Loading...
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 <linux/cleancache.h>
43#include "compat.h"
44#include "delayed-inode.h"
45#include "ctree.h"
46#include "disk-io.h"
47#include "transaction.h"
48#include "btrfs_inode.h"
49#include "ioctl.h"
50#include "print-tree.h"
51#include "xattr.h"
52#include "volumes.h"
53#include "version.h"
54#include "export.h"
55#include "compression.h"
56
57#define CREATE_TRACE_POINTS
58#include <trace/events/btrfs.h>
59
60static const struct super_operations btrfs_super_ops;
61
62static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
63 char nbuf[16])
64{
65 char *errstr = NULL;
66
67 switch (errno) {
68 case -EIO:
69 errstr = "IO failure";
70 break;
71 case -ENOMEM:
72 errstr = "Out of memory";
73 break;
74 case -EROFS:
75 errstr = "Readonly filesystem";
76 break;
77 default:
78 if (nbuf) {
79 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
80 errstr = nbuf;
81 }
82 break;
83 }
84
85 return errstr;
86}
87
88static void __save_error_info(struct btrfs_fs_info *fs_info)
89{
90 /*
91 * today we only save the error info into ram. Long term we'll
92 * also send it down to the disk
93 */
94 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
95}
96
97/* NOTE:
98 * We move write_super stuff at umount in order to avoid deadlock
99 * for umount hold all lock.
100 */
101static void save_error_info(struct btrfs_fs_info *fs_info)
102{
103 __save_error_info(fs_info);
104}
105
106/* btrfs handle error by forcing the filesystem readonly */
107static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
108{
109 struct super_block *sb = fs_info->sb;
110
111 if (sb->s_flags & MS_RDONLY)
112 return;
113
114 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
115 sb->s_flags |= MS_RDONLY;
116 printk(KERN_INFO "btrfs is forced readonly\n");
117 }
118}
119
120/*
121 * __btrfs_std_error decodes expected errors from the caller and
122 * invokes the approciate error response.
123 */
124void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
125 unsigned int line, int errno)
126{
127 struct super_block *sb = fs_info->sb;
128 char nbuf[16];
129 const char *errstr;
130
131 /*
132 * Special case: if the error is EROFS, and we're already
133 * under MS_RDONLY, then it is safe here.
134 */
135 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
136 return;
137
138 errstr = btrfs_decode_error(fs_info, errno, nbuf);
139 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
140 sb->s_id, function, line, errstr);
141 save_error_info(fs_info);
142
143 btrfs_handle_error(fs_info);
144}
145
146static void btrfs_put_super(struct super_block *sb)
147{
148 struct btrfs_root *root = btrfs_sb(sb);
149 int ret;
150
151 ret = close_ctree(root);
152 sb->s_fs_info = NULL;
153
154 (void)ret; /* FIXME: need to fix VFS to return error? */
155}
156
157enum {
158 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
159 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
160 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
161 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
162 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
163 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
164 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
165 Opt_inode_cache, Opt_err,
166};
167
168static match_table_t tokens = {
169 {Opt_degraded, "degraded"},
170 {Opt_subvol, "subvol=%s"},
171 {Opt_subvolid, "subvolid=%d"},
172 {Opt_device, "device=%s"},
173 {Opt_nodatasum, "nodatasum"},
174 {Opt_nodatacow, "nodatacow"},
175 {Opt_nobarrier, "nobarrier"},
176 {Opt_max_inline, "max_inline=%s"},
177 {Opt_alloc_start, "alloc_start=%s"},
178 {Opt_thread_pool, "thread_pool=%d"},
179 {Opt_compress, "compress"},
180 {Opt_compress_type, "compress=%s"},
181 {Opt_compress_force, "compress-force"},
182 {Opt_compress_force_type, "compress-force=%s"},
183 {Opt_ssd, "ssd"},
184 {Opt_ssd_spread, "ssd_spread"},
185 {Opt_nossd, "nossd"},
186 {Opt_noacl, "noacl"},
187 {Opt_notreelog, "notreelog"},
188 {Opt_flushoncommit, "flushoncommit"},
189 {Opt_ratio, "metadata_ratio=%d"},
190 {Opt_discard, "discard"},
191 {Opt_space_cache, "space_cache"},
192 {Opt_clear_cache, "clear_cache"},
193 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
194 {Opt_enospc_debug, "enospc_debug"},
195 {Opt_subvolrootid, "subvolrootid=%d"},
196 {Opt_defrag, "autodefrag"},
197 {Opt_inode_cache, "inode_cache"},
198 {Opt_err, NULL},
199};
200
201/*
202 * Regular mount options parser. Everything that is needed only when
203 * reading in a new superblock is parsed here.
204 */
205int btrfs_parse_options(struct btrfs_root *root, char *options)
206{
207 struct btrfs_fs_info *info = root->fs_info;
208 substring_t args[MAX_OPT_ARGS];
209 char *p, *num, *orig;
210 int intarg;
211 int ret = 0;
212 char *compress_type;
213 bool compress_force = false;
214
215 if (!options)
216 return 0;
217
218 /*
219 * strsep changes the string, duplicate it because parse_options
220 * gets called twice
221 */
222 options = kstrdup(options, GFP_NOFS);
223 if (!options)
224 return -ENOMEM;
225
226 orig = options;
227
228 while ((p = strsep(&options, ",")) != NULL) {
229 int token;
230 if (!*p)
231 continue;
232
233 token = match_token(p, tokens, args);
234 switch (token) {
235 case Opt_degraded:
236 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
237 btrfs_set_opt(info->mount_opt, DEGRADED);
238 break;
239 case Opt_subvol:
240 case Opt_subvolid:
241 case Opt_subvolrootid:
242 case Opt_device:
243 /*
244 * These are parsed by btrfs_parse_early_options
245 * and can be happily ignored here.
246 */
247 break;
248 case Opt_nodatasum:
249 printk(KERN_INFO "btrfs: setting nodatasum\n");
250 btrfs_set_opt(info->mount_opt, NODATASUM);
251 break;
252 case Opt_nodatacow:
253 printk(KERN_INFO "btrfs: setting nodatacow\n");
254 btrfs_set_opt(info->mount_opt, NODATACOW);
255 btrfs_set_opt(info->mount_opt, NODATASUM);
256 break;
257 case Opt_compress_force:
258 case Opt_compress_force_type:
259 compress_force = true;
260 case Opt_compress:
261 case Opt_compress_type:
262 if (token == Opt_compress ||
263 token == Opt_compress_force ||
264 strcmp(args[0].from, "zlib") == 0) {
265 compress_type = "zlib";
266 info->compress_type = BTRFS_COMPRESS_ZLIB;
267 } else if (strcmp(args[0].from, "lzo") == 0) {
268 compress_type = "lzo";
269 info->compress_type = BTRFS_COMPRESS_LZO;
270 } else {
271 ret = -EINVAL;
272 goto out;
273 }
274
275 btrfs_set_opt(info->mount_opt, COMPRESS);
276 if (compress_force) {
277 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
278 pr_info("btrfs: force %s compression\n",
279 compress_type);
280 } else
281 pr_info("btrfs: use %s compression\n",
282 compress_type);
283 break;
284 case Opt_ssd:
285 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
286 btrfs_set_opt(info->mount_opt, SSD);
287 break;
288 case Opt_ssd_spread:
289 printk(KERN_INFO "btrfs: use spread ssd "
290 "allocation scheme\n");
291 btrfs_set_opt(info->mount_opt, SSD);
292 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
293 break;
294 case Opt_nossd:
295 printk(KERN_INFO "btrfs: not using ssd allocation "
296 "scheme\n");
297 btrfs_set_opt(info->mount_opt, NOSSD);
298 btrfs_clear_opt(info->mount_opt, SSD);
299 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
300 break;
301 case Opt_nobarrier:
302 printk(KERN_INFO "btrfs: turning off barriers\n");
303 btrfs_set_opt(info->mount_opt, NOBARRIER);
304 break;
305 case Opt_thread_pool:
306 intarg = 0;
307 match_int(&args[0], &intarg);
308 if (intarg) {
309 info->thread_pool_size = intarg;
310 printk(KERN_INFO "btrfs: thread pool %d\n",
311 info->thread_pool_size);
312 }
313 break;
314 case Opt_max_inline:
315 num = match_strdup(&args[0]);
316 if (num) {
317 info->max_inline = memparse(num, NULL);
318 kfree(num);
319
320 if (info->max_inline) {
321 info->max_inline = max_t(u64,
322 info->max_inline,
323 root->sectorsize);
324 }
325 printk(KERN_INFO "btrfs: max_inline at %llu\n",
326 (unsigned long long)info->max_inline);
327 }
328 break;
329 case Opt_alloc_start:
330 num = match_strdup(&args[0]);
331 if (num) {
332 info->alloc_start = memparse(num, NULL);
333 kfree(num);
334 printk(KERN_INFO
335 "btrfs: allocations start at %llu\n",
336 (unsigned long long)info->alloc_start);
337 }
338 break;
339 case Opt_noacl:
340 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
341 break;
342 case Opt_notreelog:
343 printk(KERN_INFO "btrfs: disabling tree log\n");
344 btrfs_set_opt(info->mount_opt, NOTREELOG);
345 break;
346 case Opt_flushoncommit:
347 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
348 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
349 break;
350 case Opt_ratio:
351 intarg = 0;
352 match_int(&args[0], &intarg);
353 if (intarg) {
354 info->metadata_ratio = intarg;
355 printk(KERN_INFO "btrfs: metadata ratio %d\n",
356 info->metadata_ratio);
357 }
358 break;
359 case Opt_discard:
360 btrfs_set_opt(info->mount_opt, DISCARD);
361 break;
362 case Opt_space_cache:
363 printk(KERN_INFO "btrfs: enabling disk space caching\n");
364 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
365 break;
366 case Opt_inode_cache:
367 printk(KERN_INFO "btrfs: enabling inode map caching\n");
368 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
369 break;
370 case Opt_clear_cache:
371 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
372 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
373 break;
374 case Opt_user_subvol_rm_allowed:
375 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
376 break;
377 case Opt_enospc_debug:
378 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
379 break;
380 case Opt_defrag:
381 printk(KERN_INFO "btrfs: enabling auto defrag");
382 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
383 break;
384 case Opt_err:
385 printk(KERN_INFO "btrfs: unrecognized mount option "
386 "'%s'\n", p);
387 ret = -EINVAL;
388 goto out;
389 default:
390 break;
391 }
392 }
393out:
394 kfree(orig);
395 return ret;
396}
397
398/*
399 * Parse mount options that are required early in the mount process.
400 *
401 * All other options will be parsed on much later in the mount process and
402 * only when we need to allocate a new super block.
403 */
404static int btrfs_parse_early_options(const char *options, fmode_t flags,
405 void *holder, char **subvol_name, u64 *subvol_objectid,
406 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
407{
408 substring_t args[MAX_OPT_ARGS];
409 char *opts, *orig, *p;
410 int error = 0;
411 int intarg;
412
413 if (!options)
414 goto out;
415
416 /*
417 * strsep changes the string, duplicate it because parse_options
418 * gets called twice
419 */
420 opts = kstrdup(options, GFP_KERNEL);
421 if (!opts)
422 return -ENOMEM;
423 orig = opts;
424
425 while ((p = strsep(&opts, ",")) != NULL) {
426 int token;
427 if (!*p)
428 continue;
429
430 token = match_token(p, tokens, args);
431 switch (token) {
432 case Opt_subvol:
433 *subvol_name = match_strdup(&args[0]);
434 break;
435 case Opt_subvolid:
436 intarg = 0;
437 error = match_int(&args[0], &intarg);
438 if (!error) {
439 /* we want the original fs_tree */
440 if (!intarg)
441 *subvol_objectid =
442 BTRFS_FS_TREE_OBJECTID;
443 else
444 *subvol_objectid = intarg;
445 }
446 break;
447 case Opt_subvolrootid:
448 intarg = 0;
449 error = match_int(&args[0], &intarg);
450 if (!error) {
451 /* we want the original fs_tree */
452 if (!intarg)
453 *subvol_rootid =
454 BTRFS_FS_TREE_OBJECTID;
455 else
456 *subvol_rootid = intarg;
457 }
458 break;
459 case Opt_device:
460 error = btrfs_scan_one_device(match_strdup(&args[0]),
461 flags, holder, fs_devices);
462 if (error)
463 goto out_free_opts;
464 break;
465 default:
466 break;
467 }
468 }
469
470 out_free_opts:
471 kfree(orig);
472 out:
473 /*
474 * If no subvolume name is specified we use the default one. Allocate
475 * a copy of the string "." here so that code later in the
476 * mount path doesn't care if it's the default volume or another one.
