL) {
295 		int token;
296 		if (!*p)
297 			continue;
298 
299 		token = match_token(p, tokens, args);
300 		switch (token) {
301 		case Opt_subvol:
302 			*subvol_name = match_strdup(&args[0]);
303 			break;
304 		case Opt_subvolid:
305 			intarg = 0;
306 			error = match_int(&args[0], &intarg);
307 			if (!error) {
308 				/* we want the original fs_tree */
309 				if (!intarg)
310 					*subvol_objectid =
311 						BTRFS_FS_TREE_OBJECTID;
312 				else
313 					*subvol_objectid = intarg;
314 			}
315 			break;
316 		case Opt_device:
317 			error = btrfs_scan_one_device(match_strdup(&args[0]),
318 					flags, holder, fs_devices);
319 			if (error)
320 				goto out_free_opts;
321 			break;
322 		default:
323 			break;
324 		}
325 	}
326 
327  out_free_opts:
328 	kfree(opts);
329  out:
330 	/*
331 	 * If no subvolume name is specified we use the default one.  Allocate
332 	 * a copy of the string "." here so that code later in the
333 	 * mount path doesn't care if it's the default volume or another one.
334 	 */
335 	if (!*subvol_name) {
336 		*subvol_name = kstrdup(".", GFP_KERNEL);
337 		if (!*subvol_name)
338 			return -ENOMEM;
339 	}
340 	return error;
341 }
342 
343 static struct dentry *get_default_root(struct super_block *sb,
344 				       u64 subvol_objectid)
345 {
346 	struct btrfs_root *root = sb->s_fs_info;
347 	struct btrfs_root *new_root;
348 	struct btrfs_dir_item *di;
349 	struct btrfs_path *path;
350 	struct btrfs_key location;
351 	struct inode *inode;
352 	struct dentry *dentry;
353 	u64 dir_id;
354 	int new = 0;
355 
356 	/*
357 	 * We have a specific subvol we want to mount, just setup location and
358 	 * go look up the root.
359 	 */
360 	if (subvol_objectid) {
361 		location.objectid = subvol_objectid;
362 		location.type = BTRFS_ROOT_ITEM_KEY;
363 		location.offset = (u64)-1;
364 		goto find_root;
365 	}
366 
367 	path = btrfs_alloc_path();
368 	if (!path)
369 		return ERR_PTR(-ENOMEM);
370 	path->leave_spinning = 1;
371 
372 	/*
373 	 * Find the "default" dir item which points to the root item that we
374 	 * will mount by default if we haven't been given a specific subvolume
375 	 * to mount.
376 	 */
377 	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
378 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
379 	if (IS_ERR(di))
380 		return ERR_CAST(di);
381 	if (!di) {
382 		/*
383 		 * Ok the default dir item isn't there.  This is weird since
384 		 * it's always been there, but don't freak out, just try and
385 		 * mount to root most subvolume.
386 		 */
387 		btrfs_free_path(path);
388 		dir_id = BTRFS_FIRST_FREE_OBJECTID;
389 		new_root = root->fs_info->fs_root;
390 		goto setup_root;
391 	}
392 
393 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
394 	btrfs_free_path(path);
395 
396 find_root:
397 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
398 	if (IS_ERR(new_root))
399 		return ERR_CAST(new_root);
400 
401 	if (btrfs_root_refs(&new_root->root_item) == 0)
402 		return ERR_PTR(-ENOENT);
403 
404 	dir_id = btrfs_root_dirid(&new_root->root_item);
405 setup_root:
406 	location.objectid = dir_id;
407 	location.type = BTRFS_INODE_ITEM_KEY;
408 	location.offset = 0;
409 
410 	inode = btrfs_iget(sb, &location, new_root, &new);
411 	if (IS_ERR(inode))
412 		return ERR_CAST(inode);
413 
414 	/*
415 	 * If we're just mounting the root most subvol put the inode and return
416 	 * a reference to the dentry.  We will have already gotten a reference
417 	 * to the inode in btrfs_fill_super so we're good to go.
