projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[deliverable/linux.git] / fs / btrfs / super.c
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 "xattr.h"
52 #include "volumes.h"
53 #include "export.h"
54 #include "compression.h"
55 #include "rcu-string.h"
56 #include "dev-replace.h"
57 #include "free-space-cache.h"
58 #include "backref.h"
59 #include "tests/btrfs-tests.h"
60
61 #define CREATE_TRACE_POINTS
62 #include <trace/events/btrfs.h>
63
64 static const struct super_operations btrfs_super_ops;
65 static struct file_system_type btrfs_fs_type;
66
67 static const char *btrfs_decode_error(int errno)
68 {
69 char *errstr = "unknown";
70
71 switch (errno) {
72 case -EIO:
73 errstr = "IO failure";
74 break;
75 case -ENOMEM:
76 errstr = "Out of memory";
77 break;
78 case -EROFS:
79 errstr = "Readonly filesystem";
80 break;
81 case -EEXIST:
82 errstr = "Object already exists";
83 break;
84 case -ENOSPC:
85 errstr = "No space left";
86 break;
87 case -ENOENT:
88 errstr = "No such entry";
89 break;
90 }
91
92 return errstr;
93 }
94
95 static void save_error_info(struct btrfs_fs_info *fs_info)
96 {
97 /*
98 * today we only save the error info into ram. Long term we'll
99 * also send it down to the disk
100 */
101 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
102 }
103
104 /* btrfs handle error by forcing the filesystem readonly */
105 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
106 {
107 struct super_block *sb = fs_info->sb;
108
109 if (sb->s_flags & MS_RDONLY)
110 return;
111
112 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
113 sb->s_flags |= MS_RDONLY;
114 btrfs_info(fs_info, "forced readonly");
115 /*
116 * Note that a running device replace operation is not
117 * canceled here although there is no way to update
118 * the progress. It would add the risk of a deadlock,
119 * therefore the canceling is ommited. The only penalty
120 * is that some I/O remains active until the procedure
121 * completes. The next time when the filesystem is
122 * mounted writeable again, the device replace
123 * operation continues.
124 */
125 }
126 }
127
128 #ifdef CONFIG_PRINTK
129 /*
130 * __btrfs_std_error decodes expected errors from the caller and
131 * invokes the approciate error response.
132 */
133 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
134 unsigned int line, int errno, const char *fmt, ...)
135 {
136 struct super_block *sb = fs_info->sb;
137 const char *errstr;
138
139 /*
140 * Special case: if the error is EROFS, and we're already
141 * under MS_RDONLY, then it is safe here.
142 */
143 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
144 return;
145
146 errstr = btrfs_decode_error(errno);
147 if (fmt) {
148 struct va_format vaf;
149 va_list args;
150
151 va_start(args, fmt);
152 vaf.fmt = fmt;
153 vaf.va = &args;
154
155 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s (%pV)\n",
156 sb->s_id, function, line, errno, errstr, &vaf);
157 va_end(args);
158 } else {
159 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s\n",
160 sb->s_id, function, line, errno, errstr);
161 }
162
163 /* Don't go through full error handling during mount */
164 save_error_info(fs_info);
165 if (sb->s_flags & MS_BORN)
166 btrfs_handle_error(fs_info);
167 }
168
169 static const char * const logtypes[] = {
170 "emergency",
171 "alert",
172 "critical",
173 "error",
174 "warning",
175 "notice",
176 "info",
177 "debug",
178 };
179
180 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
181 {
182 struct super_block *sb = fs_info->sb;
183 char lvl[4];
184 struct va_format vaf;
185 va_list args;
186 const char *type = logtypes[4];
187 int kern_level;
188
189 va_start(args, fmt);
190
191 kern_level = printk_get_level(fmt);
192 if (kern_level) {
193 size_t size = printk_skip_level(fmt) - fmt;
194 memcpy(lvl, fmt, size);
195 lvl[size] = '\0';
196 fmt += size;
197 type = logtypes[kern_level - '0'];
198 } else
199 *lvl = '\0';
200
201 vaf.fmt = fmt;
202 vaf.va = &args;
203
204 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
205
206 va_end(args);
207 }
208
209 #else
210
211 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
212 unsigned int line, int errno, const char *fmt, ...)
213 {
214 struct super_block *sb = fs_info->sb;
215
216 /*
217 * Special case: if the error is EROFS, and we're already
218 * under MS_RDONLY, then it is safe here.
219 */
220 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
221 return;
222
223 /* Don't go through full error handling during mount */
224 if (sb->s_flags & MS_BORN) {
225 save_error_info(fs_info);
226 btrfs_handle_error(fs_info);
227 }
228 }
229 #endif
230
231 /*
232 * We only mark the transaction aborted and then set the file system read-only.
233 * This will prevent new transactions from starting or trying to join this
234 * one.
235 *
236 * This means that error recovery at the call site is limited to freeing
237 * any local memory allocations and passing the error code up without
238 * further cleanup. The transaction should complete as it normally would
239 * in the call path but will return -EIO.
240 *
241 * We'll complete the cleanup in btrfs_end_transaction and
242 * btrfs_commit_transaction.
243 */
244 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
245 struct btrfs_root *root, const char *function,
246 unsigned int line, int errno)
247 {
248 /*
249 * Report first abort since mount
250 */
251 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
252 &root->fs_info->fs_state)) {
253 WARN(1, KERN_DEBUG "btrfs: Transaction aborted (error %d)\n",
254 errno);
255 }
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) {
260 const char *errstr;
261
262 errstr = btrfs_decode_error(errno);
263 btrfs_warn(root->fs_info,
264 "%s:%d: Aborting unused transaction(%s).",
265 function, line, errstr);
266 return;
267 }
268 ACCESS_ONCE(trans->transaction->aborted) = errno;
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&root->fs_info->transaction_wait);
271 wake_up(&root->fs_info->transaction_blocked_wait);
272 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
273 }
274 /*
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
277 */
278 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
279 unsigned int line, int errno, const char *fmt, ...)
280 {
281 char *s_id = "<unknown>";
282 const char *errstr;
283 struct va_format vaf = { .fmt = fmt };
284 va_list args;
285
286 if (fs_info)
287 s_id = fs_info->sb->s_id;
288
289 va_start(args, fmt);
290 vaf.va = &args;
291
292 errstr = btrfs_decode_error(errno);
293 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
294 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
295 s_id, function, line, &vaf, errno, errstr);
296
297 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
298 s_id, function, line, &vaf, errno, errstr);
299 va_end(args);
300 /* Caller calls BUG() */
301 }
302
303 static void btrfs_put_super(struct super_block *sb)
304 {
305 (void)close_ctree(btrfs_sb(sb)->tree_root);
306 /* FIXME: need to fix VFS to return error? */
307 /* AV: return it _where_? ->put_super() can be triggered by any number
308 * of async events, up to and including delivery of SIGKILL to the
309 * last process that kept it busy. Or segfault in the aforementioned
310 * process... Whom would you report that to?
