]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/btrfs/super.c
Btrfs: fix the deadlock between the transaction start/attach and commit
[mirror_ubuntu-bionic-kernel.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>
90a887c9 42#include <linux/cleancache.h>
22c44fe6 43#include <linux/ratelimit.h>
55e301fd 44#include <linux/btrfs.h>
4b4e25f2 45#include "compat.h"
16cdcec7 46#include "delayed-inode.h"
2e635a27 47#include "ctree.h"
e20d96d6 48#include "disk-io.h"
d5719762 49#include "transaction.h"
2c90e5d6 50#include "btrfs_inode.h"
3a686375 51#include "print-tree.h"
5103e947 52#include "xattr.h"
8a4b83cc 53#include "volumes.h"
b3c3da71 54#include "version.h"
be6e8dc0 55#include "export.h"
c8b97818 56#include "compression.h"
9c5085c1 57#include "rcu-string.h"
8dabb742 58#include "dev-replace.h"
2e635a27 59
1abe9b8a 60#define CREATE_TRACE_POINTS
61#include <trace/events/btrfs.h>
62
b87221de 63static const struct super_operations btrfs_super_ops;
830c4adb 64static struct file_system_type btrfs_fs_type;
75dfe396 65
5a016047 66static const char *btrfs_decode_error(int errno, char nbuf[16])
acce952b 67{
68 char *errstr = NULL;
69
70 switch (errno) {
71 case -EIO:
72 errstr = "IO failure";
73 break;
74 case -ENOMEM:
75 errstr = "Out of memory";
76 break;
77 case -EROFS:
78 errstr = "Readonly filesystem";
79 break;
8c342930
JM
80 case -EEXIST:
81 errstr = "Object already exists";
82 break;
acce952b 83 default:
84 if (nbuf) {
85 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
86 errstr = nbuf;
87 }
88 break;
89 }
90
91 return errstr;
92}
93
94static void __save_error_info(struct btrfs_fs_info *fs_info)
95{
96 /*
97 * today we only save the error info into ram. Long term we'll
98 * also send it down to the disk
99 */
87533c47 100 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 101}
102
acce952b 103static void save_error_info(struct btrfs_fs_info *fs_info)
104{
105 __save_error_info(fs_info);
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
87533c47 116 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
acce952b 117 sb->s_flags |= MS_RDONLY;
118 printk(KERN_INFO "btrfs is forced readonly\n");
1acd6831
SB
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 */
49b25e05 129// WARN_ON(1);
acce952b 130 }
131}
132
533574c6 133#ifdef CONFIG_PRINTK
acce952b 134/*
135 * __btrfs_std_error decodes expected errors from the caller and
136 * invokes the approciate error response.
137 */
138void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
4da35113 139 unsigned int line, int errno, const char *fmt, ...)
acce952b 140{
141 struct super_block *sb = fs_info->sb;
142 char nbuf[16];
143 const char *errstr;
144
145 /*
146 * Special case: if the error is EROFS, and we're already
147 * under MS_RDONLY, then it is safe here.
148 */
149 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
4da35113
JM
150 return;
151
5a016047 152 errstr = btrfs_decode_error(errno, nbuf);
4da35113 153 if (fmt) {
37252a66
ES
154 struct va_format vaf;
155 va_list args;
156
157 va_start(args, fmt);
158 vaf.fmt = fmt;
159 vaf.va = &args;
4da35113
JM
160
161 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s (%pV)\n",
162 sb->s_id, function, line, errstr, &vaf);
37252a66 163 va_end(args);
4da35113
JM
164 } else {
165 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
166 sb->s_id, function, line, errstr);
167 }
acce952b 168
4da35113
JM
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN) {
171 save_error_info(fs_info);
172 btrfs_handle_error(fs_info);
173 }
4da35113 174}
acce952b 175
533574c6 176static const char * const logtypes[] = {
4da35113
JM
177 "emergency",
178 "alert",
179 "critical",
180 "error",
181 "warning",
182 "notice",
183 "info",
184 "debug",
185};
186
187void btrfs_printk(struct btrfs_fs_info *fs_info, const char *fmt, ...)
188{
189 struct super_block *sb = fs_info->sb;
190 char lvl[4];
191 struct va_format vaf;
192 va_list args;
193 const char *type = logtypes[4];
533574c6 194 int kern_level;
4da35113
JM
195
196 va_start(args, fmt);
197
533574c6
JP
198 kern_level = printk_get_level(fmt);
199 if (kern_level) {
200 size_t size = printk_skip_level(fmt) - fmt;
201 memcpy(lvl, fmt, size);
202 lvl[size] = '\0';
203 fmt += size;
204 type = logtypes[kern_level - '0'];
4da35113
JM
205 } else
206 *lvl = '\0';
207
208 vaf.fmt = fmt;
209 vaf.va = &args;
533574c6 210
4da35113 211 printk("%sBTRFS %s (device %s): %pV", lvl, type, sb->s_id, &vaf);
533574c6
JP
212
213 va_end(args);
214}
215
216#else
217
218void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
219 unsigned int line, int errno, const char *fmt, ...)
220{
221 struct super_block *sb = fs_info->sb;
222
223 /*
224 * Special case: if the error is EROFS, and we're already
225 * under MS_RDONLY, then it is safe here.
226 */
227 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
228 return;
229
230 /* Don't go through full error handling during mount */
231 if (sb->s_flags & MS_BORN) {
232 save_error_info(fs_info);
233 btrfs_handle_error(fs_info);
234 }
acce952b 235}
533574c6 236#endif
acce952b 237
49b25e05
JM
238/*
239 * We only mark the transaction aborted and then set the file system read-only.
240 * This will prevent new transactions from starting or trying to join this
241 * one.
242 *
243 * This means that error recovery at the call site is limited to freeing
244 * any local memory allocations and passing the error code up without
245 * further cleanup. The transaction should complete as it normally would
246 * in the call path but will return -EIO.
247 *
248 * We'll complete the cleanup in btrfs_end_transaction and
249 * btrfs_commit_transaction.
250 */
251void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
252 struct btrfs_root *root, const char *function,
253 unsigned int line, int errno)
254{
48940662 255 WARN_ONCE(1, KERN_DEBUG "btrfs: Transaction aborted\n");
49b25e05
JM
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->blocks_used) {
69ce977a
MX
260 char nbuf[16];
261 const char *errstr;
262
5a016047 263 errstr = btrfs_decode_error(errno, nbuf);
69ce977a
MX
264 btrfs_printk(root->fs_info,
265 "%s:%d: Aborting unused transaction(%s).\n",
266 function, line, errstr);
acce952b 267 return;
49b25e05 268 }
8d25a086 269 ACCESS_ONCE(trans->transaction->aborted) = errno;
49b25e05
JM
270 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
271}
8c342930
JM
272/*
273 * __btrfs_panic decodes unexpected, fatal errors from the caller,
274 * issues an alert, and either panics or BUGs, depending on mount options.
275 */
276void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
277 unsigned int line, int errno, const char *fmt, ...)
278{
279 char nbuf[16];
280 char *s_id = "<unknown>";
281 const char *errstr;
282 struct va_format vaf = { .fmt = fmt };
283 va_list args;
acce952b 284
8c342930
JM
285 if (fs_info)
286 s_id = fs_info->sb->s_id;
acce952b 287
8c342930
JM
288 va_start(args, fmt);
289 vaf.va = &args;
290
5a016047 291 errstr = btrfs_decode_error(errno, nbuf);
aa43a17c 292 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
8c342930
JM
293 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
294 s_id, function, line, &vaf, errstr);
295
296 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
297 s_id, function, line, &vaf, errstr);
298 va_end(args);
299 /* Caller calls BUG() */
acce952b 300}
301
d397712b 302static void btrfs_put_super(struct super_block *sb)
b18c6685 303{
815745cf 304 (void)close_ctree(btrfs_sb(sb)->tree_root);
aea52e19
AV
305 /* FIXME: need to fix VFS to return error? */
306 /* AV: return it _where_? ->put_super() can be triggered by any number
307 * of async events, up to and including delivery of SIGKILL to the
308 * last process that kept it busy. Or segfault in the aforementioned
309 * process... Whom would you report that to?
