projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'spi/fix/core' into spi-linus
[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 info->compress_type = BTRFS_COMPRESS_NONE;
436 btrfs_clear_opt(info->mount_opt, COMPRESS);
437 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
438 btrfs_set_opt(info->mount_opt, NODATACOW);
439 btrfs_set_opt(info->mount_opt, NODATASUM);
440 break;
441 case Opt_compress_force:
442 case Opt_compress_force_type:
443 compress_force = true;
444 /* Fallthrough */
445 case Opt_compress:
446 case Opt_compress_type:
447 if (token == Opt_compress ||
448 token == Opt_compress_force ||
449 strcmp(args[0].from, "zlib") == 0) {
450 compress_type = "zlib";
451 info->compress_type = BTRFS_COMPRESS_ZLIB;
452 btrfs_set_opt(info->mount_opt, COMPRESS);
453 btrfs_clear_opt(info->mount_opt, NODATACOW);
454 btrfs_clear_opt(info->mount_opt, NODATASUM);
455 } else if (strcmp(args[0].from, "lzo") == 0) {
456 compress_type = "lzo";
457 info->compress_type = BTRFS_COMPRESS_LZO;
458 btrfs_set_opt(info->mount_opt, COMPRESS);
459 btrfs_clear_opt(info->mount_opt, NODATACOW);
460 btrfs_clear_opt(info->mount_opt, NODATASUM);
461 btrfs_set_fs_incompat(info, COMPRESS_LZO);
462 } else if (strncmp(args[0].from, "no", 2) == 0) {
463 compress_type = "no";
464 info->compress_type = BTRFS_COMPRESS_NONE;
465 btrfs_clear_opt(info->mount_opt, COMPRESS);
466 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
467 compress_force = false;
468 } else {
469 ret = -EINVAL;
470 goto out;
471 }
472
473 if (compress_force) {
474 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
475 pr_info("btrfs: force %s compression\n",
476 compress_type);
477 } else
478 pr_info("btrfs: use %s compression\n",
479 compress_type);
480 break;
481 case Opt_ssd:
482 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
483 btrfs_set_opt(info->mount_opt, SSD);
484 break;
485 case Opt_ssd_spread:
486 printk(KERN_INFO "btrfs: use spread ssd "
487 "allocation scheme\n");
488 btrfs_set_opt(info->mount_opt, SSD);
489 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
490 break;
491 case Opt_nossd:
492 printk(KERN_INFO "btrfs: not using ssd allocation "
493 "scheme\n");
494 btrfs_set_opt(info->mount_opt, NOSSD);
495 btrfs_clear_opt(info->mount_opt, SSD);
496 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
497 break;
498 case Opt_nobarrier:
499 printk(KERN_INFO "btrfs: turning off barriers\n");
500 btrfs_set_opt(info->mount_opt, NOBARRIER);
501 break;
502 case Opt_thread_pool:
503 ret = match_int(&args[0], &intarg);
504 if (ret) {
505 goto out;
506 } else if (intarg > 0) {
507 info->thread_pool_size = intarg;
508 } else {
509 ret = -EINVAL;
510 goto out;
511 }
512 break;
513 case Opt_max_inline:
514 num = match_strdup(&args[0]);
515 if (num) {
516 info->max_inline = memparse(num, NULL);
517 kfree(num);
518
519 if (info->max_inline) {
520 info->max_inline = max_t(u64,
521 info->max_inline,
522 root->sectorsize);
523 }
524 printk(KERN_INFO "btrfs: max_inline at %llu\n",
525 info->max_inline);
526 } else {
527 ret = -ENOMEM;
528 goto out;
529 }
530 break;
531 case Opt_alloc_start:
532 num = match_strdup(&args[0]);
533 if (num) {
534 mutex_lock(&info->chunk_mutex);
535 info->alloc_start = memparse(num, NULL);
536 mutex_unlock(&info->chunk_mutex);
537 kfree(num);
538 printk(KERN_INFO
539 "btrfs: allocations start at %llu\n",
540 info->alloc_start);
541 } else {
542 ret = -ENOMEM;
543 goto out;
544 }
545 break;
546 case Opt_noacl:
547 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
548 break;
549 case Opt_notreelog:
550 printk(KERN_INFO "btrfs: disabling tree log\n");
551 btrfs_set_opt(info->mount_opt, NOTREELOG);
552 break;
553 case Opt_flushoncommit:
554 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
555 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
556 break;
557 case Opt_ratio:
558 ret = match_int(&args[0], &intarg);
559 if (ret) {
560 goto out;
561 } else if (intarg >= 0) {
562 info->metadata_ratio = intarg;
563 printk(KERN_INFO "btrfs: metadata ratio %d\n",
564 info->metadata_ratio);
565 } else {
566 ret = -EINVAL;
567 goto out;
568 }
569 break;
570 case Opt_discard:
571 btrfs_set_opt(info->mount_opt, DISCARD);
572 break;
573 case Opt_space_cache:
574 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
575 break;
576 case Opt_rescan_uuid_tree:
577 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
578 break;
579 case Opt_no_space_cache:
580 printk(KERN_INFO "btrfs: disabling disk space caching\n");
581 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
582 break;
583 case Opt_inode_cache:
584 printk(KERN_INFO "btrfs: enabling inode map caching\n");
585 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
586 break;
587 case Opt_clear_cache:
588 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
589 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
590 break;
591 case Opt_user_subvol_rm_allowed:
592 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
593 break;
594 case Opt_enospc_debug:
595 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
596 break;
597 case Opt_defrag:
598 printk(KERN_INFO "btrfs: enabling auto defrag\n");
599 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
600 break;
601 case Opt_recovery:
602 printk(KERN_INFO "btrfs: enabling auto recovery\n");
603 btrfs_set_opt(info->mount_opt, RECOVERY);
