]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/btrfs/super.c
Btrfs: Integrate metadata reservation with start_transaction
[net-next-2.6.git] / fs / btrfs / super.c
CommitLineData
6cbd5570
CM
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
4b82d6e4 19#include <linux/blkdev.h>
2e635a27 20#include <linux/module.h>
e20d96d6 21#include <linux/buffer_head.h>
2e635a27
CM
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>
a9572a15 27#include <linux/seq_file.h>
2e635a27 28#include <linux/string.h>
2e635a27 29#include <linux/backing-dev.h>
4b82d6e4 30#include <linux/mount.h>
dee26a9f 31#include <linux/mpage.h>
75dfe396
CM
32#include <linux/swap.h>
33#include <linux/writeback.h>
8fd17795 34#include <linux/statfs.h>
08607c1b 35#include <linux/compat.h>
95e05289 36#include <linux/parser.h>
c59f8951 37#include <linux/ctype.h>
6da6abae 38#include <linux/namei.h>
a9218f6b 39#include <linux/miscdevice.h>
1bcbf313 40#include <linux/magic.h>
5a0e3ad6 41#include <linux/slab.h>
4b4e25f2 42#include "compat.h"
2e635a27 43#include "ctree.h"
e20d96d6 44#include "disk-io.h"
d5719762 45#include "transaction.h"
2c90e5d6 46#include "btrfs_inode.h"
c5739bba 47#include "ioctl.h"
3a686375 48#include "print-tree.h"
5103e947 49#include "xattr.h"
8a4b83cc 50#include "volumes.h"
b3c3da71 51#include "version.h"
be6e8dc0 52#include "export.h"
c8b97818 53#include "compression.h"
2e635a27 54
b87221de 55static const struct super_operations btrfs_super_ops;
75dfe396 56
d397712b 57static void btrfs_put_super(struct super_block *sb)
b18c6685 58{
39279cc3 59 struct btrfs_root *root = btrfs_sb(sb);
b18c6685 60 int ret;
b18c6685 61
39279cc3 62 ret = close_ctree(root);
39279cc3 63 sb->s_fs_info = NULL;
75dfe396
CM
64}
65
95e05289 66enum {
73f73415 67 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
287a0ab9
JB
68 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
69 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
70 Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
e244a0ae 71 Opt_discard, Opt_err,
95e05289
CM
72};
73
74static match_table_t tokens = {
dfe25020 75 {Opt_degraded, "degraded"},
95e05289 76 {Opt_subvol, "subvol=%s"},
73f73415 77 {Opt_subvolid, "subvolid=%d"},
43e570b0 78 {Opt_device, "device=%s"},
b6cda9bc 79 {Opt_nodatasum, "nodatasum"},
be20aa9d 80 {Opt_nodatacow, "nodatacow"},
21ad10cf 81 {Opt_nobarrier, "nobarrier"},
6f568d35 82 {Opt_max_inline, "max_inline=%s"},
8f662a76 83 {Opt_alloc_start, "alloc_start=%s"},
4543df7e 84 {Opt_thread_pool, "thread_pool=%d"},
c8b97818 85 {Opt_compress, "compress"},
a555f810 86 {Opt_compress_force, "compress-force"},
e18e4809 87 {Opt_ssd, "ssd"},
451d7585 88 {Opt_ssd_spread, "ssd_spread"},
3b30c22f 89 {Opt_nossd, "nossd"},
33268eaf 90 {Opt_noacl, "noacl"},
3a5e1404 91 {Opt_notreelog, "notreelog"},
dccae999 92 {Opt_flushoncommit, "flushoncommit"},
97e728d4 93 {Opt_ratio, "metadata_ratio=%d"},
e244a0ae 94 {Opt_discard, "discard"},
33268eaf 95 {Opt_err, NULL},
95e05289
CM
96};
97
edf24abe
CH
98/*
99 * Regular mount options parser. Everything that is needed only when
100 * reading in a new superblock is parsed here.
