Btrfs: add a log of past tree roots
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / btrfs / super.c
blob57080dffdfc6f101246dfa222f26ecfdc6f4b818
1 /*
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>
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 <linux/cleancache.h>
43 #include <linux/mnt_namespace.h>
44 #include "compat.h"
45 #include "delayed-inode.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "ioctl.h"
51 #include "print-tree.h"
52 #include "xattr.h"
53 #include "volumes.h"
54 #include "version.h"
55 #include "export.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,
65 char nbuf[16])
67 char *errstr = NULL;
69 switch (errno) {
70 case -EIO:
71 errstr = "IO failure";
72 break;
73 case -ENOMEM:
74 errstr = "Out of memory";
75 break;
76 case -EROFS:
77 errstr = "Readonly filesystem";
78 break;
79 default:
80 if (nbuf) {
81 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
82 errstr = nbuf;
84 break;
87 return errstr;
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;
99 /* NOTE:
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)
114 return;
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;
130 char nbuf[16];
131 const char *errstr;
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))
138 return;
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);
151 int ret;
153 ret = close_ctree(root);
154 sb->s_fs_info = NULL;
156 (void)ret; /* FIXME: need to fix VFS to return error? */
159 enum {
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_recovery, 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"},
185 {Opt_ssd, "ssd"},
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"},
201 {Opt_recovery, "recovery"},
202 {Opt_err, NULL},
206 * Regular mount options parser. Everything that is needed only when
207 * reading in a new superblock is parsed here.
209 int btrfs_parse_options(struct btrfs_root *root, char *options)
211 struct btrfs_fs_info *info = root->fs_info;
212 substring_t args[MAX_OPT_ARGS];
213 char *p, *num, *orig = NULL;
214 u64 cache_gen;
215 int intarg;
216 int ret = 0;
217 char *compress_type;
218 bool compress_force = false;
220 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
221 if (cache_gen)
222 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
224 if (!options)
225 goto out;
228 * strsep changes the string, duplicate it because parse_options
229 * gets called twice
231 options = kstrdup(options, GFP_NOFS);
232 if (!options)
233 return -ENOMEM;
235 orig = options;
237 while ((p = strsep(&options, ",")) != NULL) {
238 int token;
239 if (!*p)
240 continue;
242 token = match_token(p, tokens, args);
243 switch (token) {
244 case Opt_degraded:
245 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
246 btrfs_set_opt(info->mount_opt, DEGRADED);
247 break;
248 case Opt_subvol:
249 case Opt_subvolid:
250 case Opt_subvolrootid:
251 case Opt_device:
253 * These are parsed by btrfs_parse_early_options
254 * and can be happily ignored here.
256 break;
257 case Opt_nodatasum:
258 printk(KERN_INFO "btrfs: setting nodatasum\n");
259 btrfs_set_opt(info->mount_opt, NODATASUM);
260 break;
261 case Opt_nodatacow:
262 printk(KERN_INFO "btrfs: setting nodatacow\n");
263 btrfs_set_opt(info->mount_opt, NODATACOW);
264 btrfs_set_opt(info->mount_opt, NODATASUM);
265 break;
266 case Opt_compress_force:
267 case Opt_compress_force_type:
268 compress_force = true;
269 case Opt_compress:
270 case Opt_compress_type:
271 if (token == Opt_compress ||
272 token == Opt_compress_force ||
273 strcmp(args[0].from, "zlib") == 0) {
274 compress_type = "zlib";
275 info->compress_type = BTRFS_COMPRESS_ZLIB;
276 } else if (strcmp(args[0].from, "lzo") == 0) {
277 compress_type = "lzo";
278 info->compress_type = BTRFS_COMPRESS_LZO;
279 } else {
280 ret = -EINVAL;
281 goto out;
284 btrfs_set_opt(info->mount_opt, COMPRESS);
285 if (compress_force) {
286 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