477 */
478 if (!*subvol_name) {
479 *subvol_name = kstrdup(".", GFP_KERNEL);
480 if (!*subvol_name)
481 return -ENOMEM;
482 }
483 return error;
484}
485
486static struct dentry *get_default_root(struct super_block *sb,
487 u64 subvol_objectid)
488{
489 struct btrfs_root *root = sb->s_fs_info;
490 struct btrfs_root *new_root;
491 struct btrfs_dir_item *di;
492 struct btrfs_path *path;
493 struct btrfs_key location;
494 struct inode *inode;
495 struct dentry *dentry;
496 u64 dir_id;
497 int new = 0;
498
499 /*
500 * We have a specific subvol we want to mount, just setup location and
501 * go look up the root.
502 */
503 if (subvol_objectid) {
504 location.objectid = subvol_objectid;
505 location.type = BTRFS_ROOT_ITEM_KEY;
506 location.offset = (u64)-1;
507 goto find_root;
508 }
509
510 path = btrfs_alloc_path();
511 if (!path)
512 return ERR_PTR(-ENOMEM);
513 path->leave_spinning = 1;
514
515 /*
516 * Find the "default" dir item which points to the root item that we
517 * will mount by default if we haven't been given a specific subvolume
518 * to mount.
519 */
520 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
521 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
522 if (IS_ERR(di)) {
523 btrfs_free_path(path);
524 return ERR_CAST(di);
525 }
526 if (!di) {
527 /*
528 * Ok the default dir item isn't there. This is weird since
529 * it's always been there, but don't freak out, just try and
530 * mount to root most subvolume.
531 */
532 btrfs_free_path(path);
533 dir_id = BTRFS_FIRST_FREE_OBJECTID;
534 new_root = root->fs_info->fs_root;
535 goto setup_root;
536 }
537
538 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
539 btrfs_free_path(path);
540
541find_root:
542 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
543 if (IS_ERR(new_root))
544 return ERR_CAST(new_root);
545
546 if (btrfs_root_refs(&new_root->root_item) == 0)
547 return ERR_PTR(-ENOENT);
548
549 dir_id = btrfs_root_dirid(&new_root->root_item);
550setup_root:
551 location.objectid = dir_id;
552 location.type = BTRFS_INODE_ITEM_KEY;
553 location.offset = 0;
554
555 inode = btrfs_iget(sb, &location, new_root, &new);
556 if (IS_ERR(inode))
557 return ERR_CAST(inode);
558
559 /*
560 * If we're just mounting the root most subvol put the inode and return
561 * a reference to the dentry. We will have already gotten a reference
562 * to the inode in btrfs_fill_super so we're good to go.
563 */
564 if (!new && sb->s_root->d_inode == inode) {
565 iput(inode);
566 return dget(sb->s_root);
567 }
568
569 if (new) {
570 const struct qstr name = { .name = "/", .len = 1 };
571
572 /*
573 * New inode, we need to make the dentry a sibling of s_root so
574 * everything gets cleaned up properly on unmount.
575 */
576 dentry = d_alloc(sb->s_root, &name);
577 if (!dentry) {
578 iput(inode);
579 return ERR_PTR(-ENOMEM);
580 }
581 d_splice_alias(inode, dentry);
582 } else {
583 /*
584 * We found the inode in cache, just find a dentry for it and
585 * put the reference to the inode we just got.
586 */
587 dentry = d_find_alias(inode);
588 iput(inode);
589 }
590
591 return dentry;
592}
593
594static int btrfs_fill_super(struct super_block *sb,
595 struct btrfs_fs_devices *fs_devices,
596 void *data, int silent)
597{
598 struct inode *inode;
599 struct dentry *root_dentry;
600 struct btrfs_root *tree_root;
601 struct btrfs_key key;
602 int err;
603
604 sb->s_maxbytes = MAX_LFS_FILESIZE;
605 sb->s_magic = BTRFS_SUPER_MAGIC;
606 sb->s_op = &btrfs_super_ops;
607 sb->s_d_op = &btrfs_dentry_operations;
608 sb->s_export_op = &btrfs_export_ops;
609 sb->s_xattr = btrfs_xattr_handlers;
610 sb->s_time_gran = 1;
611#ifdef CONFIG_BTRFS_FS_POSIX_ACL
612 sb->s_flags |= MS_POSIXACL;
613#endif
614
615 tree_root = open_ctree(sb, fs_devices, (char *)data);
616
617 if (IS_ERR(tree_root)) {
618 printk("btrfs: open_ctree failed\n");
619 return PTR_ERR(tree_root);
620 }
621 sb->s_fs_info = tree_root;
622
623 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
624 key.type = BTRFS_INODE_ITEM_KEY;
625 key.offset = 0;
626 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
627 if (IS_ERR(inode)) {
628 err = PTR_ERR(inode);
629 goto fail_close;
630 }
631
632 root_dentry = d_alloc_root(inode);
633 if (!root_dentry) {
634 iput(inode);
635 err = -ENOMEM;
636 goto fail_close;
637 }
638
639 sb->s_root = root_dentry;
640
641 save_mount_options(sb, data);
642 cleancache_init_fs(sb);
643 return 0;
644
645fail_close:
646 close_ctree(tree_root);
647 return err;
648}
649
650int btrfs_sync_fs(struct super_block *sb, int wait)
651{
652 struct btrfs_trans_handle *trans;
653 struct btrfs_root *root = btrfs_sb(sb);
654 int ret;
655
656 trace_btrfs_sync_fs(wait);
657
658 if (!wait) {
659 filemap_flush(root->fs_info->btree_inode->i_mapping);
660 return 0;
661 }
662
663 btrfs_start_delalloc_inodes(root, 0);
664 btrfs_wait_ordered_extents(root, 0, 0);
665
666 trans = btrfs_start_transaction(root, 0);
667 if (IS_ERR(trans))
668 return PTR_ERR(trans);
669 ret = btrfs_commit_transaction(trans, root);
670 return ret;
671}
672
673static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
674{
675 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
676 struct btrfs_fs_info *info = root->fs_info;
677 char *compress_type;
678
679 if (btrfs_test_opt(root, DEGRADED))
680 seq_puts(seq, ",degraded");
681 if (btrfs_test_opt(root, NODATASUM))
682 seq_puts(seq, ",nodatasum");
683 if (btrfs_test_opt(root, NODATACOW))
684 seq_puts(seq, ",nodatacow");
685 if (btrfs_test_opt(root, NOBARRIER))
686 seq_puts(seq, ",nobarrier");
687 if (info->max_inline != 8192 * 1024)
688 seq_printf(seq, ",max_inline=%llu",
689 (unsigned long long)info->max_inline);
690 if (info->alloc_start != 0)
691 seq_printf(seq, ",alloc_start=%llu",
692 (unsigned long long)info->alloc_start);
693 if (info->thread_pool_size != min_t(unsigned long,
694 num_online_cpus() + 2, 8))
695 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
696 if (btrfs_test_opt(root, COMPRESS)) {
697 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
698 compress_type = "zlib";
699 else
700 compress_type = "lzo";
701 if (btrfs_test_opt(root, FORCE_COMPRESS))
702 seq_printf(seq, ",compress-force=%s", compress_type);
703 else
704 seq_printf(seq, ",compress=%s", compress_type);
705 }
706 if (btrfs_test_opt(root, NOSSD))
707 seq_puts(seq, ",nossd");
708 if (btrfs_test_opt(root, SSD_SPREAD))
709 seq_puts(seq, ",ssd_spread");
710 else if (btrfs_test_opt(root, SSD))
711 seq_puts(seq, ",ssd");
712 if (btrfs_test_opt(root, NOTREELOG))
713 seq_puts(seq, ",notreelog");
714 if (btrfs_test_opt(root, FLUSHONCOMMIT))
715 seq_puts(seq, ",flushoncommit");
716 if (btrfs_test_opt(root, DISCARD))
717 seq_puts(seq, ",discard");
718 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
719 seq_puts(seq, ",noacl");
720 if (btrfs_test_opt(root, SPACE_CACHE))
721 seq_puts(seq, ",space_cache");
722 if (btrfs_test_opt(root, CLEAR_CACHE))
723 seq_puts(seq, ",clear_cache");
724 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
725 seq_puts(seq, ",user_subvol_rm_allowed");
726 if (btrfs_test_opt(root, ENOSPC_DEBUG))
727 seq_puts(seq, ",enospc_debug");
728 if (btrfs_test_opt(root, AUTO_DEFRAG))
729 seq_puts(seq, ",autodefrag");
730 if (btrfs_test_opt(root, INODE_MAP_CACHE))
731 seq_puts(seq, ",inode_cache");
732 return 0;
733}
734
735static int btrfs_test_super(struct super_block *s, void *data)
736{
737 struct btrfs_root *test_root = data;
738 struct btrfs_root *root = btrfs_sb(s);
739
740 /*
741 * If this super block is going away, return false as it
742 * can't match as an existing super block.
743 */
744 if (!atomic_read(&s->s_active))
745 return 0;
746 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
747}
748
749static int btrfs_set_super(struct super_block *s, void *data)
750{
751 s->s_fs_info = data;
752
753 return set_anon_super(s, data);
754}
755
756
757/*
758 * Find a superblock for the given device / mount point.
759 *
760 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
761 * for multiple device setup. Make sure to keep it in sync.
762 */
763static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
764 const char *device_name, void *data)
765{
766 struct block_device *bdev = NULL;
767 struct super_block *s;
768 struct dentry *root;
769 struct btrfs_fs_devices *fs_devices = NULL;
770 struct btrfs_root *tree_root = NULL;
771 struct btrfs_fs_info *fs_info = NULL;
772 fmode_t mode = FMODE_READ;
773 char *subvol_name = NULL;
774 u64 subvol_objectid = 0;
775 u64 subvol_rootid = 0;
776 int error = 0;
777
778 if (!(flags & MS_RDONLY))
779 mode |= FMODE_WRITE;
780
781 error = btrfs_parse_early_options(data, mode, fs_type,
782 &subvol_name, &subvol_objectid,
783 &subvol_rootid, &fs_devices);
784 if (error)
785 return ERR_PTR(error);
786
787 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
788 if (error)
789 goto error_free_subvol_name;
790
791 error = btrfs_open_devices(fs_devices, mode, fs_type);
792 if (error)
793 goto error_free_subvol_name;
794
795 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
796 error = -EACCES;
797 goto error_close_devices;
798 }
799
800 /*
801 * Setup a dummy root and fs_info for test/set super. This is because
802 * we don't actually fill this stuff out until open_ctree, but we need
803 * it for searching for existing supers, so this lets us do that and
804 * then open_ctree will properly initialize everything later.