311 */
312 }
313
314 enum {
315 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
316 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
317 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
318 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
319 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
320 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
321 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
322 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
323 Opt_check_integrity, Opt_check_integrity_including_extent_data,
324 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
325 Opt_commit_interval,
326 Opt_err,
327 };
328
329 static match_table_t tokens = {
330 {Opt_degraded, "degraded"},
331 {Opt_subvol, "subvol=%s"},
332 {Opt_subvolid, "subvolid=%s"},
333 {Opt_device, "device=%s"},
334 {Opt_nodatasum, "nodatasum"},
335 {Opt_nodatacow, "nodatacow"},
336 {Opt_nobarrier, "nobarrier"},
337 {Opt_max_inline, "max_inline=%s"},
338 {Opt_alloc_start, "alloc_start=%s"},
339 {Opt_thread_pool, "thread_pool=%d"},
340 {Opt_compress, "compress"},
341 {Opt_compress_type, "compress=%s"},
342 {Opt_compress_force, "compress-force"},
343 {Opt_compress_force_type, "compress-force=%s"},
344 {Opt_ssd, "ssd"},
345 {Opt_ssd_spread, "ssd_spread"},
346 {Opt_nossd, "nossd"},
347 {Opt_noacl, "noacl"},
348 {Opt_notreelog, "notreelog"},
349 {Opt_flushoncommit, "flushoncommit"},
350 {Opt_ratio, "metadata_ratio=%d"},
351 {Opt_discard, "discard"},
352 {Opt_space_cache, "space_cache"},
353 {Opt_clear_cache, "clear_cache"},
354 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
355 {Opt_enospc_debug, "enospc_debug"},
356 {Opt_subvolrootid, "subvolrootid=%d"},
357 {Opt_defrag, "autodefrag"},
358 {Opt_inode_cache, "inode_cache"},
359 {Opt_no_space_cache, "nospace_cache"},
360 {Opt_recovery, "recovery"},
361 {Opt_skip_balance, "skip_balance"},
362 {Opt_check_integrity, "check_int"},
363 {Opt_check_integrity_including_extent_data, "check_int_data"},
364 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
365 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
366 {Opt_fatal_errors, "fatal_errors=%s"},
367 {Opt_commit_interval, "commit=%d"},
368 {Opt_err, NULL},
369 };
370
371 /*
372 * Regular mount options parser. Everything that is needed only when
373 * reading in a new superblock is parsed here.
374 * XXX JDM: This needs to be cleaned up for remount.
375 */
376 int btrfs_parse_options(struct btrfs_root *root, char *options)
377 {
378 struct btrfs_fs_info *info = root->fs_info;
379 substring_t args[MAX_OPT_ARGS];
380 char *p, *num, *orig = NULL;
381 u64 cache_gen;
382 int intarg;
383 int ret = 0;
384 char *compress_type;
385 bool compress_force = false;
386
387 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
388 if (cache_gen)
389 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
390
391 if (!options)
392 goto out;
393
394 /*
395 * strsep changes the string, duplicate it because parse_options
396 * gets called twice
397 */
398 options = kstrdup(options, GFP_NOFS);
399 if (!options)
400 return -ENOMEM;
401
402 orig = options;
403
404 while ((p = strsep(&options, ",")) != NULL) {
405 int token;
406 if (!*p)
407 continue;
408
409 token = match_token(p, tokens, args);
410 switch (token) {
411 case Opt_degraded:
412 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
413 btrfs_set_opt(info->mount_opt, DEGRADED);
414 break;
415 case Opt_subvol:
416 case Opt_subvolid:
417 case Opt_subvolrootid:
418 case Opt_device:
419 /*
420 * These are parsed by btrfs_parse_early_options
421 * and can be happily ignored here.
422 */
423 break;
424 case Opt_nodatasum:
425 printk(KERN_INFO "btrfs: setting nodatasum\n");
426 btrfs_set_opt(info->mount_opt, NODATASUM);
427 break;
428 case Opt_nodatacow:
429 if (!btrfs_test_opt(root, COMPRESS) ||
430 !btrfs_test_opt(root, FORCE_COMPRESS)) {
431 printk(KERN_INFO "btrfs: setting nodatacow, compression disabled\n");
432 } else {
433 printk(KERN_INFO "btrfs: setting nodatacow\n");
434 }
435 btrfs_clear_opt(info->mount_opt, COMPRESS);
436 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
437 btrfs_set_opt(info->mount_opt, NODATACOW);
438 btrfs_set_opt(info->mount_opt, NODATASUM);
439 break;
440 case Opt_compress_force:
441 case Opt_compress_force_type:
442 compress_force = true;
443 /* Fallthrough */
444 case Opt_compress:
445 case Opt_compress_type:
446 if (token == Opt_compress ||
447 token == Opt_compress_force ||
448 strcmp(args[0].from, "zlib") == 0) {
449 compress_type = "zlib";
450 info->compress_type = BTRFS_COMPRESS_ZLIB;
451 btrfs_set_opt(info->mount_opt, COMPRESS);
452 btrfs_clear_opt(info->mount_opt, NODATACOW);
453 btrfs_clear_opt(info->mount_opt, NODATASUM);
454 } else if (strcmp(args[0].from, "lzo") == 0) {
455 compress_type = "lzo";
456 info->compress_type = BTRFS_COMPRESS_LZO;
457 btrfs_set_opt(info->mount_opt, COMPRESS);
458 btrfs_clear_opt(info->mount_opt, NODATACOW);
459 btrfs_clear_opt(info->mount_opt, NODATASUM);
460 btrfs_set_fs_incompat(info, COMPRESS_LZO);
461 } else if (strncmp(args[0].from, "no", 2) == 0) {
462 compress_type = "no";
463 btrfs_clear_opt(info->mount_opt, COMPRESS);
464 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
465 compress_force = false;
466 } else {
467 ret = -EINVAL;
468 goto out;
469 }
470
471 if (compress_force) {
472 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
473 pr_info("btrfs: force %s compression\n",
474 compress_type);
475 } else if (btrfs_test_opt(root, COMPRESS)) {
476 pr_info("btrfs: use %s compression\n",
477 compress_type);
478 }
479 break;
480 case Opt_ssd:
481 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
482 btrfs_set_opt(info->mount_opt, SSD);
483 break;
484 case Opt_ssd_spread:
485 printk(KERN_INFO "btrfs: use spread ssd "
486 "allocation scheme\n");
487 btrfs_set_opt(info->mount_opt, SSD);
488 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
489 break;
490 case Opt_nossd:
491 printk(KERN_INFO "btrfs: not using ssd allocation "
492 "scheme\n");
493 btrfs_set_opt(info->mount_opt, NOSSD);
494 btrfs_clear_opt(info->mount_opt, SSD);
495 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
496 break;
497 case Opt_nobarrier:
498 printk(KERN_INFO "btrfs: turning off barriers\n");
499 btrfs_set_opt(info->mount_opt, NOBARRIER);
500 break;
501 case Opt_thread_pool:
502 ret = match_int(&args[0], &intarg);
503 if (ret) {
504 goto out;
505 } else if (intarg > 0) {
506 info->thread_pool_size = intarg;
507 } else {
508 ret = -EINVAL;
509 goto out;
510 }
511 break;
512 case Opt_max_inline:
513 num = match_strdup(&args[0]);
514 if (num) {
515 info->max_inline = memparse(num, NULL);
516 kfree(num);
517
518 if (info->max_inline) {
519 info->max_inline = max_t(u64,
520 info->max_inline,
521 root->sectorsize);
522 }
523 printk(KERN_INFO "btrfs: max_inline at %llu\n",
524 info->max_inline);
525 } else {
526 ret = -ENOMEM;
527 goto out;
528 }
529 break;
530 case Opt_alloc_start:
531 num = match_strdup(&args[0]);
532 if (num) {
533 mutex_lock(&info->chunk_mutex);