101 */
102int btrfs_parse_options(struct btrfs_root *root, char *options)
95e05289 103{
edf24abe 104 struct btrfs_fs_info *info = root->fs_info;
95e05289 105 substring_t args[MAX_OPT_ARGS];
da495ecc 106 char *p, *num, *orig;
4543df7e 107 int intarg;
a7a3f7ca 108 int ret = 0;
b6cda9bc 109
95e05289 110 if (!options)
edf24abe 111 return 0;
95e05289 112
be20aa9d
CM
113 /*
114 * strsep changes the string, duplicate it because parse_options
115 * gets called twice
116 */
117 options = kstrdup(options, GFP_NOFS);
118 if (!options)
119 return -ENOMEM;
120
da495ecc 121 orig = options;
be20aa9d 122
edf24abe 123 while ((p = strsep(&options, ",")) != NULL) {
95e05289
CM
124 int token;
125 if (!*p)
126 continue;
127
128 token = match_token(p, tokens, args);
129 switch (token) {
dfe25020 130 case Opt_degraded:
edf24abe
CH
131 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
132 btrfs_set_opt(info->mount_opt, DEGRADED);
dfe25020 133 break;
95e05289 134 case Opt_subvol:
73f73415 135 case Opt_subvolid:
43e570b0 136 case Opt_device:
edf24abe 137 /*
43e570b0 138 * These are parsed by btrfs_parse_early_options
edf24abe
CH
139 * and can be happily ignored here.
140 */
b6cda9bc
CM
141 break;
142 case Opt_nodatasum:
067c28ad 143 printk(KERN_INFO "btrfs: setting nodatasum\n");
edf24abe 144 btrfs_set_opt(info->mount_opt, NODATASUM);
be20aa9d
CM
145 break;
146 case Opt_nodatacow:
edf24abe
CH
147 printk(KERN_INFO "btrfs: setting nodatacow\n");
148 btrfs_set_opt(info->mount_opt, NODATACOW);
149 btrfs_set_opt(info->mount_opt, NODATASUM);
95e05289 150 break;
c8b97818
CM
151 case Opt_compress:
152 printk(KERN_INFO "btrfs: use compression\n");
153 btrfs_set_opt(info->mount_opt, COMPRESS);
154 break;
a555f810
CM
155 case Opt_compress_force:
156 printk(KERN_INFO "btrfs: forcing compression\n");
157 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
158 btrfs_set_opt(info->mount_opt, COMPRESS);
159 break;
e18e4809 160 case Opt_ssd:
edf24abe
CH
161 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
162 btrfs_set_opt(info->mount_opt, SSD);
e18e4809 163 break;
451d7585
CM
164 case Opt_ssd_spread:
165 printk(KERN_INFO "btrfs: use spread ssd "
166 "allocation scheme\n");
167 btrfs_set_opt(info->mount_opt, SSD);
168 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
169 break;
3b30c22f 170 case Opt_nossd:
451d7585
CM
171 printk(KERN_INFO "btrfs: not using ssd allocation "
172 "scheme\n");
c289811c 173 btrfs_set_opt(info->mount_opt, NOSSD);
3b30c22f 174 btrfs_clear_opt(info->mount_opt, SSD);
451d7585 175 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
3b30c22f 176 break;
21ad10cf 177 case Opt_nobarrier:
edf24abe
CH
178 printk(KERN_INFO "btrfs: turning off barriers\n");
179 btrfs_set_opt(info->mount_opt, NOBARRIER);
21ad10cf 180 break;
4543df7e
CM
181 case Opt_thread_pool:
182 intarg = 0;
183 match_int(&args[0], &intarg);
184 if (intarg) {
185 info->thread_pool_size = intarg;
186 printk(KERN_INFO "btrfs: thread pool %d\n",
187 info->thread_pool_size);
188 }
189 break;
6f568d35 190 case Opt_max_inline:
edf24abe
CH
191 num = match_strdup(&args[0]);
192 if (num) {
91748467 193 info->max_inline = memparse(num, NULL);
edf24abe
CH
194 kfree(num);
195
15ada040
CM
196 if (info->max_inline) {
197 info->max_inline = max_t(u64,
198 info->max_inline,
199 root->sectorsize);
200 }
edf24abe 201 printk(KERN_INFO "btrfs: max_inline at %llu\n",
21380931 202 (unsigned long long)info->max_inline);
6f568d35
CM
203 }
204 break;
8f662a76 205 case Opt_alloc_start:
edf24abe
CH
206 num = match_strdup(&args[0]);
207 if (num) {
91748467 208 info->alloc_start = memparse(num, NULL);
edf24abe
CH
209 kfree(num);
210 printk(KERN_INFO