418 	 */
419 	if (!new && sb->s_root->d_inode == inode) {
420 		iput(inode);
421 		return dget(sb->s_root);
422 	}
423 
424 	if (new) {
425 		const struct qstr name = { .name = "/", .len = 1 };
426 
427 		/*
428 		 * New inode, we need to make the dentry a sibling of s_root so
429 		 * everything gets cleaned up properly on unmount.
430 		 */
431 		dentry = d_alloc(sb->s_root, &name);
432 		if (!dentry) {
433 			iput(inode);
434 			return ERR_PTR(-ENOMEM);
435 		}
436 		d_splice_alias(inode, dentry);
437 	} else {
438 		/*
439 		 * We found the inode in cache, just find a dentry for it and
440 		 * put the reference to the inode we just got.
441 		 */
442 		dentry = d_find_alias(inode);
443 		iput(inode);
444 	}
445 
446 	return dentry;
447 }
448 
449 static int btrfs_fill_super(struct super_block *sb,
450 			    struct btrfs_fs_devices *fs_devices,
451 			    void *data, int silent)
452 {
453 	struct inode *inode;
454 	struct dentry *root_dentry;
455 	struct btrfs_root *tree_root;
456 	struct btrfs_key key;
457 	int err;
458 
459 	sb->s_maxbytes = MAX_LFS_FILESIZE;
460 	sb->s_magic = BTRFS_SUPER_MAGIC;
461 	sb->s_op = &btrfs_super_ops;
462 	sb->s_export_op = &btrfs_export_ops;
463 	sb->s_xattr = btrfs_xattr_handlers;
464 	sb->s_time_gran = 1;
465 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
466 	sb->s_flags |= MS_POSIXACL;
467 #endif
468 
469 	tree_root = open_ctree(sb, fs_devices, (char *)data);
470 
471 	if (IS_ERR(tree_root)) {
472 		printk("btrfs: open_ctree failed\n");
473 		return PTR_ERR(tree_root);
474 	}
475 	sb->s_fs_info = tree_root;
476 
477 	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
478 	key.type = BTRFS_INODE_ITEM_KEY;
479 	key.offset = 0;
480 	inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
481 	if (IS_ERR(inode)) {
482 		err = PTR_ERR(inode);
483 		goto fail_close;
484 	}
485 
486 	root_dentry = d_alloc_root(inode);
487 	if (!root_dentry) {
488 		iput(inode);
489 		err = -ENOMEM;
490 		goto fail_close;
491 	}
492 
493 	sb->s_root = root_dentry;
494 
495 	save_mount_options(sb, data);
496 	return 0;
497 
498 fail_close:
499 	close_ctree(tree_root);
500 	return err;
501 }
502 
503 int btrfs_sync_fs(struct super_block *sb, int wait)
504 {
505 	struct btrfs_trans_handle *trans;
506 	struct btrfs_root *root = btrfs_sb(sb);
507 	int ret;
508 
509 	if (!wait) {
510 		filemap_flush(root->fs_info->btree_inode->i_mapping);
511 		return 0;
512 	}
513 
514 	btrfs_start_delalloc_inodes(root, 0);
515 	btrfs_wait_ordered_extents(root, 0, 0);
516 
517 	trans = btrfs_start_transaction(root, 0);
518 	ret = btrfs_commit_transaction(trans, root);
519 	return ret;
520 }
521 
522 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
523 {
524 	struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
525 	struct btrfs_fs_info *info = root->fs_info;
526 
527 	if (btrfs_test_opt(root, DEGRADED))
528 		seq_puts(seq, ",degraded");
529 	if (btrfs_test_opt(root, NODATASUM))
530 		seq_puts(seq, ",nodatasum");
531 	if (btrfs_test_opt(root, NODATACOW))
532 		seq_puts(seq, ",nodatacow");
533 	if (btrfs_test_opt(root, NOBARRIER))
534 		seq_puts(seq, ",nobarrier");
535 	if (info->max_inline != 8192 * 1024)
536 		seq_printf(seq, ",max_inline=%llu",
537 			   (unsigned long long)info->max_inline);
538 	if (info->alloc_start != 0)
539 		seq_printf(seq, ",alloc_start=%llu",
540 			   (unsigned long long)info->alloc_start);
541 	if (info->thread_pool_size !=  min_t(unsigned long,
542 					     num_online_cpus() + 2, 8))