805 */
806 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
807 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
808 if (!fs_info || !tree_root) {
809 error = -ENOMEM;
810 goto error_close_devices;
811 }
812 fs_info->tree_root = tree_root;
813 fs_info->fs_devices = fs_devices;
814 tree_root->fs_info = fs_info;
815
816 bdev = fs_devices->latest_bdev;
817 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
818 if (IS_ERR(s))
819 goto error_s;
820
821 if (s->s_root) {
822 if ((flags ^ s->s_flags) & MS_RDONLY) {
823 deactivate_locked_super(s);
824 error = -EBUSY;
825 goto error_close_devices;
826 }
827
828 btrfs_close_devices(fs_devices);
829 kfree(fs_info);
830 kfree(tree_root);
831 } else {
832 char b[BDEVNAME_SIZE];
833
834 s->s_flags = flags | MS_NOSEC;
835 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
836 error = btrfs_fill_super(s, fs_devices, data,
837 flags & MS_SILENT ? 1 : 0);
838 if (error) {
839 deactivate_locked_super(s);
840 goto error_free_subvol_name;
841 }
842
843 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
844 s->s_flags |= MS_ACTIVE;
845 }
846
847 /* if they gave us a subvolume name bind mount into that */
848 if (strcmp(subvol_name, ".")) {
849 struct dentry *new_root;
850
851 root = get_default_root(s, subvol_rootid);
852 if (IS_ERR(root)) {
853 error = PTR_ERR(root);
854 deactivate_locked_super(s);
855 goto error_free_subvol_name;
856 }
857
858 mutex_lock(&root->d_inode->i_mutex);
859 new_root = lookup_one_len(subvol_name, root,
860 strlen(subvol_name));
861 mutex_unlock(&root->d_inode->i_mutex);
862
863 if (IS_ERR(new_root)) {
864 dput(root);
865 deactivate_locked_super(s);
866 error = PTR_ERR(new_root);
867 goto error_free_subvol_name;
868 }
869 if (!new_root->d_inode) {
870 dput(root);
871 dput(new_root);
872 deactivate_locked_super(s);
873 error = -ENXIO;
874 goto error_free_subvol_name;
875 }
876 dput(root);
877 root = new_root;
878 } else {
879 root = get_default_root(s, subvol_objectid);
880 if (IS_ERR(root)) {
881 error = PTR_ERR(root);
882 deactivate_locked_super(s);
883 goto error_free_subvol_name;
884 }
885 }
886
887 kfree(subvol_name);
888 return root;
889
890error_s:
891 error = PTR_ERR(s);
892error_close_devices:
893 btrfs_close_devices(fs_devices);
894 kfree(fs_info);
895 kfree(tree_root);
896error_free_subvol_name:
897 kfree(subvol_name);
898 return ERR_PTR(error);
899}
900
901static int btrfs_remount(struct super_block *sb, int *flags, char *data)
902{
903 struct btrfs_root *root = btrfs_sb(sb);
904 int ret;
905
906 ret = btrfs_parse_options(root, data);
907 if (ret)
908 return -EINVAL;
909
910 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
911 return 0;
912
913 if (*flags & MS_RDONLY) {
914 sb->s_flags |= MS_RDONLY;
915
916 ret = btrfs_commit_super(root);
917 WARN_ON(ret);
918 } else {
919 if (root->fs_info->fs_devices->rw_devices == 0)
920 return -EACCES;
921
922 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
923 return -EINVAL;
924
925 ret = btrfs_cleanup_fs_roots(root->fs_info);
926 WARN_ON(ret);
927
928 /* recover relocation */
929 ret = btrfs_recover_relocation(root);
930 WARN_ON(ret);
931
932 sb->s_flags &= ~MS_RDONLY;
933 }
934
935 return 0;
936}
937
938/* Used to sort the devices by max_avail(descending sort) */
939static int btrfs_cmp_device_free_bytes(const void *dev_info1,
940 const void *dev_info2)
941{
942 if (((struct btrfs_device_info *)dev_info1)->max_avail >
943 ((struct btrfs_device_info *)dev_info2)->max_avail)
944 return -1;
945 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
946 ((struct btrfs_device_info *)dev_info2)->max_avail)
947 return 1;
948 else
949 return 0;
950}
951
952/*
953 * sort the devices by max_avail, in which max free extent size of each device
954 * is stored.(Descending Sort)
955 */
956static inline void btrfs_descending_sort_devices(
957 struct btrfs_device_info *devices,
958 size_t nr_devices)
959{
960 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
961 btrfs_cmp_device_free_bytes, NULL);
962}
963
964/*
965 * The helper to calc the free space on the devices that can be used to store
966 * file data.
967 */
968static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
969{
970 struct btrfs_fs_info *fs_info = root->fs_info;
971 struct btrfs_device_info *devices_info;
972 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
973 struct btrfs_device *device;
974 u64 skip_space;
975 u64 type;
976 u64 avail_space;
977 u64 used_space;
978 u64 min_stripe_size;
979 int min_stripes = 1;
980 int i = 0, nr_devices;
981 int ret;
982
983 nr_devices = fs_info->fs_devices->rw_devices;
984 BUG_ON(!nr_devices);
985
986 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
987 GFP_NOFS);
988 if (!devices_info)
989 return -ENOMEM;
990
991 /* calc min stripe number for data space alloction */
992 type = btrfs_get_alloc_profile(root, 1);
993 if (type & BTRFS_BLOCK_GROUP_RAID0)
994 min_stripes = 2;
995 else if (type & BTRFS_BLOCK_GROUP_RAID1)
996 min_stripes = 2;
997 else if (type & BTRFS_BLOCK_GROUP_RAID10)
998 min_stripes = 4;
999
1000 if (type & BTRFS_BLOCK_GROUP_DUP)
1001 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1002 else
1003 min_stripe_size = BTRFS_STRIPE_LEN;
1004
1005 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
1006 if (!device->in_fs_metadata)
1007 continue;
1008
1009 avail_space = device->total_bytes - device->bytes_used;
1010
1011 /* align with stripe_len */
1012 do_div(avail_space, BTRFS_STRIPE_LEN);
1013 avail_space *= BTRFS_STRIPE_LEN;
1014
1015 /*
1016 * In order to avoid overwritting the superblock on the drive,
1017 * btrfs starts at an offset of at least 1MB when doing chunk
1018 * allocation.
1019 */
1020 skip_space = 1024 * 1024;
1021
1022 /* user can set the offset in fs_info->alloc_start. */
1023 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1024 device->total_bytes)
1025 skip_space = max(fs_info->alloc_start, skip_space);
1026
1027 /*
1028 * btrfs can not use the free space in [0, skip_space - 1],
1029 * we must subtract it from the total. In order to implement
1030 * it, we account the used space in this range first.
1031 */
1032 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1033 &used_space);
1034 if (ret) {
1035 kfree(devices_info);
1036 return ret;
1037 }
1038
1039 /* calc the free space in [0, skip_space - 1] */
1040 skip_space -= used_space;
1041
1042 /*
1043 * we can use the free space in [0, skip_space - 1], subtract
1044 * it from the total.
1045 */
1046 if (avail_space && avail_space >= skip_space)
1047 avail_space -= skip_space;
1048 else
1049 avail_space = 0;
1050
1051 if (avail_space < min_stripe_size)
1052 continue;
1053
1054 devices_info[i].dev = device;
1055 devices_info[i].max_avail = avail_space;
1056
1057 i++;
1058 }
1059
1060 nr_devices = i;
1061
1062 btrfs_descending_sort_devices(devices_info, nr_devices);
1063
1064 i = nr_devices - 1;
1065 avail_space = 0;
1066 while (nr_devices >= min_stripes) {
1067 if (devices_info[i].max_avail >= min_stripe_size) {
1068 int j;
1069 u64 alloc_size;
1070
1071 avail_space += devices_info[i].max_avail * min_stripes;
1072 alloc_size = devices_info[i].max_avail;
1073 for (j = i + 1 - min_stripes; j <= i; j++)
1074 devices_info[j].max_avail -= alloc_size;
1075 }
1076 i--;
1077 nr_devices--;
1078 }
1079
1080 kfree(devices_info);
1081 *free_bytes = avail_space;
1082 return 0;
1083}
1084
1085static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1086{
1087 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
1088 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
1089 struct list_head *head = &root->fs_info->space_info;
1090 struct btrfs_space_info *found;
1091 u64 total_used = 0;
1092 u64 total_free_data = 0;
1093 int bits = dentry->d_sb->s_blocksize_bits;
1094 __be32 *fsid = (__be32 *)root->fs_info->fsid;
1095 int ret;
1096
1097 /* holding chunk_muext to avoid allocating new chunks */
1098 mutex_lock(&root->fs_info->chunk_mutex);
1099 rcu_read_lock();
1100 list_for_each_entry_rcu(found, head, list) {
1101 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1102 total_free_data += found->disk_total - found->disk_used;
1103 total_free_data -=
1104 btrfs_account_ro_block_groups_free_space(found);
1105 }
1106
1107 total_used += found->disk_used;
1108 }
1109 rcu_read_unlock();
1110
1111 buf->f_namelen = BTRFS_NAME_LEN;
1112 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1113 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1114 buf->f_bsize = dentry->d_sb->s_blocksize;
1115 buf->f_type = BTRFS_SUPER_MAGIC;
1116 buf->f_bavail = total_free_data;
1117 ret = btrfs_calc_avail_data_space(root, &total_free_data);
1118 if (ret) {
1119 mutex_unlock(&root->fs_info->chunk_mutex);
1120 return ret;
1121 }
1122 buf->f_bavail += total_free_data;
1123 buf->f_bavail = buf->f_bavail >> bits;
1124 mutex_unlock(&root->fs_info->chunk_mutex);
1125
1126 /* We treat it as constant endianness (it doesn't matter _which_)
1127 because we want the fsid to come out the same whether mounted
1128 on a big-endian or little-endian host */
1129 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1130 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1131 /* Mask in the root object ID too, to disambiguate subvols */
1132 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1133 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1134
1135 return 0;
1136}
1137
1138static struct file_system_type btrfs_fs_type = {
1139 .owner = THIS_MODULE,
1140 .name = "btrfs",
1141 .mount = btrfs_mount,
1142 .kill_sb = kill_anon_super,
1143 .fs_flags = FS_REQUIRES_DEV,
1144};
1145
1146/*
1147 * used by btrfsctl to scan devices when no FS is mounted
1148 */
1149static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1150 unsigned long arg)
1151{
1152 struct btrfs_ioctl_vol_args *vol;
1153 struct btrfs_fs_devices *fs_devices;
1154 int ret = -ENOTTY;
1155
1156 if (!capable(CAP_SYS_ADMIN))
1157 return -EPERM;
1158
1159 vol = memdup_user((void __user *)arg, sizeof(*vol));
1160 if (IS_ERR(vol))
1161 return PTR_ERR(vol);
1162
1163 switch (cmd) {
1164 case BTRFS_IOC_SCAN_DEV:
1165 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1166 &btrfs_fs_type, &fs_devices);
1167 break;
1168 }
1169
1170 kfree(vol);
1171 return ret;
1172}
1173
1174static int btrfs_freeze(struct super_block *sb)
1175{
1176 struct btrfs_root *root = btrfs_sb(sb);
1177 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1178 mutex_lock(&root->fs_info->cleaner_mutex);
1179 return 0;
1180}
1181
1182static int btrfs_unfreeze(struct super_block *sb)
1183{
1184 struct btrfs_root *root = btrfs_sb(sb);
1185 mutex_unlock(&root->fs_info->cleaner_mutex);
1186 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1187 return 0;
1188}
1189
1190static const struct super_operations btrfs_super_ops = {
1191 .drop_inode = btrfs_drop_inode,
1192 .evict_inode = btrfs_evict_inode,
1193 .put_super = btrfs_put_super,
1194 .sync_fs = btrfs_sync_fs,
1195 .show_options = btrfs_show_options,
1196 .write_inode = btrfs_write_inode,
1197 .dirty_inode = btrfs_dirty_inode,
1198 .alloc_inode = btrfs_alloc_inode,
1199 .destroy_inode = btrfs_destroy_inode,
1200 .statfs = btrfs_statfs,
1201 .remount_fs = btrfs_remount,
1202 .freeze_fs = btrfs_freeze,
1203 .unfreeze_fs = btrfs_unfreeze,
1204};
1205
1206static const struct file_operations btrfs_ctl_fops = {
1207 .unlocked_ioctl = btrfs_control_ioctl,
1208 .compat_ioctl = btrfs_control_ioctl,
1209 .owner = THIS_MODULE,
1210 .llseek = noop_llseek,
1211};
1212
1213static struct miscdevice btrfs_misc = {
1214 .minor = BTRFS_MINOR,
1215 .name = "btrfs-control",
1216 .fops = &btrfs_ctl_fops
1217};
1218
1219MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1220MODULE_ALIAS("devname:btrfs-control");
1221
1222static int btrfs_interface_init(void)
1223{
1224 return misc_register(&btrfs_misc);
1225}
1226
1227static void btrfs_interface_exit(void)
1228{
1229 if (misc_deregister(&btrfs_misc) < 0)
1230 printk(KERN_INFO "misc_deregister failed for control device");
1231}
1232
1233static int __init init_btrfs_fs(void)
1234{
1235 int err;
1236
1237 err = btrfs_init_sysfs();
1238 if (err)
1239 return err;
1240
1241 err = btrfs_init_compress();
1242 if (err)
1243 goto free_sysfs;
1244
1245 err = btrfs_init_cachep();
1246 if (err)
1247 goto free_compress;
1248
1249 err = extent_io_init();
1250 if (err)
1251 goto free_cachep;
1252
1253 err = extent_map_init();
1254 if (err)
1255 goto free_extent_io;
1256
1257 err = btrfs_delayed_inode_init();
1258 if (err)
1259 goto free_extent_map;
1260
1261 err = btrfs_interface_init();
1262 if (err)