604 break;
605 case Opt_skip_balance:
606 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
607 break;
608 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
609 case Opt_check_integrity_including_extent_data:
610 printk(KERN_INFO "btrfs: enabling check integrity"
611 " including extent data\n");
612 btrfs_set_opt(info->mount_opt,
613 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
614 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
615 break;
616 case Opt_check_integrity:
617 printk(KERN_INFO "btrfs: enabling check integrity\n");
618 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
619 break;
620 case Opt_check_integrity_print_mask:
621 ret = match_int(&args[0], &intarg);
622 if (ret) {
623 goto out;
624 } else if (intarg >= 0) {
625 info->check_integrity_print_mask = intarg;
626 printk(KERN_INFO "btrfs:"
627 " check_integrity_print_mask 0x%x\n",
628 info->check_integrity_print_mask);
629 } else {
630 ret = -EINVAL;
631 goto out;
632 }
633 break;
634 #else
635 case Opt_check_integrity_including_extent_data:
636 case Opt_check_integrity:
637 case Opt_check_integrity_print_mask:
638 printk(KERN_ERR "btrfs: support for check_integrity*"
639 " not compiled in!\n");
640 ret = -EINVAL;
641 goto out;
642 #endif
643 case Opt_fatal_errors:
644 if (strcmp(args[0].from, "panic") == 0)
645 btrfs_set_opt(info->mount_opt,
646 PANIC_ON_FATAL_ERROR);
647 else if (strcmp(args[0].from, "bug") == 0)
648 btrfs_clear_opt(info->mount_opt,
649 PANIC_ON_FATAL_ERROR);
650 else {
651 ret = -EINVAL;
652 goto out;
653 }
654 break;
655 case Opt_commit_interval:
656 intarg = 0;
657 ret = match_int(&args[0], &intarg);
658 if (ret < 0) {
659 printk(KERN_ERR
660 "btrfs: invalid commit interval\n");
661 ret = -EINVAL;
662 goto out;
663 }
664 if (intarg > 0) {
665 if (intarg > 300) {
666 printk(KERN_WARNING
667 "btrfs: excessive commit interval %d\n",
668 intarg);
669 }
670 info->commit_interval = intarg;
671 } else {
672 printk(KERN_INFO
673 "btrfs: using default commit interval %ds\n",
674 BTRFS_DEFAULT_COMMIT_INTERVAL);
675 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
676 }
677 break;
678 case Opt_err:
679 printk(KERN_INFO "btrfs: unrecognized mount option "
680 "'%s'\n", p);
681 ret = -EINVAL;
682 goto out;
683 default:
684 break;
685 }
686 }
687 out:
688 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
689 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
690 kfree(orig);
691 return ret;
692 }
693
694 /*
695 * Parse mount options that are required early in the mount process.
696 *
697 * All other options will be parsed on much later in the mount process and
698 * only when we need to allocate a new super block.
699 */
700 static int btrfs_parse_early_options(const char *options, fmode_t flags,
701 void *holder, char **subvol_name, u64 *subvol_objectid,
702 struct btrfs_fs_devices **fs_devices)
703 {
704 substring_t args[MAX_OPT_ARGS];
705 char *device_name, *opts, *orig, *p;
706 char *num = NULL;
707 int error = 0;
708
709 if (!options)
710 return 0;
711
712 /*
713 * strsep changes the string, duplicate it because parse_options
714 * gets called twice
715 */
716 opts = kstrdup(options, GFP_KERNEL);
717 if (!opts)
718 return -ENOMEM;
719 orig = opts;
720
721 while ((p = strsep(&opts, ",")) != NULL) {
722 int token;
723 if (!*p)
724 continue;
725
726 token = match_token(p, tokens, args);
727 switch (token) {
728 case Opt_subvol:
729 kfree(*subvol_name);
730 *subvol_name = match_strdup(&args[0]);
731 if (!*subvol_name) {
732 error = -ENOMEM;
733 goto out;
734 }
735 break;
736 case Opt_subvolid:
737 num = match_strdup(&args[0]);
738 if (num) {
739 *subvol_objectid = memparse(num, NULL);
740 kfree(num);
741 /* we want the original fs_tree */
742 if (!*subvol_objectid)
743 *subvol_objectid =
744 BTRFS_FS_TREE_OBJECTID;
745 } else {
746 error = -EINVAL;
747 goto out;
748 }
749 break;
750 case Opt_subvolrootid:
751 printk(KERN_WARNING
752 "btrfs: 'subvolrootid' mount option is deprecated and has no effect\n");
753 break;
754 case Opt_device:
755 device_name = match_strdup(&args[0]);
756 if (!device_name) {
757 error = -ENOMEM;
758 goto out;
759 }
760 error = btrfs_scan_one_device(device_name,
761 flags, holder, fs_devices);
762 kfree(device_name);
763 if (error)
764 goto out;
765 break;
766 default:
767 break;
768 }
769 }
770
771 out:
772 kfree(orig);
773 return error;
774 }
775
776 static struct dentry *get_default_root(struct super_block *sb,
777 u64 subvol_objectid)
778 {
779 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
780 struct btrfs_root *root = fs_info->tree_root;
781 struct btrfs_root *new_root;
782 struct btrfs_dir_item *di;
783 struct btrfs_path *path;
784 struct btrfs_key location;
785 struct inode *inode;
786 u64 dir_id;
787 int new = 0;
788
789 /*
790 * We have a specific subvol we want to mount, just setup location and
791 * go look up the root.
792 */
793 if (subvol_objectid) {
794 location.objectid = subvol_objectid;
795 location.type = BTRFS_ROOT_ITEM_KEY;
796 location.offset = (u64)-1;
797 goto find_root;
798 }
799
800 path = btrfs_alloc_path();
801 if (!path)
802 return ERR_PTR(-ENOMEM);
803 path->leave_spinning = 1;
804
805 /*
806 * Find the "default" dir item which points to the root item that we
807 * will mount by default if we haven't been given a specific subvolume
808 * to mount.