287 pr_info("btrfs: force %s compression\n",
288 compress_type);
289 } else
290 pr_info("btrfs: use %s compression\n",
291 compress_type);
292 break;
293 case Opt_ssd:
294 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
295 btrfs_set_opt(info->mount_opt, SSD);
296 break;
297 case Opt_ssd_spread:
298 printk(KERN_INFO "btrfs: use spread ssd "
299 "allocation scheme\n");
300 btrfs_set_opt(info->mount_opt, SSD);
301 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
302 break;
303 case Opt_nossd:
304 printk(KERN_INFO "btrfs: not using ssd allocation "
305 "scheme\n");
306 btrfs_set_opt(info->mount_opt, NOSSD);
307 btrfs_clear_opt(info->mount_opt, SSD);
308 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
309 break;
310 case Opt_nobarrier:
311 printk(KERN_INFO "btrfs: turning off barriers\n");
312 btrfs_set_opt(info->mount_opt, NOBARRIER);
313 break;
314 case Opt_thread_pool:
315 intarg = 0;
316 match_int(&args[0], &intarg);
317 if (intarg) {
318 info->thread_pool_size = intarg;
319 printk(KERN_INFO "btrfs: thread pool %d\n",
320 info->thread_pool_size);
322 break;
323 case Opt_max_inline:
324 num = match_strdup(&args[0]);
325 if (num) {
326 info->max_inline = memparse(num, NULL);
327 kfree(num);
329 if (info->max_inline) {
330 info->max_inline = max_t(u64,
331 info->max_inline,
332 root->sectorsize);
334 printk(KERN_INFO "btrfs: max_inline at %llu\n",
335 (unsigned long long)info->max_inline);
337 break;
338 case Opt_alloc_start:
339 num = match_strdup(&args[0]);
340 if (num) {
341 info->alloc_start = memparse(num, NULL);
342 kfree(num);
343 printk(KERN_INFO
344 "btrfs: allocations start at %llu\n",
345 (unsigned long long)info->alloc_start);
347 break;
348 case Opt_noacl:
349 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
350 break;
351 case Opt_notreelog:
352 printk(KERN_INFO "btrfs: disabling tree log\n");
353 btrfs_set_opt(info->mount_opt, NOTREELOG);
354 break;
355 case Opt_flushoncommit:
356 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
357 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
358 break;
359 case Opt_ratio:
360 intarg = 0;
361 match_int(&args[0], &intarg);
362 if (intarg) {
363 info->metadata_ratio = intarg;
364 printk(KERN_INFO "btrfs: metadata ratio %d\n",
365 info->metadata_ratio);
367 break;
368 case Opt_discard:
369 btrfs_set_opt(info->mount_opt, DISCARD);
370 break;
371 case Opt_space_cache:
372 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
373 break;
374 case Opt_no_space_cache:
375 printk(KERN_INFO "btrfs: disabling disk space caching\n");
376 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
377 break;
378 case Opt_inode_cache:
379 printk(KERN_INFO "btrfs: enabling inode map caching\n");
380 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
381 break;
382 case Opt_clear_cache:
383 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
384 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
385 break;
386 case Opt_user_subvol_rm_allowed:
387 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
388 break;
389 case Opt_enospc_debug:
390 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
391 break;
392 case Opt_defrag:
393 printk(KERN_INFO "btrfs: enabling auto defrag");
394 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
395 break;
396 case Opt_recovery:
397 printk(KERN_INFO "btrfs: enabling auto recovery");
398 btrfs_set_opt(info->mount_opt, RECOVERY);
399 break;
400 case Opt_err:
401 printk(KERN_INFO "btrfs: unrecognized mount option "
402 "'%s'\n", p);
403 ret = -EINVAL;
404 goto out;
405 default:
406 break;
409 out:
410 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
411 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
412 kfree(orig);
413 return ret;
417 * Parse mount options that are required early in the mount process.
419 * All other options will be parsed on much later in the mount process and
420 * only when we need to allocate a new super block.