310 */
75dfe396
CM
311}
312
95e05289 313enum {
73f73415 314 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
287a0ab9
JB
315 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
316 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
261507a0
LZ
317 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
318 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
91435650 319 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
9555c6c1
ID
320 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
321 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
21adbd5c 322 Opt_check_integrity, Opt_check_integrity_including_extent_data,
8c342930 323 Opt_check_integrity_print_mask, Opt_fatal_errors,
9555c6c1 324 Opt_err,
95e05289
CM
325};
326
327static match_table_t tokens = {
dfe25020 328 {Opt_degraded, "degraded"},
95e05289 329 {Opt_subvol, "subvol=%s"},
73f73415 330 {Opt_subvolid, "subvolid=%d"},
43e570b0 331 {Opt_device, "device=%s"},
b6cda9bc 332 {Opt_nodatasum, "nodatasum"},
be20aa9d 333 {Opt_nodatacow, "nodatacow"},
21ad10cf 334 {Opt_nobarrier, "nobarrier"},
6f568d35 335 {Opt_max_inline, "max_inline=%s"},
8f662a76 336 {Opt_alloc_start, "alloc_start=%s"},
4543df7e 337 {Opt_thread_pool, "thread_pool=%d"},
c8b97818 338 {Opt_compress, "compress"},
261507a0 339 {Opt_compress_type, "compress=%s"},
a555f810 340 {Opt_compress_force, "compress-force"},
261507a0 341 {Opt_compress_force_type, "compress-force=%s"},
e18e4809 342 {Opt_ssd, "ssd"},
451d7585 343 {Opt_ssd_spread, "ssd_spread"},
3b30c22f 344 {Opt_nossd, "nossd"},
33268eaf 345 {Opt_noacl, "noacl"},
3a5e1404 346 {Opt_notreelog, "notreelog"},
dccae999 347 {Opt_flushoncommit, "flushoncommit"},
97e728d4 348 {Opt_ratio, "metadata_ratio=%d"},
e244a0ae 349 {Opt_discard, "discard"},
0af3d00b 350 {Opt_space_cache, "space_cache"},
88c2ba3b 351 {Opt_clear_cache, "clear_cache"},
4260f7c7 352 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
91435650 353 {Opt_enospc_debug, "enospc_debug"},
e15d0542 354 {Opt_subvolrootid, "subvolrootid=%d"},
4cb5300b 355 {Opt_defrag, "autodefrag"},
4b9465cb 356 {Opt_inode_cache, "inode_cache"},
8965593e 357 {Opt_no_space_cache, "nospace_cache"},
af31f5e5 358 {Opt_recovery, "recovery"},
9555c6c1 359 {Opt_skip_balance, "skip_balance"},
21adbd5c
SB
360 {Opt_check_integrity, "check_int"},
361 {Opt_check_integrity_including_extent_data, "check_int_data"},
362 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
8c342930 363 {Opt_fatal_errors, "fatal_errors=%s"},
33268eaf 364 {Opt_err, NULL},
95e05289
CM
365};
366
edf24abe
CH
367/*
368 * Regular mount options parser. Everything that is needed only when
369 * reading in a new superblock is parsed here.
49b25e05 370 * XXX JDM: This needs to be cleaned up for remount.
edf24abe
CH
371 */
372int btrfs_parse_options(struct btrfs_root *root, char *options)
95e05289 373{
edf24abe 374 struct btrfs_fs_info *info = root->fs_info;
95e05289 375 substring_t args[MAX_OPT_ARGS];
73bc1876
JB
376 char *p, *num, *orig = NULL;
377 u64 cache_gen;
4543df7e 378 int intarg;
a7a3f7ca 379 int ret = 0;
261507a0
LZ
380 char *compress_type;
381 bool compress_force = false;
b6cda9bc 382
6c41761f 383 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
73bc1876
JB
384 if (cache_gen)
385 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
386
95e05289 387 if (!options)
73bc1876 388 goto out;
95e05289 389
be20aa9d
CM
390 /*
391 * strsep changes the string, duplicate it because parse_options
392 * gets called twice
393 */
394 options = kstrdup(options, GFP_NOFS);
395 if (!options)
396 return -ENOMEM;
397
da495ecc 398 orig = options;
be20aa9d 399
edf24abe 400 while ((p = strsep(&options, ",")) != NULL) {
95e05289
CM
401 int token;
402 if (!*p)
403 continue;
404
405 token = match_token(p, tokens, args);
406 switch (token) {
dfe25020 407 case Opt_degraded:
edf24abe
CH
408 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
409 btrfs_set_opt(info->mount_opt, DEGRADED);
dfe25020 410 break;
95e05289 411 case Opt_subvol:
73f73415 412 case Opt_subvolid:
e15d0542 413 case Opt_subvolrootid:
43e570b0 414 case Opt_device:
edf24abe 415 /*
43e570b0 416 * These are parsed by btrfs_parse_early_options
edf24abe
CH
417 * and can be happily ignored here.
418 */
b6cda9bc
CM
419 break;
420 case Opt_nodatasum:
067c28ad 421 printk(KERN_INFO "btrfs: setting nodatasum\n");
edf24abe 422 btrfs_set_opt(info->mount_opt, NODATASUM);
be20aa9d
CM
423 break;
424 case Opt_nodatacow:
bedb2cca
AP
425 if (!btrfs_test_opt(root, COMPRESS) ||
426 !btrfs_test_opt(root, FORCE_COMPRESS)) {
427 printk(KERN_INFO "btrfs: setting nodatacow, compression disabled\n");
428 } else {
429 printk(KERN_INFO "btrfs: setting nodatacow\n");
430 }
431 info->compress_type = BTRFS_COMPRESS_NONE;
432 btrfs_clear_opt(info->mount_opt, COMPRESS);
433 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
edf24abe
CH
434 btrfs_set_opt(info->mount_opt, NODATACOW);
435 btrfs_set_opt(info->mount_opt, NODATASUM);
95e05289 436 break;
a555f810 437 case Opt_compress_force:
261507a0
LZ
438 case Opt_compress_force_type:
439 compress_force = true;
1c697d4a 440 /* Fallthrough */
261507a0
LZ
441 case Opt_compress:
442 case Opt_compress_type:
443 if (token == Opt_compress ||
444 token == Opt_compress_force ||
445 strcmp(args[0].from, "zlib") == 0) {
446 compress_type = "zlib";
447 info->compress_type = BTRFS_COMPRESS_ZLIB;
063849ea 448 btrfs_set_opt(info->mount_opt, COMPRESS);
bedb2cca
AP
449 btrfs_clear_opt(info->mount_opt, NODATACOW);
450 btrfs_clear_opt(info->mount_opt, NODATASUM);
a6fa6fae
LZ
451 } else if (strcmp(args[0].from, "lzo") == 0) {
452 compress_type = "lzo";
453 info->compress_type = BTRFS_COMPRESS_LZO;
063849ea 454 btrfs_set_opt(info->mount_opt, COMPRESS);
bedb2cca
AP
455 btrfs_clear_opt(info->mount_opt, NODATACOW);
456 btrfs_clear_opt(info->mount_opt, NODATASUM);
2b0ce2c2 457 btrfs_set_fs_incompat(info, COMPRESS_LZO);
063849ea
AH
458 } else if (strncmp(args[0].from, "no", 2) == 0) {
459 compress_type = "no";
460 info->compress_type = BTRFS_COMPRESS_NONE;
461 btrfs_clear_opt(info->mount_opt, COMPRESS);
462 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
463 compress_force = false;
261507a0
LZ
464 } else {
465 ret = -EINVAL;
466 goto out;
467 }
468
261507a0
LZ
469 if (compress_force) {
470 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
471 pr_info("btrfs: force %s compression\n",
472 compress_type);
473 } else
474 pr_info("btrfs: use %s compression\n",
475 compress_type);
a555f810 476 break;
e18e4809 477 case Opt_ssd:
edf24abe
CH
478 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
479 btrfs_set_opt(info->mount_opt, SSD);
e18e4809 480 break;
451d7585
CM
481 case Opt_ssd_spread:
482 printk(KERN_INFO "btrfs: use spread ssd "
483 "allocation scheme\n");
484 btrfs_set_opt(info->mount_opt, SSD);
485 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
486 break;
3b30c22f 487 case Opt_nossd:
451d7585
CM
488 printk(KERN_INFO "btrfs: not using ssd allocation "
489 "scheme\n");
c289811c 490 btrfs_set_opt(info->mount_opt, NOSSD);
3b30c22f 491 btrfs_clear_opt(info->mount_opt, SSD);
451d7585 492 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
3b30c22f 493 break;
21ad10cf 494 case Opt_nobarrier:
edf24abe
CH
495 printk(KERN_INFO "btrfs: turning off barriers\n");
496 btrfs_set_opt(info->mount_opt, NOBARRIER);
21ad10cf 497 break;
4543df7e
CM
498 case Opt_thread_pool:
499 intarg = 0;
500 match_int(&args[0], &intarg);
0d2450ab 501 if (intarg)
4543df7e 502 info->thread_pool_size = intarg;
4543df7e 503 break;