809 */
810 dir_id = btrfs_super_root_dir(fs_info->super_copy);
811 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
812 if (IS_ERR(di)) {
813 btrfs_free_path(path);
814 return ERR_CAST(di);
815 }
816 if (!di) {
817 /*
818 * Ok the default dir item isn't there. This is weird since
819 * it's always been there, but don't freak out, just try and
820 * mount to root most subvolume.
821 */
822 btrfs_free_path(path);
823 dir_id = BTRFS_FIRST_FREE_OBJECTID;
824 new_root = fs_info->fs_root;
825 goto setup_root;
826 }
827
828 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
829 btrfs_free_path(path);
830
831 find_root:
832 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
833 if (IS_ERR(new_root))
834 return ERR_CAST(new_root);
835
836 dir_id = btrfs_root_dirid(&new_root->root_item);
837 setup_root:
838 location.objectid = dir_id;
839 location.type = BTRFS_INODE_ITEM_KEY;
840 location.offset = 0;
841
842 inode = btrfs_iget(sb, &location, new_root, &new);
843 if (IS_ERR(inode))
844 return ERR_CAST(inode);
845
846 /*
847 * If we're just mounting the root most subvol put the inode and return
848 * a reference to the dentry. We will have already gotten a reference
849 * to the inode in btrfs_fill_super so we're good to go.
850 */
851 if (!new && sb->s_root->d_inode == inode) {
852 iput(inode);
853 return dget(sb->s_root);
854 }
855
856 return d_obtain_alias(inode);
857 }
858
859 static int btrfs_fill_super(struct super_block *sb,
860 struct btrfs_fs_devices *fs_devices,
861 void *data, int silent)
862 {
863 struct inode *inode;
864 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
865 struct btrfs_key key;
866 int err;
867
868 sb->s_maxbytes = MAX_LFS_FILESIZE;
869 sb->s_magic = BTRFS_SUPER_MAGIC;
870 sb->s_op = &btrfs_super_ops;
871 sb->s_d_op = &btrfs_dentry_operations;
872 sb->s_export_op = &btrfs_export_ops;
873 sb->s_xattr = btrfs_xattr_handlers;
874 sb->s_time_gran = 1;
875 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
876 sb->s_flags |= MS_POSIXACL;
877 #endif
878 sb->s_flags |= MS_I_VERSION;
879 err = open_ctree(sb, fs_devices, (char *)data);
880 if (err) {
881 printk("btrfs: open_ctree failed\n");
882 return err;
883 }
884
885 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
886 key.type = BTRFS_INODE_ITEM_KEY;
887 key.offset = 0;
888 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
889 if (IS_ERR(inode)) {
890 err = PTR_ERR(inode);
891 goto fail_close;
892 }
893
894 sb->s_root = d_make_root(inode);
895 if (!sb->s_root) {
896 err = -ENOMEM;
897 goto fail_close;
898 }
899
900 save_mount_options(sb, data);
901 cleancache_init_fs(sb);
902 sb->s_flags |= MS_ACTIVE;
903 return 0;
904
905 fail_close:
906 close_ctree(fs_info->tree_root);
907 return err;
908 }
909
910 int btrfs_sync_fs(struct super_block *sb, int wait)
911 {
912 struct btrfs_trans_handle *trans;
913 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
914 struct btrfs_root *root = fs_info->tree_root;
915
916 trace_btrfs_sync_fs(wait);
917
918 if (!wait) {
919 filemap_flush(fs_info->btree_inode->i_mapping);
920 return 0;
921 }
922
923 btrfs_wait_ordered_roots(fs_info, -1);
924
925 trans = btrfs_attach_transaction_barrier(root);
926 if (IS_ERR(trans)) {
927 /* no transaction, don't bother */
928 if (PTR_ERR(trans) == -ENOENT)
929 return 0;
930 return PTR_ERR(trans);
931 }
932 return btrfs_commit_transaction(trans, root);
933 }
934
935 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
936 {
937 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
938 struct btrfs_root *root = info->tree_root;
939 char *compress_type;
940
941 if (btrfs_test_opt(root, DEGRADED))
942 seq_puts(seq, ",degraded");
943 if (btrfs_test_opt(root, NODATASUM))
944 seq_puts(seq, ",nodatasum");
945 if (btrfs_test_opt(root, NODATACOW))
946 seq_puts(seq, ",nodatacow");
947 if (btrfs_test_opt(root, NOBARRIER))
948 seq_puts(seq, ",nobarrier");
949 if (info->max_inline != 8192 * 1024)
950 seq_printf(seq, ",max_inline=%llu", info->max_inline);
951 if (info->alloc_start != 0)
952 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
953 if (info->thread_pool_size != min_t(unsigned long,
954 num_online_cpus() + 2, 8))
955 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
956 if (btrfs_test_opt(root, COMPRESS)) {
957 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
958 compress_type = "zlib";
959 else
960 compress_type = "lzo";
961 if (btrfs_test_opt(root, FORCE_COMPRESS))
962 seq_printf(seq, ",compress-force=%s", compress_type);
963 else
964 seq_printf(seq, ",compress=%s", compress_type);
965 }
966 if (btrfs_test_opt(root, NOSSD))
967 seq_puts(seq, ",nossd");
968 if (btrfs_test_opt(root, SSD_SPREAD))
969 seq_puts(seq, ",ssd_spread");
970 else if (btrfs_test_opt(root, SSD))
971 seq_puts(seq, ",ssd");
972 if (btrfs_test_opt(root, NOTREELOG))
973 seq_puts(seq, ",notreelog");
974 if (btrfs_test_opt(root, FLUSHONCOMMIT))
975 seq_puts(seq, ",flushoncommit");
976 if (btrfs_test_opt(root, DISCARD))
977 seq_puts(seq, ",discard");
978 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
979 seq_puts(seq, ",noacl");
980 if (btrfs_test_opt(root, SPACE_CACHE))
981 seq_puts(seq, ",space_cache");
982 else
983 seq_puts(seq, ",nospace_cache");
984 if (btrfs_test_opt(root, RESCAN_UUID_TREE))
985 seq_puts(seq, ",rescan_uuid_tree");
986 if (btrfs_test_opt(root, CLEAR_CACHE))
987 seq_puts(seq, ",clear_cache");