1263 goto free_delayed_inode;
1264
1265 err = register_filesystem(&btrfs_fs_type);
1266 if (err)
1267 goto unregister_ioctl;
1268
1269 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1270 return 0;
1271
1272unregister_ioctl:
1273 btrfs_interface_exit();
1274free_delayed_inode:
1275 btrfs_delayed_inode_exit();
1276free_extent_map:
1277 extent_map_exit();
1278free_extent_io:
1279 extent_io_exit();
1280free_cachep:
1281 btrfs_destroy_cachep();
1282free_compress:
1283 btrfs_exit_compress();
1284free_sysfs:
1285 btrfs_exit_sysfs();
1286 return err;
1287}
1288
1289static void __exit exit_btrfs_fs(void)
1290{
1291 btrfs_destroy_cachep();
1292 btrfs_delayed_inode_exit();
1293 extent_map_exit();
1294 extent_io_exit();
1295 btrfs_interface_exit();
1296 unregister_filesystem(&btrfs_fs_type);
1297 btrfs_exit_sysfs();
1298 btrfs_cleanup_fs_uuids();
1299 btrfs_exit_compress();
1300}
1301
1302module_init(init_btrfs_fs)
1303module_exit(exit_btrfs_fs)
1304
1305MODULE_LICENSE("GPL");
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 <linux/cleancache.h>
43#include <linux/ratelimit.h>
44#include <linux/btrfs.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 "print-tree.h"
51#include "hash.h"
52#include "props.h"
53#include "xattr.h"
54#include "volumes.h"
55#include "export.h"
56#include "compression.h"
57#include "rcu-string.h"
58#include "dev-replace.h"
59#include "free-space-cache.h"
60#include "backref.h"
61#include "tests/btrfs-tests.h"
62
63#define CREATE_TRACE_POINTS
64#include <trace/events/btrfs.h>
65
66static const struct super_operations btrfs_super_ops;
67static struct file_system_type btrfs_fs_type;
68
69static int btrfs_remount(struct super_block *sb, int *flags, char *data);
70
71static const char *btrfs_decode_error(int errno)
72{
73 char *errstr = "unknown";
74
75 switch (errno) {
76 case -EIO:
77 errstr = "IO failure";
78 break;
79 case -ENOMEM:
80 errstr = "Out of memory";
81 break;
82 case -EROFS:
83 errstr = "Readonly filesystem";
84 break;
85 case -EEXIST:
86 errstr = "Object already exists";
87 break;
88 case -ENOSPC:
89 errstr = "No space left";
90 break;
91 case -ENOENT:
92 errstr = "No such entry";
93 break;
94 }
95
96 return errstr;
97}
98
99static void save_error_info(struct btrfs_fs_info *fs_info)
100{
101 /*
102 * today we only save the error info into ram. Long term we'll
103 * also send it down to the disk
104 */
105 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
106}
107
108/* btrfs handle error by forcing the filesystem readonly */
109static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
110{
111 struct super_block *sb = fs_info->sb;
112
113 if (sb->s_flags & MS_RDONLY)
114 return;
115
116 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
117 sb->s_flags |= MS_RDONLY;
118 btrfs_info(fs_info, "forced readonly");
119 /*
120 * Note that a running device replace operation is not
121 * canceled here although there is no way to update
122 * the progress. It would add the risk of a deadlock,
123 * therefore the canceling is ommited. The only penalty
124 * is that some I/O remains active until the procedure
125 * completes. The next time when the filesystem is
126 * mounted writeable again, the device replace
127 * operation continues.
128 */
129 }
130}
131
132#ifdef CONFIG_PRINTK
133/*
134 * __btrfs_std_error decodes expected errors from the caller and
135 * invokes the approciate error response.
136 */
137void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
138 unsigned int line, int errno, const char *fmt, ...)
139{
140 struct super_block *sb = fs_info->sb;
141 const char *errstr;
142
143 /*
144 * Special case: if the error is EROFS, and we're already
145 * under MS_RDONLY, then it is safe here.
146 */
147 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
148 return;
149
150 errstr = btrfs_decode_error(errno);
151 if (fmt) {
152 struct va_format vaf;
153 va_list args;
154
155 va_start(args, fmt);
156 vaf.fmt = fmt;
157 vaf.va = &args;
158
159 printk(KERN_CRIT
160 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
161 sb->s_id, function, line, errno, errstr, &vaf);
162 va_end(args);
163 } else {
164 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
165 sb->s_id, function, line, errno, errstr);
166 }
167
168 /* Don't go through full error handling during mount */
169 save_error_info(fs_info);
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
172}
173
174static const char * const logtypes[] = {
175 "emergency",
176 "alert",
177 "critical",
178 "error",
179 "warning",
180 "notice",
181 "info",
182 "debug",
183};
184
185void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
186{
187 struct super_block *sb = fs_info->sb;
188 char lvl[4];
189 struct va_format vaf;
190 va_list args;
191 const char *type = logtypes[4];
192 int kern_level;
193
194 va_start(args, fmt);
195
196 kern_level = printk_get_level(fmt);
197 if (kern_level) {
198 size_t size = printk_skip_level(fmt) - fmt;
199 memcpy(lvl, fmt, size);
200 lvl[size] = '\0';
201 fmt += size;
202 type = logtypes[kern_level - '0'];
203 } else
204 *lvl = '\0';
205
206 vaf.fmt = fmt;
207 vaf.va = &args;
208
209 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
210
211 va_end(args);
212}
213
214#else
215
216void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
217 unsigned int line, int errno, const char *fmt, ...)
218{
219 struct super_block *sb = fs_info->sb;
220
221 /*
222 * Special case: if the error is EROFS, and we're already
223 * under MS_RDONLY, then it is safe here.
224 */
225 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
226 return;
227
228 /* Don't go through full error handling during mount */
229 if (sb->s_flags & MS_BORN) {
230 save_error_info(fs_info);
231 btrfs_handle_error(fs_info);
232 }
233}
234#endif
235
236/*
237 * We only mark the transaction aborted and then set the file system read-only.
238 * This will prevent new transactions from starting or trying to join this
239 * one.
240 *
241 * This means that error recovery at the call site is limited to freeing
242 * any local memory allocations and passing the error code up without
243 * further cleanup. The transaction should complete as it normally would
244 * in the call path but will return -EIO.
245 *
246 * We'll complete the cleanup in btrfs_end_transaction and
247 * btrfs_commit_transaction.
248 */
249void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
250 struct btrfs_root *root, const char *function,
251 unsigned int line, int errno)
252{
253 /*
254 * Report first abort since mount
255 */
256 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
257 &root->fs_info->fs_state)) {
258 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
259 errno);
260 }
261 trans->aborted = errno;
262 /* Nothing used. The other threads that have joined this
263 * transaction may be able to continue. */
264 if (!trans->blocks_used) {
265 const char *errstr;
266
267 errstr = btrfs_decode_error(errno);
268 btrfs_warn(root->fs_info,
269 "%s:%d: Aborting unused transaction(%s).",
270 function, line, errstr);
271 return;
272 }
273 ACCESS_ONCE(trans->transaction->aborted) = errno;
274 /* Wake up anybody who may be waiting on this transaction */
275 wake_up(&root->fs_info->transaction_wait);
276 wake_up(&root->fs_info->transaction_blocked_wait);
277 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
278}
279/*
280 * __btrfs_panic decodes unexpected, fatal errors from the caller,
281 * issues an alert, and either panics or BUGs, depending on mount options.
282 */
283void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
284 unsigned int line, int errno, const char *fmt, ...)
285{
286 char *s_id = "<unknown>";
287 const char *errstr;
288 struct va_format vaf = { .fmt = fmt };
289 va_list args;
290
291 if (fs_info)
292 s_id = fs_info->sb->s_id;
293
294 va_start(args, fmt);
295 vaf.va = &args;
296
297 errstr = btrfs_decode_error(errno);
298 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
299 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
300 s_id, function, line, &vaf, errno, errstr);
301
302 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
303 function, line, &vaf, errno, errstr);
304 va_end(args);
305 /* Caller calls BUG() */
306}
307
308static void btrfs_put_super(struct super_block *sb)
309{
310 (void)close_ctree(btrfs_sb(sb)->tree_root);
311 /* FIXME: need to fix VFS to return error? */
312 /* AV: return it _where_? ->put_super() can be triggered by any number
313 * of async events, up to and including delivery of SIGKILL to the
314 * last process that kept it busy. Or segfault in the aforementioned
315 * process... Whom would you report that to?
316 */
317}
318
319enum {
320 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
321 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
322 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
323 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
324 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
325 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
326 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
327 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
328 Opt_check_integrity, Opt_check_integrity_including_extent_data,
329 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
330 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
331 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
332 Opt_datasum, Opt_treelog, Opt_noinode_cache,
333 Opt_err,
334};
335
336static match_table_t tokens = {
337 {Opt_degraded, "degraded"},
338 {Opt_subvol, "subvol=%s"},
339 {Opt_subvolid, "subvolid=%s"},
340 {Opt_device, "device=%s"},
341 {Opt_nodatasum, "nodatasum"},
342 {Opt_datasum, "datasum"},
343 {Opt_nodatacow, "nodatacow"},
344 {Opt_datacow, "datacow"},
345 {Opt_nobarrier, "nobarrier"},
346 {Opt_barrier, "barrier"},
347 {Opt_max_inline, "max_inline=%s"},
348 {Opt_alloc_start, "alloc_start=%s"},
349 {Opt_thread_pool, "thread_pool=%d"},
350 {Opt_compress, "compress"},
351 {Opt_compress_type, "compress=%s"},
352 {Opt_compress_force, "compress-force"},
353 {Opt_compress_force_type, "compress-force=%s"},
354 {Opt_ssd, "ssd"},
355 {Opt_ssd_spread, "ssd_spread"},
356 {Opt_nossd, "nossd"},
357 {Opt_acl, "acl"},
358 {Opt_noacl, "noacl"},
359 {Opt_notreelog, "notreelog"},
360 {Opt_treelog, "treelog"},
361 {Opt_flushoncommit, "flushoncommit"},
362 {Opt_noflushoncommit, "noflushoncommit"},
363 {Opt_ratio, "metadata_ratio=%d"},
364 {Opt_discard, "discard"},
365 {Opt_nodiscard, "nodiscard"},
366 {Opt_space_cache, "space_cache"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug, "enospc_debug"},
370 {Opt_noenospc_debug, "noenospc_debug"},
371 {Opt_subvolrootid, "subvolrootid=%d"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_inode_cache, "inode_cache"},
375 {Opt_noinode_cache, "noinode_cache"},
376 {Opt_no_space_cache, "nospace_cache"},
377 {Opt_recovery, "recovery"},
378 {Opt_skip_balance, "skip_balance"},
379 {Opt_check_integrity, "check_int"},
380 {Opt_check_integrity_including_extent_data, "check_int_data"},
381 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_fatal_errors, "fatal_errors=%s"},
384 {Opt_commit_interval, "commit=%d"},
385 {Opt_err, NULL},
386};
387
388/*
389 * Regular mount options parser. Everything that is needed only when
390 * reading in a new superblock is parsed here.
391 * XXX JDM: This needs to be cleaned up for remount.
392 */
393int btrfs_parse_options(struct btrfs_root *root, char *options)
394{
395 struct btrfs_fs_info *info = root->fs_info;
396 substring_t args[MAX_OPT_ARGS];
397 char *p, *num, *orig = NULL;
398 u64 cache_gen;
399 int intarg;
400 int ret = 0;
401 char *compress_type;
402 bool compress_force = false;
403 bool compress = false;
404
405 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
406 if (cache_gen)
407 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
408
409 if (!options)
410 goto out;
411
412 /*
413 * strsep changes the string, duplicate it because parse_options
414 * gets called twice
415 */
416 options = kstrdup(options, GFP_NOFS);
417 if (!options)
418 return -ENOMEM;
419
420 orig = options;
421
422 while ((p = strsep(&options, ",")) != NULL) {
423 int token;
424 if (!*p)
425 continue;
426
427 token = match_token(p, tokens, args);
428 switch (token) {
429 case Opt_degraded:
430 btrfs_info(root->fs_info, "allowing degraded mounts");
431 btrfs_set_opt(info->mount_opt, DEGRADED);
432 break;
433 case Opt_subvol:
434 case Opt_subvolid:
435 case Opt_subvolrootid:
436 case Opt_device:
437 /*
438 * These are parsed by btrfs_parse_early_options
439 * and can be happily ignored here.