6f568d35 504 case Opt_max_inline:
edf24abe
CH
505 num = match_strdup(&args[0]);
506 if (num) {
91748467 507 info->max_inline = memparse(num, NULL);
edf24abe
CH
508 kfree(num);
509
15ada040
CM
510 if (info->max_inline) {
511 info->max_inline = max_t(u64,
512 info->max_inline,
513 root->sectorsize);
514 }
edf24abe 515 printk(KERN_INFO "btrfs: max_inline at %llu\n",
21380931 516 (unsigned long long)info->max_inline);
6f568d35
CM
517 }
518 break;
8f662a76 519 case Opt_alloc_start:
edf24abe
CH
520 num = match_strdup(&args[0]);
521 if (num) {
c018daec 522 mutex_lock(&info->chunk_mutex);
91748467 523 info->alloc_start = memparse(num, NULL);
c018daec 524 mutex_unlock(&info->chunk_mutex);
edf24abe
CH
525 kfree(num);
526 printk(KERN_INFO
527 "btrfs: allocations start at %llu\n",
21380931 528 (unsigned long long)info->alloc_start);
8f662a76
CM
529 }
530 break;
33268eaf
JB
531 case Opt_noacl:
532 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
533 break;
3a5e1404
SW
534 case Opt_notreelog:
535 printk(KERN_INFO "btrfs: disabling tree log\n");
536 btrfs_set_opt(info->mount_opt, NOTREELOG);
537 break;
dccae999
SW
538 case Opt_flushoncommit:
539 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
540 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
541 break;
97e728d4
JB
542 case Opt_ratio:
543 intarg = 0;
544 match_int(&args[0], &intarg);
545 if (intarg) {
546 info->metadata_ratio = intarg;
547 printk(KERN_INFO "btrfs: metadata ratio %d\n",
548 info->metadata_ratio);
549 }
550 break;
e244a0ae
CH
551 case Opt_discard:
552 btrfs_set_opt(info->mount_opt, DISCARD);
553 break;
0af3d00b 554 case Opt_space_cache:
0af3d00b 555 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
0de90876 556 break;
73bc1876
JB
557 case Opt_no_space_cache:
558 printk(KERN_INFO "btrfs: disabling disk space caching\n");
559 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
560 break;
4b9465cb
CM
561 case Opt_inode_cache:
562 printk(KERN_INFO "btrfs: enabling inode map caching\n");
563 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
564 break;
88c2ba3b
JB
565 case Opt_clear_cache:
566 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
567 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
0af3d00b 568 break;
4260f7c7
SW
569 case Opt_user_subvol_rm_allowed:
570 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
571 break;
91435650
CM
572 case Opt_enospc_debug:
573 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
574 break;
4cb5300b 575 case Opt_defrag:
48940662 576 printk(KERN_INFO "btrfs: enabling auto defrag\n");
4cb5300b
CM
577 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
578 break;
af31f5e5 579 case Opt_recovery:
48940662 580 printk(KERN_INFO "btrfs: enabling auto recovery\n");
af31f5e5
CM
581 btrfs_set_opt(info->mount_opt, RECOVERY);
582 break;
9555c6c1
ID
583 case Opt_skip_balance:
584 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
585 break;
21adbd5c
SB
586#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
587 case Opt_check_integrity_including_extent_data:
588 printk(KERN_INFO "btrfs: enabling check integrity"
589 " including extent data\n");
590 btrfs_set_opt(info->mount_opt,
591 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
592 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
593 break;
594 case Opt_check_integrity:
595 printk(KERN_INFO "btrfs: enabling check integrity\n");
596 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
597 break;
598 case Opt_check_integrity_print_mask:
599 intarg = 0;
600 match_int(&args[0], &intarg);
601 if (intarg) {
602 info->check_integrity_print_mask = intarg;
603 printk(KERN_INFO "btrfs:"
604 " check_integrity_print_mask 0x%x\n",
605 info->check_integrity_print_mask);
606 }
607 break;
608#else
609 case Opt_check_integrity_including_extent_data:
610 case Opt_check_integrity:
611 case Opt_check_integrity_print_mask:
612 printk(KERN_ERR "btrfs: support for check_integrity*"
613 " not compiled in!\n");
614 ret = -EINVAL;
615 goto out;
616#endif
8c342930
JM
617 case Opt_fatal_errors:
618 if (strcmp(args[0].from, "panic") == 0)
619 btrfs_set_opt(info->mount_opt,
620 PANIC_ON_FATAL_ERROR);
621 else if (strcmp(args[0].from, "bug") == 0)
622 btrfs_clear_opt(info->mount_opt,
623 PANIC_ON_FATAL_ERROR);
624 else {
625 ret = -EINVAL;
626 goto out;
627 }
628 break;
a7a3f7ca
SW
629 case Opt_err:
630 printk(KERN_INFO "btrfs: unrecognized mount option "
631 "'%s'\n", p);
632 ret = -EINVAL;
633 goto out;
95e05289 634 default:
be20aa9d 635 break;
95e05289
CM
636 }
637 }
a7a3f7ca 638out:
73bc1876
JB
639 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
640 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
da495ecc 641 kfree(orig);
a7a3f7ca 642 return ret;
edf24abe
CH
643}
644
645/*
646 * Parse mount options that are required early in the mount process.
647 *
648 * All other options will be parsed on much later in the mount process and
649 * only when we need to allocate a new super block.
650 */
97288f2c 651static int btrfs_parse_early_options(const char *options, fmode_t flags,
73f73415 652 void *holder, char **subvol_name, u64 *subvol_objectid,
e15d0542 653 u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
edf24abe
CH
654{
655 substring_t args[MAX_OPT_ARGS];
83c8c9bd 656 char *device_name, *opts, *orig, *p;
edf24abe 657 int error = 0;
73f73415 658 int intarg;
edf24abe
CH
659
660 if (!options)
830c4adb 661 return 0;
edf24abe
CH
662
663 /*
664 * strsep changes the string, duplicate it because parse_options
665 * gets called twice
666 */
667 opts = kstrdup(options, GFP_KERNEL);
668 if (!opts)
669 return -ENOMEM;
3f3d0bc0 670 orig = opts;
edf24abe
CH
671
672 while ((p = strsep(&opts, ",")) != NULL) {
673 int token;
674 if (!*p)
675 continue;
676
677 token = match_token(p, tokens, args);
678 switch (token) {
679 case Opt_subvol:
a90e8b6f 680 kfree(*subvol_name);
edf24abe
CH
681 *subvol_name = match_strdup(&args[0]);
682 break;
73f73415
JB
683 case Opt_subvolid:
684 intarg = 0;
4849f01d
JB
685 error = match_int(&args[0], &intarg);
686 if (!error) {
687 /* we want the original fs_tree */
688 if (!intarg)
689 *subvol_objectid =
690 BTRFS_FS_TREE_OBJECTID;
691 else
692 *subvol_objectid = intarg;
693 }
73f73415 694 break;
e15d0542
XZ
695 case Opt_subvolrootid:
696 intarg = 0;
697 error = match_int(&args[0], &intarg);
698 if (!error) {
699 /* we want the original fs_tree */
700 if (!intarg)
701 *subvol_rootid =
702 BTRFS_FS_TREE_OBJECTID;
703 else
704 *subvol_rootid = intarg;
705 }
706 break;
43e570b0 707 case Opt_device:
83c8c9bd
JL
708 device_name = match_strdup(&args[0]);
709 if (!device_name) {
710 error = -ENOMEM;
711 goto out;
712 }
713 error = btrfs_scan_one_device(device_name,
43e570b0 714 flags, holder, fs_devices);
83c8c9bd 715 kfree(device_name);
43e570b0 716 if (error)
830c4adb 717 goto out;
43e570b0 718 break;
edf24abe
CH
719 default:
720 break;
721 }
722 }
723
830c4adb 724out:
3f3d0bc0 725 kfree(orig);
edf24abe 726 return error;
95e05289
CM
727}
728
73f73415
JB
729static struct dentry *get_default_root(struct super_block *sb,
730 u64 subvol_objectid)
731{
815745cf
AV
732 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
733 struct btrfs_root *root = fs_info->tree_root;
73f73415
JB
734 struct btrfs_root *new_root;
735 struct btrfs_dir_item *di;
736 struct btrfs_path *path;
737 struct btrfs_key location;
738 struct inode *inode;
73f73415
JB
739 u64 dir_id;
740 int new = 0;
741
742 /*
743 * We have a specific subvol we want to mount, just setup location and
744 * go look up the root.