988 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
989 seq_puts(seq, ",user_subvol_rm_allowed");
990 if (btrfs_test_opt(root, ENOSPC_DEBUG))
991 seq_puts(seq, ",enospc_debug");
992 if (btrfs_test_opt(root, AUTO_DEFRAG))
993 seq_puts(seq, ",autodefrag");
994 if (btrfs_test_opt(root, INODE_MAP_CACHE))
995 seq_puts(seq, ",inode_cache");
996 if (btrfs_test_opt(root, SKIP_BALANCE))
997 seq_puts(seq, ",skip_balance");
998 if (btrfs_test_opt(root, RECOVERY))
999 seq_puts(seq, ",recovery");
1000 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1001 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1002 seq_puts(seq, ",check_int_data");
1003 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1004 seq_puts(seq, ",check_int");
1005 if (info->check_integrity_print_mask)
1006 seq_printf(seq, ",check_int_print_mask=%d",
1007 info->check_integrity_print_mask);
1008 #endif
1009 if (info->metadata_ratio)
1010 seq_printf(seq, ",metadata_ratio=%d",
1011 info->metadata_ratio);
1012 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1013 seq_puts(seq, ",fatal_errors=panic");
1014 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1015 seq_printf(seq, ",commit=%d", info->commit_interval);
1016 return 0;
1017 }
1018
1019 static int btrfs_test_super(struct super_block *s, void *data)
1020 {
1021 struct btrfs_fs_info *p = data;
1022 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1023
1024 return fs_info->fs_devices == p->fs_devices;
1025 }
1026
1027 static int btrfs_set_super(struct super_block *s, void *data)
1028 {
1029 int err = set_anon_super(s, data);
1030 if (!err)
1031 s->s_fs_info = data;
1032 return err;
1033 }
1034
1035 /*
1036 * subvolumes are identified by ino 256
1037 */
1038 static inline int is_subvolume_inode(struct inode *inode)
1039 {
1040 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1041 return 1;
1042 return 0;
1043 }
1044
1045 /*
1046 * This will strip out the subvol=%s argument for an argument string and add
1047 * subvolid=0 to make sure we get the actual tree root for path walking to the
1048 * subvol we want.
1049 */
1050 static char *setup_root_args(char *args)
1051 {
1052 unsigned len = strlen(args) + 2 + 1;
1053 char *src, *dst, *buf;
1054
1055 /*
1056 * We need the same args as before, but with this substitution:
1057 * s!subvol=[^,]+!subvolid=0!
1058 *
1059 * Since the replacement string is up to 2 bytes longer than the
1060 * original, allocate strlen(args) + 2 + 1 bytes.
1061 */
1062
1063 src = strstr(args, "subvol=");
1064 /* This shouldn't happen, but just in case.. */
1065 if (!src)
1066 return NULL;
1067
1068 buf = dst = kmalloc(len, GFP_NOFS);
1069 if (!buf)
1070 return NULL;
1071
1072 /*
1073 * If the subvol= arg is not at the start of the string,
1074 * copy whatever precedes it into buf.
1075 */
1076 if (src != args) {
1077 *src++ = '\0';
1078 strcpy(buf, args);
1079 dst += strlen(args);
1080 }
1081
1082 strcpy(dst, "subvolid=0");
1083 dst += strlen("subvolid=0");
1084
1085 /*
1086 * If there is a "," after the original subvol=... string,
1087 * copy that suffix into our buffer. Otherwise, we're done.
1088 */
1089 src = strchr(src, ',');
1090 if (src)
1091 strcpy(dst, src);
1092
1093 return buf;
1094 }
1095
1096 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1097 const char *device_name, char *data)
1098 {
1099 struct dentry *root;
1100 struct vfsmount *mnt;
1101 char *newargs;
1102
1103 newargs = setup_root_args(data);
1104 if (!newargs)
1105 return ERR_PTR(-ENOMEM);
1106 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1107 newargs);
1108 kfree(newargs);
1109 if (IS_ERR(mnt))
1110 return ERR_CAST(mnt);
1111
1112 root = mount_subtree(mnt, subvol_name);
1113
1114 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1115 struct super_block *s = root->d_sb;
1116 dput(root);
1117 root = ERR_PTR(-EINVAL);
1118 deactivate_locked_super(s);
1119 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1120 subvol_name);
1121 }
1122
1123 return root;
1124 }
1125
1126 /*
1127 * Find a superblock for the given device / mount point.
1128 *
1129 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1130 * for multiple device setup. Make sure to keep it in sync.
1131 */
1132 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1133 const char *device_name, void *data)
1134 {
1135 struct block_device *bdev = NULL;
1136 struct super_block *s;
1137 struct dentry *root;
1138 struct btrfs_fs_devices *fs_devices = NULL;
1139 struct btrfs_fs_info *fs_info = NULL;
1140 fmode_t mode = FMODE_READ;
1141 char *subvol_name = NULL;
1142 u64 subvol_objectid = 0;
1143 int error = 0;
1144
1145 if (!(flags & MS_RDONLY))
1146 mode |= FMODE_WRITE;
1147
1148 error = btrfs_parse_early_options(data, mode, fs_type,
1149 &subvol_name, &subvol_objectid,
1150 &fs_devices);
1151 if (error) {
1152 kfree(subvol_name);
1153 return ERR_PTR(error);
1154 }
1155
1156 if (subvol_name) {
1157 root = mount_subvol(subvol_name, flags, device_name, data);
1158 kfree(subvol_name);
1159 return root;
1160 }
1161
1162 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1163 if (error)
1164 return ERR_PTR(error);
1165
1166 /*
1167 * Setup a dummy root and fs_info for test/set super. This is because
1168 * we don't actually fill this stuff out until open_ctree, but we need
1169 * it for searching for existing supers, so this lets us do that and
1170 * then open_ctree will properly initialize everything later.