211 "btrfs: allocations start at %llu\n",
21380931 212 (unsigned long long)info->alloc_start);
8f662a76
CM
213 }
214 break;
33268eaf
JB
215 case Opt_noacl:
216 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
217 break;
3a5e1404
SW
218 case Opt_notreelog:
219 printk(KERN_INFO "btrfs: disabling tree log\n");
220 btrfs_set_opt(info->mount_opt, NOTREELOG);
221 break;
dccae999
SW
222 case Opt_flushoncommit:
223 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
224 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
225 break;
97e728d4
JB
226 case Opt_ratio:
227 intarg = 0;
228 match_int(&args[0], &intarg);
229 if (intarg) {
230 info->metadata_ratio = intarg;
231 printk(KERN_INFO "btrfs: metadata ratio %d\n",
232 info->metadata_ratio);
233 }
234 break;
e244a0ae
CH
235 case Opt_discard:
236 btrfs_set_opt(info->mount_opt, DISCARD);
237 break;
a7a3f7ca
SW
238 case Opt_err:
239 printk(KERN_INFO "btrfs: unrecognized mount option "
240 "'%s'\n", p);
241 ret = -EINVAL;
242 goto out;
95e05289 243 default:
be20aa9d 244 break;
95e05289
CM
245 }
246 }
a7a3f7ca 247out:
da495ecc 248 kfree(orig);
a7a3f7ca 249 return ret;
edf24abe
CH
250}
251
252/*
253 * Parse mount options that are required early in the mount process.
254 *
255 * All other options will be parsed on much later in the mount process and
256 * only when we need to allocate a new super block.
257 */
97288f2c 258static int btrfs_parse_early_options(const char *options, fmode_t flags,
73f73415 259 void *holder, char **subvol_name, u64 *subvol_objectid,
43e570b0 260 struct btrfs_fs_devices **fs_devices)
edf24abe
CH
261{
262 substring_t args[MAX_OPT_ARGS];
263 char *opts, *p;
264 int error = 0;
73f73415 265 int intarg;
edf24abe
CH
266
267 if (!options)
268 goto out;
269
270 /*
271 * strsep changes the string, duplicate it because parse_options
272 * gets called twice
273 */
274 opts = kstrdup(options, GFP_KERNEL);
275 if (!opts)
276 return -ENOMEM;
277
278 while ((p = strsep(&opts, ",")) != NULL) {
279 int token;
280 if (!*p)
281 continue;
282
283 token = match_token(p, tokens, args);
284 switch (token) {
285 case Opt_subvol:
286 *subvol_name = match_strdup(&args[0]);
287 break;
73f73415
JB
288 case Opt_subvolid:
289 intarg = 0;
4849f01d
JB
290 error = match_int(&args[0], &intarg);
291 if (!error) {
292 /* we want the original fs_tree */
293 if (!intarg)
294 *subvol_objectid =
295 BTRFS_FS_TREE_OBJECTID;
296 else
297 *subvol_objectid = intarg;
298 }
73f73415 299 break;
43e570b0
CH
300 case Opt_device:
301 error = btrfs_scan_one_device(match_strdup(&args[0]),
302 flags, holder, fs_devices);
303 if (error)
304 goto out_free_opts;
305 break;
edf24abe
CH
306 default:
307 break;
308 }
309 }
310
43e570b0 311 out_free_opts:
edf24abe
CH
312 kfree(opts);
313 out:
314 /*
315 * If no subvolume name is specified we use the default one. Allocate
3de4586c 316 * a copy of the string "." here so that code later in the
edf24abe
CH
317 * mount path doesn't care if it's the default volume or another one.
318 */
319 if (!*subvol_name) {
3de4586c 320 *subvol_name = kstrdup(".", GFP_KERNEL);
edf24abe
CH
321 if (!*subvol_name)
322 return -ENOMEM;
323 }
324 return error;
95e05289
CM
325}
326
73f73415
JB
327static struct dentry *get_default_root(struct super_block *sb,
328 u64 subvol_objectid)
329{
330 struct btrfs_root *root = sb->s_fs_info;
331 struct btrfs_root *new_root;
332 struct btrfs_dir_item *di;
333 struct btrfs_path *path;
334 struct btrfs_key location;
335 struct inode *inode;
336 struct dentry *dentry;
337 u64 dir_id;
338 int new = 0;
339
340 /*
341 * We have a specific subvol we want to mount, just setup location and
342 * go look up the root.