534 info->alloc_start = memparse(num, NULL);
535 mutex_unlock(&info->chunk_mutex);
536 kfree(num);
537 printk(KERN_INFO
538 "btrfs: allocations start at %llu\n",
539 info->alloc_start);
540 } else {
541 ret = -ENOMEM;
542 goto out;
543 }
544 break;
545 case Opt_noacl:
546 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
547 break;
548 case Opt_notreelog:
549 printk(KERN_INFO "btrfs: disabling tree log\n");
550 btrfs_set_opt(info->mount_opt, NOTREELOG);
551 break;
552 case Opt_flushoncommit:
553 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
554 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
555 break;
556 case Opt_ratio:
557 ret = match_int(&args[0], &intarg);
558 if (ret) {
559 goto out;
560 } else if (intarg >= 0) {
561 info->metadata_ratio = intarg;
562 printk(KERN_INFO "btrfs: metadata ratio %d\n",
563 info->metadata_ratio);
564 } else {
565 ret = -EINVAL;
566 goto out;
567 }
568 break;
569 case Opt_discard:
570 btrfs_set_opt(info->mount_opt, DISCARD);
571 break;
572 case Opt_space_cache:
573 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
574 break;
575 case Opt_rescan_uuid_tree:
576 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
577 break;
578 case Opt_no_space_cache:
579 printk(KERN_INFO "btrfs: disabling disk space caching\n");
580 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
581 break;
582 case Opt_inode_cache:
583 printk(KERN_INFO "btrfs: enabling inode map caching\n");
584 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
585 break;
586 case Opt_clear_cache:
587 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
588 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
589 break;
590 case Opt_user_subvol_rm_allowed:
591 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
592 break;
593 case Opt_enospc_debug:
594 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
595 break;
596 case Opt_defrag:
597 printk(KERN_INFO "btrfs: enabling auto defrag\n");
598 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
599 break;
600 case Opt_recovery:
601 printk(KERN_INFO "btrfs: enabling auto recovery\n");
602 btrfs_set_opt(info->mount_opt, RECOVERY);
603 break;
604 case Opt_skip_balance:
605 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
606 break;
607 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
608 case Opt_check_integrity_including_extent_data:
609 printk(KERN_INFO "btrfs: enabling check integrity"
610 " including extent data\n");
611 btrfs_set_opt(info->mount_opt,
612 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
613 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
614 break;
615 case Opt_check_integrity:
616 printk(KERN_INFO "btrfs: enabling check integrity\n");
617 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
618 break;
619 case Opt_check_integrity_print_mask:
620 ret = match_int(&args[0], &intarg);
621 if (ret) {
622 goto out;
623 } else if (intarg >= 0) {
624 info->check_integrity_print_mask = intarg;
625 printk(KERN_INFO "btrfs:"
626 " check_integrity_print_mask 0x%x\n",
627 info->check_integrity_print_mask);
628 } else {
629 ret = -EINVAL;
630 goto out;
631 }
632 break;
633 #else
634 case Opt_check_integrity_including_extent_data:
635 case Opt_check_integrity:
636 case Opt_check_integrity_print_mask:
637 printk(KERN_ERR "btrfs: support for check_integrity*"
638 " not compiled in!\n");
639 ret = -EINVAL;
640 goto out;
641 #endif
642 case Opt_fatal_errors:
643 if (strcmp(args[0].from, "panic") == 0)
644 btrfs_set_opt(info->mount_opt,
645 PANIC_ON_FATAL_ERROR);
646 else if (strcmp(args[0].from, "bug") == 0)
647 btrfs_clear_opt(info->mount_opt,
648 PANIC_ON_FATAL_ERROR);
649 else {
650 ret = -EINVAL;
651 goto out;
652 }
653 break;
654 case Opt_commit_interval:
655 intarg = 0;
656 ret = match_int(&args[0], &intarg);
657 if (ret < 0) {
658 printk(KERN_ERR
659 "btrfs: invalid commit interval\n");
660 ret = -EINVAL;
661 goto out;
662 }
663 if (intarg > 0) {
664 if (intarg > 300) {
665 printk(KERN_WARNING
666 "btrfs: excessive commit interval %d\n",
667 intarg);
668 }
669 info->commit_interval = intarg;
670 } else {
671 printk(KERN_INFO
672 "btrfs: using default commit interval %ds\n",
673 BTRFS_DEFAULT_COMMIT_INTERVAL);
674 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
675 }
676 break;
677 case Opt_err:
678 printk(KERN_INFO "btrfs: unrecognized mount option "
679 "'%s'\n", p);
680 ret = -EINVAL;
681 goto out;
682 default:
683 break;
684 }
685 }
686 out:
687 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
688 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
689 kfree(orig);
690 return ret;
691 }
692
693 /*
694 * Parse mount options that are required early in the mount process.
695 *
696 * All other options will be parsed on much later in the mount process and
697 * only when we need to allocate a new super block.
698 */
699 static int btrfs_parse_early_options(const char *options, fmode_t flags,
700 void *holder, char **subvol_name, u64 *subvol_objectid,
701 struct btrfs_fs_devices **fs_devices)
702 {
703 substring_t args[MAX_OPT_ARGS];
704 char *device_name, *opts, *orig, *p;
705 char *num = NULL;
706 int error = 0;
707
708 if (!options)
709 return 0;
710
711 /*
712 * strsep changes the string, duplicate it because parse_options
713 * gets called twice
714 */
715 opts = kstrdup(options, GFP_KERNEL);
716 if (!opts)
717 return -ENOMEM;
718 orig = opts;
719
720 while ((p = strsep(&opts, ",")) != NULL) {
721 int token;
722 if (!*p)
723 continue;
724
725 token = match_token(p, tokens, args);
726 switch (token) {
727 case Opt_subvol:
728 kfree(*subvol_name);
729 *subvol_name = match_strdup(&args[0]);
730 if (!*subvol_name) {
731 error = -ENOMEM;
732 goto out;
733 }
734 break;
735 case Opt_subvolid:
736 num = match_strdup(&args[0]);
737 if (num) {
738 *subvol_objectid = memparse(num, NULL);
739 kfree(num);
740 /* we want the original fs_tree */
741 if (!*subvol_objectid)
742 *subvol_objectid =
743 BTRFS_FS_TREE_OBJECTID;
744 } else {
745 error = -EINVAL;
746 goto out;
747 }
748 break;
749 case Opt_subvolrootid:
750 printk(KERN_WARNING
751 "btrfs: 'subvolrootid' mount option is deprecated and has no effect\n");
752 break;
753 case Opt_device:
754 device_name = match_strdup(&args[0]);
755 if (!device_name) {
756 error = -ENOMEM;
757 goto out;
758 }
759 error = btrfs_scan_one_device(device_name,
760 flags, holder, fs_devices);
761 kfree(device_name);
762 if (error)
763 goto out;
764 break;
765 default:
766 break;
767 }
768 }
769
770 out:
771 kfree(orig);
772 return error;
773 }
774
775 static struct dentry *get_default_root(struct super_block *sb,
776 u64 subvol_objectid)
777 {
778 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
779 struct btrfs_root *root = fs_info->tree_root;
780 struct btrfs_root *new_root;
781 struct btrfs_dir_item *di;
782 struct btrfs_path *path;
783 struct btrfs_key location;
784 struct inode *inode;
785 u64 dir_id;
786 int new = 0;
787
788 /*
789 * We have a specific subvol we want to mount, just setup location and
790 * go look up the root.