1171 */
1172 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1173 if (!fs_info)
1174 return ERR_PTR(-ENOMEM);
1175
1176 fs_info->fs_devices = fs_devices;
1177
1178 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1179 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1180 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1181 error = -ENOMEM;
1182 goto error_fs_info;
1183 }
1184
1185 error = btrfs_open_devices(fs_devices, mode, fs_type);
1186 if (error)
1187 goto error_fs_info;
1188
1189 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1190 error = -EACCES;
1191 goto error_close_devices;
1192 }
1193
1194 bdev = fs_devices->latest_bdev;
1195 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1196 fs_info);
1197 if (IS_ERR(s)) {
1198 error = PTR_ERR(s);
1199 goto error_close_devices;
1200 }
1201
1202 if (s->s_root) {
1203 btrfs_close_devices(fs_devices);
1204 free_fs_info(fs_info);
1205 if ((flags ^ s->s_flags) & MS_RDONLY)
1206 error = -EBUSY;
1207 } else {
1208 char b[BDEVNAME_SIZE];
1209
1210 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1211 btrfs_sb(s)->bdev_holder = fs_type;
1212 error = btrfs_fill_super(s, fs_devices, data,
1213 flags & MS_SILENT ? 1 : 0);
1214 }
1215
1216 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1217 if (IS_ERR(root))
1218 deactivate_locked_super(s);
1219
1220 return root;
1221
1222 error_close_devices:
1223 btrfs_close_devices(fs_devices);
1224 error_fs_info:
1225 free_fs_info(fs_info);
1226 return ERR_PTR(error);
1227 }
1228
1229 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1230 {
1231 spin_lock_irq(&workers->lock);
1232 workers->max_workers = new_limit;
1233 spin_unlock_irq(&workers->lock);
1234 }
1235
1236 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1237 int new_pool_size, int old_pool_size)
1238 {
1239 if (new_pool_size == old_pool_size)
1240 return;
1241
1242 fs_info->thread_pool_size = new_pool_size;
1243
1244 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1245 old_pool_size, new_pool_size);
1246
1247 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1248 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1249 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1250 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1251 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1252 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1253 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1254 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1255 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1256 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1257 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1258 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1259 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1260 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1261 new_pool_size);
1262 }
1263
1264 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1265 {
1266 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1267 }
1268
1269 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1270 unsigned long old_opts, int flags)
1271 {
1272 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1273 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1274 (flags & MS_RDONLY))) {
1275 /* wait for any defraggers to finish */
1276 wait_event(fs_info->transaction_wait,
1277 (atomic_read(&fs_info->defrag_running) == 0));
1278 if (flags & MS_RDONLY)
1279 sync_filesystem(fs_info->sb);
1280 }
1281 }
1282
1283 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1284 unsigned long old_opts)
1285 {
1286 /*
1287 * We need cleanup all defragable inodes if the autodefragment is
1288 * close or the fs is R/O.
1289 */
1290 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1291 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1292 (fs_info->sb->s_flags & MS_RDONLY))) {
1293 btrfs_cleanup_defrag_inodes(fs_info);
1294 }
1295
1296 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1297 }
1298
1299 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1300 {
1301 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1302 struct btrfs_root *root = fs_info->tree_root;
1303 unsigned old_flags = sb->s_flags;
1304 unsigned long old_opts = fs_info->mount_opt;
1305 unsigned long old_compress_type = fs_info->compress_type;
1306 u64 old_max_inline = fs_info->max_inline;
1307 u64 old_alloc_start = fs_info->alloc_start;
1308 int old_thread_pool_size = fs_info->thread_pool_size;
1309 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1310 int ret;
1311
1312 btrfs_remount_prepare(fs_info);
1313
1314 ret = btrfs_parse_options(root, data);
1315 if (ret) {
1316 ret = -EINVAL;
1317 goto restore;
1318 }
1319
1320 btrfs_remount_begin(fs_info, old_opts, *flags);
1321 btrfs_resize_thread_pool(fs_info,
1322 fs_info->thread_pool_size, old_thread_pool_size);
1323
1324 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1325 goto out;
1326
1327 if (*flags & MS_RDONLY) {
1328 /*
1329 * this also happens on 'umount -rf' or on shutdown, when
1330 * the filesystem is busy.