543 		seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
544 	if (btrfs_test_opt(root, COMPRESS))
545 		seq_puts(seq, ",compress");
546 	if (btrfs_test_opt(root, NOSSD))
547 		seq_puts(seq, ",nossd");
548 	if (btrfs_test_opt(root, SSD_SPREAD))
549 		seq_puts(seq, ",ssd_spread");
550 	else if (btrfs_test_opt(root, SSD))
551 		seq_puts(seq, ",ssd");
552 	if (btrfs_test_opt(root, NOTREELOG))
553 		seq_puts(seq, ",notreelog");
554 	if (btrfs_test_opt(root, FLUSHONCOMMIT))
555 		seq_puts(seq, ",flushoncommit");
556 	if (btrfs_test_opt(root, DISCARD))
557 		seq_puts(seq, ",discard");
558 	if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
559 		seq_puts(seq, ",noacl");
560 	return 0;
561 }
562 
563 static int btrfs_test_super(struct super_block *s, void *data)
564 {
565 	struct btrfs_fs_devices *test_fs_devices = data;
566 	struct btrfs_root *root = btrfs_sb(s);
567 
568 	return root->fs_info->fs_devices == test_fs_devices;
569 }
570 
571 /*
572  * Find a superblock for the given device / mount point.
573  *
574  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
575  *	  for multiple device setup.  Make sure to keep it in sync.
576  */
577 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
578 		const char *dev_name, void *data)
579 {
580 	struct block_device *bdev = NULL;
581 	struct super_block *s;
582 	struct dentry *root;
583 	struct btrfs_fs_devices *fs_devices = NULL;
584 	fmode_t mode = FMODE_READ;
585 	char *subvol_name = NULL;
586 	u64 subvol_objectid = 0;
587 	int error = 0;
588 
589 	if (!(flags & MS_RDONLY))
590 		mode |= FMODE_WRITE;
591 
592 	error = btrfs_parse_early_options(data, mode, fs_type,
593 					  &subvol_name, &subvol_objectid,
594 					  &fs_devices);
595 	if (error)
596 		return ERR_PTR(error);
597 
598 	error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
599 	if (error)
600 		goto error_free_subvol_name;
601 
602 	error = btrfs_open_devices(fs_devices, mode, fs_type);
603 	if (error)
604 		goto error_free_subvol_name;
605 
606 	if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
607 		error = -EACCES;
608 		goto error_close_devices;
609 	}
610 
611 	bdev = fs_devices->latest_bdev;
612 	s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
613 	if (IS_ERR(s))
614 		goto error_s;
615 
616 	if (s->s_root) {
617 		if ((flags ^ s->s_flags) & MS_RDONLY) {
618 			deactivate_locked_super(s);
619 			error = -EBUSY;
620 			goto error_close_devices;
621 		}
622 
623 		btrfs_close_devices(fs_devices);
624 	} else {
625 		char b[BDEVNAME_SIZE];
626 
627 		s->s_flags = flags;
628 		strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
629 		error = btrfs_fill_super(s, fs_devices, data,
630 					 flags & MS_SILENT ? 1 : 0);
631 		if (error) {
632 			deactivate_locked_super(s);
633 			goto error_free_subvol_name;
634 		}
635 
636 		btrfs_sb(s)->fs_info->bdev_holder = fs_type;
637 		s->s_flags |= MS_ACTIVE;
638 	}
639 
640 	root = get_default_root(s, subvol_objectid);
641 	if (IS_ERR(root)) {
642 		error = PTR_ERR(root);
643 		deactivate_locked_super(s);
644 		goto error_free_subvol_name;
645 	}
646 	/* if they gave us a subvolume name bind mount into that */
647 	if (strcmp(subvol_name, ".")) {
648 		struct dentry *new_root;
649 		mutex_lock(&root->d_inode->i_mutex);
650 		new_root = lookup_one_len(subvol_name, root,
651 				      strlen(subvol_name));
652 		mutex_unlock(&root->d_inode->i_mutex);
653 
654 		if (IS_ERR(new_root)) {
655 			deactivate_locked_super(s);