343 */
344 if (subvol_objectid) {
345 location.objectid = subvol_objectid;
346 location.type = BTRFS_ROOT_ITEM_KEY;
347 location.offset = (u64)-1;
348 goto find_root;
349 }
350
351 path = btrfs_alloc_path();
352 if (!path)
353 return ERR_PTR(-ENOMEM);
354 path->leave_spinning = 1;
355
356 /*
357 * Find the "default" dir item which points to the root item that we
358 * will mount by default if we haven't been given a specific subvolume
359 * to mount.
360 */
361 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
362 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
363 if (!di) {
364 /*
365 * Ok the default dir item isn't there. This is weird since
366 * it's always been there, but don't freak out, just try and
367 * mount to root most subvolume.
368 */
369 btrfs_free_path(path);
370 dir_id = BTRFS_FIRST_FREE_OBJECTID;
371 new_root = root->fs_info->fs_root;
372 goto setup_root;
373 }
374
375 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
376 btrfs_free_path(path);
377
378find_root:
379 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
380 if (IS_ERR(new_root))
381 return ERR_PTR(PTR_ERR(new_root));
382
383 if (btrfs_root_refs(&new_root->root_item) == 0)
384 return ERR_PTR(-ENOENT);
385
386 dir_id = btrfs_root_dirid(&new_root->root_item);
387setup_root:
388 location.objectid = dir_id;
389 location.type = BTRFS_INODE_ITEM_KEY;
390 location.offset = 0;
391
392 inode = btrfs_iget(sb, &location, new_root, &new);
393 if (!inode)
394 return ERR_PTR(-ENOMEM);
395
396 /*
397 * If we're just mounting the root most subvol put the inode and return
398 * a reference to the dentry. We will have already gotten a reference
399 * to the inode in btrfs_fill_super so we're good to go.
400 */
401 if (!new && sb->s_root->d_inode == inode) {
402 iput(inode);
403 return dget(sb->s_root);
404 }
405
406 if (new) {
407 const struct qstr name = { .name = "/", .len = 1 };
408
409 /*
410 * New inode, we need to make the dentry a sibling of s_root so
411 * everything gets cleaned up properly on unmount.
412 */
413 dentry = d_alloc(sb->s_root, &name);
414 if (!dentry) {
415 iput(inode);
416 return ERR_PTR(-ENOMEM);
417 }
418 d_splice_alias(inode, dentry);
419 } else {
420 /*
421 * We found the inode in cache, just find a dentry for it and
422 * put the reference to the inode we just got.
423 */
424 dentry = d_find_alias(inode);
425 iput(inode);
426 }
427
428 return dentry;
429}
430
d397712b 431static int btrfs_fill_super(struct super_block *sb,
8a4b83cc 432 struct btrfs_fs_devices *fs_devices,
d397712b 433 void *data, int silent)
75dfe396 434{
d397712b
CM
435 struct inode *inode;
436 struct dentry *root_dentry;
39279cc3
CM
437 struct btrfs_super_block *disk_super;
438 struct btrfs_root *tree_root;
5d4f98a2 439 struct btrfs_key key;
39279cc3 440 int err;
a429e513 441
39279cc3
CM
442 sb->s_maxbytes = MAX_LFS_FILESIZE;
443 sb->s_magic = BTRFS_SUPER_MAGIC;
444 sb->s_op = &btrfs_super_ops;
be6e8dc0 445 sb->s_export_op = &btrfs_export_ops;
5103e947 446 sb->s_xattr = btrfs_xattr_handlers;
39279cc3 447 sb->s_time_gran = 1;
0eda294d 448#ifdef CONFIG_BTRFS_FS_POSIX_ACL
33268eaf 449 sb->s_flags |= MS_POSIXACL;
49cf6f45 450#endif
a429e513 451
dfe25020 452 tree_root = open_ctree(sb, fs_devices, (char *)data);
6567e837 453
e58ca020 454 if (IS_ERR(tree_root)) {
39279cc3 455 printk("btrfs: open_ctree failed\n");
e58ca020 456 return PTR_ERR(tree_root);
a429e513 457 }
39279cc3 458 sb->s_fs_info = tree_root;