791 */
792 if (subvol_objectid) {
793 location.objectid = subvol_objectid;
794 location.type = BTRFS_ROOT_ITEM_KEY;
795 location.offset = (u64)-1;
796 goto find_root;
797 }
798
799 path = btrfs_alloc_path();
800 if (!path)
801 return ERR_PTR(-ENOMEM);
802 path->leave_spinning = 1;
803
804 /*
805 * Find the "default" dir item which points to the root item that we
806 * will mount by default if we haven't been given a specific subvolume
807 * to mount.
808 */
809 dir_id = btrfs_super_root_dir(fs_info->super_copy);
810 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
811 if (IS_ERR(di)) {
812 btrfs_free_path(path);
813 return ERR_CAST(di);
814 }
815 if (!di) {
816 /*
817 * Ok the default dir item isn't there. This is weird since
818 * it's always been there, but don't freak out, just try and
819 * mount to root most subvolume.
820 */
821 btrfs_free_path(path);
822 dir_id = BTRFS_FIRST_FREE_OBJECTID;
823 new_root = fs_info->fs_root;
824 goto setup_root;
825 }
826
827 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
828 btrfs_free_path(path);
829
830 find_root:
831 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
832 if (IS_ERR(new_root))
833 return ERR_CAST(new_root);
834
835 dir_id = btrfs_root_dirid(&new_root->root_item);
836 setup_root:
837 location.objectid = dir_id;
838 location.type = BTRFS_INODE_ITEM_KEY;
839 location.offset = 0;
840
841 inode = btrfs_iget(sb, &location, new_root, &new);
842 if (IS_ERR(inode))
843 return ERR_CAST(inode);
844
845 /*
846 * If we're just mounting the root most subvol put the inode and return
847 * a reference to the dentry. We will have already gotten a reference
848 * to the inode in btrfs_fill_super so we're good to go.
849 */
850 if (!new && sb->s_root->d_inode == inode) {
851 iput(inode);
852 return dget(sb->s_root);
853 }
854
855 return d_obtain_alias(inode);
856 }
857
858 static int btrfs_fill_super(struct super_block *sb,
859 struct btrfs_fs_devices *fs_devices,
860 void *data, int silent)
861 {
862 struct inode *inode;
863 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
864 struct btrfs_key key;
865 int err;
866
867 sb->s_maxbytes = MAX_LFS_FILESIZE;
868 sb->s_magic = BTRFS_SUPER_MAGIC;
869 sb->s_op = &btrfs_super_ops;
870 sb->s_d_op = &btrfs_dentry_operations;
871 sb->s_export_op = &btrfs_export_ops;
872 sb->s_xattr = btrfs_xattr_handlers;
873 sb->s_time_gran = 1;
874 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
875 sb->s_flags |= MS_POSIXACL;
876 #endif
877 sb->s_flags |= MS_I_VERSION;
878 err = open_ctree(sb, fs_devices, (char *)data);
879 if (err) {
880 printk("btrfs: open_ctree failed\n");
881 return err;
882 }
883
884 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
885 key.type = BTRFS_INODE_ITEM_KEY;
886 key.offset = 0;
887 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
888 if (IS_ERR(inode)) {
889 err = PTR_ERR(inode);
890 goto fail_close;
891 }
892
893 sb->s_root = d_make_root(inode);
894 if (!sb->s_root) {
895 err = -ENOMEM;
896 goto fail_close;
897 }
898
899 save_mount_options(sb, data);
900 cleancache_init_fs(sb);
901 sb->s_flags |= MS_ACTIVE;
902 return 0;
903
904 fail_close:
905 close_ctree(fs_info->tree_root);
906 return err;
907 }
908
909 int btrfs_sync_fs(struct super_block *sb, int wait)
910 {
911 struct btrfs_trans_handle *trans;
912 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
913 struct btrfs_root *root = fs_info->tree_root;
914
915 trace_btrfs_sync_fs(wait);
916
917 if (!wait) {
918 filemap_flush(fs_info->btree_inode->i_mapping);
919 return 0;
920 }
921
922 btrfs_wait_ordered_roots(fs_info, -1);
923
924 trans = btrfs_attach_transaction_barrier(root);
925 if (IS_ERR(trans)) {
926 /* no transaction, don't bother */
927 if (PTR_ERR(trans) == -ENOENT)
928 return 0;
929 return PTR_ERR(trans);
930 }
931 return btrfs_commit_transaction(trans, root);
932 }
933
934 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
935 {
936 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
937 struct btrfs_root *root = info->tree_root;
938 char *compress_type;
939
940 if (btrfs_test_opt(root, DEGRADED))
941 seq_puts(seq, ",degraded");
942 if (btrfs_test_opt(root, NODATASUM))
943 seq_puts(seq, ",nodatasum");
944 if (btrfs_test_opt(root, NODATACOW))
945 seq_puts(seq, ",nodatacow");
946 if (btrfs_test_opt(root, NOBARRIER))
947 seq_puts(seq, ",nobarrier");
948 if (info->max_inline != 8192 * 1024)
949 seq_printf(seq, ",max_inline=%llu", info->max_inline);
950 if (info->alloc_start != 0)
951 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
952 if (info->thread_pool_size != min_t(unsigned long,
953 num_online_cpus() + 2, 8))
954 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
955 if (btrfs_test_opt(root, COMPRESS)) {
956 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
957 compress_type = "zlib";
958 else
959 compress_type = "lzo";
960 if (btrfs_test_opt(root, FORCE_COMPRESS))
961 seq_printf(seq, ",compress-force=%s", compress_type);
962 else
963 seq_printf(seq, ",compress=%s", compress_type);
964 }
965 if (btrfs_test_opt(root, NOSSD))
966 seq_puts(seq, ",nossd");
967 if (btrfs_test_opt(root, SSD_SPREAD))
968 seq_puts(seq, ",ssd_spread");
969 else if (btrfs_test_opt(root, SSD))
970 seq_puts(seq, ",ssd");
971 if (btrfs_test_opt(root, NOTREELOG))
972 seq_puts(seq, ",notreelog");
973 if (btrfs_test_opt(root, FLUSHONCOMMIT))
974 seq_puts(seq, ",flushoncommit");
975 if (btrfs_test_opt(root, DISCARD))
976 seq_puts(seq, ",discard");
977 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
978 seq_puts(seq, ",noacl");
979 if (btrfs_test_opt(root, SPACE_CACHE))
980 seq_puts(seq, ",space_cache");
981 else