745 */
746 if (subvol_objectid) {
747 location.objectid = subvol_objectid;
748 location.type = BTRFS_ROOT_ITEM_KEY;
749 location.offset = (u64)-1;
750 goto find_root;
751 }
752
753 path = btrfs_alloc_path();
754 if (!path)
755 return ERR_PTR(-ENOMEM);
756 path->leave_spinning = 1;
757
758 /*
759 * Find the "default" dir item which points to the root item that we
760 * will mount by default if we haven't been given a specific subvolume
761 * to mount.
762 */
815745cf 763 dir_id = btrfs_super_root_dir(fs_info->super_copy);
73f73415 764 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
b0839166
JL
765 if (IS_ERR(di)) {
766 btrfs_free_path(path);
fb4f6f91 767 return ERR_CAST(di);
b0839166 768 }
73f73415
JB
769 if (!di) {
770 /*
771 * Ok the default dir item isn't there. This is weird since
772 * it's always been there, but don't freak out, just try and
773 * mount to root most subvolume.
774 */
775 btrfs_free_path(path);
776 dir_id = BTRFS_FIRST_FREE_OBJECTID;
815745cf 777 new_root = fs_info->fs_root;
73f73415
JB
778 goto setup_root;
779 }
780
781 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
782 btrfs_free_path(path);
783
784find_root:
815745cf 785 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
73f73415 786 if (IS_ERR(new_root))
d0b678cb 787 return ERR_CAST(new_root);
73f73415
JB
788
789 if (btrfs_root_refs(&new_root->root_item) == 0)
790 return ERR_PTR(-ENOENT);
791
792 dir_id = btrfs_root_dirid(&new_root->root_item);
793setup_root:
794 location.objectid = dir_id;
795 location.type = BTRFS_INODE_ITEM_KEY;
796 location.offset = 0;
797
798 inode = btrfs_iget(sb, &location, new_root, &new);
4cbd1149
DC
799 if (IS_ERR(inode))
800 return ERR_CAST(inode);
73f73415
JB
801
802 /*
803 * If we're just mounting the root most subvol put the inode and return
804 * a reference to the dentry. We will have already gotten a reference
805 * to the inode in btrfs_fill_super so we're good to go.
806 */
807 if (!new && sb->s_root->d_inode == inode) {
808 iput(inode);
809 return dget(sb->s_root);
810 }
811
ba5b8958 812 return d_obtain_alias(inode);
73f73415
JB
813}
814
d397712b 815static int btrfs_fill_super(struct super_block *sb,
8a4b83cc 816 struct btrfs_fs_devices *fs_devices,
d397712b 817 void *data, int silent)
75dfe396 818{
d397712b 819 struct inode *inode;
815745cf 820 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
5d4f98a2 821 struct btrfs_key key;
39279cc3 822 int err;
a429e513 823
39279cc3
CM
824 sb->s_maxbytes = MAX_LFS_FILESIZE;
825 sb->s_magic = BTRFS_SUPER_MAGIC;
826 sb->s_op = &btrfs_super_ops;
af53d29a 827 sb->s_d_op = &btrfs_dentry_operations;
be6e8dc0 828 sb->s_export_op = &btrfs_export_ops;
5103e947 829 sb->s_xattr = btrfs_xattr_handlers;
39279cc3 830 sb->s_time_gran = 1;
0eda294d 831#ifdef CONFIG_BTRFS_FS_POSIX_ACL
33268eaf 832 sb->s_flags |= MS_POSIXACL;
49cf6f45 833#endif
0c4d2d95 834 sb->s_flags |= MS_I_VERSION;
ad2b2c80
AV
835 err = open_ctree(sb, fs_devices, (char *)data);
836 if (err) {
39279cc3 837 printk("btrfs: open_ctree failed\n");
ad2b2c80 838 return err;
a429e513
CM
839 }
840
5d4f98a2
YZ
841 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
842 key.type = BTRFS_INODE_ITEM_KEY;
843 key.offset = 0;
98c7089c 844 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
5d4f98a2
YZ
845 if (IS_ERR(inode)) {
846 err = PTR_ERR(inode);
39279cc3 847 goto fail_close;
f254e52c 848 }
f254e52c 849
48fde701
AV
850 sb->s_root = d_make_root(inode);
851 if (!sb->s_root) {
39279cc3
CM
852 err = -ENOMEM;
853 goto fail_close;
f254e52c 854 }
58176a96 855
6885f308 856 save_mount_options(sb, data);
90a887c9 857 cleancache_init_fs(sb);
59553edf 858 sb->s_flags |= MS_ACTIVE;
2619ba1f 859 return 0;
39279cc3
CM
860
861fail_close:
815745cf 862 close_ctree(fs_info->tree_root);
39279cc3 863 return err;
2619ba1f
CM
864}
865
6bf13c0c 866int btrfs_sync_fs(struct super_block *sb, int wait)
c5739bba
CM
867{
868 struct btrfs_trans_handle *trans;
815745cf
AV
869 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
870 struct btrfs_root *root = fs_info->tree_root;
2619ba1f 871
1abe9b8a 872 trace_btrfs_sync_fs(wait);
873
39279cc3 874 if (!wait) {
815745cf 875 filemap_flush(fs_info->btree_inode->i_mapping);
39279cc3
CM
876 return 0;
877 }
771ed689 878
6bbe3a9c 879 btrfs_wait_ordered_extents(root, 0);
771ed689 880
354aa0fb 881 trans = btrfs_attach_transaction(root);
60376ce4 882 if (IS_ERR(trans)) {
354aa0fb
MX
883 /* no transaction, don't bother */
884 if (PTR_ERR(trans) == -ENOENT)
60376ce4 885 return 0;
98d5dc13 886 return PTR_ERR(trans);
60376ce4 887 }
bd7de2c9 888 return btrfs_commit_transaction(trans, root);
2c90e5d6
CM
889}
890
34c80b1d 891static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
a9572a15 892{
815745cf
AV
893 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
894 struct btrfs_root *root = info->tree_root;
200da64e 895 char *compress_type;
a9572a15
EP
896
897 if (btrfs_test_opt(root, DEGRADED))
898 seq_puts(seq, ",degraded");
899 if (btrfs_test_opt(root, NODATASUM))
900 seq_puts(seq, ",nodatasum");
901 if (btrfs_test_opt(root, NODATACOW))
902 seq_puts(seq, ",nodatacow");
903 if (btrfs_test_opt(root, NOBARRIER))
904 seq_puts(seq, ",nobarrier");
a9572a15 905 if (info->max_inline != 8192 * 1024)
21380931
JB
906 seq_printf(seq, ",max_inline=%llu",
907 (unsigned long long)info->max_inline);
a9572a15 908 if (info->alloc_start != 0)
21380931
JB
909 seq_printf(seq, ",alloc_start=%llu",
910 (unsigned long long)info->alloc_start);
a9572a15
EP
911 if (info->thread_pool_size != min_t(unsigned long,
912 num_online_cpus() + 2, 8))
913 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
200da64e
TI
914 if (btrfs_test_opt(root, COMPRESS)) {
915 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
916 compress_type = "zlib";
917 else
918 compress_type = "lzo";
919 if (btrfs_test_opt(root, FORCE_COMPRESS))
920 seq_printf(seq, ",compress-force=%s", compress_type);
921 else
922 seq_printf(seq, ",compress=%s", compress_type);
923 }
c289811c
CM
924 if (btrfs_test_opt(root, NOSSD))
925 seq_puts(seq, ",nossd");
451d7585
CM
926 if (btrfs_test_opt(root, SSD_SPREAD))
927 seq_puts(seq, ",ssd_spread");
928 else if (btrfs_test_opt(root, SSD))