1331 */
1332
1333 /* wait for the uuid_scan task to finish */
1334 down(&fs_info->uuid_tree_rescan_sem);
1335 /* avoid complains from lockdep et al. */
1336 up(&fs_info->uuid_tree_rescan_sem);
1337
1338 sb->s_flags |= MS_RDONLY;
1339
1340 btrfs_dev_replace_suspend_for_unmount(fs_info);
1341 btrfs_scrub_cancel(fs_info);
1342 btrfs_pause_balance(fs_info);
1343
1344 ret = btrfs_commit_super(root);
1345 if (ret)
1346 goto restore;
1347 } else {
1348 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1349 btrfs_err(fs_info,
1350 "Remounting read-write after error is not allowed\n");
1351 ret = -EINVAL;
1352 goto restore;
1353 }
1354 if (fs_info->fs_devices->rw_devices == 0) {
1355 ret = -EACCES;
1356 goto restore;
1357 }
1358
1359 if (fs_info->fs_devices->missing_devices >
1360 fs_info->num_tolerated_disk_barrier_failures &&
1361 !(*flags & MS_RDONLY)) {
1362 printk(KERN_WARNING
1363 "Btrfs: too many missing devices, writeable remount is not allowed\n");
1364 ret = -EACCES;
1365 goto restore;
1366 }
1367
1368 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1369 ret = -EINVAL;
1370 goto restore;
1371 }
1372
1373 ret = btrfs_cleanup_fs_roots(fs_info);
1374 if (ret)
1375 goto restore;
1376
1377 /* recover relocation */
1378 ret = btrfs_recover_relocation(root);
1379 if (ret)
1380 goto restore;
1381
1382 ret = btrfs_resume_balance_async(fs_info);
1383 if (ret)
1384 goto restore;
1385
1386 ret = btrfs_resume_dev_replace_async(fs_info);
1387 if (ret) {
1388 pr_warn("btrfs: failed to resume dev_replace\n");
1389 goto restore;
1390 }
1391
1392 if (!fs_info->uuid_root) {
1393 pr_info("btrfs: creating UUID tree\n");
1394 ret = btrfs_create_uuid_tree(fs_info);
1395 if (ret) {
1396 pr_warn("btrfs: failed to create the uuid tree"
1397 "%d\n", ret);
1398 goto restore;
1399 }
1400 }
1401 sb->s_flags &= ~MS_RDONLY;
1402 }
1403 out:
1404 btrfs_remount_cleanup(fs_info, old_opts);
1405 return 0;
1406
1407 restore:
1408 /* We've hit an error - don't reset MS_RDONLY */
1409 if (sb->s_flags & MS_RDONLY)
1410 old_flags |= MS_RDONLY;
1411 sb->s_flags = old_flags;
1412 fs_info->mount_opt = old_opts;
1413 fs_info->compress_type = old_compress_type;
1414 fs_info->max_inline = old_max_inline;
1415 mutex_lock(&fs_info->chunk_mutex);
1416 fs_info->alloc_start = old_alloc_start;
1417 mutex_unlock(&fs_info->chunk_mutex);
1418 btrfs_resize_thread_pool(fs_info,
1419 old_thread_pool_size, fs_info->thread_pool_size);
1420 fs_info->metadata_ratio = old_metadata_ratio;
1421 btrfs_remount_cleanup(fs_info, old_opts);
1422 return ret;
1423 }
1424
1425 /* Used to sort the devices by max_avail(descending sort) */
1426 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1427 const void *dev_info2)
1428 {
1429 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1430 ((struct btrfs_device_info *)dev_info2)->max_avail)
1431 return -1;
1432 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1433 ((struct btrfs_device_info *)dev_info2)->max_avail)
1434 return 1;
1435 else
1436 return 0;
1437 }
1438
1439 /*
1440 * sort the devices by max_avail, in which max free extent size of each device
1441 * is stored.(Descending Sort)
1442 */
1443 static inline void btrfs_descending_sort_devices(
1444 struct btrfs_device_info *devices,
1445 size_t nr_devices)
1446 {
1447 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1448 btrfs_cmp_device_free_bytes, NULL);
1449 }
1450
1451 /*
1452 * The helper to calc the free space on the devices that can be used to store
1453 * file data.
1454 */
1455 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1456 {
1457 struct btrfs_fs_info *fs_info = root->fs_info;
1458 struct btrfs_device_info *devices_info;
1459 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1460 struct btrfs_device *device;
1461 u64 skip_space;
1462 u64 type;
1463 u64 avail_space;
1464 u64 used_space;
1465 u64 min_stripe_size;
1466 int min_stripes = 1, num_stripes = 1;
1467 int i = 0, nr_devices;
1468 int ret;
1469
1470 nr_devices = fs_info->fs_devices->open_devices;
1471 BUG_ON(!nr_devices);
1472
1473 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1474 GFP_NOFS);
1475 if (!devices_info)
1476 return -ENOMEM;
1477
1478 /* calc min stripe number for data space alloction */
1479 type = btrfs_get_alloc_profile(root, 1);
1480 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1481 min_stripes = 2;
1482 num_stripes = nr_devices;
1483 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1484 min_stripes = 2;
1485 num_stripes = 2;
1486 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1487 min_stripes = 4;
1488 num_stripes = 4;
1489 }
1490
1491 if (type & BTRFS_BLOCK_GROUP_DUP)
1492 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1493 else
1494 min_stripe_size = BTRFS_STRIPE_LEN;
1495
1496 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1497 if (!device->in_fs_metadata || !device->bdev ||
1498 device->is_tgtdev_for_dev_replace)
1499 continue;
1500
1501 avail_space = device->total_bytes - device->bytes_used;
1502
1503 /* align with stripe_len */
1504 do_div(avail_space, BTRFS_STRIPE_LEN);
1505 avail_space *= BTRFS_STRIPE_LEN;
1506
1507 /*
1508 * In order to avoid overwritting the superblock on the drive,
1509 * btrfs starts at an offset of at least 1MB when doing chunk
1510 * allocation.
1511 */
1512 skip_space = 1024 * 1024;
1513
1514 /* user can set the offset in fs_info->alloc_start. */
1515 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1516 device->total_bytes)
1517 skip_space = max(fs_info->alloc_start, skip_space);
1518
1519 /*
1520 * btrfs can not use the free space in [0, skip_space - 1],
1521 * we must subtract it from the total. In order to implement
1522 * it, we account the used space in this range first.
1523 */
1524 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1525 &used_space);
1526 if (ret) {
1527 kfree(devices_info);
1528 return ret;
1529 }
1530
1531 /* calc the free space in [0, skip_space - 1] */
1532 skip_space -= used_space;
1533
1534 /*
1535 * we can use the free space in [0, skip_space - 1], subtract
1536 * it from the total.