422 static int btrfs_parse_early_options(const char *options, fmode_t flags,
423 void *holder, char **subvol_name, u64 *subvol_objectid,
424 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
426 substring_t args[MAX_OPT_ARGS];
427 char *device_name, *opts, *orig, *p;
428 int error = 0;
429 int intarg;
431 if (!options)
432 return 0;
435 * strsep changes the string, duplicate it because parse_options
436 * gets called twice
438 opts = kstrdup(options, GFP_KERNEL);
439 if (!opts)
440 return -ENOMEM;
441 orig = opts;
443 while ((p = strsep(&opts, ",")) != NULL) {
444 int token;
445 if (!*p)
446 continue;
448 token = match_token(p, tokens, args);
449 switch (token) {
450 case Opt_subvol:
451 *subvol_name = match_strdup(&args[0]);
452 break;
453 case Opt_subvolid:
454 intarg = 0;
455 error = match_int(&args[0], &intarg);
456 if (!error) {
457 /* we want the original fs_tree */
458 if (!intarg)
459 *subvol_objectid =
460 BTRFS_FS_TREE_OBJECTID;
461 else
462 *subvol_objectid = intarg;
464 break;
465 case Opt_subvolrootid:
466 intarg = 0;
467 error = match_int(&args[0], &intarg);
468 if (!error) {
469 /* we want the original fs_tree */
470 if (!intarg)
471 *subvol_rootid =
472 BTRFS_FS_TREE_OBJECTID;
473 else
474 *subvol_rootid = intarg;
476 break;
477 case Opt_device:
478 device_name = match_strdup(&args[0]);
479 if (!device_name) {
480 error = -ENOMEM;
481 goto out;
483 error = btrfs_scan_one_device(device_name,
484 flags, holder, fs_devices);
485 kfree(device_name);
486 if (error)
487 goto out;
488 break;
489 default:
490 break;
494 out:
495 kfree(orig);
496 return error;
499 static struct dentry *get_default_root(struct super_block *sb,
500 u64 subvol_objectid)
502 struct btrfs_root *root = sb->s_fs_info;
503 struct btrfs_root *new_root;
504 struct btrfs_dir_item *di;
505 struct btrfs_path *path;
506 struct btrfs_key location;
507 struct inode *inode;
508 u64 dir_id;
509 int new = 0;
512 * We have a specific subvol we want to mount, just setup location and
513 * go look up the root.
515 if (subvol_objectid) {
516 location.objectid = subvol_objectid;
517 location.type = BTRFS_ROOT_ITEM_KEY;
518 location.offset = (u64)-1;
519 goto find_root;
522 path = btrfs_alloc_path();
523 if (!path)
524 return ERR_PTR(-ENOMEM);
525 path->leave_spinning = 1;
528 * Find the "default" dir item which points to the root item that we
529 * will mount by default if we haven't been given a specific subvolume
530 * to mount.
532 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
533 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
534 if (IS_ERR(di)) {
535 btrfs_free_path(path);
536 return ERR_CAST(di);
538 if (!di) {
540 * Ok the default dir item isn't there. This is weird since
541 * it's always been there, but don't freak out, just try and
542 * mount to root most subvolume.
544 btrfs_free_path(path);
545 dir_id = BTRFS_FIRST_FREE_OBJECTID;
546 new_root = root->fs_info->fs_root;
547 goto setup_root;
550 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
551 btrfs_free_path(path);
553 find_root:
554 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
555 if (IS_ERR(new_root))
556 return ERR_CAST(new_root);
558 if (btrfs_root_refs(&new_root->root_item) == 0)
559 return ERR_PTR(-ENOENT);
561 dir_id = btrfs_root_dirid(&new_root->root_item);
562 setup_root:
563 location.objectid = dir_id;
564 location.type = BTRFS_INODE_ITEM_KEY;
565 location.offset = 0;
567 inode = btrfs_iget(sb, &location, new_root, &new);
568 if (IS_ERR(inode))
569 return ERR_CAST(inode);
572 * If we're just mounting the root most subvol put the inode and return
573 * a reference to the dentry. We will have already gotten a reference
574 * to the inode in btrfs_fill_super so we're good to go.
576 if (!new && sb->s_root->d_inode == inode) {
577 iput(inode);
578 return dget(sb->s_root);
581 return d_obtain_alias(inode);
584 static int btrfs_fill_super(struct super_block *sb,
585 struct btrfs_fs_devices *fs_devices,
586 void *data, int silent)
588 struct inode *inode;
589 struct dentry *root_dentry;
590 struct btrfs_root *tree_root;
591 struct btrfs_key key;
592 int err;
594 sb->s_maxbytes = MAX_LFS_FILESIZE;
595 sb->s_magic = BTRFS_SUPER_MAGIC;
596 sb->s_op = &btrfs_super_ops;
597 sb->s_d_op = &btrfs_dentry_operations;
598 sb->s_export_op = &btrfs_export_ops;
599 sb->s_xattr = btrfs_xattr_handlers;
600 sb->s_time_gran = 1;
601 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