a9572a15 929 seq_puts(seq, ",ssd");
3a5e1404 930 if (btrfs_test_opt(root, NOTREELOG))
6b65c5c6 931 seq_puts(seq, ",notreelog");
dccae999 932 if (btrfs_test_opt(root, FLUSHONCOMMIT))
6b65c5c6 933 seq_puts(seq, ",flushoncommit");
20a5239a
MW
934 if (btrfs_test_opt(root, DISCARD))
935 seq_puts(seq, ",discard");
a9572a15
EP
936 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
937 seq_puts(seq, ",noacl");
200da64e
TI
938 if (btrfs_test_opt(root, SPACE_CACHE))
939 seq_puts(seq, ",space_cache");
73bc1876 940 else
8965593e 941 seq_puts(seq, ",nospace_cache");
200da64e
TI
942 if (btrfs_test_opt(root, CLEAR_CACHE))
943 seq_puts(seq, ",clear_cache");
944 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
945 seq_puts(seq, ",user_subvol_rm_allowed");
0942caa3
DS
946 if (btrfs_test_opt(root, ENOSPC_DEBUG))
947 seq_puts(seq, ",enospc_debug");
948 if (btrfs_test_opt(root, AUTO_DEFRAG))
949 seq_puts(seq, ",autodefrag");
950 if (btrfs_test_opt(root, INODE_MAP_CACHE))
951 seq_puts(seq, ",inode_cache");
9555c6c1
ID
952 if (btrfs_test_opt(root, SKIP_BALANCE))
953 seq_puts(seq, ",skip_balance");
8c342930
JM
954 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
955 seq_puts(seq, ",fatal_errors=panic");
a9572a15
EP
956 return 0;
957}
958
a061fc8d 959static int btrfs_test_super(struct super_block *s, void *data)
4b82d6e4 960{
815745cf
AV
961 struct btrfs_fs_info *p = data;
962 struct btrfs_fs_info *fs_info = btrfs_sb(s);
4b82d6e4 963
815745cf 964 return fs_info->fs_devices == p->fs_devices;
4b82d6e4
Y
965}
966
450ba0ea
JB
967static int btrfs_set_super(struct super_block *s, void *data)
968{
6de1d09d
AV
969 int err = set_anon_super(s, data);
970 if (!err)
971 s->s_fs_info = data;
972 return err;
4b82d6e4
Y
973}
974
f9d9ef62
DS
975/*
976 * subvolumes are identified by ino 256
977 */
978static inline int is_subvolume_inode(struct inode *inode)
979{
980 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
981 return 1;
982 return 0;
983}
984
830c4adb
JB
985/*
986 * This will strip out the subvol=%s argument for an argument string and add
987 * subvolid=0 to make sure we get the actual tree root for path walking to the
988 * subvol we want.
989 */
990static char *setup_root_args(char *args)
991{
f60d16a8
JM
992 unsigned len = strlen(args) + 2 + 1;
993 char *src, *dst, *buf;
830c4adb
JB
994
995 /*
f60d16a8
JM
996 * We need the same args as before, but with this substitution:
997 * s!subvol=[^,]+!subvolid=0!
830c4adb 998 *
f60d16a8
JM
999 * Since the replacement string is up to 2 bytes longer than the
1000 * original, allocate strlen(args) + 2 + 1 bytes.
830c4adb 1001 */
830c4adb 1002
f60d16a8 1003 src = strstr(args, "subvol=");
830c4adb 1004 /* This shouldn't happen, but just in case.. */
f60d16a8
JM
1005 if (!src)
1006 return NULL;
1007
1008 buf = dst = kmalloc(len, GFP_NOFS);
1009 if (!buf)
830c4adb 1010 return NULL;
830c4adb
JB
1011
1012 /*
f60d16a8
JM
1013 * If the subvol= arg is not at the start of the string,
1014 * copy whatever precedes it into buf.
830c4adb 1015 */
f60d16a8
JM
1016 if (src != args) {
1017 *src++ = '\0';
1018 strcpy(buf, args);
1019 dst += strlen(args);
830c4adb
JB
1020 }
1021
f60d16a8
JM
1022 strcpy(dst, "subvolid=0");
1023 dst += strlen("subvolid=0");
830c4adb
JB
1024
1025 /*
f60d16a8
JM
1026 * If there is a "," after the original subvol=... string,
1027 * copy that suffix into our buffer. Otherwise, we're done.
830c4adb 1028 */
f60d16a8
JM
1029 src = strchr(src, ',');
1030 if (src)
1031 strcpy(dst, src);
830c4adb 1032
f60d16a8 1033 return buf;
830c4adb
JB
1034}
1035
1036static struct dentry *mount_subvol(const char *subvol_name, int flags,
1037 const char *device_name, char *data)
1038{
830c4adb
JB
1039 struct dentry *root;
1040 struct vfsmount *mnt;
830c4adb 1041 char *newargs;
830c4adb
JB
1042
1043 newargs = setup_root_args(data);
1044 if (!newargs)
1045 return ERR_PTR(-ENOMEM);
1046 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1047 newargs);
1048 kfree(newargs);
1049 if (IS_ERR(mnt))
1050 return ERR_CAST(mnt);
1051
ea441d11 1052 root = mount_subtree(mnt, subvol_name);
830c4adb 1053
ea441d11
AV
1054 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1055 struct super_block *s = root->d_sb;
1056 dput(root);
1057 root = ERR_PTR(-EINVAL);
1058 deactivate_locked_super(s);
f9d9ef62
DS
1059 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1060 subvol_name);
f9d9ef62
DS
1061 }
1062
830c4adb
JB
1063 return root;
1064}
450ba0ea 1065
edf24abe
CH
1066/*
1067 * Find a superblock for the given device / mount point.
1068 *
1069 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1070 * for multiple device setup. Make sure to keep it in sync.
1071 */
061dbc6b 1072static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
306e16ce 1073 const char *device_name, void *data)
4b82d6e4
Y
1074{
1075 struct block_device *bdev = NULL;
1076 struct super_block *s;
1077 struct dentry *root;
8a4b83cc 1078 struct btrfs_fs_devices *fs_devices = NULL;
450ba0ea 1079 struct btrfs_fs_info *fs_info = NULL;
97288f2c 1080 fmode_t mode = FMODE_READ;
73f73415
JB
1081 char *subvol_name = NULL;
1082 u64 subvol_objectid = 0;
e15d0542 1083 u64 subvol_rootid = 0;
4b82d6e4
Y
1084 int error = 0;
1085
97288f2c
CH
1086 if (!(flags & MS_RDONLY))
1087 mode |= FMODE_WRITE;
1088
1089 error = btrfs_parse_early_options(data, mode, fs_type,
73f73415 1090 &subvol_name, &subvol_objectid,
e15d0542 1091 &subvol_rootid, &fs_devices);
f23c8af8
ID
1092 if (error) {
1093 kfree(subvol_name);
061dbc6b 1094 return ERR_PTR(error);
f23c8af8 1095 }
edf24abe 1096
830c4adb
JB
1097 if (subvol_name) {
1098 root = mount_subvol(subvol_name, flags, device_name, data);
1099 kfree(subvol_name);
1100 return root;
1101 }
1102
306e16ce 1103 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
8a4b83cc 1104 if (error)
830c4adb 1105 return ERR_PTR(error);
4b82d6e4 1106
450ba0ea
JB
1107 /*
1108 * Setup a dummy root and fs_info for test/set super. This is because
1109 * we don't actually fill this stuff out until open_ctree, but we need
1110 * it for searching for existing supers, so this lets us do that and
1111 * then open_ctree will properly initialize everything later.