982 seq_puts(seq, ",nospace_cache");
983 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
984 seq_puts(seq, ",rescan_uuid_tree");
985 if (btrfs_test_opt(root, CLEAR_CACHE))
986 seq_puts(seq, ",clear_cache");
987 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
988 seq_puts(seq, ",user_subvol_rm_allowed");
989 if (btrfs_test_opt(root, ENOSPC_DEBUG))
990 seq_puts(seq, ",enospc_debug");
991 if (btrfs_test_opt(root, AUTO_DEFRAG))
992 seq_puts(seq, ",autodefrag");
993 if (btrfs_test_opt(root, INODE_MAP_CACHE))
994 seq_puts(seq, ",inode_cache");
995 if (btrfs_test_opt(root, SKIP_BALANCE))
996 seq_puts(seq, ",skip_balance");
997 if (btrfs_test_opt(root, RECOVERY))
998 seq_puts(seq, ",recovery");
999 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1000 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1001 seq_puts(seq, ",check_int_data");
1002 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1003 seq_puts(seq, ",check_int");
1004 if (info->check_integrity_print_mask)
1005 seq_printf(seq, ",check_int_print_mask=%d",
1006 info->check_integrity_print_mask);
1007 #endif
1008 if (info->metadata_ratio)
1009 seq_printf(seq, ",metadata_ratio=%d",
1010 info->metadata_ratio);
1011 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1012 seq_puts(seq, ",fatal_errors=panic");
1013 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1014 seq_printf(seq, ",commit=%d", info->commit_interval);
1015 return 0;
1016 }
1017
1018 static int btrfs_test_super(struct super_block *s, void *data)
1019 {
1020 struct btrfs_fs_info *p = data;
1021 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1022
1023 return fs_info->fs_devices == p->fs_devices;
1024 }
1025
1026 static int btrfs_set_super(struct super_block *s, void *data)
1027 {
1028 int err = set_anon_super(s, data);
1029 if (!err)
1030 s->s_fs_info = data;
1031 return err;
1032 }
1033
1034 /*
1035 * subvolumes are identified by ino 256
1036 */
1037 static inline int is_subvolume_inode(struct inode *inode)
1038 {
1039 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1040 return 1;
1041 return 0;
1042 }
1043
1044 /*
1045 * This will strip out the subvol=%s argument for an argument string and add
1046 * subvolid=0 to make sure we get the actual tree root for path walking to the
1047 * subvol we want.
1048 */
1049 static char *setup_root_args(char *args)
1050 {
1051 unsigned len = strlen(args) + 2 + 1;
1052 char *src, *dst, *buf;
1053
1054 /*
1055 * We need the same args as before, but with this substitution:
1056 * s!subvol=[^,]+!subvolid=0!
1057 *
1058 * Since the replacement string is up to 2 bytes longer than the
1059 * original, allocate strlen(args) + 2 + 1 bytes.
1060 */
1061
1062 src = strstr(args, "subvol=");
1063 /* This shouldn't happen, but just in case.. */
1064 if (!src)
1065 return NULL;
1066
1067 buf = dst = kmalloc(len, GFP_NOFS);
1068 if (!buf)
1069 return NULL;
1070
1071 /*
1072 * If the subvol= arg is not at the start of the string,
1073 * copy whatever precedes it into buf.
1074 */
1075 if (src != args) {
1076 *src++ = '\0';
1077 strcpy(buf, args);
1078 dst += strlen(args);
1079 }
1080
1081 strcpy(dst, "subvolid=0");
1082 dst += strlen("subvolid=0");
1083
1084 /*
1085 * If there is a "," after the original subvol=... string,
1086 * copy that suffix into our buffer. Otherwise, we're done.
1087 */
1088 src = strchr(src, ',');
1089 if (src)
1090 strcpy(dst, src);
1091
1092 return buf;
1093 }
1094
1095 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1096 const char *device_name, char *data)
1097 {
1098 struct dentry *root;
1099 struct vfsmount *mnt;
1100 char *newargs;
1101
1102 newargs = setup_root_args(data);
1103 if (!newargs)
1104 return ERR_PTR(-ENOMEM);
1105 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1106 newargs);
1107 kfree(newargs);
1108 if (IS_ERR(mnt))
1109 return ERR_CAST(mnt);
1110
1111 root = mount_subtree(mnt, subvol_name);
1112
1113 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1114 struct super_block *s = root->d_sb;
1115 dput(root);
1116 root = ERR_PTR(-EINVAL);
1117 deactivate_locked_super(s);
1118 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1119 subvol_name);
1120 }
1121
1122 return root;
1123 }
1124
1125 /*
1126 * Find a superblock for the given device / mount point.
1127 *
1128 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1129 * for multiple device setup. Make sure to keep it in sync.
1130 */
1131 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1132 const char *device_name, void *data)
1133 {
1134 struct block_device *bdev = NULL;
1135 struct super_block *s;
1136 struct dentry *root;
1137 struct btrfs_fs_devices *fs_devices = NULL;
1138 struct btrfs_fs_info *fs_info = NULL;
1139 fmode_t mode = FMODE_READ;
1140 char *subvol_name = NULL;
1141 u64 subvol_objectid = 0;
1142 int error = 0;
1143
1144 if (!(flags & MS_RDONLY))
1145 mode |= FMODE_WRITE;
1146
1147 error = btrfs_parse_early_options(data, mode, fs_type,
1148 &subvol_name, &subvol_objectid,
1149 &fs_devices);
1150 if (error) {
1151 kfree(subvol_name);
1152 return ERR_PTR(error);
1153 }
1154
1155 if (subvol_name) {
1156 root = mount_subvol(subvol_name, flags, device_name, data);
1157 kfree(subvol_name);
1158 return root;
1159 }
1160
1161 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1162 if (error)
1163 return ERR_PTR(error);
1164
1165 /*
1166 * Setup a dummy root and fs_info for test/set super. This is because
1167 * we don't actually fill this stuff out until open_ctree, but we need
1168 * it for searching for existing supers, so this lets us do that and
1169 * then open_ctree will properly initialize everything later.