602 sb->s_flags |= MS_POSIXACL;
603 #endif
605 tree_root = open_ctree(sb, fs_devices, (char *)data);
607 if (IS_ERR(tree_root)) {
608 printk("btrfs: open_ctree failed\n");
609 return PTR_ERR(tree_root);
611 sb->s_fs_info = tree_root;
613 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
614 key.type = BTRFS_INODE_ITEM_KEY;
615 key.offset = 0;
616 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
617 if (IS_ERR(inode)) {
618 err = PTR_ERR(inode);
619 goto fail_close;
622 root_dentry = d_alloc_root(inode);
623 if (!root_dentry) {
624 iput(inode);
625 err = -ENOMEM;
626 goto fail_close;
629 sb->s_root = root_dentry;
631 save_mount_options(sb, data);
632 cleancache_init_fs(sb);
633 return 0;
635 fail_close:
636 close_ctree(tree_root);
637 return err;
640 int btrfs_sync_fs(struct super_block *sb, int wait)
642 struct btrfs_trans_handle *trans;
643 struct btrfs_root *root = btrfs_sb(sb);
644 int ret;
646 trace_btrfs_sync_fs(wait);
648 if (!wait) {
649 filemap_flush(root->fs_info->btree_inode->i_mapping);
650 return 0;
653 btrfs_start_delalloc_inodes(root, 0);
654 btrfs_wait_ordered_extents(root, 0, 0);
656 trans = btrfs_start_transaction(root, 0);
657 if (IS_ERR(trans))
658 return PTR_ERR(trans);
659 ret = btrfs_commit_transaction(trans, root);
660 return ret;
663 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
665 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
666 struct btrfs_fs_info *info = root->fs_info;
667 char *compress_type;
669 if (btrfs_test_opt(root, DEGRADED))
670 seq_puts(seq, ",degraded");
671 if (btrfs_test_opt(root, NODATASUM))
672 seq_puts(seq, ",nodatasum");
673 if (btrfs_test_opt(root, NODATACOW))
674 seq_puts(seq, ",nodatacow");
675 if (btrfs_test_opt(root, NOBARRIER))
676 seq_puts(seq, ",nobarrier");
677 if (info->max_inline != 8192 * 1024)
678 seq_printf(seq, ",max_inline=%llu",
679 (unsigned long long)info->max_inline);
680 if (info->alloc_start != 0)
681 seq_printf(seq, ",alloc_start=%llu",
682 (unsigned long long)info->alloc_start);
683 if (info->thread_pool_size != min_t(unsigned long,
684 num_online_cpus() + 2, 8))
685 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
686 if (btrfs_test_opt(root, COMPRESS)) {
687 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
688 compress_type = "zlib";
689 else
690 compress_type = "lzo";
691 if (btrfs_test_opt(root, FORCE_COMPRESS))
692 seq_printf(seq, ",compress-force=%s", compress_type);
693 else
694 seq_printf(seq, ",compress=%s", compress_type);
696 if (btrfs_test_opt(root, NOSSD))
697 seq_puts(seq, ",nossd");
698 if (btrfs_test_opt(root, SSD_SPREAD))
699 seq_puts(seq, ",ssd_spread");
700 else if (btrfs_test_opt(root, SSD))
701 seq_puts(seq, ",ssd");
702 if (btrfs_test_opt(root, NOTREELOG))
703 seq_puts(seq, ",notreelog");
704 if (btrfs_test_opt(root, FLUSHONCOMMIT))
705 seq_puts(seq, ",flushoncommit");
706 if (btrfs_test_opt(root, DISCARD))
707 seq_puts(seq, ",discard");
708 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
709 seq_puts(seq, ",noacl");
710 if (btrfs_test_opt(root, SPACE_CACHE))
711 seq_puts(seq, ",space_cache");
712 else
713 seq_puts(seq, ",no_space_cache");
714 if (btrfs_test_opt(root, CLEAR_CACHE))
715 seq_puts(seq, ",clear_cache");
716 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
717 seq_puts(seq, ",user_subvol_rm_allowed");
718 if (btrfs_test_opt(root, ENOSPC_DEBUG))
719 seq_puts(seq, ",enospc_debug");
720 if (btrfs_test_opt(root, AUTO_DEFRAG))
721 seq_puts(seq, ",autodefrag");
722 if (btrfs_test_opt(root, INODE_MAP_CACHE))
723 seq_puts(seq, ",inode_cache");
724 return 0;
727 static int btrfs_test_super(struct super_block *s, void *data)
729 struct btrfs_root *test_root = data;
730 struct btrfs_root *root = btrfs_sb(s);
733 * If this super block is going away, return false as it
734 * can't match as an existing super block.
736 if (!atomic_read(&s->s_active))
737 return 0;
738 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
741 static int btrfs_set_super(struct super_block *s, void *data)
743 s->s_fs_info = data;
745 return set_anon_super(s, data);
749 * subvolumes are identified by ino 256
751 static inline int is_subvolume_inode(struct inode *inode)
753 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
754 return 1;
755 return 0;
759 * This will strip out the subvol=%s argument for an argument string and add
760 * subvolid=0 to make sure we get the actual tree root for path walking to the
761 * subvol we want.