656 			error = PTR_ERR(new_root);
657 			dput(root);
658 			goto error_free_subvol_name;
659 		}
660 		if (!new_root->d_inode) {
661 			dput(root);
662 			dput(new_root);
663 			deactivate_locked_super(s);
664 			error = -ENXIO;
665 			goto error_free_subvol_name;
666 		}
667 		dput(root);
668 		root = new_root;
669 	}
670 
671 	kfree(subvol_name);
672 	return root;
673 
674 error_s:
675 	error = PTR_ERR(s);
676 error_close_devices:
677 	btrfs_close_devices(fs_devices);
678 error_free_subvol_name:
679 	kfree(subvol_name);
680 	return ERR_PTR(error);
681 }
682 
683 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
684 {
685 	struct btrfs_root *root = btrfs_sb(sb);
686 	int ret;
687 
688 	ret = btrfs_parse_options(root, data);
689 	if (ret)
690 		return -EINVAL;
691 
692 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
693 		return 0;
694 
695 	if (*flags & MS_RDONLY) {
696 		sb->s_flags |= MS_RDONLY;
697 
698 		ret =  btrfs_commit_super(root);
699 		WARN_ON(ret);
700 	} else {
701 		if (root->fs_info->fs_devices->rw_devices == 0)
702 			return -EACCES;
703 
704 		if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
705 			return -EINVAL;
706 
707 		ret = btrfs_cleanup_fs_roots(root->fs_info);
708 		WARN_ON(ret);
709 
710 		/* recover relocation */
711 		ret = btrfs_recover_relocation(root);
712 		WARN_ON(ret);
713 
714 		sb->s_flags &= ~MS_RDONLY;
715 	}
716 
717 	return 0;
718 }
719 
720 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
721 {
722 	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
723 	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
724 	struct list_head *head = &root->fs_info->space_info;
725 	struct btrfs_space_info *found;
726 	u64 total_used = 0;
727 	u64 total_used_data = 0;
728 	int bits = dentry->d_sb->s_blocksize_bits;
729 	__be32 *fsid = (__be32 *)root->fs_info->fsid;
730 
731 	rcu_read_lock();
732 	list_for_each_entry_rcu(found, head, list) {
733 		if (found->flags & (BTRFS_BLOCK_GROUP_METADATA |
734 				    BTRFS_BLOCK_GROUP_SYSTEM))
735 			total_used_data += found->disk_total;
736 		else
737 			total_used_data += found->disk_used;
738 		total_used += found->disk_used;
739 	}
740 	rcu_read_unlock();
741 
742 	buf->f_namelen = BTRFS_NAME_LEN;
743 	buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
744 	buf->f_bfree = buf->f_blocks - (total_used >> bits);
745 	buf->f_bavail = buf->f_blocks - (total_used_data >> bits);
746 	buf->f_bsize = dentry->d_sb->s_blocksize;
747 	buf->f_type = BTRFS_SUPER_MAGIC;
748 
749 	/* We treat it as constant endianness (it doesn't matter _which_)
750 	   because we want the fsid to come out the same whether mounted
751 	   on a big-endian or little-endian host */
752 	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
753 	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
754 	/* Mask in the root object ID too, to disambiguate subvols */
755 	buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
756 	buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
757 
758 	return 0;
759 }
760 
761 static struct file_system_type btrfs_fs_type = {
762 	.owner		= THIS_MODULE,
763 	.name		= "btrfs",
764 	.mount		= btrfs_mount,
765 	.kill_sb	= kill_anon_super,
766 	.fs_flags	= FS_REQUIRES_DEV,
767 };
768 
769 /*
770  * used by btrfsctl to scan devices when no FS is mounted
771  */
772 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
773 				unsigned long arg)
774 {
775 	struct btrfs_ioctl_vol_args *vol;