440 */
441 break;
442 case Opt_nodatasum:
443 btrfs_set_and_info(root, NODATASUM,
444 "setting nodatasum");
445 break;
446 case Opt_datasum:
447 if (btrfs_test_opt(root, NODATASUM)) {
448 if (btrfs_test_opt(root, NODATACOW))
449 btrfs_info(root->fs_info, "setting datasum, datacow enabled");
450 else
451 btrfs_info(root->fs_info, "setting datasum");
452 }
453 btrfs_clear_opt(info->mount_opt, NODATACOW);
454 btrfs_clear_opt(info->mount_opt, NODATASUM);
455 break;
456 case Opt_nodatacow:
457 if (!btrfs_test_opt(root, NODATACOW)) {
458 if (!btrfs_test_opt(root, COMPRESS) ||
459 !btrfs_test_opt(root, FORCE_COMPRESS)) {
460 btrfs_info(root->fs_info,
461 "setting nodatacow, compression disabled");
462 } else {
463 btrfs_info(root->fs_info, "setting nodatacow");
464 }
465 }
466 btrfs_clear_opt(info->mount_opt, COMPRESS);
467 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
468 btrfs_set_opt(info->mount_opt, NODATACOW);
469 btrfs_set_opt(info->mount_opt, NODATASUM);
470 break;
471 case Opt_datacow:
472 btrfs_clear_and_info(root, NODATACOW,
473 "setting datacow");
474 break;
475 case Opt_compress_force:
476 case Opt_compress_force_type:
477 compress_force = true;
478 /* Fallthrough */
479 case Opt_compress:
480 case Opt_compress_type:
481 compress = true;
482 if (token == Opt_compress ||
483 token == Opt_compress_force ||
484 strcmp(args[0].from, "zlib") == 0) {
485 compress_type = "zlib";
486 info->compress_type = BTRFS_COMPRESS_ZLIB;
487 btrfs_set_opt(info->mount_opt, COMPRESS);
488 btrfs_clear_opt(info->mount_opt, NODATACOW);
489 btrfs_clear_opt(info->mount_opt, NODATASUM);
490 } else if (strcmp(args[0].from, "lzo") == 0) {
491 compress_type = "lzo";
492 info->compress_type = BTRFS_COMPRESS_LZO;
493 btrfs_set_opt(info->mount_opt, COMPRESS);
494 btrfs_clear_opt(info->mount_opt, NODATACOW);
495 btrfs_clear_opt(info->mount_opt, NODATASUM);
496 btrfs_set_fs_incompat(info, COMPRESS_LZO);
497 } else if (strncmp(args[0].from, "no", 2) == 0) {
498 compress_type = "no";
499 btrfs_clear_opt(info->mount_opt, COMPRESS);
500 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
501 compress_force = false;
502 } else {
503 ret = -EINVAL;
504 goto out;
505 }
506
507 if (compress_force) {
508 btrfs_set_and_info(root, FORCE_COMPRESS,
509 "force %s compression",
510 compress_type);
511 } else if (compress) {
512 if (!btrfs_test_opt(root, COMPRESS))
513 btrfs_info(root->fs_info,
514 "btrfs: use %s compression\n",
515 compress_type);
516 }
517 break;
518 case Opt_ssd:
519 btrfs_set_and_info(root, SSD,
520 "use ssd allocation scheme");
521 break;
522 case Opt_ssd_spread:
523 btrfs_set_and_info(root, SSD_SPREAD,
524 "use spread ssd allocation scheme");
525 break;
526 case Opt_nossd:
527 btrfs_clear_and_info(root, NOSSD,
528 "not using ssd allocation scheme");
529 btrfs_clear_opt(info->mount_opt, SSD);
530 break;
531 case Opt_barrier:
532 btrfs_clear_and_info(root, NOBARRIER,
533 "turning on barriers");
534 break;
535 case Opt_nobarrier:
536 btrfs_set_and_info(root, NOBARRIER,
537 "turning off barriers");
538 break;
539 case Opt_thread_pool:
540 ret = match_int(&args[0], &intarg);
541 if (ret) {
542 goto out;
543 } else if (intarg > 0) {
544 info->thread_pool_size = intarg;
545 } else {
546 ret = -EINVAL;
547 goto out;
548 }
549 break;
550 case Opt_max_inline:
551 num = match_strdup(&args[0]);
552 if (num) {
553 info->max_inline = memparse(num, NULL);
554 kfree(num);
555
556 if (info->max_inline) {
557 info->max_inline = min_t(u64,
558 info->max_inline,
559 root->sectorsize);
560 }
561 btrfs_info(root->fs_info, "max_inline at %llu",
562 info->max_inline);
563 } else {
564 ret = -ENOMEM;
565 goto out;
566 }
567 break;
568 case Opt_alloc_start:
569 num = match_strdup(&args[0]);
570 if (num) {
571 mutex_lock(&info->chunk_mutex);
572 info->alloc_start = memparse(num, NULL);
573 mutex_unlock(&info->chunk_mutex);
574 kfree(num);
575 btrfs_info(root->fs_info, "allocations start at %llu",
576 info->alloc_start);
577 } else {
578 ret = -ENOMEM;
579 goto out;
580 }
581 break;
582 case Opt_acl:
583 root->fs_info->sb->s_flags |= MS_POSIXACL;
584 break;
585 case Opt_noacl:
586 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
587 break;
588 case Opt_notreelog:
589 btrfs_set_and_info(root, NOTREELOG,
590 "disabling tree log");
591 break;
592 case Opt_treelog:
593 btrfs_clear_and_info(root, NOTREELOG,
594 "enabling tree log");
595 break;
596 case Opt_flushoncommit:
597 btrfs_set_and_info(root, FLUSHONCOMMIT,
598 "turning on flush-on-commit");
599 break;
600 case Opt_noflushoncommit:
601 btrfs_clear_and_info(root, FLUSHONCOMMIT,
602 "turning off flush-on-commit");
603 break;
604 case Opt_ratio:
605 ret = match_int(&args[0], &intarg);
606 if (ret) {
607 goto out;
608 } else if (intarg >= 0) {
609 info->metadata_ratio = intarg;
610 btrfs_info(root->fs_info, "metadata ratio %d",
611 info->metadata_ratio);
612 } else {
613 ret = -EINVAL;
614 goto out;
615 }
616 break;
617 case Opt_discard:
618 btrfs_set_and_info(root, DISCARD,
619 "turning on discard");
620 break;
621 case Opt_nodiscard:
622 btrfs_clear_and_info(root, DISCARD,
623 "turning off discard");
624 break;
625 case Opt_space_cache:
626 btrfs_set_and_info(root, SPACE_CACHE,
627 "enabling disk space caching");
628 break;
629 case Opt_rescan_uuid_tree:
630 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
631 break;
632 case Opt_no_space_cache:
633 btrfs_clear_and_info(root, SPACE_CACHE,
634 "disabling disk space caching");
635 break;
636 case Opt_inode_cache:
637 btrfs_set_and_info(root, CHANGE_INODE_CACHE,
638 "enabling inode map caching");
639 break;
640 case Opt_noinode_cache:
641 btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
642 "disabling inode map caching");
643 break;
644 case Opt_clear_cache:
645 btrfs_set_and_info(root, CLEAR_CACHE,
646 "force clearing of disk cache");
647 break;
648 case Opt_user_subvol_rm_allowed:
649 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
650 break;
651 case Opt_enospc_debug:
652 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
653 break;
654 case Opt_noenospc_debug:
655 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
656 break;
657 case Opt_defrag:
658 btrfs_set_and_info(root, AUTO_DEFRAG,
659 "enabling auto defrag");
660 break;
661 case Opt_nodefrag:
662 btrfs_clear_and_info(root, AUTO_DEFRAG,
663 "disabling auto defrag");
664 break;
665 case Opt_recovery:
666 btrfs_info(root->fs_info, "enabling auto recovery");
667 btrfs_set_opt(info->mount_opt, RECOVERY);
668 break;
669 case Opt_skip_balance:
670 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
671 break;
672#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
673 case Opt_check_integrity_including_extent_data:
674 btrfs_info(root->fs_info,
675 "enabling check integrity including extent data");
676 btrfs_set_opt(info->mount_opt,
677 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
678 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
679 break;
680 case Opt_check_integrity:
681 btrfs_info(root->fs_info, "enabling check integrity");
682 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
683 break;
684 case Opt_check_integrity_print_mask:
685 ret = match_int(&args[0], &intarg);
686 if (ret) {
687 goto out;
688 } else if (intarg >= 0) {
689 info->check_integrity_print_mask = intarg;
690 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
691 info->check_integrity_print_mask);
692 } else {
693 ret = -EINVAL;
694 goto out;
695 }
696 break;
697#else
698 case Opt_check_integrity_including_extent_data:
699 case Opt_check_integrity:
700 case Opt_check_integrity_print_mask:
701 btrfs_err(root->fs_info,
702 "support for check_integrity* not compiled in!");
703 ret = -EINVAL;
704 goto out;
705#endif
706 case Opt_fatal_errors:
707 if (strcmp(args[0].from, "panic") == 0)
708 btrfs_set_opt(info->mount_opt,
709 PANIC_ON_FATAL_ERROR);
710 else if (strcmp(args[0].from, "bug") == 0)
711 btrfs_clear_opt(info->mount_opt,
712 PANIC_ON_FATAL_ERROR);
713 else {
714 ret = -EINVAL;
715 goto out;
716 }
717 break;
718 case Opt_commit_interval:
719 intarg = 0;
720 ret = match_int(&args[0], &intarg);
721 if (ret < 0) {
722 btrfs_err(root->fs_info, "invalid commit interval");
723 ret = -EINVAL;
724 goto out;
725 }
726 if (intarg > 0) {
727 if (intarg > 300) {
728 btrfs_warn(root->fs_info, "excessive commit interval %d",
729 intarg);
730 }
731 info->commit_interval = intarg;
732 } else {
733 btrfs_info(root->fs_info, "using default commit interval %ds",
734 BTRFS_DEFAULT_COMMIT_INTERVAL);
735 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
736 }
737 break;
738 case Opt_err:
739 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
740 ret = -EINVAL;
741 goto out;
742 default:
743 break;
744 }
745 }
746out:
747 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
748 btrfs_info(root->fs_info, "disk space caching is enabled");
749 kfree(orig);
750 return ret;
751}
752
753/*
754 * Parse mount options that are required early in the mount process.
755 *
756 * All other options will be parsed on much later in the mount process and
757 * only when we need to allocate a new super block.
758 */
759static int btrfs_parse_early_options(const char *options, fmode_t flags,
760 void *holder, char **subvol_name, u64 *subvol_objectid,
761 struct btrfs_fs_devices **fs_devices)
762{
763 substring_t args[MAX_OPT_ARGS];
764 char *device_name, *opts, *orig, *p;
765 char *num = NULL;
766 int error = 0;
767
768 if (!options)
769 return 0;
770
771 /*
772 * strsep changes the string, duplicate it because parse_options
773 * gets called twice
774 */
775 opts = kstrdup(options, GFP_KERNEL);
776 if (!opts)
777 return -ENOMEM;
778 orig = opts;
779
780 while ((p = strsep(&opts, ",")) != NULL) {
781 int token;
782 if (!*p)
783 continue;
784
785 token = match_token(p, tokens, args);
786 switch (token) {
787 case Opt_subvol:
788 kfree(*subvol_name);
789 *subvol_name = match_strdup(&args[0]);
790 if (!*subvol_name) {
791 error = -ENOMEM;
792 goto out;
793 }
794 break;
795 case Opt_subvolid:
796 num = match_strdup(&args[0]);
797 if (num) {
798 *subvol_objectid = memparse(num, NULL);
799 kfree(num);
800 /* we want the original fs_tree */
801 if (!*subvol_objectid)
802 *subvol_objectid =
803 BTRFS_FS_TREE_OBJECTID;
804 } else {
805 error = -EINVAL;
806 goto out;
807 }
808 break;
809 case Opt_subvolrootid:
810 printk(KERN_WARNING
811 "BTRFS: 'subvolrootid' mount option is deprecated and has "
812 "no effect\n");
813 break;
814 case Opt_device:
815 device_name = match_strdup(&args[0]);
816 if (!device_name) {
817 error = -ENOMEM;
818 goto out;
819 }
820 error = btrfs_scan_one_device(device_name,
821 flags, holder, fs_devices);
822 kfree(device_name);
823 if (error)
824 goto out;
825 break;
826 default:
827 break;
828 }
829 }
830
831out:
832 kfree(orig);
833 return error;
834}
835
836static struct dentry *get_default_root(struct super_block *sb,
837 u64 subvol_objectid)
838{
839 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
840 struct btrfs_root *root = fs_info->tree_root;
841 struct btrfs_root *new_root;
842 struct btrfs_dir_item *di;
843 struct btrfs_path *path;
844 struct btrfs_key location;
845 struct inode *inode;
846 struct dentry *dentry;
847 u64 dir_id;
848 int new = 0;
849
850 /*
851 * We have a specific subvol we want to mount, just setup location and
852 * go look up the root.