763 static char *setup_root_args(char *args)
765 unsigned copied = 0;
766 unsigned len = strlen(args) + 2;
767 char *pos;
768 char *ret;
771 * We need the same args as before, but minus
773 * subvol=a
775 * and add
777 * subvolid=0
779 * which is a difference of 2 characters, so we allocate strlen(args) +
780 * 2 characters.
782 ret = kzalloc(len * sizeof(char), GFP_NOFS);
783 if (!ret)
784 return NULL;
785 pos = strstr(args, "subvol=");
787 /* This shouldn't happen, but just in case.. */
788 if (!pos) {
789 kfree(ret);
790 return NULL;
794 * The subvol=<> arg is not at the front of the string, copy everybody
795 * up to that into ret.
797 if (pos != args) {
798 *pos = '\0';
799 strcpy(ret, args);
800 copied += strlen(args);
801 pos++;
804 strncpy(ret + copied, "subvolid=0", len - copied);
806 /* Length of subvolid=0 */
807 copied += 10;
810 * If there is no , after the subvol= option then we know there's no
811 * other options and we can just return.
813 pos = strchr(pos, ',');
814 if (!pos)
815 return ret;
817 /* Copy the rest of the arguments into our buffer */
818 strncpy(ret + copied, pos, len - copied);
819 copied += strlen(pos);
821 return ret;
824 static struct dentry *mount_subvol(const char *subvol_name, int flags,
825 const char *device_name, char *data)
827 struct super_block *s;
828 struct dentry *root;
829 struct vfsmount *mnt;
830 struct mnt_namespace *ns_private;
831 char *newargs;
832 struct path path;
833 int error;
835 newargs = setup_root_args(data);
836 if (!newargs)
837 return ERR_PTR(-ENOMEM);
838 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
839 newargs);
840 kfree(newargs);
841 if (IS_ERR(mnt))
842 return ERR_CAST(mnt);
844 ns_private = create_mnt_ns(mnt);
845 if (IS_ERR(ns_private)) {
846 mntput(mnt);
847 return ERR_CAST(ns_private);
851 * This will trigger the automount of the subvol so we can just
852 * drop the mnt we have here and return the dentry that we
853 * found.
855 error = vfs_path_lookup(mnt->mnt_root, mnt, subvol_name,
856 LOOKUP_FOLLOW, &path);
857 put_mnt_ns(ns_private);
858 if (error)
859 return ERR_PTR(error);
861 if (!is_subvolume_inode(path.dentry->d_inode)) {
862 path_put(&path);
863 mntput(mnt);
864 error = -EINVAL;
865 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
866 subvol_name);
867 return ERR_PTR(-EINVAL);
870 /* Get a ref to the sb and the dentry we found and return it */
871 s = path.mnt->mnt_sb;
872 atomic_inc(&s->s_active);
873 root = dget(path.dentry);
874 path_put(&path);
875 down_write(&s->s_umount);
877 return root;
881 * Find a superblock for the given device / mount point.
883 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
884 * for multiple device setup. Make sure to keep it in sync.
886 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
887 const char *device_name, void *data)
889 struct block_device *bdev = NULL;
890 struct super_block *s;
891 struct dentry *root;
892 struct btrfs_fs_devices *fs_devices = NULL;
893 struct btrfs_root *tree_root = NULL;
894 struct btrfs_fs_info *fs_info = NULL;
895 fmode_t mode = FMODE_READ;
896 char *subvol_name = NULL;
897 u64 subvol_objectid = 0;
898 u64 subvol_rootid = 0;
899 int error = 0;
901 if (!(flags & MS_RDONLY))
902 mode |= FMODE_WRITE;
904 error = btrfs_parse_early_options(data, mode, fs_type,
905 &subvol_name, &subvol_objectid,
906 &subvol_rootid, &fs_devices);
907 if (error)
908 return ERR_PTR(error);
910 if (subvol_name) {
911 root = mount_subvol(subvol_name, flags, device_name, data);
912 kfree(subvol_name);
913 return root;
916 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
917 if (error)
918 return ERR_PTR(error);
920 error = btrfs_open_devices(fs_devices, mode, fs_type);
921 if (error)
922 return ERR_PTR(error);
924 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
925 error = -EACCES;
926 goto error_close_devices;
930 * Setup a dummy root and fs_info for test/set super. This is because
931 * we don't actually fill this stuff out until open_ctree, but we need
932 * it for searching for existing supers, so this lets us do that and
933 * then open_ctree will properly initialize everything later.