776 	struct btrfs_fs_devices *fs_devices;
777 	int ret = -ENOTTY;
778 
779 	if (!capable(CAP_SYS_ADMIN))
780 		return -EPERM;
781 
782 	vol = memdup_user((void __user *)arg, sizeof(*vol));
783 	if (IS_ERR(vol))
784 		return PTR_ERR(vol);
785 
786 	switch (cmd) {
787 	case BTRFS_IOC_SCAN_DEV:
788 		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
789 					    &btrfs_fs_type, &fs_devices);
790 		break;
791 	}
792 
793 	kfree(vol);
794 	return ret;
795 }
796 
797 static int btrfs_freeze(struct super_block *sb)
798 {
799 	struct btrfs_root *root = btrfs_sb(sb);
800 	mutex_lock(&root->fs_info->transaction_kthread_mutex);
801 	mutex_lock(&root->fs_info->cleaner_mutex);
802 	return 0;
803 }
804 
805 static int btrfs_unfreeze(struct super_block *sb)
806 {
807 	struct btrfs_root *root = btrfs_sb(sb);
808 	mutex_unlock(&root->fs_info->cleaner_mutex);
809 	mutex_unlock(&root->fs_info->transaction_kthread_mutex);
810 	return 0;
811 }
812 
813 static const struct super_operations btrfs_super_ops = {
814 	.drop_inode	= btrfs_drop_inode,
815 	.evict_inode	= btrfs_evict_inode,
816 	.put_super	= btrfs_put_super,
817 	.sync_fs	= btrfs_sync_fs,
818 	.show_options	= btrfs_show_options,
819 	.write_inode	= btrfs_write_inode,
820 	.dirty_inode	= btrfs_dirty_inode,
821 	.alloc_inode	= btrfs_alloc_inode,
822 	.destroy_inode	= btrfs_destroy_inode,
823 	.statfs		= btrfs_statfs,
824 	.remount_fs	= btrfs_remount,
825 	.freeze_fs	= btrfs_freeze,
826 	.unfreeze_fs	= btrfs_unfreeze,
827 };
828 
829 static const struct file_operations btrfs_ctl_fops = {
830 	.unlocked_ioctl	 = btrfs_control_ioctl,
831 	.compat_ioctl = btrfs_control_ioctl,
832 	.owner	 = THIS_MODULE,
833 	.llseek = noop_llseek,
834 };
835 
836 static struct miscdevice btrfs_misc = {
837 	.minor		= BTRFS_MINOR,
838 	.name		= "btrfs-control",
839 	.fops		= &btrfs_ctl_fops
840 };
841 
842 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
843 MODULE_ALIAS("devname:btrfs-control");
844 
845 static int btrfs_interface_init(void)
846 {
847 	return misc_register(&btrfs_misc);
848 }
849 
850 static void btrfs_interface_exit(void)
851 {
852 	if (misc_deregister(&btrfs_misc) < 0)
853 		printk(KERN_INFO "misc_deregister failed for control device");
854 }
855 
856 static int __init init_btrfs_fs(void)
857 {
858 	int err;
859 
860 	err = btrfs_init_sysfs();
861 	if (err)
862 		return err;
863 
864 	err = btrfs_init_cachep();
865 	if (err)
866 		goto free_sysfs;
867 
868 	err = extent_io_init();
869 	if (err)
870 		goto free_cachep;
871 
872 	err = extent_map_init();
873 	if (err)
874 		goto free_extent_io;
875 
876 	err = btrfs_interface_init();
877 	if (err)
878 		goto free_extent_map;
879 
880 	err = register_filesystem(&btrfs_fs_type);
881 	if (err)
882 		goto unregister_ioctl;
883 
884 	printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
885 	return 0;
886 
887 unregister_ioctl:
888 	btrfs_interface_exit();
889 free_extent_map:
890 	extent_map_exit();
891 free_extent_io:
892 	extent_io_exit();
893 free_cachep:
894 	btrfs_destroy_cachep();
895 free_sysfs:
896 	btrfs_exit_sysfs();
897 	return err;
898 }
899 
900 static void __exit exit_btrfs_fs(void)
901 {
902 	btrfs_destroy_cachep();
903 	extent_map_exit();
904 	extent_io_exit();
905 	btrfs_interface_exit();
906 	unregister_filesystem(&btrfs_fs_type);
907 	btrfs_exit_sysfs();
908 	btrfs_cleanup_fs_uuids();
909 	btrfs_zlib_exit();
910 }
911 
912 module_init(init_btrfs_fs)
913 module_exit(exit_btrfs_fs)
914 
915 MODULE_LICENSE("GPL");
916