1537 */
1538 if (avail_space && avail_space >= skip_space)
1539 avail_space -= skip_space;
1540 else
1541 avail_space = 0;
1542
1543 if (avail_space < min_stripe_size)
1544 continue;
1545
1546 devices_info[i].dev = device;
1547 devices_info[i].max_avail = avail_space;
1548
1549 i++;
1550 }
1551
1552 nr_devices = i;
1553
1554 btrfs_descending_sort_devices(devices_info, nr_devices);
1555
1556 i = nr_devices - 1;
1557 avail_space = 0;
1558 while (nr_devices >= min_stripes) {
1559 if (num_stripes > nr_devices)
1560 num_stripes = nr_devices;
1561
1562 if (devices_info[i].max_avail >= min_stripe_size) {
1563 int j;
1564 u64 alloc_size;
1565
1566 avail_space += devices_info[i].max_avail * num_stripes;
1567 alloc_size = devices_info[i].max_avail;
1568 for (j = i + 1 - num_stripes; j <= i; j++)
1569 devices_info[j].max_avail -= alloc_size;
1570 }
1571 i--;
1572 nr_devices--;
1573 }
1574
1575 kfree(devices_info);
1576 *free_bytes = avail_space;
1577 return 0;
1578 }
1579
1580 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1581 {
1582 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1583 struct btrfs_super_block *disk_super = fs_info->super_copy;
1584 struct list_head *head = &fs_info->space_info;
1585 struct btrfs_space_info *found;
1586 u64 total_used = 0;
1587 u64 total_free_data = 0;
1588 int bits = dentry->d_sb->s_blocksize_bits;
1589 __be32 *fsid = (__be32 *)fs_info->fsid;
1590 int ret;
1591
1592 /* holding chunk_muext to avoid allocating new chunks */
1593 mutex_lock(&fs_info->chunk_mutex);
1594 rcu_read_lock();
1595 list_for_each_entry_rcu(found, head, list) {
1596 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1597 total_free_data += found->disk_total - found->disk_used;
1598 total_free_data -=
1599 btrfs_account_ro_block_groups_free_space(found);
1600 }
1601
1602 total_used += found->disk_used;
1603 }
1604 rcu_read_unlock();
1605
1606 buf->f_namelen = BTRFS_NAME_LEN;
1607 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1608 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1609 buf->f_bsize = dentry->d_sb->s_blocksize;
1610 buf->f_type = BTRFS_SUPER_MAGIC;
1611 buf->f_bavail = total_free_data;
1612 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1613 if (ret) {
1614 mutex_unlock(&fs_info->chunk_mutex);
1615 return ret;
1616 }
1617 buf->f_bavail += total_free_data;
1618 buf->f_bavail = buf->f_bavail >> bits;
1619 mutex_unlock(&fs_info->chunk_mutex);
1620
1621 /* We treat it as constant endianness (it doesn't matter _which_)
1622 because we want the fsid to come out the same whether mounted
1623 on a big-endian or little-endian host */
1624 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1625 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1626 /* Mask in the root object ID too, to disambiguate subvols */
1627 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1628 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1629
1630 return 0;
1631 }
1632
1633 static void btrfs_kill_super(struct super_block *sb)
1634 {
1635 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1636 kill_anon_super(sb);
1637 free_fs_info(fs_info);
1638 }
1639
1640 static struct file_system_type btrfs_fs_type = {
1641 .owner = THIS_MODULE,
1642 .name = "btrfs",
1643 .mount = btrfs_mount,
1644 .kill_sb = btrfs_kill_super,
1645 .fs_flags = FS_REQUIRES_DEV,
1646 };
1647 MODULE_ALIAS_FS("btrfs");
1648
1649 /*
1650 * used by btrfsctl to scan devices when no FS is mounted
1651 */
1652 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1653 unsigned long arg)
1654 {
1655 struct btrfs_ioctl_vol_args *vol;
1656 struct btrfs_fs_devices *fs_devices;
1657 int ret = -ENOTTY;
1658
1659 if (!capable(CAP_SYS_ADMIN))
1660 return -EPERM;
1661
1662 vol = memdup_user((void __user *)arg, sizeof(*vol));
1663 if (IS_ERR(vol))
1664 return PTR_ERR(vol);
1665
1666 switch (cmd) {
1667 case BTRFS_IOC_SCAN_DEV:
1668 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1669 &btrfs_fs_type, &fs_devices);
1670 break;
1671 case BTRFS_IOC_DEVICES_READY:
1672 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1673 &btrfs_fs_type, &fs_devices);
1674 if (ret)
1675 break;
1676 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1677 break;
1678 }
1679
1680 kfree(vol);
1681 return ret;
1682 }
1683
1684 static int btrfs_freeze(struct super_block *sb)
1685 {
1686 struct btrfs_trans_handle *trans;
1687 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1688
1689 trans = btrfs_attach_transaction_barrier(root);
1690 if (IS_ERR(trans)) {
1691 /* no transaction, don't bother */
1692 if (PTR_ERR(trans) == -ENOENT)
1693 return 0;
1694 return PTR_ERR(trans);
1695 }
1696 return btrfs_commit_transaction(trans, root);
1697 }
1698
1699 static int btrfs_unfreeze(struct super_block *sb)
1700 {
1701 return 0;
1702 }
1703
1704 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1705 {
1706 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1707 struct btrfs_fs_devices *cur_devices;
1708 struct btrfs_device *dev, *first_dev = NULL;
1709 struct list_head *head;
1710 struct rcu_string *name;
1711