5f39d397 459 disk_super = &tree_root->fs_info->super_copy;
a429e513 460
5d4f98a2
YZ
461 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
462 key.type = BTRFS_INODE_ITEM_KEY;
463 key.offset = 0;
73f73415 464 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
5d4f98a2
YZ
465 if (IS_ERR(inode)) {
466 err = PTR_ERR(inode);
39279cc3 467 goto fail_close;
f254e52c 468 }
f254e52c 469
39279cc3
CM
470 root_dentry = d_alloc_root(inode);
471 if (!root_dentry) {
472 iput(inode);
473 err = -ENOMEM;
474 goto fail_close;
f254e52c 475 }
58176a96 476
39279cc3 477 sb->s_root = root_dentry;
6885f308 478
6885f308 479 save_mount_options(sb, data);
2619ba1f 480 return 0;
39279cc3
CM
481
482fail_close:
483 close_ctree(tree_root);
484 return err;
2619ba1f
CM
485}
486
6bf13c0c 487int btrfs_sync_fs(struct super_block *sb, int wait)
c5739bba
CM
488{
489 struct btrfs_trans_handle *trans;
dccae999 490 struct btrfs_root *root = btrfs_sb(sb);
c5739bba 491 int ret;
2619ba1f 492
39279cc3
CM
493 if (!wait) {
494 filemap_flush(root->fs_info->btree_inode->i_mapping);
495 return 0;
496 }
771ed689 497
24bbcf04
YZ
498 btrfs_start_delalloc_inodes(root, 0);
499 btrfs_wait_ordered_extents(root, 0, 0);
771ed689 500
a22285a6 501 trans = btrfs_start_transaction(root, 0);
c5739bba 502 ret = btrfs_commit_transaction(trans, root);
54aa1f4d 503 return ret;
2c90e5d6
CM
504}
505
a9572a15
EP
506static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
507{
508 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
509 struct btrfs_fs_info *info = root->fs_info;
510
511 if (btrfs_test_opt(root, DEGRADED))
512 seq_puts(seq, ",degraded");
513 if (btrfs_test_opt(root, NODATASUM))
514 seq_puts(seq, ",nodatasum");
515 if (btrfs_test_opt(root, NODATACOW))
516 seq_puts(seq, ",nodatacow");
517 if (btrfs_test_opt(root, NOBARRIER))
518 seq_puts(seq, ",nobarrier");
a9572a15 519 if (info->max_inline != 8192 * 1024)
21380931
JB
520 seq_printf(seq, ",max_inline=%llu",
521 (unsigned long long)info->max_inline);
a9572a15 522 if (info->alloc_start != 0)
21380931
JB
523 seq_printf(seq, ",alloc_start=%llu",
524 (unsigned long long)info->alloc_start);
a9572a15
EP
525 if (info->thread_pool_size != min_t(unsigned long,
526 num_online_cpus() + 2, 8))
527 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
528 if (btrfs_test_opt(root, COMPRESS))
529 seq_puts(seq, ",compress");
c289811c
CM
530 if (btrfs_test_opt(root, NOSSD))
531 seq_puts(seq, ",nossd");
451d7585
CM
532 if (btrfs_test_opt(root, SSD_SPREAD))
533 seq_puts(seq, ",ssd_spread");
534 else if (btrfs_test_opt(root, SSD))
a9572a15 535 seq_puts(seq, ",ssd");
3a5e1404 536 if (btrfs_test_opt(root, NOTREELOG))
6b65c5c6 537 seq_puts(seq, ",notreelog");
dccae999 538 if (btrfs_test_opt(root, FLUSHONCOMMIT))
6b65c5c6 539 seq_puts(seq, ",flushoncommit");
20a5239a
MW
540 if (btrfs_test_opt(root, DISCARD))
541 seq_puts(seq, ",discard");
a9572a15
EP
542 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
543 seq_puts(seq, ",noacl");
544 return 0;
545}
546
a061fc8d 547static int btrfs_test_super(struct super_block *s, void *data)
4b82d6e4 548{
a061fc8d
CM
549 struct btrfs_fs_devices *test_fs_devices = data;
550 struct btrfs_root *root = btrfs_sb(s);
4b82d6e4 551
a061fc8d 552 return root->fs_info->fs_devices == test_fs_devices;
4b82d6e4
Y
553}
554
edf24abe
CH
555/*
556 * Find a superblock for the given device / mount point.
557 *
558 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
559 * for multiple device setup. Make sure to keep it in sync.