1112 */
1113 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
04d21a24
ID
1114 if (!fs_info)
1115 return ERR_PTR(-ENOMEM);
1116
450ba0ea 1117 fs_info->fs_devices = fs_devices;
450ba0ea 1118
6c41761f
DS
1119 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1120 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1121 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1122 error = -ENOMEM;
04d21a24
ID
1123 goto error_fs_info;
1124 }
1125
1126 error = btrfs_open_devices(fs_devices, mode, fs_type);
1127 if (error)
1128 goto error_fs_info;
1129
1130 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1131 error = -EACCES;
6c41761f
DS
1132 goto error_close_devices;
1133 }
1134
dfe25020 1135 bdev = fs_devices->latest_bdev;
9249e17f
DH
1136 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1137 fs_info);
830c4adb
JB
1138 if (IS_ERR(s)) {
1139 error = PTR_ERR(s);
1140 goto error_close_devices;
1141 }
4b82d6e4
Y
1142
1143 if (s->s_root) {
2b82032c 1144 btrfs_close_devices(fs_devices);
6c41761f 1145 free_fs_info(fs_info);
59553edf
AV
1146 if ((flags ^ s->s_flags) & MS_RDONLY)
1147 error = -EBUSY;
4b82d6e4
Y
1148 } else {
1149 char b[BDEVNAME_SIZE];
1150
4b82d6e4 1151 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
815745cf 1152 btrfs_sb(s)->bdev_holder = fs_type;
8a4b83cc
CM
1153 error = btrfs_fill_super(s, fs_devices, data,
1154 flags & MS_SILENT ? 1 : 0);
4b82d6e4
Y
1155 }
1156
59553edf
AV
1157 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1158 if (IS_ERR(root))
830c4adb 1159 deactivate_locked_super(s);
4b82d6e4 1160
061dbc6b 1161 return root;
4b82d6e4 1162
c146afad 1163error_close_devices:
8a4b83cc 1164 btrfs_close_devices(fs_devices);
04d21a24 1165error_fs_info:
6c41761f 1166 free_fs_info(fs_info);
061dbc6b 1167 return ERR_PTR(error);
4b82d6e4 1168}
2e635a27 1169
0d2450ab
ST
1170static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1171{
1172 spin_lock_irq(&workers->lock);
1173 workers->max_workers = new_limit;
1174 spin_unlock_irq(&workers->lock);
1175}
1176
1177static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1178 int new_pool_size, int old_pool_size)
1179{
1180 if (new_pool_size == old_pool_size)
1181 return;
1182
1183 fs_info->thread_pool_size = new_pool_size;
1184
1185 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1186 old_pool_size, new_pool_size);
1187
1188 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1189 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1190 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1191 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1192 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1193 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1194 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1195 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1196 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1197 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1198 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1199 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1200 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
ff023aac
SB
1201 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1202 new_pool_size);
0d2450ab
ST
1203}
1204
c146afad
YZ
1205static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1206{
815745cf
AV
1207 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1208 struct btrfs_root *root = fs_info->tree_root;
49b25e05
JM
1209 unsigned old_flags = sb->s_flags;
1210 unsigned long old_opts = fs_info->mount_opt;
1211 unsigned long old_compress_type = fs_info->compress_type;
1212 u64 old_max_inline = fs_info->max_inline;
1213 u64 old_alloc_start = fs_info->alloc_start;
1214 int old_thread_pool_size = fs_info->thread_pool_size;
1215 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
c146afad
YZ
1216 int ret;
1217
b288052e 1218 ret = btrfs_parse_options(root, data);
49b25e05
JM
1219 if (ret) {
1220 ret = -EINVAL;
1221 goto restore;
1222 }
b288052e 1223
0d2450ab
ST
1224 btrfs_resize_thread_pool(fs_info,
1225 fs_info->thread_pool_size, old_thread_pool_size);
1226
c146afad
YZ
1227 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1228 return 0;
1229
1230 if (*flags & MS_RDONLY) {
8dabb742
SB
1231 /*
1232 * this also happens on 'umount -rf' or on shutdown, when
1233 * the filesystem is busy.
1234 */
c146afad
YZ
1235 sb->s_flags |= MS_RDONLY;
1236
8dabb742
SB
1237 btrfs_dev_replace_suspend_for_unmount(fs_info);
1238 btrfs_scrub_cancel(fs_info);
1239
49b25e05
JM
1240 ret = btrfs_commit_super(root);
1241 if (ret)
1242 goto restore;
c146afad 1243 } else {
8a3db184 1244 if (fs_info->fs_devices->rw_devices == 0) {
49b25e05
JM
1245 ret = -EACCES;
1246 goto restore;
8a3db184 1247 }
2b82032c 1248
292fd7fc
SB
1249 if (fs_info->fs_devices->missing_devices >
1250 fs_info->num_tolerated_disk_barrier_failures &&
1251 !(*flags & MS_RDONLY)) {
1252 printk(KERN_WARNING
1253 "Btrfs: too many missing devices, writeable remount is not allowed\n");
1254 ret = -EACCES;
1255 goto restore;
1256 }
1257
8a3db184 1258 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
49b25e05
JM
1259 ret = -EINVAL;
1260 goto restore;
8a3db184 1261 }
c146afad 1262
815745cf 1263 ret = btrfs_cleanup_fs_roots(fs_info);
49b25e05
JM
1264 if (ret)
1265 goto restore;
c146afad 1266
d68fc57b
YZ
1267 /* recover relocation */
1268 ret = btrfs_recover_relocation(root);
49b25e05
JM
1269 if (ret)
1270 goto restore;
c146afad 1271
2b6ba629
ID
1272 ret = btrfs_resume_balance_async(fs_info);
1273 if (ret)
1274 goto restore;
1275
8dabb742
SB
1276 ret = btrfs_resume_dev_replace_async(fs_info);
1277 if (ret) {
1278 pr_warn("btrfs: failed to resume dev_replace\n");
1279 goto restore;
1280 }
c146afad
YZ
1281 sb->s_flags &= ~MS_RDONLY;
1282 }
1283
1284 return 0;
49b25e05
JM
1285
1286restore:
1287 /* We've hit an error - don't reset MS_RDONLY */
1288 if (sb->s_flags & MS_RDONLY)
1289 old_flags |= MS_RDONLY;
1290 sb->s_flags = old_flags;
1291 fs_info->mount_opt = old_opts;
1292 fs_info->compress_type = old_compress_type;
1293 fs_info->max_inline = old_max_inline;
c018daec 1294 mutex_lock(&fs_info->chunk_mutex);
49b25e05 1295 fs_info->alloc_start = old_alloc_start;
c018daec 1296 mutex_unlock(&fs_info->chunk_mutex);
0d2450ab
ST
1297 btrfs_resize_thread_pool(fs_info,
1298 old_thread_pool_size, fs_info->thread_pool_size);
49b25e05
JM
1299 fs_info->metadata_ratio = old_metadata_ratio;
1300 return ret;
c146afad
YZ
1301}
1302
bcd53741
AJ
1303/* Used to sort the devices by max_avail(descending sort) */
1304static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1305 const void *dev_info2)
1306{
1307 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1308 ((struct btrfs_device_info *)dev_info2)->max_avail)
1309 return -1;
1310 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1311 ((struct btrfs_device_info *)dev_info2)->max_avail)
1312 return 1;
1313 else
1314 return 0;
1315}
1316
1317/*
1318 * sort the devices by max_avail, in which max free extent size of each device
1319 * is stored.(Descending Sort)
1320 */
1321static inline void btrfs_descending_sort_devices(
1322 struct btrfs_device_info *devices,
1323 size_t nr_devices)
1324{
1325 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1326 btrfs_cmp_device_free_bytes, NULL);
1327}
1328
6d07bcec
MX
1329/*
1330 * The helper to calc the free space on the devices that can be used to store
1331 * file data.