1170 */
1171 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1172 if (!fs_info)
1173 return ERR_PTR(-ENOMEM);
1174
1175 fs_info->fs_devices = fs_devices;
1176
1177 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1178 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1179 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1180 error = -ENOMEM;
1181 goto error_fs_info;
1182 }
1183
1184 error = btrfs_open_devices(fs_devices, mode, fs_type);
1185 if (error)
1186 goto error_fs_info;
1187
1188 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1189 error = -EACCES;
1190 goto error_close_devices;
1191 }
1192
1193 bdev = fs_devices->latest_bdev;
1194 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1195 fs_info);
1196 if (IS_ERR(s)) {
1197 error = PTR_ERR(s);
1198 goto error_close_devices;
1199 }
1200
1201 if (s->s_root) {
1202 btrfs_close_devices(fs_devices);
1203 free_fs_info(fs_info);
1204 if ((flags ^ s->s_flags) & MS_RDONLY)
1205 error = -EBUSY;
1206 } else {
1207 char b[BDEVNAME_SIZE];
1208
1209 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1210 btrfs_sb(s)->bdev_holder = fs_type;
1211 error = btrfs_fill_super(s, fs_devices, data,
1212 flags & MS_SILENT ? 1 : 0);
1213 }
1214
1215 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1216 if (IS_ERR(root))
1217 deactivate_locked_super(s);
1218
1219 return root;
1220
1221 error_close_devices:
1222 btrfs_close_devices(fs_devices);
1223 error_fs_info:
1224 free_fs_info(fs_info);
1225 return ERR_PTR(error);
1226 }
1227
1228 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1229 {
1230 spin_lock_irq(&workers->lock);
1231 workers->max_workers = new_limit;
1232 spin_unlock_irq(&workers->lock);
1233 }
1234
1235 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1236 int new_pool_size, int old_pool_size)
1237 {
1238 if (new_pool_size == old_pool_size)
1239 return;
1240
1241 fs_info->thread_pool_size = new_pool_size;
1242
1243 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1244 old_pool_size, new_pool_size);
1245
1246 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1247 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1248 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1249 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1250 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1251 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1252 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1253 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1254 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1255 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1256 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1257 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1258 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1259 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1260 new_pool_size);
1261 }
1262
1263 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1264 {
1265 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1266 }
1267
1268 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1269 unsigned long old_opts, int flags)
1270 {
1271 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1272 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1273 (flags & MS_RDONLY))) {
1274 /* wait for any defraggers to finish */
1275 wait_event(fs_info->transaction_wait,
1276 (atomic_read(&fs_info->defrag_running) == 0));
1277 if (flags & MS_RDONLY)
1278 sync_filesystem(fs_info->sb);
1279 }
1280 }
1281
1282 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1283 unsigned long old_opts)
1284 {
1285 /*
1286 * We need cleanup all defragable inodes if the autodefragment is
1287 * close or the fs is R/O.
1288 */
1289 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1290 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1291 (fs_info->sb->s_flags & MS_RDONLY))) {
1292 btrfs_cleanup_defrag_inodes(fs_info);
1293 }
1294
1295 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1296 }
1297
1298 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1299 {
1300 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1301 struct btrfs_root *root = fs_info->tree_root;
1302 unsigned old_flags = sb->s_flags;
1303 unsigned long old_opts = fs_info->mount_opt;
1304 unsigned long old_compress_type = fs_info->compress_type;
1305 u64 old_max_inline = fs_info->max_inline;
1306 u64 old_alloc_start = fs_info->alloc_start;
1307 int old_thread_pool_size = fs_info->thread_pool_size;
1308 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1309 int ret;
1310
1311 btrfs_remount_prepare(fs_info);
1312
1313 ret = btrfs_parse_options(root, data);
1314 if (ret) {
1315 ret = -EINVAL;
1316 goto restore;
1317 }
1318
1319 btrfs_remount_begin(fs_info, old_opts, *flags);
1320 btrfs_resize_thread_pool(fs_info,
1321 fs_info->thread_pool_size, old_thread_pool_size);
1322
1323 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1324 goto out;
1325
1326 if (*flags & MS_RDONLY) {
1327 /*
1328 * this also happens on 'umount -rf' or on shutdown, when
1329 * the filesystem is busy.
1330 */
1331
1332 /* wait for the uuid_scan task to finish */
1333 down(&fs_info->uuid_tree_rescan_sem);
1334 /* avoid complains from lockdep et al. */
1335 up(&fs_info->uuid_tree_rescan_sem);
1336
1337 sb->s_flags |= MS_RDONLY;
1338
1339 btrfs_dev_replace_suspend_for_unmount(fs_info);
1340 btrfs_scrub_cancel(fs_info);
1341 btrfs_pause_balance(fs_info);
1342
1343 ret = btrfs_commit_super(root);
1344 if (ret)
1345 goto restore;
1346 } else {
1347 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1348 btrfs_err(fs_info,
1349 "Remounting read-write after error is not allowed\n");
1350 ret = -EINVAL;
1351 goto restore;
1352 }
1353 if (fs_info->fs_devices->rw_devices == 0) {
1354 ret = -EACCES;
1355 goto restore;
1356 }
1357
1358 if (fs_info->fs_devices->missing_devices >
1359 fs_info->num_tolerated_disk_barrier_failures &&
1360 !(*flags & MS_RDONLY)) {
1361 printk(KERN_WARNING
1362 "Btrfs: too many missing devices, writeable remount is not allowed\n");
1363 ret = -EACCES;
1364 goto restore;
1365 }
1366
1367 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1368 ret = -EINVAL;
1369 goto restore;
1370 }
1371
1372 ret = btrfs_cleanup_fs_roots(fs_info);
1373 if (ret)
1374 goto restore;
1375
1376 /* recover relocation */
1377 ret = btrfs_recover_relocation(root);
1378 if (ret)
1379 goto restore;
1380
1381 ret = btrfs_resume_balance_async(fs_info);
1382 if (ret)
1383 goto restore;
1384
1385 ret = btrfs_resume_dev_replace_async(fs_info);
1386 if (ret) {
1387 pr_warn("btrfs: failed to resume dev_replace\n");
1388 goto restore;
1389 }
1390
1391 if (!fs_info->uuid_root) {
1392 pr_info("btrfs: creating UUID tree\n");
1393 ret = btrfs_create_uuid_tree(fs_info);
1394 if (ret) {
1395 pr_warn("btrfs: failed to create the uuid tree"
1396 "%d\n", ret);
1397 goto restore;
1398 }
1399 }
1400 sb->s_flags &= ~MS_RDONLY;
1401 }
1402 out:
1403 btrfs_remount_cleanup(fs_info, old_opts);
1404 return 0;
1405
1406 restore:
1407 /* We've hit an error - don't reset MS_RDONLY */
1408 if (sb->s_flags & MS_RDONLY)
1409 old_flags |= MS_RDONLY;
1410 sb->s_flags = old_flags;
1411 fs_info->mount_opt = old_opts;
1412 fs_info->compress_type = old_compress_type;
1413 fs_info->max_inline = old_max_inline;
1414 mutex_lock(&fs_info->chunk_mutex);
1415 fs_info->alloc_start = old_alloc_start;
1416 mutex_unlock(&fs_info->chunk_mutex);
1417 btrfs_resize_thread_pool(fs_info,
1418 old_thread_pool_size, fs_info->thread_pool_size);
1419 fs_info->metadata_ratio = old_metadata_ratio;
1420 btrfs_remount_cleanup(fs_info, old_opts);
1421 return ret;
1422 }
1423
1424 /* Used to sort the devices by max_avail(descending sort) */
1425 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1426 const void *dev_info2)
1427 {
1428 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1429 ((struct btrfs_device_info *)dev_info2)->max_avail)
1430 return -1;
1431 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1432 ((struct btrfs_device_info *)dev_info2)->max_avail)
1433 return 1;
1434 else
1435 return 0;
1436 }
1437
1438 /*
1439 * sort the devices by max_avail, in which max free extent size of each device
1440 * is stored.(Descending Sort)
1441 */
1442 static inline void btrfs_descending_sort_devices(
1443 struct btrfs_device_info *devices,
1444 size_t nr_devices)
1445 {
1446 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1447 btrfs_cmp_device_free_bytes, NULL);
1448 }
1449
1450 /*
1451 * The helper to calc the free space on the devices that can be used to store
1452 * file data.