1712 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1713 cur_devices = fs_info->fs_devices;
1714 while (cur_devices) {
1715 head = &cur_devices->devices;
1716 list_for_each_entry(dev, head, dev_list) {
1717 if (dev->missing)
1718 continue;
1719 if (!first_dev || dev->devid < first_dev->devid)
1720 first_dev = dev;
1721 }
1722 cur_devices = cur_devices->seed;
1723 }
1724
1725 if (first_dev) {
1726 rcu_read_lock();
1727 name = rcu_dereference(first_dev->name);
1728 seq_escape(m, name->str, " \t\n\\");
1729 rcu_read_unlock();
1730 } else {
1731 WARN_ON(1);
1732 }
1733 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1734 return 0;
1735 }
1736
1737 static const struct super_operations btrfs_super_ops = {
1738 .drop_inode = btrfs_drop_inode,
1739 .evict_inode = btrfs_evict_inode,
1740 .put_super = btrfs_put_super,
1741 .sync_fs = btrfs_sync_fs,
1742 .show_options = btrfs_show_options,
1743 .show_devname = btrfs_show_devname,
1744 .write_inode = btrfs_write_inode,
1745 .alloc_inode = btrfs_alloc_inode,
1746 .destroy_inode = btrfs_destroy_inode,
1747 .statfs = btrfs_statfs,
1748 .remount_fs = btrfs_remount,
1749 .freeze_fs = btrfs_freeze,
1750 .unfreeze_fs = btrfs_unfreeze,
1751 };
1752
1753 static const struct file_operations btrfs_ctl_fops = {
1754 .unlocked_ioctl = btrfs_control_ioctl,
1755 .compat_ioctl = btrfs_control_ioctl,
1756 .owner = THIS_MODULE,
1757 .llseek = noop_llseek,
1758 };
1759
1760 static struct miscdevice btrfs_misc = {
1761 .minor = BTRFS_MINOR,
1762 .name = "btrfs-control",
1763 .fops = &btrfs_ctl_fops
1764 };
1765
1766 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1767 MODULE_ALIAS("devname:btrfs-control");
1768
1769 static int btrfs_interface_init(void)
1770 {
1771 return misc_register(&btrfs_misc);
1772 }
1773
1774 static void btrfs_interface_exit(void)
1775 {
1776 if (misc_deregister(&btrfs_misc) < 0)
1777 printk(KERN_INFO "btrfs: misc_deregister failed for control device\n");
1778 }
1779
1780 static void btrfs_print_info(void)
1781 {
1782 printk(KERN_INFO "Btrfs loaded"
1783 #ifdef CONFIG_BTRFS_DEBUG
1784 ", debug=on"
1785 #endif
1786 #ifdef CONFIG_BTRFS_ASSERT
1787 ", assert=on"
1788 #endif
1789 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1790 ", integrity-checker=on"
1791 #endif
1792 "\n");
1793 }
1794
1795 static int btrfs_run_sanity_tests(void)
1796 {
1797 int ret;
1798
1799 ret = btrfs_init_test_fs();
1800 if (ret)
1801 return ret;
1802
1803 ret = btrfs_test_free_space_cache();
1804 if (ret)
1805 goto out;
1806 ret = btrfs_test_extent_buffer_operations();
1807 if (ret)
1808 goto out;
1809 ret = btrfs_test_extent_io();
1810 if (ret)
1811 goto out;
1812 ret = btrfs_test_inodes();
1813 out:
1814 btrfs_destroy_test_fs();
1815 return ret;
1816 }
1817
1818 static int __init init_btrfs_fs(void)
1819 {
1820 int err;
1821
1822 err = btrfs_init_sysfs();
1823 if (err)
1824 return err;
1825
1826 btrfs_init_compress();
1827
1828 err = btrfs_init_cachep();
1829 if (err)
1830 goto free_compress;
1831
1832 err = extent_io_init();
1833 if (err)
1834 goto free_cachep;
1835
1836 err = extent_map_init();
1837 if (err)
1838 goto free_extent_io;
1839
1840 err = ordered_data_init();
1841 if (err)
1842 goto free_extent_map;
1843
1844 err = btrfs_delayed_inode_init();
1845 if (err)
1846 goto free_ordered_data;
1847
1848 err = btrfs_auto_defrag_init();
1849 if (err)
1850 goto free_delayed_inode;
1851
1852 err = btrfs_delayed_ref_init();
1853 if (err)
1854 goto free_auto_defrag;
1855
1856 err = btrfs_prelim_ref_init();
1857 if (err)
1858 goto free_prelim_ref;
1859
1860 err = btrfs_interface_init();
1861 if (err)
1862 goto free_delayed_ref;
1863
1864 btrfs_init_lockdep();
1865
1866 btrfs_print_info();
1867
1868 err = btrfs_run_sanity_tests();
1869 if (err)
1870 goto unregister_ioctl;
1871
1872 err = register_filesystem(&btrfs_fs_type);
1873 if (err)
1874 goto unregister_ioctl;
1875
1876 return 0;
1877
1878 unregister_ioctl:
1879 btrfs_interface_exit();
1880 free_prelim_ref:
1881 btrfs_prelim_ref_exit();
1882 free_delayed_ref:
1883 btrfs_delayed_ref_exit();
1884 free_auto_defrag:
1885 btrfs_auto_defrag_exit();
1886 free_delayed_inode:
1887 btrfs_delayed_inode_exit();
1888 free_ordered_data:
1889 ordered_data_exit();
1890 free_extent_map:
1891 extent_map_exit();
1892 free_extent_io:
1893 extent_io_exit();
1894 free_cachep:
1895 btrfs_destroy_cachep();
1896 free_compress:
1897 btrfs_exit_compress();
1898 btrfs_exit_sysfs();
1899 return err;
1900 }
1901
1902 static void __exit exit_btrfs_fs(void)
1903 {
1904 btrfs_destroy_cachep();
1905 btrfs_delayed_ref_exit();
1906 btrfs_auto_defrag_exit();
1907 btrfs_delayed_inode_exit();
1908 btrfs_prelim_ref_exit();
1909 ordered_data_exit();
1910 extent_map_exit();
1911 extent_io_exit();
1912 btrfs_interface_exit();
1913 unregister_filesystem(&btrfs_fs_type);
1914 btrfs_exit_sysfs();
1915 btrfs_cleanup_fs_uuids();
1916 btrfs_exit_compress();
1917 }
1918
1919 module_init(init_btrfs_fs)
1920 module_exit(exit_btrfs_fs)
1921
1922 MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.069451 seconds and 6 git commands to generate.