1332 */
1333static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1334{
1335 struct btrfs_fs_info *fs_info = root->fs_info;
1336 struct btrfs_device_info *devices_info;
1337 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1338 struct btrfs_device *device;
1339 u64 skip_space;
1340 u64 type;
1341 u64 avail_space;
1342 u64 used_space;
1343 u64 min_stripe_size;
39fb26c3 1344 int min_stripes = 1, num_stripes = 1;
6d07bcec
MX
1345 int i = 0, nr_devices;
1346 int ret;
1347
b772a86e 1348 nr_devices = fs_info->fs_devices->open_devices;
6d07bcec
MX
1349 BUG_ON(!nr_devices);
1350
1351 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1352 GFP_NOFS);
1353 if (!devices_info)
1354 return -ENOMEM;
1355
1356 /* calc min stripe number for data space alloction */
1357 type = btrfs_get_alloc_profile(root, 1);
39fb26c3 1358 if (type & BTRFS_BLOCK_GROUP_RAID0) {
6d07bcec 1359 min_stripes = 2;
39fb26c3
MX
1360 num_stripes = nr_devices;
1361 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
6d07bcec 1362 min_stripes = 2;
39fb26c3
MX
1363 num_stripes = 2;
1364 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
6d07bcec 1365 min_stripes = 4;
39fb26c3
MX
1366 num_stripes = 4;
1367 }
6d07bcec
MX
1368
1369 if (type & BTRFS_BLOCK_GROUP_DUP)
1370 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1371 else
1372 min_stripe_size = BTRFS_STRIPE_LEN;
1373
b772a86e 1374 list_for_each_entry(device, &fs_devices->devices, dev_list) {
63a212ab
SB
1375 if (!device->in_fs_metadata || !device->bdev ||
1376 device->is_tgtdev_for_dev_replace)
6d07bcec
MX
1377 continue;
1378
1379 avail_space = device->total_bytes - device->bytes_used;
1380
1381 /* align with stripe_len */
1382 do_div(avail_space, BTRFS_STRIPE_LEN);
1383 avail_space *= BTRFS_STRIPE_LEN;
1384
1385 /*
1386 * In order to avoid overwritting the superblock on the drive,
1387 * btrfs starts at an offset of at least 1MB when doing chunk
1388 * allocation.
1389 */
1390 skip_space = 1024 * 1024;
1391
1392 /* user can set the offset in fs_info->alloc_start. */
1393 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1394 device->total_bytes)
1395 skip_space = max(fs_info->alloc_start, skip_space);
1396
1397 /*
1398 * btrfs can not use the free space in [0, skip_space - 1],
1399 * we must subtract it from the total. In order to implement
1400 * it, we account the used space in this range first.
1401 */
1402 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1403 &used_space);
1404 if (ret) {
1405 kfree(devices_info);
1406 return ret;
1407 }
1408
1409 /* calc the free space in [0, skip_space - 1] */
1410 skip_space -= used_space;
1411
1412 /*
1413 * we can use the free space in [0, skip_space - 1], subtract
1414 * it from the total.
1415 */
1416 if (avail_space && avail_space >= skip_space)
1417 avail_space -= skip_space;
1418 else
1419 avail_space = 0;
1420
1421 if (avail_space < min_stripe_size)
1422 continue;
1423
1424 devices_info[i].dev = device;
1425 devices_info[i].max_avail = avail_space;
1426
1427 i++;
1428 }
1429
1430 nr_devices = i;
1431
1432 btrfs_descending_sort_devices(devices_info, nr_devices);
1433
1434 i = nr_devices - 1;
1435 avail_space = 0;
1436 while (nr_devices >= min_stripes) {
39fb26c3
MX
1437 if (num_stripes > nr_devices)
1438 num_stripes = nr_devices;
1439
6d07bcec
MX
1440 if (devices_info[i].max_avail >= min_stripe_size) {
1441 int j;
1442 u64 alloc_size;
1443
39fb26c3 1444 avail_space += devices_info[i].max_avail * num_stripes;
6d07bcec 1445 alloc_size = devices_info[i].max_avail;
39fb26c3 1446 for (j = i + 1 - num_stripes; j <= i; j++)
6d07bcec
MX
1447 devices_info[j].max_avail -= alloc_size;
1448 }
1449 i--;
1450 nr_devices--;
1451 }
1452
1453 kfree(devices_info);
1454 *free_bytes = avail_space;
1455 return 0;
1456}
1457
8fd17795
CM
1458static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1459{
815745cf
AV
1460 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1461 struct btrfs_super_block *disk_super = fs_info->super_copy;
1462 struct list_head *head = &fs_info->space_info;
bd4d1088
JB
1463 struct btrfs_space_info *found;
1464 u64 total_used = 0;
6d07bcec 1465 u64 total_free_data = 0;
db94535d 1466 int bits = dentry->d_sb->s_blocksize_bits;
815745cf 1467 __be32 *fsid = (__be32 *)fs_info->fsid;
6d07bcec 1468 int ret;
8fd17795 1469
6d07bcec 1470 /* holding chunk_muext to avoid allocating new chunks */
815745cf 1471 mutex_lock(&fs_info->chunk_mutex);
bd4d1088 1472 rcu_read_lock();
89a55897 1473 list_for_each_entry_rcu(found, head, list) {
6d07bcec
MX
1474 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1475 total_free_data += found->disk_total - found->disk_used;
1476 total_free_data -=
1477 btrfs_account_ro_block_groups_free_space(found);
1478 }
1479
b742bb82 1480 total_used += found->disk_used;
89a55897 1481 }
bd4d1088
JB
1482 rcu_read_unlock();
1483
8fd17795 1484 buf->f_namelen = BTRFS_NAME_LEN;
db94535d 1485 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
bd4d1088 1486 buf->f_bfree = buf->f_blocks - (total_used >> bits);
8fd17795
CM
1487 buf->f_bsize = dentry->d_sb->s_blocksize;
1488 buf->f_type = BTRFS_SUPER_MAGIC;
6d07bcec 1489 buf->f_bavail = total_free_data;
815745cf 1490 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
6d07bcec 1491 if (ret) {
815745cf 1492 mutex_unlock(&fs_info->chunk_mutex);
6d07bcec
MX
1493 return ret;
1494 }
1495 buf->f_bavail += total_free_data;
1496 buf->f_bavail = buf->f_bavail >> bits;
815745cf 1497 mutex_unlock(&fs_info->chunk_mutex);
d397712b 1498
9d03632e 1499 /* We treat it as constant endianness (it doesn't matter _which_)
d397712b 1500 because we want the fsid to come out the same whether mounted
9d03632e
DW
1501 on a big-endian or little-endian host */
1502 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1503 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
32d48fa1
DW
1504 /* Mask in the root object ID too, to disambiguate subvols */
1505 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1506 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1507
8fd17795
CM
1508 return 0;
1509}
b5133862 1510
aea52e19
AV
1511static void btrfs_kill_super(struct super_block *sb)
1512{
815745cf 1513 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
aea52e19 1514 kill_anon_super(sb);
d22ca7de 1515 free_fs_info(fs_info);
aea52e19
AV
1516}
1517
2e635a27
CM
1518static struct file_system_type btrfs_fs_type = {
1519 .owner = THIS_MODULE,
1520 .name = "btrfs",
061dbc6b 1521 .mount = btrfs_mount,
aea52e19 1522 .kill_sb = btrfs_kill_super,
2e635a27
CM
1523 .fs_flags = FS_REQUIRES_DEV,
1524};
a9218f6b 1525
d352ac68
CM
1526/*
1527 * used by btrfsctl to scan devices when no FS is mounted