935 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
936 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
937 if (!fs_info || !tree_root) {
938 error = -ENOMEM;
939 goto error_close_devices;
941 fs_info->tree_root = tree_root;
942 fs_info->fs_devices = fs_devices;
943 tree_root->fs_info = fs_info;
945 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
946 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
947 if (!fs_info->super_copy || !fs_info->super_for_commit) {
948 error = -ENOMEM;
949 goto error_close_devices;
952 bdev = fs_devices->latest_bdev;
953 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
954 if (IS_ERR(s)) {
955 error = PTR_ERR(s);
956 goto error_close_devices;
959 if (s->s_root) {
960 if ((flags ^ s->s_flags) & MS_RDONLY) {
961 deactivate_locked_super(s);
962 return ERR_PTR(-EBUSY);
965 btrfs_close_devices(fs_devices);
966 free_fs_info(fs_info);
967 kfree(tree_root);
968 } else {
969 char b[BDEVNAME_SIZE];
971 s->s_flags = flags | MS_NOSEC;
972 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
973 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
974 error = btrfs_fill_super(s, fs_devices, data,
975 flags & MS_SILENT ? 1 : 0);
976 if (error) {
977 deactivate_locked_super(s);
978 return ERR_PTR(error);
981 s->s_flags |= MS_ACTIVE;
984 root = get_default_root(s, subvol_objectid);
985 if (IS_ERR(root)) {
986 deactivate_locked_super(s);
987 return root;
990 return root;
992 error_close_devices:
993 btrfs_close_devices(fs_devices);
994 free_fs_info(fs_info);
995 kfree(tree_root);
996 return ERR_PTR(error);
999 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1001 struct btrfs_root *root = btrfs_sb(sb);
1002 int ret;
1004 ret = btrfs_parse_options(root, data);
1005 if (ret)
1006 return -EINVAL;
1008 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1009 return 0;
1011 if (*flags & MS_RDONLY) {
1012 sb->s_flags |= MS_RDONLY;
1014 ret = btrfs_commit_super(root);
1015 WARN_ON(ret);
1016 } else {
1017 if (root->fs_info->fs_devices->rw_devices == 0)
1018 return -EACCES;
1020 if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
1021 return -EINVAL;
1023 ret = btrfs_cleanup_fs_roots(root->fs_info);
1024 WARN_ON(ret);
1026 /* recover relocation */
1027 ret = btrfs_recover_relocation(root);
1028 WARN_ON(ret);
1030 sb->s_flags &= ~MS_RDONLY;
1033 return 0;
1036 /* Used to sort the devices by max_avail(descending sort) */
1037 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1038 const void *dev_info2)
1040 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1041 ((struct btrfs_device_info *)dev_info2)->max_avail)
1042 return -1;
1043 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1044 ((struct btrfs_device_info *)dev_info2)->max_avail)
1045 return 1;
1046 else
1047 return 0;
1051 * sort the devices by max_avail, in which max free extent size of each device
1052 * is stored.(Descending Sort)
1054 static inline void btrfs_descending_sort_devices(
1055 struct btrfs_device_info *devices,
1056 size_t nr_devices)
1058 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1059 btrfs_cmp_device_free_bytes, NULL);
1063 * The helper to calc the free space on the devices that can be used to store
1064 * file data.
1066 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1068 struct btrfs_fs_info *fs_info = root->fs_info;
1069 struct btrfs_device_info *devices_info;
1070 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1071 struct btrfs_device *device;
1072 u64 skip_space;
1073 u64 type;
1074 u64 avail_space;
1075 u64 used_space;
1076 u64 min_stripe_size;
1077 int min_stripes = 1;
1078 int i = 0, nr_devices;
1079 int ret;
1081 nr_devices = fs_info->fs_devices->rw_devices;
1082 BUG_ON(!nr_devices);
1084 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1085 GFP_NOFS);
1086 if (!devices_info)
1087 return -ENOMEM;
1089 /* calc min stripe number for data space alloction */
1090 type = btrfs_get_alloc_profile(root, 1);
1091 if (type & BTRFS_BLOCK_GROUP_RAID0)
1092 min_stripes = 2;
1093 else if (type & BTRFS_BLOCK_GROUP_RAID1)
1094 min_stripes = 2;
1095 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1096 min_stripes = 4;
1098 if (type & BTRFS_BLOCK_GROUP_DUP)
1099 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1100 else
1101 min_stripe_size = BTRFS_STRIPE_LEN;
1103 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
1104 if (!device->in_fs_metadata)
1105 continue;
1107 avail_space = device->total_bytes - device->bytes_used;
1109 /* align with stripe_len */
1110 do_div(avail_space, BTRFS_STRIPE_LEN);
1111 avail_space *= BTRFS_STRIPE_LEN;
1114 * In order to avoid overwritting the superblock on the drive,
1115 * btrfs starts at an offset of at least 1MB when doing chunk
1116 * allocation.