560 */
561static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
562 const char *dev_name, void *data, struct vfsmount *mnt)
4b82d6e4
Y
563{
564 struct block_device *bdev = NULL;
565 struct super_block *s;
566 struct dentry *root;
8a4b83cc 567 struct btrfs_fs_devices *fs_devices = NULL;
97288f2c 568 fmode_t mode = FMODE_READ;
73f73415
JB
569 char *subvol_name = NULL;
570 u64 subvol_objectid = 0;
4b82d6e4 571 int error = 0;
73f73415 572 int found = 0;
4b82d6e4 573
97288f2c
CH
574 if (!(flags & MS_RDONLY))
575 mode |= FMODE_WRITE;
576
577 error = btrfs_parse_early_options(data, mode, fs_type,
73f73415
JB
578 &subvol_name, &subvol_objectid,
579 &fs_devices);
edf24abe 580 if (error)
1f483660 581 return error;
edf24abe 582
97288f2c 583 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
8a4b83cc 584 if (error)
edf24abe 585 goto error_free_subvol_name;
4b82d6e4 586
97288f2c 587 error = btrfs_open_devices(fs_devices, mode, fs_type);
8a4b83cc 588 if (error)
edf24abe 589 goto error_free_subvol_name;
8a4b83cc 590
2b82032c
YZ
591 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
592 error = -EACCES;
593 goto error_close_devices;
594 }
595
dfe25020 596 bdev = fs_devices->latest_bdev;
a061fc8d 597 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
4b82d6e4
Y
598 if (IS_ERR(s))
599 goto error_s;
600
601 if (s->s_root) {
602 if ((flags ^ s->s_flags) & MS_RDONLY) {
6f5bbff9 603 deactivate_locked_super(s);
4b82d6e4 604 error = -EBUSY;
c146afad 605 goto error_close_devices;
4b82d6e4
Y
606 }
607
73f73415 608 found = 1;
2b82032c 609 btrfs_close_devices(fs_devices);
4b82d6e4
Y
610 } else {
611 char b[BDEVNAME_SIZE];
612
613 s->s_flags = flags;
614 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
8a4b83cc
CM
615 error = btrfs_fill_super(s, fs_devices, data,
616 flags & MS_SILENT ? 1 : 0);
4b82d6e4 617 if (error) {
6f5bbff9 618 deactivate_locked_super(s);
1f483660 619 goto error_free_subvol_name;
4b82d6e4
Y
620 }
621
788f20eb 622 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
4b82d6e4
Y
623 s->s_flags |= MS_ACTIVE;
624 }
625
73f73415
JB
626 root = get_default_root(s, subvol_objectid);
627 if (IS_ERR(root)) {
628 error = PTR_ERR(root);
629 deactivate_locked_super(s);
630 goto error;
631 }
632 /* if they gave us a subvolume name bind mount into that */
633 if (strcmp(subvol_name, ".")) {
634 struct dentry *new_root;
635 mutex_lock(&root->d_inode->i_mutex);
636 new_root = lookup_one_len(subvol_name, root,
d397712b 637 strlen(subvol_name));
73f73415 638 mutex_unlock(&root->d_inode->i_mutex);
d397712b 639
73f73415 640 if (IS_ERR(new_root)) {
6f5bbff9 641 deactivate_locked_super(s);
73f73415
JB
642 error = PTR_ERR(new_root);
643 dput(root);
644 goto error_close_devices;
76fcef19 645 }
73f73415 646 if (!new_root->d_inode) {
76fcef19 647 dput(root);
73f73415 648 dput(new_root);
6f5bbff9 649 deactivate_locked_super(s);
76fcef19 650 error = -ENXIO;
73f73415 651 goto error_close_devices;
76fcef19 652 }
73f73415
JB
653 dput(root);
654 root = new_root;
4b82d6e4
Y
655 }
656
657 mnt->mnt_sb = s;
658 mnt->mnt_root = root;
edf24abe
CH
659
660 kfree(subvol_name);
4b82d6e4
Y
661 return 0;
662
663error_s:
664 error = PTR_ERR(s);
c146afad 665error_close_devices:
8a4b83cc 666 btrfs_close_devices(fs_devices);
edf24abe
CH
667error_free_subvol_name:
668 kfree(subvol_name);
73f73415 669error:
4b82d6e4
Y
670 return error;
671}
2e635a27 672
c146afad
YZ