1453 */
1454 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1455 {
1456 struct btrfs_fs_info *fs_info = root->fs_info;
1457 struct btrfs_device_info *devices_info;
1458 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1459 struct btrfs_device *device;
1460 u64 skip_space;
1461 u64 type;
1462 u64 avail_space;
1463 u64 used_space;
1464 u64 min_stripe_size;
1465 int min_stripes = 1, num_stripes = 1;
1466 int i = 0, nr_devices;
1467 int ret;
1468
1469 nr_devices = fs_info->fs_devices->open_devices;
1470 BUG_ON(!nr_devices);
1471
1472 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1473 GFP_NOFS);
1474 if (!devices_info)
1475 return -ENOMEM;
1476
1477 /* calc min stripe number for data space alloction */
1478 type = btrfs_get_alloc_profile(root, 1);
1479 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1480 min_stripes = 2;
1481 num_stripes = nr_devices;
1482 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1483 min_stripes = 2;
1484 num_stripes = 2;
1485 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1486 min_stripes = 4;
1487 num_stripes = 4;
1488 }
1489
1490 if (type & BTRFS_BLOCK_GROUP_DUP)
1491 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1492 else
1493 min_stripe_size = BTRFS_STRIPE_LEN;
1494
1495 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1496 if (!device->in_fs_metadata || !device->bdev ||
1497 device->is_tgtdev_for_dev_replace)
1498 continue;
1499
1500 avail_space = device->total_bytes - device->bytes_used;
1501
1502 /* align with stripe_len */
1503 do_div(avail_space, BTRFS_STRIPE_LEN);
1504 avail_space *= BTRFS_STRIPE_LEN;
1505
1506 /*
1507 * In order to avoid overwritting the superblock on the drive,
1508 * btrfs starts at an offset of at least 1MB when doing chunk
1509 * allocation.
1510 */
1511 skip_space = 1024 * 1024;
1512
1513 /* user can set the offset in fs_info->alloc_start. */
1514 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1515 device->total_bytes)
1516 skip_space = max(fs_info->alloc_start, skip_space);
1517
1518 /*
1519 * btrfs can not use the free space in [0, skip_space - 1],
1520 * we must subtract it from the total. In order to implement
1521 * it, we account the used space in this range first.
1522 */
1523 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1524 &used_space);
1525 if (ret) {
1526 kfree(devices_info);
1527 return ret;
1528 }
1529
1530 /* calc the free space in [0, skip_space - 1] */
1531 skip_space -= used_space;
1532
1533 /*
1534 * we can use the free space in [0, skip_space - 1], subtract
1535 * it from the total.
1536 */
1537 if (avail_space && avail_space >= skip_space)
1538 avail_space -= skip_space;
1539 else
1540 avail_space = 0;
1541
1542 if (avail_space < min_stripe_size)
1543 continue;
1544
1545 devices_info[i].dev = device;
1546 devices_info[i].max_avail = avail_space;
1547
1548 i++;
1549 }
1550
1551 nr_devices = i;
1552
1553 btrfs_descending_sort_devices(devices_info, nr_devices);
1554
1555 i = nr_devices - 1;
1556 avail_space = 0;
1557 while (nr_devices >= min_stripes) {
1558 if (num_stripes > nr_devices)
1559 num_stripes = nr_devices;
1560
1561 if (devices_info[i].max_avail >= min_stripe_size) {
1562 int j;
1563 u64 alloc_size;
1564
1565 avail_space += devices_info[i].max_avail * num_stripes;
1566 alloc_size = devices_info[i].max_avail;
1567 for (j = i + 1 - num_stripes; j <= i; j++)
1568 devices_info[j].max_avail -= alloc_size;
1569 }
1570 i--;
1571 nr_devices--;
1572 }
1573
1574 kfree(devices_info);
1575 *free_bytes = avail_space;
1576 return 0;
1577 }
1578
1579 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1580 {
1581 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1582 struct btrfs_super_block *disk_super = fs_info->super_copy;
1583 struct list_head *head = &fs_info->space_info;
1584 struct btrfs_space_info *found;
1585 u64 total_used = 0;
1586 u64 total_free_data = 0;
1587 int bits = dentry->d_sb->s_blocksize_bits;
1588 __be32 *fsid = (__be32 *)fs_info->fsid;
1589 int ret;
1590
1591 /* holding chunk_muext to avoid allocating new chunks */
1592 mutex_lock(&fs_info->chunk_mutex);
1593 rcu_read_lock();
1594 list_for_each_entry_rcu(found, head, list) {
1595 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1596 total_free_data += found->disk_total - found->disk_used;
1597 total_free_data -=
1598 btrfs_account_ro_block_groups_free_space(found);
1599 }
1600
1601 total_used += found->disk_used;
1602 }
1603 rcu_read_unlock();
1604
1605 buf->f_namelen = BTRFS_NAME_LEN;
1606 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1607 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1608 buf->f_bsize = dentry->d_sb->s_blocksize;
1609 buf->f_type = BTRFS_SUPER_MAGIC;
1610 buf->f_bavail = total_free_data;
1611 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1612 if (ret) {
1613 mutex_unlock(&fs_info->chunk_mutex);
1614 return ret;
1615 }
1616 buf->f_bavail += total_free_data;
1617 buf->f_bavail = buf->f_bavail >> bits;
1618 mutex_unlock(&fs_info->chunk_mutex);
1619
1620 /* We treat it as constant endianness (it doesn't matter _which_)
1621 because we want the fsid to come out the same whether mounted
1622 on a big-endian or little-endian host */
1623 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1624 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1625 /* Mask in the root object ID too, to disambiguate subvols */
1626 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1627 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1628
1629 return 0;
1630 }
1631
1632 static void btrfs_kill_super(struct super_block *sb)
1633 {
1634 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1635 kill_anon_super(sb);
1636 free_fs_info(fs_info);
1637 }
1638
1639 static struct file_system_type btrfs_fs_type = {
1640 .owner = THIS_MODULE,
1641 .name = "btrfs",
1642 .mount = btrfs_mount,
1643 .kill_sb = btrfs_kill_super,
1644 .fs_flags = FS_REQUIRES_DEV,
1645 };
1646 MODULE_ALIAS_FS("btrfs");
1647
1648 /*
1649 * used by btrfsctl to scan devices when no FS is mounted
1650 */
1651 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1652 unsigned long arg)
1653 {
1654 struct btrfs_ioctl_vol_args *vol;
1655 struct btrfs_fs_devices *fs_devices;
1656 int ret = -ENOTTY;
1657
1658 if (!capable(CAP_SYS_ADMIN))
1659 return -EPERM;
1660
1661 vol = memdup_user((void __user *)arg, sizeof(*vol));
1662 if (IS_ERR(vol))