1118 skip_space = 1024 * 1024;
1120 /* user can set the offset in fs_info->alloc_start. */
1121 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1122 device->total_bytes)
1123 skip_space = max(fs_info->alloc_start, skip_space);
1126 * btrfs can not use the free space in [0, skip_space - 1],
1127 * we must subtract it from the total. In order to implement
1128 * it, we account the used space in this range first.
1130 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1131 &used_space);
1132 if (ret) {
1133 kfree(devices_info);
1134 return ret;
1137 /* calc the free space in [0, skip_space - 1] */
1138 skip_space -= used_space;
1141 * we can use the free space in [0, skip_space - 1], subtract
1142 * it from the total.
1144 if (avail_space && avail_space >= skip_space)
1145 avail_space -= skip_space;
1146 else
1147 avail_space = 0;
1149 if (avail_space < min_stripe_size)
1150 continue;
1152 devices_info[i].dev = device;
1153 devices_info[i].max_avail = avail_space;
1155 i++;
1158 nr_devices = i;
1160 btrfs_descending_sort_devices(devices_info, nr_devices);
1162 i = nr_devices - 1;
1163 avail_space = 0;
1164 while (nr_devices >= min_stripes) {
1165 if (devices_info[i].max_avail >= min_stripe_size) {
1166 int j;
1167 u64 alloc_size;
1169 avail_space += devices_info[i].max_avail * min_stripes;
1170 alloc_size = devices_info[i].max_avail;
1171 for (j = i + 1 - min_stripes; j <= i; j++)
1172 devices_info[j].max_avail -= alloc_size;
1174 i--;
1175 nr_devices--;
1178 kfree(devices_info);
1179 *free_bytes = avail_space;
1180 return 0;
1183 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1185 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1186 struct btrfs_super_block *disk_super = root->fs_info->super_copy;
1187 struct list_head *head = &root->fs_info->space_info;
1188 struct btrfs_space_info *found;
1189 u64 total_used = 0;
1190 u64 total_free_data = 0;
1191 int bits = dentry->d_sb->s_blocksize_bits;
1192 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1193 int ret;
1195 /* holding chunk_muext to avoid allocating new chunks */
1196 mutex_lock(&root->fs_info->chunk_mutex);
1197 rcu_read_lock();
1198 list_for_each_entry_rcu(found, head, list) {
1199 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1200 total_free_data += found->disk_total - found->disk_used;
1201 total_free_data -=
1202 btrfs_account_ro_block_groups_free_space(found);
1205 total_used += found->disk_used;
1207 rcu_read_unlock();
1209 buf->f_namelen = BTRFS_NAME_LEN;
1210 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1211 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1212 buf->f_bsize = dentry->d_sb->s_blocksize;
1213 buf->f_type = BTRFS_SUPER_MAGIC;
1214 buf->f_bavail = total_free_data;
1215 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1216 if (ret) {
1217 mutex_unlock(&root->fs_info->chunk_mutex);
1218 return ret;
1220 buf->f_bavail += total_free_data;
1221 buf->f_bavail = buf->f_bavail >> bits;
1222 mutex_unlock(&root->fs_info->chunk_mutex);
1224 /* We treat it as constant endianness (it doesn't matter _which_)
1225 because we want the fsid to come out the same whether mounted
1226 on a big-endian or little-endian host */
1227 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1228 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1229 /* Mask in the root object ID too, to disambiguate subvols */
1230 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1231 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1233 return 0;
1236 static struct file_system_type btrfs_fs_type = {
1237 .owner = THIS_MODULE,
1238 .name = "btrfs",
1239 .mount = btrfs_mount,
1240 .kill_sb = kill_anon_super,
1241 .fs_flags = FS_REQUIRES_DEV,