673static int btrfs_remount(struct super_block *sb, int *flags, char *data)
674{
675 struct btrfs_root *root = btrfs_sb(sb);
676 int ret;
677
b288052e
CM
678 ret = btrfs_parse_options(root, data);
679 if (ret)
680 return -EINVAL;
681
c146afad
YZ
682 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
683 return 0;
684
685 if (*flags & MS_RDONLY) {
686 sb->s_flags |= MS_RDONLY;
687
688 ret = btrfs_commit_super(root);
689 WARN_ON(ret);
690 } else {
2b82032c
YZ
691 if (root->fs_info->fs_devices->rw_devices == 0)
692 return -EACCES;
693
c146afad
YZ
694 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
695 return -EINVAL;
696
5d4f98a2
YZ
697 /* recover relocation */
698 ret = btrfs_recover_relocation(root);
c146afad
YZ
699 WARN_ON(ret);
700
701 ret = btrfs_cleanup_fs_roots(root->fs_info);
702 WARN_ON(ret);
703
704 sb->s_flags &= ~MS_RDONLY;
705 }
706
707 return 0;
708}
709
8fd17795
CM
710static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
711{
712 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
4b52dff6 713 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
bd4d1088
JB
714 struct list_head *head = &root->fs_info->space_info;
715 struct btrfs_space_info *found;
716 u64 total_used = 0;
db94535d 717 int bits = dentry->d_sb->s_blocksize_bits;
9d03632e 718 __be32 *fsid = (__be32 *)root->fs_info->fsid;
8fd17795 719
bd4d1088 720 rcu_read_lock();
b742bb82
YZ
721 list_for_each_entry_rcu(found, head, list)
722 total_used += found->disk_used;
bd4d1088
JB
723 rcu_read_unlock();
724
8fd17795 725 buf->f_namelen = BTRFS_NAME_LEN;
db94535d 726 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
bd4d1088 727 buf->f_bfree = buf->f_blocks - (total_used >> bits);
b742bb82 728 buf->f_bavail = buf->f_bfree;
8fd17795
CM
729 buf->f_bsize = dentry->d_sb->s_blocksize;
730 buf->f_type = BTRFS_SUPER_MAGIC;
d397712b 731
9d03632e 732 /* We treat it as constant endianness (it doesn't matter _which_)
d397712b 733 because we want the fsid to come out the same whether mounted
9d03632e
DW
734 on a big-endian or little-endian host */
735 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
736 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
32d48fa1
DW
737 /* Mask in the root object ID too, to disambiguate subvols */
738 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
739 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
740
8fd17795
CM
741 return 0;
742}
b5133862 743
2e635a27
CM
744static struct file_system_type btrfs_fs_type = {
745 .owner = THIS_MODULE,
746 .name = "btrfs",
747 .get_sb = btrfs_get_sb,
a061fc8d 748 .kill_sb = kill_anon_super,
2e635a27
CM
749 .fs_flags = FS_REQUIRES_DEV,
750};
a9218f6b 751
d352ac68
CM
752/*
753 * used by btrfsctl to scan devices when no FS is mounted
754 */
8a4b83cc
CM
755static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
756 unsigned long arg)
757{
758 struct btrfs_ioctl_vol_args *vol;
759 struct btrfs_fs_devices *fs_devices;
c071fcfd 760 int ret = -ENOTTY;
8a4b83cc 761
e441d54d
CM
762 if (!capable(CAP_SYS_ADMIN))
763 return -EPERM;
764
dae7b665
LZ
765 vol = memdup_user((void __user *)arg, sizeof(*vol));
766 if (IS_ERR(vol))
767 return PTR_ERR(vol);
c071fcfd 768
8a4b83cc
CM
769 switch (cmd) {
770 case BTRFS_IOC_SCAN_DEV:
97288f2c 771 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
8a4b83cc
CM
772 &btrfs_fs_type, &fs_devices);
773 break;
774 }
dae7b665 775
8a4b83cc 776 kfree(vol);
f819d837 777 return ret;