1663 return PTR_ERR(vol);
1664
1665 switch (cmd) {
1666 case BTRFS_IOC_SCAN_DEV:
1667 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1668 &btrfs_fs_type, &fs_devices);
1669 break;
1670 case BTRFS_IOC_DEVICES_READY:
1671 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1672 &btrfs_fs_type, &fs_devices);
1673 if (ret)
1674 break;
1675 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1676 break;
1677 }
1678
1679 kfree(vol);
1680 return ret;
1681 }
1682
1683 static int btrfs_freeze(struct super_block *sb)
1684 {
1685 struct btrfs_trans_handle *trans;
1686 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1687
1688 trans = btrfs_attach_transaction_barrier(root);
1689 if (IS_ERR(trans)) {
1690 /* no transaction, don't bother */
1691 if (PTR_ERR(trans) == -ENOENT)
1692 return 0;
1693 return PTR_ERR(trans);
1694 }
1695 return btrfs_commit_transaction(trans, root);
1696 }
1697
1698 static int btrfs_unfreeze(struct super_block *sb)
1699 {
1700 return 0;
1701 }
1702
1703 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1704 {
1705 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1706 struct btrfs_fs_devices *cur_devices;
1707 struct btrfs_device *dev, *first_dev = NULL;
1708 struct list_head *head;
1709 struct rcu_string *name;
1710
1711 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1712 cur_devices = fs_info->fs_devices;
1713 while (cur_devices) {
1714 head = &cur_devices->devices;
1715 list_for_each_entry(dev, head, dev_list) {
1716 if (dev->missing)
1717 continue;
1718 if (!first_dev || dev->devid < first_dev->devid)
1719 first_dev = dev;
1720 }
1721 cur_devices = cur_devices->seed;
1722 }
1723
1724 if (first_dev) {
1725 rcu_read_lock();
1726 name = rcu_dereference(first_dev->name);
1727 seq_escape(m, name->str, " \t\n\\");
1728 rcu_read_unlock();
1729 } else {
1730 WARN_ON(1);
1731 }
1732 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1733 return 0;
1734 }
1735
1736 static const struct super_operations btrfs_super_ops = {
1737 .drop_inode = btrfs_drop_inode,
1738 .evict_inode = btrfs_evict_inode,
1739 .put_super = btrfs_put_super,
1740 .sync_fs = btrfs_sync_fs,
1741 .show_options = btrfs_show_options,
1742 .show_devname = btrfs_show_devname,
1743 .write_inode = btrfs_write_inode,
1744 .alloc_inode = btrfs_alloc_inode,
1745 .destroy_inode = btrfs_destroy_inode,
1746 .statfs = btrfs_statfs,
1747 .remount_fs = btrfs_remount,
1748 .freeze_fs = btrfs_freeze,
1749 .unfreeze_fs = btrfs_unfreeze,
1750 };
1751
1752 static const struct file_operations btrfs_ctl_fops = {
1753 .unlocked_ioctl = btrfs_control_ioctl,
1754 .compat_ioctl = btrfs_control_ioctl,
1755 .owner = THIS_MODULE,
1756 .llseek = noop_llseek,
1757 };
1758
1759 static struct miscdevice btrfs_misc = {
1760 .minor = BTRFS_MINOR,
1761 .name = "btrfs-control",
1762 .fops = &btrfs_ctl_fops
1763 };
1764
1765 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1766 MODULE_ALIAS("devname:btrfs-control");
1767
1768 static int btrfs_interface_init(void)
1769 {
1770 return misc_register(&btrfs_misc);
1771 }
1772
1773 static void btrfs_interface_exit(void)
1774 {
1775 if (misc_deregister(&btrfs_misc) < 0)
1776 printk(KERN_INFO "btrfs: misc_deregister failed for control device\n");
1777 }
1778
1779 static void btrfs_print_info(void)
1780 {
1781 printk(KERN_INFO "Btrfs loaded"
1782 #ifdef CONFIG_BTRFS_DEBUG
1783 ", debug=on"
1784 #endif
1785 #ifdef CONFIG_BTRFS_ASSERT
1786 ", assert=on"
1787 #endif
1788 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1789 ", integrity-checker=on"
1790 #endif
1791 "\n");
1792 }
1793
1794 static int btrfs_run_sanity_tests(void)
1795 {
1796 int ret;
1797
1798 ret = btrfs_init_test_fs();
1799 if (ret)
1800 return ret;
1801
1802 ret = btrfs_test_free_space_cache();
1803 if (ret)
1804 goto out;
1805 ret = btrfs_test_extent_buffer_operations();
1806 if (ret)
1807 goto out;
1808 ret = btrfs_test_extent_io();
1809 if (ret)
1810 goto out;
1811 ret = btrfs_test_inodes();
1812 out:
1813 btrfs_destroy_test_fs();
1814 return ret;
1815 }
1816
1817 static int __init init_btrfs_fs(void)
1818 {
1819 int err;
1820
1821 err = btrfs_init_sysfs();
1822 if (err)
1823 return err;
1824
1825 btrfs_init_compress();
1826
1827 err = btrfs_init_cachep();
1828 if (err)
1829 goto free_compress;
1830
1831 err = extent_io_init();
1832 if (err)
1833 goto free_cachep;
1834
1835 err = extent_map_init();
1836 if (err)
1837 goto free_extent_io;
1838
1839 err = ordered_data_init();
1840 if (err)
1841 goto free_extent_map;
1842
1843 err = btrfs_delayed_inode_init();
1844 if (err)
1845 goto free_ordered_data;
1846
1847 err = btrfs_auto_defrag_init();
1848 if (err)
1849 goto free_delayed_inode;
1850
1851 err = btrfs_delayed_ref_init();
1852 if (err)
1853 goto free_auto_defrag;
1854
1855 err = btrfs_prelim_ref_init();
1856 if (err)
1857 goto free_prelim_ref;
1858
1859 err = btrfs_interface_init();
1860 if (err)
1861 goto free_delayed_ref;
1862
1863 btrfs_init_lockdep();
1864
1865 btrfs_print_info();
1866
1867 err = btrfs_run_sanity_tests();
1868 if (err)
1869 goto unregister_ioctl;
1870
1871 err = register_filesystem(&btrfs_fs_type);
1872 if (err)
1873 goto unregister_ioctl;
1874
1875 return 0;
1876
1877 unregister_ioctl:
1878 btrfs_interface_exit();
1879 free_prelim_ref:
1880 btrfs_prelim_ref_exit();
1881 free_delayed_ref:
1882 btrfs_delayed_ref_exit();
1883 free_auto_defrag:
1884 btrfs_auto_defrag_exit();
1885 free_delayed_inode:
1886 btrfs_delayed_inode_exit();
1887 free_ordered_data:
1888 ordered_data_exit();
1889 free_extent_map:
1890 extent_map_exit();
1891 free_extent_io:
1892 extent_io_exit();
1893 free_cachep:
1894 btrfs_destroy_cachep();
1895 free_compress:
1896 btrfs_exit_compress();
1897 btrfs_exit_sysfs();
1898 return err;
1899 }
1900
1901 static void __exit exit_btrfs_fs(void)
1902 {
1903 btrfs_destroy_cachep();
1904 btrfs_delayed_ref_exit();
1905 btrfs_auto_defrag_exit();
1906 btrfs_delayed_inode_exit();
1907 btrfs_prelim_ref_exit();
1908 ordered_data_exit();
1909 extent_map_exit();
1910 extent_io_exit();
1911 btrfs_interface_exit();
1912 unregister_filesystem(&btrfs_fs_type);
1913 btrfs_exit_sysfs();
1914 btrfs_cleanup_fs_uuids();
1915 btrfs_exit_compress();
1916 }
1917
1918 module_init(init_btrfs_fs)
1919 module_exit(exit_btrfs_fs)
1920
1921 MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.256571 seconds and 6 git commands to generate.