853 */
854 if (subvol_objectid) {
855 location.objectid = subvol_objectid;
856 location.type = BTRFS_ROOT_ITEM_KEY;
857 location.offset = (u64)-1;
858 goto find_root;
859 }
860
861 path = btrfs_alloc_path();
862 if (!path)
863 return ERR_PTR(-ENOMEM);
864 path->leave_spinning = 1;
865
866 /*
867 * Find the "default" dir item which points to the root item that we
868 * will mount by default if we haven't been given a specific subvolume
869 * to mount.
870 */
871 dir_id = btrfs_super_root_dir(fs_info->super_copy);
872 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
873 if (IS_ERR(di)) {
874 btrfs_free_path(path);
875 return ERR_CAST(di);
876 }
877 if (!di) {
878 /*
879 * Ok the default dir item isn't there. This is weird since
880 * it's always been there, but don't freak out, just try and
881 * mount to root most subvolume.
882 */
883 btrfs_free_path(path);
884 dir_id = BTRFS_FIRST_FREE_OBJECTID;
885 new_root = fs_info->fs_root;
886 goto setup_root;
887 }
888
889 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
890 btrfs_free_path(path);
891
892find_root:
893 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
894 if (IS_ERR(new_root))
895 return ERR_CAST(new_root);
896
897 dir_id = btrfs_root_dirid(&new_root->root_item);
898setup_root:
899 location.objectid = dir_id;
900 location.type = BTRFS_INODE_ITEM_KEY;
901 location.offset = 0;
902
903 inode = btrfs_iget(sb, &location, new_root, &new);
904 if (IS_ERR(inode))
905 return ERR_CAST(inode);
906
907 /*
908 * If we're just mounting the root most subvol put the inode and return
909 * a reference to the dentry. We will have already gotten a reference
910 * to the inode in btrfs_fill_super so we're good to go.
911 */
912 if (!new && sb->s_root->d_inode == inode) {
913 iput(inode);
914 return dget(sb->s_root);
915 }
916
917 dentry = d_obtain_alias(inode);
918 if (!IS_ERR(dentry)) {
919 spin_lock(&dentry->d_lock);
920 dentry->d_flags &= ~DCACHE_DISCONNECTED;
921 spin_unlock(&dentry->d_lock);
922 }
923 return dentry;
924}
925
926static int btrfs_fill_super(struct super_block *sb,
927 struct btrfs_fs_devices *fs_devices,
928 void *data, int silent)
929{
930 struct inode *inode;
931 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
932 struct btrfs_key key;
933 int err;
934
935 sb->s_maxbytes = MAX_LFS_FILESIZE;
936 sb->s_magic = BTRFS_SUPER_MAGIC;
937 sb->s_op = &btrfs_super_ops;
938 sb->s_d_op = &btrfs_dentry_operations;
939 sb->s_export_op = &btrfs_export_ops;
940 sb->s_xattr = btrfs_xattr_handlers;
941 sb->s_time_gran = 1;
942#ifdef CONFIG_BTRFS_FS_POSIX_ACL
943 sb->s_flags |= MS_POSIXACL;
944#endif
945 sb->s_flags |= MS_I_VERSION;
946 err = open_ctree(sb, fs_devices, (char *)data);
947 if (err) {
948 printk(KERN_ERR "BTRFS: open_ctree failed\n");
949 return err;
950 }
951
952 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
953 key.type = BTRFS_INODE_ITEM_KEY;
954 key.offset = 0;
955 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
956 if (IS_ERR(inode)) {
957 err = PTR_ERR(inode);
958 goto fail_close;
959 }
960
961 sb->s_root = d_make_root(inode);
962 if (!sb->s_root) {
963 err = -ENOMEM;
964 goto fail_close;
965 }
966
967 save_mount_options(sb, data);
968 cleancache_init_fs(sb);
969 sb->s_flags |= MS_ACTIVE;
970 return 0;
971
972fail_close:
973 close_ctree(fs_info->tree_root);
974 return err;
975}
976
977int btrfs_sync_fs(struct super_block *sb, int wait)
978{
979 struct btrfs_trans_handle *trans;
980 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
981 struct btrfs_root *root = fs_info->tree_root;
982
983 trace_btrfs_sync_fs(wait);
984
985 if (!wait) {
986 filemap_flush(fs_info->btree_inode->i_mapping);
987 return 0;
988 }
989
990 btrfs_wait_ordered_roots(fs_info, -1);
991
992 trans = btrfs_attach_transaction_barrier(root);
993 if (IS_ERR(trans)) {
994 /* no transaction, don't bother */
995 if (PTR_ERR(trans) == -ENOENT)
996 return 0;
997 return PTR_ERR(trans);
998 }
999 return btrfs_commit_transaction(trans, root);
1000}
1001
1002static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1003{
1004 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1005 struct btrfs_root *root = info->tree_root;
1006 char *compress_type;
1007
1008 if (btrfs_test_opt(root, DEGRADED))
1009 seq_puts(seq, ",degraded");
1010 if (btrfs_test_opt(root, NODATASUM))
1011 seq_puts(seq, ",nodatasum");
1012 if (btrfs_test_opt(root, NODATACOW))
1013 seq_puts(seq, ",nodatacow");
1014 if (btrfs_test_opt(root, NOBARRIER))
1015 seq_puts(seq, ",nobarrier");
1016 if (info->max_inline != 8192 * 1024)
1017 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1018 if (info->alloc_start != 0)
1019 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1020 if (info->thread_pool_size != min_t(unsigned long,
1021 num_online_cpus() + 2, 8))
1022 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1023 if (btrfs_test_opt(root, COMPRESS)) {
1024 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1025 compress_type = "zlib";
1026 else
1027 compress_type = "lzo";
1028 if (btrfs_test_opt(root, FORCE_COMPRESS))
1029 seq_printf(seq, ",compress-force=%s", compress_type);
1030 else
1031 seq_printf(seq, ",compress=%s", compress_type);
1032 }
1033 if (btrfs_test_opt(root, NOSSD))
1034 seq_puts(seq, ",nossd");
1035 if (btrfs_test_opt(root, SSD_SPREAD))
1036 seq_puts(seq, ",ssd_spread");
1037 else if (btrfs_test_opt(root, SSD))
1038 seq_puts(seq, ",ssd");
1039 if (btrfs_test_opt(root, NOTREELOG))
1040 seq_puts(seq, ",notreelog");
1041 if (btrfs_test_opt(root, FLUSHONCOMMIT))
1042 seq_puts(seq, ",flushoncommit");
1043 if (btrfs_test_opt(root, DISCARD))
1044 seq_puts(seq, ",discard");
1045 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1046 seq_puts(seq, ",noacl");
1047 if (btrfs_test_opt(root, SPACE_CACHE))
1048 seq_puts(seq, ",space_cache");
1049 else
1050 seq_puts(seq, ",nospace_cache");
1051 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1052 seq_puts(seq, ",rescan_uuid_tree");
1053 if (btrfs_test_opt(root, CLEAR_CACHE))
1054 seq_puts(seq, ",clear_cache");
1055 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1056 seq_puts(seq, ",user_subvol_rm_allowed");
1057 if (btrfs_test_opt(root, ENOSPC_DEBUG))
1058 seq_puts(seq, ",enospc_debug");
1059 if (btrfs_test_opt(root, AUTO_DEFRAG))
1060 seq_puts(seq, ",autodefrag");
1061 if (btrfs_test_opt(root, INODE_MAP_CACHE))
1062 seq_puts(seq, ",inode_cache");
1063 if (btrfs_test_opt(root, SKIP_BALANCE))
1064 seq_puts(seq, ",skip_balance");
1065 if (btrfs_test_opt(root, RECOVERY))
1066 seq_puts(seq, ",recovery");
1067#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1068 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1069 seq_puts(seq, ",check_int_data");
1070 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1071 seq_puts(seq, ",check_int");
1072 if (info->check_integrity_print_mask)
1073 seq_printf(seq, ",check_int_print_mask=%d",
1074 info->check_integrity_print_mask);
1075#endif
1076 if (info->metadata_ratio)
1077 seq_printf(seq, ",metadata_ratio=%d",
1078 info->metadata_ratio);
1079 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1080 seq_puts(seq, ",fatal_errors=panic");
1081 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1082 seq_printf(seq, ",commit=%d", info->commit_interval);
1083 return 0;
1084}
1085
1086static int btrfs_test_super(struct super_block *s, void *data)
1087{
1088 struct btrfs_fs_info *p = data;
1089 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1090
1091 return fs_info->fs_devices == p->fs_devices;
1092}
1093
1094static int btrfs_set_super(struct super_block *s, void *data)
1095{
1096 int err = set_anon_super(s, data);
1097 if (!err)
1098 s->s_fs_info = data;
1099 return err;
1100}
1101
1102/*
1103 * subvolumes are identified by ino 256
1104 */
1105static inline int is_subvolume_inode(struct inode *inode)
1106{
1107 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1108 return 1;
1109 return 0;
1110}
1111
1112/*
1113 * This will strip out the subvol=%s argument for an argument string and add
1114 * subvolid=0 to make sure we get the actual tree root for path walking to the
1115 * subvol we want.
1116 */
1117static char *setup_root_args(char *args)
1118{
1119 unsigned len = strlen(args) + 2 + 1;
1120 char *src, *dst, *buf;
1121
1122 /*
1123 * We need the same args as before, but with this substitution:
1124 * s!subvol=[^,]+!subvolid=0!
1125 *
1126 * Since the replacement string is up to 2 bytes longer than the
1127 * original, allocate strlen(args) + 2 + 1 bytes.
1128 */
1129
1130 src = strstr(args, "subvol=");
1131 /* This shouldn't happen, but just in case.. */
1132 if (!src)
1133 return NULL;
1134
1135 buf = dst = kmalloc(len, GFP_NOFS);
1136 if (!buf)
1137 return NULL;
1138
1139 /*
1140 * If the subvol= arg is not at the start of the string,
1141 * copy whatever precedes it into buf.
1142 */
1143 if (src != args) {
1144 *src++ = '\0';
1145 strcpy(buf, args);
1146 dst += strlen(args);
1147 }
1148
1149 strcpy(dst, "subvolid=0");
1150 dst += strlen("subvolid=0");
1151
1152 /*
1153 * If there is a "," after the original subvol=... string,
1154 * copy that suffix into our buffer. Otherwise, we're done.
1155 */
1156 src = strchr(src, ',');
1157 if (src)
1158 strcpy(dst, src);
1159
1160 return buf;
1161}
1162
1163static struct dentry *mount_subvol(const char *subvol_name, int flags,
1164 const char *device_name, char *data)
1165{
1166 struct dentry *root;
1167 struct vfsmount *mnt;
1168 char *newargs;
1169
1170 newargs = setup_root_args(data);
1171 if (!newargs)
1172 return ERR_PTR(-ENOMEM);
1173 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1174 newargs);
1175
1176 if (PTR_RET(mnt) == -EBUSY) {
1177 if (flags & MS_RDONLY) {
1178 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1179 newargs);
1180 } else {
1181 int r;
1182 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1183 newargs);
1184 if (IS_ERR(mnt)) {
1185 kfree(newargs);
1186 return ERR_CAST(mnt);
1187 }
1188
1189 r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1190 if (r < 0) {
1191 /* FIXME: release vfsmount mnt ??*/
1192 kfree(newargs);
1193 return ERR_PTR(r);
1194 }
1195 }
1196 }
1197
1198 kfree(newargs);
1199
1200 if (IS_ERR(mnt))
1201 return ERR_CAST(mnt);
1202
1203 root = mount_subtree(mnt, subvol_name);
1204
1205 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1206 struct super_block *s = root->d_sb;
1207 dput(root);
1208 root = ERR_PTR(-EINVAL);
1209 deactivate_locked_super(s);
1210 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1211 subvol_name);
1212 }
1213
1214 return root;
1215}
1216
1217/*
1218 * Find a superblock for the given device / mount point.