8a4b83cc
CM
778}
779
0176260f 780static int btrfs_freeze(struct super_block *sb)
ed0dab6b
Y
781{
782 struct btrfs_root *root = btrfs_sb(sb);
a74a4b97
CM
783 mutex_lock(&root->fs_info->transaction_kthread_mutex);
784 mutex_lock(&root->fs_info->cleaner_mutex);
0176260f 785 return 0;
ed0dab6b
Y
786}
787
0176260f 788static int btrfs_unfreeze(struct super_block *sb)
ed0dab6b
Y
789{
790 struct btrfs_root *root = btrfs_sb(sb);
a74a4b97
CM
791 mutex_unlock(&root->fs_info->cleaner_mutex);
792 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
0176260f 793 return 0;
ed0dab6b 794}
2e635a27 795
b87221de 796static const struct super_operations btrfs_super_ops = {
76dda93c 797 .drop_inode = btrfs_drop_inode,
134e9731 798 .delete_inode = btrfs_delete_inode,
e20d96d6 799 .put_super = btrfs_put_super,
d5719762 800 .sync_fs = btrfs_sync_fs,
a9572a15 801 .show_options = btrfs_show_options,
4730a4bc 802 .write_inode = btrfs_write_inode,
b5133862 803 .dirty_inode = btrfs_dirty_inode,
2c90e5d6
CM
804 .alloc_inode = btrfs_alloc_inode,
805 .destroy_inode = btrfs_destroy_inode,
8fd17795 806 .statfs = btrfs_statfs,
c146afad 807 .remount_fs = btrfs_remount,
0176260f
LT
808 .freeze_fs = btrfs_freeze,
809 .unfreeze_fs = btrfs_unfreeze,
e20d96d6 810};
a9218f6b
CM
811
812static const struct file_operations btrfs_ctl_fops = {
813 .unlocked_ioctl = btrfs_control_ioctl,
814 .compat_ioctl = btrfs_control_ioctl,
815 .owner = THIS_MODULE,
816};
817
818static struct miscdevice btrfs_misc = {
819 .minor = MISC_DYNAMIC_MINOR,
820 .name = "btrfs-control",
821 .fops = &btrfs_ctl_fops
822};
823
824static int btrfs_interface_init(void)
825{
826 return misc_register(&btrfs_misc);
827}
828
b2950863 829static void btrfs_interface_exit(void)
a9218f6b
CM
830{
831 if (misc_deregister(&btrfs_misc) < 0)
d397712b 832 printk(KERN_INFO "misc_deregister failed for control device");
a9218f6b
CM
833}
834
2e635a27
CM
835static int __init init_btrfs_fs(void)
836{
2c90e5d6 837 int err;
58176a96
JB
838
839 err = btrfs_init_sysfs();
840 if (err)
841 return err;
842
39279cc3 843 err = btrfs_init_cachep();
2c90e5d6 844 if (err)
a74a4b97 845 goto free_sysfs;
d1310b2e
CM
846
847 err = extent_io_init();
2f4cbe64
WB
848 if (err)
849 goto free_cachep;
850
d1310b2e
CM
851 err = extent_map_init();
852 if (err)
853 goto free_extent_io;
854
a9218f6b 855 err = btrfs_interface_init();
2f4cbe64
WB
856 if (err)
857 goto free_extent_map;
c8b97818 858
a9218f6b
CM
859 err = register_filesystem(&btrfs_fs_type);
860 if (err)
861 goto unregister_ioctl;
b3c3da71
CM
862
863 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
2f4cbe64
WB
864 return 0;
865
a9218f6b
CM
866unregister_ioctl:
867 btrfs_interface_exit();
2f4cbe64
WB
868free_extent_map:
869 extent_map_exit();
d1310b2e
CM
870free_extent_io:
871 extent_io_exit();
2f4cbe64
WB
872free_cachep:
873 btrfs_destroy_cachep();
a74a4b97 874free_sysfs:
2f4cbe64
WB
875 btrfs_exit_sysfs();
876 return err;
2e635a27
CM
877}
878
879static void __exit exit_btrfs_fs(void)
880{
39279cc3 881 btrfs_destroy_cachep();
a52d9a80 882 extent_map_exit();
d1310b2e 883 extent_io_exit();
a9218f6b 884 btrfs_interface_exit();
2e635a27 885 unregister_filesystem(&btrfs_fs_type);
58176a96 886 btrfs_exit_sysfs();
8a4b83cc 887 btrfs_cleanup_fs_uuids();
c8b97818 888 btrfs_zlib_exit();
2e635a27
CM
889}
890
891module_init(init_btrfs_fs)
892module_exit(exit_btrfs_fs)
893
894MODULE_LICENSE("GPL");