1245 * used by btrfsctl to scan devices when no FS is mounted
1247 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1248 unsigned long arg)
1250 struct btrfs_ioctl_vol_args *vol;
1251 struct btrfs_fs_devices *fs_devices;
1252 int ret = -ENOTTY;
1254 if (!capable(CAP_SYS_ADMIN))
1255 return -EPERM;
1257 vol = memdup_user((void __user *)arg, sizeof(*vol));
1258 if (IS_ERR(vol))
1259 return PTR_ERR(vol);
1261 switch (cmd) {
1262 case BTRFS_IOC_SCAN_DEV:
1263 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1264 &btrfs_fs_type, &fs_devices);
1265 break;
1268 kfree(vol);
1269 return ret;
1272 static int btrfs_freeze(struct super_block *sb)
1274 struct btrfs_root *root = btrfs_sb(sb);
1275 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1276 mutex_lock(&root->fs_info->cleaner_mutex);
1277 return 0;
1280 static int btrfs_unfreeze(struct super_block *sb)
1282 struct btrfs_root *root = btrfs_sb(sb);
1283 mutex_unlock(&root->fs_info->cleaner_mutex);
1284 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1285 return 0;
1288 static const struct super_operations btrfs_super_ops = {
1289 .drop_inode = btrfs_drop_inode,
1290 .evict_inode = btrfs_evict_inode,
1291 .put_super = btrfs_put_super,
1292 .sync_fs = btrfs_sync_fs,
1293 .show_options = btrfs_show_options,
1294 .write_inode = btrfs_write_inode,
1295 .dirty_inode = btrfs_dirty_inode,
1296 .alloc_inode = btrfs_alloc_inode,
1297 .destroy_inode = btrfs_destroy_inode,
1298 .statfs = btrfs_statfs,
1299 .remount_fs = btrfs_remount,
1300 .freeze_fs = btrfs_freeze,
1301 .unfreeze_fs = btrfs_unfreeze,
1304 static const struct file_operations btrfs_ctl_fops = {
1305 .unlocked_ioctl = btrfs_control_ioctl,
1306 .compat_ioctl = btrfs_control_ioctl,
1307 .owner = THIS_MODULE,
1308 .llseek = noop_llseek,
1311 static struct miscdevice btrfs_misc = {
1312 .minor = BTRFS_MINOR,
1313 .name = "btrfs-control",
1314 .fops = &btrfs_ctl_fops
1317 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1318 MODULE_ALIAS("devname:btrfs-control");
1320 static int btrfs_interface_init(void)
1322 return misc_register(&btrfs_misc);
1325 static void btrfs_interface_exit(void)
1327 if (misc_deregister(&btrfs_misc) < 0)
1328 printk(KERN_INFO "misc_deregister failed for control device");
1331 static int __init init_btrfs_fs(void)
1333 int err;
1335 err = btrfs_init_sysfs();
1336 if (err)
1337 return err;
1339 err = btrfs_init_compress();
1340 if (err)
1341 goto free_sysfs;
1343 err = btrfs_init_cachep();
1344 if (err)
1345 goto free_compress;
1347 err = extent_io_init();
1348 if (err)
1349 goto free_cachep;
1351 err = extent_map_init();
1352 if (err)
1353 goto free_extent_io;
1355 err = btrfs_delayed_inode_init();
1356 if (err)
1357 goto free_extent_map;
1359 err = btrfs_interface_init();
1360 if (err)
1361 goto free_delayed_inode;
1363 err = register_filesystem(&btrfs_fs_type);
1364 if (err)
1365 goto unregister_ioctl;
1367 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1368 return 0;
1370 unregister_ioctl:
1371 btrfs_interface_exit();
1372 free_delayed_inode:
1373 btrfs_delayed_inode_exit();
1374 free_extent_map:
1375 extent_map_exit();
1376 free_extent_io:
1377 extent_io_exit();
1378 free_cachep:
1379 btrfs_destroy_cachep();
1380 free_compress:
1381 btrfs_exit_compress();
1382 free_sysfs:
1383 btrfs_exit_sysfs();
1384 return err;
1387 static void __exit exit_btrfs_fs(void)
1389 btrfs_destroy_cachep();
1390 btrfs_delayed_inode_exit();
1391 extent_map_exit();
1392 extent_io_exit();
1393 btrfs_interface_exit();
1394 unregister_filesystem(&btrfs_fs_type);
1395 btrfs_exit_sysfs();
1396 btrfs_cleanup_fs_uuids();
1397 btrfs_exit_compress();
1400 module_init(init_btrfs_fs)
1401 module_exit(exit_btrfs_fs)
1403 MODULE_LICENSE("GPL");