1219 *
1220 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1221 * for multiple device setup. Make sure to keep it in sync.
1222 */
1223static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1224 const char *device_name, void *data)
1225{
1226 struct block_device *bdev = NULL;
1227 struct super_block *s;
1228 struct dentry *root;
1229 struct btrfs_fs_devices *fs_devices = NULL;
1230 struct btrfs_fs_info *fs_info = NULL;
1231 fmode_t mode = FMODE_READ;
1232 char *subvol_name = NULL;
1233 u64 subvol_objectid = 0;
1234 int error = 0;
1235
1236 if (!(flags & MS_RDONLY))
1237 mode |= FMODE_WRITE;
1238
1239 error = btrfs_parse_early_options(data, mode, fs_type,
1240 &subvol_name, &subvol_objectid,
1241 &fs_devices);
1242 if (error) {
1243 kfree(subvol_name);
1244 return ERR_PTR(error);
1245 }
1246
1247 if (subvol_name) {
1248 root = mount_subvol(subvol_name, flags, device_name, data);
1249 kfree(subvol_name);
1250 return root;
1251 }
1252
1253 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1254 if (error)
1255 return ERR_PTR(error);
1256
1257 /*
1258 * Setup a dummy root and fs_info for test/set super. This is because
1259 * we don't actually fill this stuff out until open_ctree, but we need
1260 * it for searching for existing supers, so this lets us do that and
1261 * then open_ctree will properly initialize everything later.
1262 */
1263 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1264 if (!fs_info)
1265 return ERR_PTR(-ENOMEM);
1266
1267 fs_info->fs_devices = fs_devices;
1268
1269 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1270 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1271 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1272 error = -ENOMEM;
1273 goto error_fs_info;
1274 }
1275
1276 error = btrfs_open_devices(fs_devices, mode, fs_type);
1277 if (error)
1278 goto error_fs_info;
1279
1280 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1281 error = -EACCES;
1282 goto error_close_devices;
1283 }
1284
1285 bdev = fs_devices->latest_bdev;
1286 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1287 fs_info);
1288 if (IS_ERR(s)) {
1289 error = PTR_ERR(s);
1290 goto error_close_devices;
1291 }
1292
1293 if (s->s_root) {
1294 btrfs_close_devices(fs_devices);
1295 free_fs_info(fs_info);
1296 if ((flags ^ s->s_flags) & MS_RDONLY)
1297 error = -EBUSY;
1298 } else {
1299 char b[BDEVNAME_SIZE];
1300
1301 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1302 btrfs_sb(s)->bdev_holder = fs_type;
1303 error = btrfs_fill_super(s, fs_devices, data,
1304 flags & MS_SILENT ? 1 : 0);
1305 }
1306
1307 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1308 if (IS_ERR(root))
1309 deactivate_locked_super(s);
1310
1311 return root;
1312
1313error_close_devices:
1314 btrfs_close_devices(fs_devices);
1315error_fs_info:
1316 free_fs_info(fs_info);
1317 return ERR_PTR(error);
1318}
1319
1320static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1321 int new_pool_size, int old_pool_size)
1322{
1323 if (new_pool_size == old_pool_size)
1324 return;
1325
1326 fs_info->thread_pool_size = new_pool_size;
1327
1328 btrfs_info(fs_info, "resize thread pool %d -> %d",
1329 old_pool_size, new_pool_size);
1330
1331 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1332 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1333 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1334 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1335 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1336 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1337 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1338 new_pool_size);
1339 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1340 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1341 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1342 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1343 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1344 new_pool_size);
1345}
1346
1347static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1348{
1349 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1350}
1351
1352static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1353 unsigned long old_opts, int flags)
1354{
1355 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1356 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1357 (flags & MS_RDONLY))) {
1358 /* wait for any defraggers to finish */
1359 wait_event(fs_info->transaction_wait,
1360 (atomic_read(&fs_info->defrag_running) == 0));
1361 if (flags & MS_RDONLY)
1362 sync_filesystem(fs_info->sb);
1363 }
1364}
1365
1366static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1367 unsigned long old_opts)
1368{
1369 /*
1370 * We need cleanup all defragable inodes if the autodefragment is
1371 * close or the fs is R/O.
1372 */
1373 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1374 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1375 (fs_info->sb->s_flags & MS_RDONLY))) {
1376 btrfs_cleanup_defrag_inodes(fs_info);
1377 }
1378
1379 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1380}
1381
1382static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1383{
1384 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1385 struct btrfs_root *root = fs_info->tree_root;
1386 unsigned old_flags = sb->s_flags;
1387 unsigned long old_opts = fs_info->mount_opt;
1388 unsigned long old_compress_type = fs_info->compress_type;
1389 u64 old_max_inline = fs_info->max_inline;
1390 u64 old_alloc_start = fs_info->alloc_start;
1391 int old_thread_pool_size = fs_info->thread_pool_size;
1392 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1393 int ret;
1394
1395 sync_filesystem(sb);
1396 btrfs_remount_prepare(fs_info);
1397
1398 ret = btrfs_parse_options(root, data);
1399 if (ret) {
1400 ret = -EINVAL;
1401 goto restore;
1402 }
1403
1404 btrfs_remount_begin(fs_info, old_opts, *flags);
1405 btrfs_resize_thread_pool(fs_info,
1406 fs_info->thread_pool_size, old_thread_pool_size);
1407
1408 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1409 goto out;
1410
1411 if (*flags & MS_RDONLY) {
1412 /*
1413 * this also happens on 'umount -rf' or on shutdown, when
1414 * the filesystem is busy.
1415 */
1416
1417 /* wait for the uuid_scan task to finish */
1418 down(&fs_info->uuid_tree_rescan_sem);
1419 /* avoid complains from lockdep et al. */
1420 up(&fs_info->uuid_tree_rescan_sem);
1421
1422 sb->s_flags |= MS_RDONLY;
1423
1424 btrfs_dev_replace_suspend_for_unmount(fs_info);
1425 btrfs_scrub_cancel(fs_info);
1426 btrfs_pause_balance(fs_info);
1427
1428 ret = btrfs_commit_super(root);
1429 if (ret)
1430 goto restore;
1431 } else {
1432 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1433 btrfs_err(fs_info,
1434 "Remounting read-write after error is not allowed");
1435 ret = -EINVAL;
1436 goto restore;
1437 }
1438 if (fs_info->fs_devices->rw_devices == 0) {
1439 ret = -EACCES;
1440 goto restore;
1441 }
1442
1443 if (fs_info->fs_devices->missing_devices >
1444 fs_info->num_tolerated_disk_barrier_failures &&
1445 !(*flags & MS_RDONLY)) {
1446 btrfs_warn(fs_info,
1447 "too many missing devices, writeable remount is not allowed");
1448 ret = -EACCES;
1449 goto restore;
1450 }
1451
1452 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1453 ret = -EINVAL;
1454 goto restore;
1455 }
1456
1457 ret = btrfs_cleanup_fs_roots(fs_info);
1458 if (ret)
1459 goto restore;
1460
1461 /* recover relocation */
1462 ret = btrfs_recover_relocation(root);
1463 if (ret)
1464 goto restore;
1465
1466 ret = btrfs_resume_balance_async(fs_info);
1467 if (ret)
1468 goto restore;
1469
1470 ret = btrfs_resume_dev_replace_async(fs_info);
1471 if (ret) {
1472 btrfs_warn(fs_info, "failed to resume dev_replace");
1473 goto restore;
1474 }
1475
1476 if (!fs_info->uuid_root) {
1477 btrfs_info(fs_info, "creating UUID tree");
1478 ret = btrfs_create_uuid_tree(fs_info);
1479 if (ret) {
1480 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1481 goto restore;
1482 }
1483 }
1484 sb->s_flags &= ~MS_RDONLY;
1485 }
1486out:
1487 wake_up_process(fs_info->transaction_kthread);
1488 btrfs_remount_cleanup(fs_info, old_opts);
1489 return 0;
1490
1491restore:
1492 /* We've hit an error - don't reset MS_RDONLY */
1493 if (sb->s_flags & MS_RDONLY)
1494 old_flags |= MS_RDONLY;
1495 sb->s_flags = old_flags;
1496 fs_info->mount_opt = old_opts;
1497 fs_info->compress_type = old_compress_type;
1498 fs_info->max_inline = old_max_inline;
1499 mutex_lock(&fs_info->chunk_mutex);
1500 fs_info->alloc_start = old_alloc_start;
1501 mutex_unlock(&fs_info->chunk_mutex);
1502 btrfs_resize_thread_pool(fs_info,
1503 old_thread_pool_size, fs_info->thread_pool_size);
1504 fs_info->metadata_ratio = old_metadata_ratio;
1505 btrfs_remount_cleanup(fs_info, old_opts);
1506 return ret;
1507}
1508
1509/* Used to sort the devices by max_avail(descending sort) */
1510static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1511 const void *dev_info2)
1512{
1513 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1514 ((struct btrfs_device_info *)dev_info2)->max_avail)
1515 return -1;
1516 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1517 ((struct btrfs_device_info *)dev_info2)->max_avail)
1518 return 1;
1519 else
1520 return 0;
1521}
1522
1523/*
1524 * sort the devices by max_avail, in which max free extent size of each device
1525 * is stored.(Descending Sort)
1526 */
1527static inline void btrfs_descending_sort_devices(
1528 struct btrfs_device_info *devices,
1529 size_t nr_devices)
1530{
1531 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1532 btrfs_cmp_device_free_bytes, NULL);
1533}
1534
1535/*
1536 * The helper to calc the free space on the devices that can be used to store
1537 * file data.
1538 */
1539static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1540{
1541 struct btrfs_fs_info *fs_info = root->fs_info;
1542 struct btrfs_device_info *devices_info;
1543 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1544 struct btrfs_device *device;
1545 u64 skip_space;
1546 u64 type;
1547 u64 avail_space;
1548 u64 used_space;
1549 u64 min_stripe_size;
1550 int min_stripes = 1, num_stripes = 1;
1551 int i = 0, nr_devices;
1552 int ret;
1553
1554 nr_devices = fs_info->fs_devices->open_devices;
1555 BUG_ON(!nr_devices);
1556
1557 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1558 GFP_NOFS);
1559 if (!devices_info)
1560 return -ENOMEM;
1561
1562 /* calc min stripe number for data space alloction */
1563 type = btrfs_get_alloc_profile(root, 1);
1564 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1565 min_stripes = 2;
1566 num_stripes = nr_devices;
1567 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1568 min_stripes = 2;
1569 num_stripes = 2;
1570 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1571 min_stripes = 4;
1572 num_stripes = 4;
1573 }
1574
1575 if (type & BTRFS_BLOCK_GROUP_DUP)
1576 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1577 else
1578 min_stripe_size = BTRFS_STRIPE_LEN;
1579
1580 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1581 if (!device->in_fs_metadata || !device->bdev ||
1582 device->is_tgtdev_for_dev_replace)
1583 continue;
1584
1585 avail_space = device->total_bytes - device->bytes_used;
1586
1587 /* align with stripe_len */
1588 do_div(avail_space, BTRFS_STRIPE_LEN);
1589 avail_space *= BTRFS_STRIPE_LEN;
1590
1591 /*
1592 * In order to avoid overwritting the superblock on the drive,
1593 * btrfs starts at an offset of at least 1MB when doing chunk
1594 * allocation.
1595 */
1596 skip_space = 1024 * 1024;
1597
1598 /* user can set the offset in fs_info->alloc_start. */
1599 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1600 device->total_bytes)
1601 skip_space = max(fs_info->alloc_start, skip_space);
1602
1603 /*
1604 * btrfs can not use the free space in [0, skip_space - 1],
1605 * we must subtract it from the total. In order to implement
1606 * it, we account the used space in this range first.