1528 */
8a4b83cc
CM
1529static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1530 unsigned long arg)
1531{
1532 struct btrfs_ioctl_vol_args *vol;
1533 struct btrfs_fs_devices *fs_devices;
c071fcfd 1534 int ret = -ENOTTY;
8a4b83cc 1535
e441d54d
CM
1536 if (!capable(CAP_SYS_ADMIN))
1537 return -EPERM;
1538
dae7b665
LZ
1539 vol = memdup_user((void __user *)arg, sizeof(*vol));
1540 if (IS_ERR(vol))
1541 return PTR_ERR(vol);
c071fcfd 1542
8a4b83cc
CM
1543 switch (cmd) {
1544 case BTRFS_IOC_SCAN_DEV:
97288f2c 1545 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
8a4b83cc
CM
1546 &btrfs_fs_type, &fs_devices);
1547 break;
02db0844
JB
1548 case BTRFS_IOC_DEVICES_READY:
1549 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1550 &btrfs_fs_type, &fs_devices);
1551 if (ret)
1552 break;
1553 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1554 break;
8a4b83cc 1555 }
dae7b665 1556
8a4b83cc 1557 kfree(vol);
f819d837 1558 return ret;
8a4b83cc
CM
1559}
1560
0176260f 1561static int btrfs_freeze(struct super_block *sb)
ed0dab6b 1562{
354aa0fb
MX
1563 struct btrfs_trans_handle *trans;
1564 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1565
1566 trans = btrfs_attach_transaction(root);
1567 if (IS_ERR(trans)) {
1568 /* no transaction, don't bother */
1569 if (PTR_ERR(trans) == -ENOENT)
1570 return 0;
1571 return PTR_ERR(trans);
1572 }
1573 return btrfs_commit_transaction(trans, root);
ed0dab6b
Y
1574}
1575
0176260f 1576static int btrfs_unfreeze(struct super_block *sb)
ed0dab6b 1577{
0176260f 1578 return 0;
ed0dab6b 1579}
2e635a27 1580
9c5085c1
JB
1581static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1582{
1583 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1584 struct btrfs_fs_devices *cur_devices;
1585 struct btrfs_device *dev, *first_dev = NULL;
1586 struct list_head *head;
1587 struct rcu_string *name;
1588
1589 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1590 cur_devices = fs_info->fs_devices;
1591 while (cur_devices) {
1592 head = &cur_devices->devices;
1593 list_for_each_entry(dev, head, dev_list) {
aa9ddcd4
JB
1594 if (dev->missing)
1595 continue;
9c5085c1
JB
1596 if (!first_dev || dev->devid < first_dev->devid)
1597 first_dev = dev;
1598 }
1599 cur_devices = cur_devices->seed;
1600 }
1601
1602 if (first_dev) {
1603 rcu_read_lock();
1604 name = rcu_dereference(first_dev->name);
1605 seq_escape(m, name->str, " \t\n\\");
1606 rcu_read_unlock();
1607 } else {
1608 WARN_ON(1);
1609 }
1610 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1611 return 0;
1612}
1613
b87221de 1614static const struct super_operations btrfs_super_ops = {
76dda93c 1615 .drop_inode = btrfs_drop_inode,
bd555975 1616 .evict_inode = btrfs_evict_inode,
e20d96d6 1617 .put_super = btrfs_put_super,
d5719762 1618 .sync_fs = btrfs_sync_fs,
a9572a15 1619 .show_options = btrfs_show_options,
9c5085c1 1620 .show_devname = btrfs_show_devname,
4730a4bc 1621 .write_inode = btrfs_write_inode,
2c90e5d6
CM
1622 .alloc_inode = btrfs_alloc_inode,
1623 .destroy_inode = btrfs_destroy_inode,
8fd17795 1624 .statfs = btrfs_statfs,
c146afad 1625 .remount_fs = btrfs_remount,
0176260f
LT
1626 .freeze_fs = btrfs_freeze,
1627 .unfreeze_fs = btrfs_unfreeze,
e20d96d6 1628};
a9218f6b
CM
1629
1630static const struct file_operations btrfs_ctl_fops = {
1631 .unlocked_ioctl = btrfs_control_ioctl,
1632 .compat_ioctl = btrfs_control_ioctl,
1633 .owner = THIS_MODULE,
6038f373 1634 .llseek = noop_llseek,
a9218f6b
CM
1635};
1636
1637static struct miscdevice btrfs_misc = {
578454ff 1638 .minor = BTRFS_MINOR,
a9218f6b
CM
1639 .name = "btrfs-control",
1640 .fops = &btrfs_ctl_fops
1641};
1642
578454ff
KS
1643MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1644MODULE_ALIAS("devname:btrfs-control");
1645
a9218f6b
CM
1646static int btrfs_interface_init(void)
1647{
1648 return misc_register(&btrfs_misc);
1649}
1650
b2950863 1651static void btrfs_interface_exit(void)
a9218f6b
CM
1652{
1653 if (misc_deregister(&btrfs_misc) < 0)
48940662 1654 printk(KERN_INFO "btrfs: misc_deregister failed for control device\n");
a9218f6b
CM
1655}
1656
2e635a27
CM
1657static int __init init_btrfs_fs(void)
1658{
2c90e5d6 1659 int err;
58176a96
JB
1660
1661 err = btrfs_init_sysfs();
1662 if (err)
1663 return err;
1664
143bede5 1665 btrfs_init_compress();
d1310b2e 1666
261507a0
LZ
1667 err = btrfs_init_cachep();
1668 if (err)
1669 goto free_compress;
1670
d1310b2e 1671 err = extent_io_init();
2f4cbe64
WB
1672 if (err)
1673 goto free_cachep;
1674
d1310b2e
CM
1675 err = extent_map_init();
1676 if (err)
1677 goto free_extent_io;
1678
6352b91d 1679 err = ordered_data_init();
2f4cbe64
WB
1680 if (err)
1681 goto free_extent_map;
c8b97818 1682
6352b91d
MX
1683 err = btrfs_delayed_inode_init();
1684 if (err)
1685 goto free_ordered_data;
1686
9247f317 1687 err = btrfs_auto_defrag_init();
16cdcec7
MX
1688 if (err)
1689 goto free_delayed_inode;
1690
78a6184a 1691 err = btrfs_delayed_ref_init();
9247f317
MX
1692 if (err)
1693 goto free_auto_defrag;
1694
78a6184a
MX
1695 err = btrfs_interface_init();
1696 if (err)
1697 goto free_delayed_ref;
1698
a9218f6b
CM
1699 err = register_filesystem(&btrfs_fs_type);
1700 if (err)
1701 goto unregister_ioctl;
b3c3da71 1702
e565d4b9
JS
1703 btrfs_init_lockdep();
1704
b3c3da71 1705 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
2f4cbe64
WB
1706 return 0;
1707
a9218f6b
CM
1708unregister_ioctl:
1709 btrfs_interface_exit();
78a6184a
MX
1710free_delayed_ref:
1711 btrfs_delayed_ref_exit();
9247f317
MX
1712free_auto_defrag:
1713 btrfs_auto_defrag_exit();
16cdcec7
MX
1714free_delayed_inode:
1715 btrfs_delayed_inode_exit();
6352b91d
MX
1716free_ordered_data:
1717 ordered_data_exit();
2f4cbe64
WB
1718free_extent_map:
1719 extent_map_exit();
d1310b2e
CM
1720free_extent_io:
1721 extent_io_exit();
2f4cbe64
WB
1722free_cachep:
1723 btrfs_destroy_cachep();
261507a0
LZ
1724free_compress:
1725 btrfs_exit_compress();
2f4cbe64
WB
1726 btrfs_exit_sysfs();
1727 return err;
2e635a27
CM
1728}
1729
1730static void __exit exit_btrfs_fs(void)
1731{
39279cc3 1732 btrfs_destroy_cachep();
78a6184a 1733 btrfs_delayed_ref_exit();
9247f317 1734 btrfs_auto_defrag_exit();
16cdcec7 1735 btrfs_delayed_inode_exit();
6352b91d 1736 ordered_data_exit();
a52d9a80 1737 extent_map_exit();
d1310b2e 1738 extent_io_exit();
a9218f6b 1739 btrfs_interface_exit();
2e635a27 1740 unregister_filesystem(&btrfs_fs_type);
58176a96 1741 btrfs_exit_sysfs();
8a4b83cc 1742 btrfs_cleanup_fs_uuids();
261507a0 1743 btrfs_exit_compress();
2e635a27
CM
1744}
1745
1746module_init(init_btrfs_fs)
1747module_exit(exit_btrfs_fs)
1748
1749MODULE_LICENSE("GPL");