1607 */
1608 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1609 &used_space);
1610 if (ret) {
1611 kfree(devices_info);
1612 return ret;
1613 }
1614
1615 /* calc the free space in [0, skip_space - 1] */
1616 skip_space -= used_space;
1617
1618 /*
1619 * we can use the free space in [0, skip_space - 1], subtract
1620 * it from the total.
1621 */
1622 if (avail_space && avail_space >= skip_space)
1623 avail_space -= skip_space;
1624 else
1625 avail_space = 0;
1626
1627 if (avail_space < min_stripe_size)
1628 continue;
1629
1630 devices_info[i].dev = device;
1631 devices_info[i].max_avail = avail_space;
1632
1633 i++;
1634 }
1635
1636 nr_devices = i;
1637
1638 btrfs_descending_sort_devices(devices_info, nr_devices);
1639
1640 i = nr_devices - 1;
1641 avail_space = 0;
1642 while (nr_devices >= min_stripes) {
1643 if (num_stripes > nr_devices)
1644 num_stripes = nr_devices;
1645
1646 if (devices_info[i].max_avail >= min_stripe_size) {
1647 int j;
1648 u64 alloc_size;
1649
1650 avail_space += devices_info[i].max_avail * num_stripes;
1651 alloc_size = devices_info[i].max_avail;
1652 for (j = i + 1 - num_stripes; j <= i; j++)
1653 devices_info[j].max_avail -= alloc_size;
1654 }
1655 i--;
1656 nr_devices--;
1657 }
1658
1659 kfree(devices_info);
1660 *free_bytes = avail_space;
1661 return 0;
1662}
1663
1664static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1665{
1666 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1667 struct btrfs_super_block *disk_super = fs_info->super_copy;
1668 struct list_head *head = &fs_info->space_info;
1669 struct btrfs_space_info *found;
1670 u64 total_used = 0;
1671 u64 total_free_data = 0;
1672 int bits = dentry->d_sb->s_blocksize_bits;
1673 __be32 *fsid = (__be32 *)fs_info->fsid;
1674 int ret;
1675
1676 /* holding chunk_muext to avoid allocating new chunks */
1677 mutex_lock(&fs_info->chunk_mutex);
1678 rcu_read_lock();
1679 list_for_each_entry_rcu(found, head, list) {
1680 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1681 total_free_data += found->disk_total - found->disk_used;
1682 total_free_data -=
1683 btrfs_account_ro_block_groups_free_space(found);
1684 }
1685
1686 total_used += found->disk_used;
1687 }
1688 rcu_read_unlock();
1689
1690 buf->f_namelen = BTRFS_NAME_LEN;
1691 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1692 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1693 buf->f_bsize = dentry->d_sb->s_blocksize;
1694 buf->f_type = BTRFS_SUPER_MAGIC;
1695 buf->f_bavail = total_free_data;
1696 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1697 if (ret) {
1698 mutex_unlock(&fs_info->chunk_mutex);
1699 return ret;
1700 }
1701 buf->f_bavail += total_free_data;
1702 buf->f_bavail = buf->f_bavail >> bits;
1703 mutex_unlock(&fs_info->chunk_mutex);
1704
1705 /* We treat it as constant endianness (it doesn't matter _which_)
1706 because we want the fsid to come out the same whether mounted
1707 on a big-endian or little-endian host */
1708 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1709 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1710 /* Mask in the root object ID too, to disambiguate subvols */
1711 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1712 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1713
1714 return 0;
1715}
1716
1717static void btrfs_kill_super(struct super_block *sb)
1718{
1719 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1720 kill_anon_super(sb);
1721 free_fs_info(fs_info);
1722}
1723
1724static struct file_system_type btrfs_fs_type = {
1725 .owner = THIS_MODULE,
1726 .name = "btrfs",
1727 .mount = btrfs_mount,
1728 .kill_sb = btrfs_kill_super,
1729 .fs_flags = FS_REQUIRES_DEV,
1730};
1731MODULE_ALIAS_FS("btrfs");
1732
1733/*
1734 * used by btrfsctl to scan devices when no FS is mounted
1735 */
1736static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1737 unsigned long arg)
1738{
1739 struct btrfs_ioctl_vol_args *vol;
1740 struct btrfs_fs_devices *fs_devices;
1741 int ret = -ENOTTY;
1742
1743 if (!capable(CAP_SYS_ADMIN))
1744 return -EPERM;
1745
1746 vol = memdup_user((void __user *)arg, sizeof(*vol));
1747 if (IS_ERR(vol))
1748 return PTR_ERR(vol);
1749
1750 switch (cmd) {
1751 case BTRFS_IOC_SCAN_DEV:
1752 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1753 &btrfs_fs_type, &fs_devices);
1754 break;
1755 case BTRFS_IOC_DEVICES_READY:
1756 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1757 &btrfs_fs_type, &fs_devices);
1758 if (ret)
1759 break;
1760 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1761 break;
1762 }
1763
1764 kfree(vol);
1765 return ret;
1766}
1767
1768static int btrfs_freeze(struct super_block *sb)
1769{
1770 struct btrfs_trans_handle *trans;
1771 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1772
1773 trans = btrfs_attach_transaction_barrier(root);
1774 if (IS_ERR(trans)) {
1775 /* no transaction, don't bother */
1776 if (PTR_ERR(trans) == -ENOENT)
1777 return 0;
1778 return PTR_ERR(trans);
1779 }
1780 return btrfs_commit_transaction(trans, root);
1781}
1782
1783static int btrfs_unfreeze(struct super_block *sb)
1784{
1785 return 0;
1786}
1787
1788static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1789{
1790 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1791 struct btrfs_fs_devices *cur_devices;
1792 struct btrfs_device *dev, *first_dev = NULL;
1793 struct list_head *head;
1794 struct rcu_string *name;
1795
1796 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1797 cur_devices = fs_info->fs_devices;
1798 while (cur_devices) {
1799 head = &cur_devices->devices;
1800 list_for_each_entry(dev, head, dev_list) {
1801 if (dev->missing)
1802 continue;
1803 if (!first_dev || dev->devid < first_dev->devid)
1804 first_dev = dev;
1805 }
1806 cur_devices = cur_devices->seed;
1807 }
1808
1809 if (first_dev) {
1810 rcu_read_lock();
1811 name = rcu_dereference(first_dev->name);
1812 seq_escape(m, name->str, " \t\n\\");
1813 rcu_read_unlock();
1814 } else {
1815 WARN_ON(1);
1816 }
1817 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1818 return 0;
1819}
1820
1821static const struct super_operations btrfs_super_ops = {
1822 .drop_inode = btrfs_drop_inode,
1823 .evict_inode = btrfs_evict_inode,
1824 .put_super = btrfs_put_super,
1825 .sync_fs = btrfs_sync_fs,
1826 .show_options = btrfs_show_options,
1827 .show_devname = btrfs_show_devname,
1828 .write_inode = btrfs_write_inode,
1829 .alloc_inode = btrfs_alloc_inode,
1830 .destroy_inode = btrfs_destroy_inode,
1831 .statfs = btrfs_statfs,
1832 .remount_fs = btrfs_remount,
1833 .freeze_fs = btrfs_freeze,
1834 .unfreeze_fs = btrfs_unfreeze,
1835};
1836
1837static const struct file_operations btrfs_ctl_fops = {
1838 .unlocked_ioctl = btrfs_control_ioctl,
1839 .compat_ioctl = btrfs_control_ioctl,
1840 .owner = THIS_MODULE,
1841 .llseek = noop_llseek,
1842};
1843
1844static struct miscdevice btrfs_misc = {
1845 .minor = BTRFS_MINOR,
1846 .name = "btrfs-control",
1847 .fops = &btrfs_ctl_fops
1848};
1849
1850MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1851MODULE_ALIAS("devname:btrfs-control");
1852
1853static int btrfs_interface_init(void)
1854{
1855 return misc_register(&btrfs_misc);
1856}
1857
1858static void btrfs_interface_exit(void)
1859{
1860 if (misc_deregister(&btrfs_misc) < 0)
1861 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
1862}
1863
1864static void btrfs_print_info(void)
1865{
1866 printk(KERN_INFO "Btrfs loaded"
1867#ifdef CONFIG_BTRFS_DEBUG
1868 ", debug=on"
1869#endif
1870#ifdef CONFIG_BTRFS_ASSERT
1871 ", assert=on"
1872#endif
1873#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1874 ", integrity-checker=on"
1875#endif
1876 "\n");
1877}
1878
1879static int btrfs_run_sanity_tests(void)
1880{
1881 int ret;
1882
1883 ret = btrfs_init_test_fs();
1884 if (ret)
1885 return ret;
1886
1887 ret = btrfs_test_free_space_cache();
1888 if (ret)
1889 goto out;
1890 ret = btrfs_test_extent_buffer_operations();
1891 if (ret)
1892 goto out;
1893 ret = btrfs_test_extent_io();
1894 if (ret)
1895 goto out;
1896 ret = btrfs_test_inodes();
1897out:
1898 btrfs_destroy_test_fs();
1899 return ret;
1900}
1901
1902static int __init init_btrfs_fs(void)
1903{
1904 int err;
1905
1906 err = btrfs_hash_init();
1907 if (err)
1908 return err;
1909
1910 btrfs_props_init();
1911
1912 err = btrfs_init_sysfs();
1913 if (err)
1914 goto free_hash;
1915
1916 btrfs_init_compress();
1917
1918 err = btrfs_init_cachep();
1919 if (err)
1920 goto free_compress;
1921
1922 err = extent_io_init();
1923 if (err)
1924 goto free_cachep;
1925
1926 err = extent_map_init();
1927 if (err)
1928 goto free_extent_io;
1929
1930 err = ordered_data_init();
1931 if (err)
1932 goto free_extent_map;
1933
1934 err = btrfs_delayed_inode_init();
1935 if (err)
1936 goto free_ordered_data;
1937
1938 err = btrfs_auto_defrag_init();
1939 if (err)
1940 goto free_delayed_inode;
1941
1942 err = btrfs_delayed_ref_init();
1943 if (err)
1944 goto free_auto_defrag;
1945
1946 err = btrfs_prelim_ref_init();
1947 if (err)
1948 goto free_prelim_ref;
1949
1950 err = btrfs_interface_init();
1951 if (err)
1952 goto free_delayed_ref;
1953
1954 btrfs_init_lockdep();
1955
1956 btrfs_print_info();
1957
1958 err = btrfs_run_sanity_tests();
1959 if (err)
1960 goto unregister_ioctl;
1961
1962 err = register_filesystem(&btrfs_fs_type);
1963 if (err)
1964 goto unregister_ioctl;
1965
1966 return 0;
1967
1968unregister_ioctl:
1969 btrfs_interface_exit();
1970free_prelim_ref:
1971 btrfs_prelim_ref_exit();
1972free_delayed_ref:
1973 btrfs_delayed_ref_exit();
1974free_auto_defrag:
1975 btrfs_auto_defrag_exit();
1976free_delayed_inode:
1977 btrfs_delayed_inode_exit();
1978free_ordered_data:
1979 ordered_data_exit();
1980free_extent_map:
1981 extent_map_exit();
1982free_extent_io:
1983 extent_io_exit();
1984free_cachep:
1985 btrfs_destroy_cachep();
1986free_compress:
1987 btrfs_exit_compress();
1988 btrfs_exit_sysfs();
1989free_hash:
1990 btrfs_hash_exit();
1991 return err;
1992}
1993
1994static void __exit exit_btrfs_fs(void)
1995{
1996 btrfs_destroy_cachep();
1997 btrfs_delayed_ref_exit();
1998 btrfs_auto_defrag_exit();
1999 btrfs_delayed_inode_exit();
2000 btrfs_prelim_ref_exit();
2001 ordered_data_exit();
2002 extent_map_exit();
2003 extent_io_exit();
2004 btrfs_interface_exit();
2005 unregister_filesystem(&btrfs_fs_type);
2006 btrfs_exit_sysfs();
2007 btrfs_cleanup_fs_uuids();
2008 btrfs_exit_compress();
2009 btrfs_hash_exit();
2010}
2011
2012late_initcall(init_btrfs_fs);
2013module_exit(exit_btrfs_fs)
2014
2015MODULE_LICENSE("GPL");