projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: fix heavy delalloc related deadlock
[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 "compat.h"
46 #include "delayed-inode.h"
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
52 #include "xattr.h"
53 #include "volumes.h"
54 #include "export.h"
55 #include "compression.h"
56 #include "rcu-string.h"
57 #include "dev-replace.h"
58 #include "free-space-cache.h"
59
60 #define CREATE_TRACE_POINTS
61 #include <trace/events/btrfs.h>
62
63 static const struct super_operations btrfs_super_ops;
64 static struct file_system_type btrfs_fs_type;
65
66 static const char *btrfs_decode_error(int errno)
67 {
68 char *errstr = "unknown";
69
70 switch (errno) {
71 case -EIO:
72 errstr = "IO failure";
73 break;
74 case -ENOMEM:
75 errstr = "Out of memory";
76 break;
77 case -EROFS:
78 errstr = "Readonly filesystem";
79 break;
80 case -EEXIST:
81 errstr = "Object already exists";
82 break;
83 case -ENOSPC:
84 errstr = "No space left";
85 break;
86 case -ENOENT:
87 errstr = "No such entry";
88 break;
89 }
90
91 return errstr;
92 }
93
94 static void save_error_info(struct btrfs_fs_info *fs_info)
95 {
96 /*
97 * today we only save the error info into ram. Long term we'll
98 * also send it down to the disk
99 */
100 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
101 }
102
103 /* btrfs handle error by forcing the filesystem readonly */
104 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
105 {
106 struct super_block *sb = fs_info->sb;
107
108 if (sb->s_flags & MS_RDONLY)
109 return;
110
111 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
112 sb->s_flags |= MS_RDONLY;
113 btrfs_info(fs_info, "forced readonly");
114 /*
115 * Note that a running device replace operation is not
116 * canceled here although there is no way to update
117 * the progress. It would add the risk of a deadlock,
118 * therefore the canceling is ommited. The only penalty
119 * is that some I/O remains active until the procedure
120 * completes. The next time when the filesystem is
121 * mounted writeable again, the device replace
122 * operation continues.
123 */
124 }
125 }
126
127 #ifdef CONFIG_PRINTK
128 /*
129 * __btrfs_std_error decodes expected errors from the caller and
130 * invokes the approciate error response.
131 */
132 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
133 unsigned int line, int errno, const char *fmt, ...)
134 {
135 struct super_block *sb = fs_info->sb;
136 const char *errstr;
137
138 /*
139 * Special case: if the error is EROFS, and we're already
140 * under MS_RDONLY, then it is safe here.
141 */
142 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
143 return;
144
145 errstr = btrfs_decode_error(errno);
146 if (fmt) {
147 struct va_format vaf;
148 va_list args;
149
150 va_start(args, fmt);
151 vaf.fmt = fmt;
152 vaf.va = &args;
153
154 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s (%pV)\n",
155 sb->s_id, function, line, errno, errstr, &vaf);
156 va_end(args);
157 } else {
158 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
160 }
161
162 /* Don't go through full error handling during mount */
163 save_error_info(fs_info);
164 if (sb->s_flags & MS_BORN)
165 btrfs_handle_error(fs_info);
166 }
167
168 static const char * const logtypes[] = {
169 "emergency",
170 "alert",
171 "critical",
172 "error",
173 "warning",
174 "notice",
175 "info",
176 "debug",
177 };
178
179 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
180 {
181 struct super_block *sb = fs_info->sb;
182 char lvl[4];
183 struct va_format vaf;
184 va_list args;
185 const char *type = logtypes[4];
186 int kern_level;
187
188 va_start(args, fmt);
189
190 kern_level = printk_get_level(fmt);
191 if (kern_level) {
192 size_t size = printk_skip_level(fmt) - fmt;
193 memcpy(lvl, fmt, size);
194 lvl[size] = '\0';
195 fmt += size;
196 type = logtypes[kern_level - '0'];
197 } else
198 *lvl = '\0';
199
200 vaf.fmt = fmt;
201 vaf.va = &args;
202
203 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
204
205 va_end(args);
206 }
207
208 #else
209
210 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
211 unsigned int line, int errno, const char *fmt, ...)
212 {
213 struct super_block *sb = fs_info->sb;
214
215 /*
216 * Special case: if the error is EROFS, and we're already
217 * under MS_RDONLY, then it is safe here.
218 */
219 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
220 return;
221
222 /* Don't go through full error handling during mount */
223 if (sb->s_flags & MS_BORN) {
224 save_error_info(fs_info);
225 btrfs_handle_error(fs_info);
226 }
227 }
228 #endif
229
230 /*
231 * We only mark the transaction aborted and then set the file system read-only.
232 * This will prevent new transactions from starting or trying to join this
233 * one.
234 *
235 * This means that error recovery at the call site is limited to freeing
236 * any local memory allocations and passing the error code up without
237 * further cleanup. The transaction should complete as it normally would
238 * in the call path but will return -EIO.
239 *
240 * We'll complete the cleanup in btrfs_end_transaction and
241 * btrfs_commit_transaction.
242 */
243 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
244 struct btrfs_root *root, const char *function,
245 unsigned int line, int errno)
246 {
247 /*
248 * Report first abort since mount
249 */
250 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
251 &root->fs_info->fs_state)) {
252 WARN(1, KERN_DEBUG "btrfs: Transaction aborted (error %d)\n",
253 errno);
254 }
255 trans->aborted = errno;
256 /* Nothing used. The other threads that have joined this
257 * transaction may be able to continue. */
258 if (!trans->blocks_used) {
259 const char *errstr;
260
261 errstr = btrfs_decode_error(errno);
262 btrfs_warn(root->fs_info,
263 "%s:%d: Aborting unused transaction(%s).",
264 function, line, errstr);
265 return;
266 }
267 ACCESS_ONCE(trans->transaction->aborted) = errno;
268 /* Wake up anybody who may be waiting on this transaction */
269 wake_up(&root->fs_info->transaction_wait);
270 wake_up(&root->fs_info->transaction_blocked_wait);
271 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
272 }
273 /*
274 * __btrfs_panic decodes unexpected, fatal errors from the caller,
275 * issues an alert, and either panics or BUGs, depending on mount options.
276 */
277 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
278 unsigned int line, int errno, const char *fmt, ...)
279 {
280 char *s_id = "<unknown>";
281 const char *errstr;
282 struct va_format vaf = { .fmt = fmt };
283 va_list args;
284
285 if (fs_info)
286 s_id = fs_info->sb->s_id;
287
288 va_start(args, fmt);
289 vaf.va = &args;
290
291 errstr = btrfs_decode_error(errno);
292 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
293 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
294 s_id, function, line, &vaf, errno, errstr);
295
296 printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
297 s_id, function, line, &vaf, errno, errstr);
298 va_end(args);
299 /* Caller calls BUG() */
300 }
301
302 static void btrfs_put_super(struct super_block *sb)
303 {
304 (void)close_ctree(btrfs_sb(sb)->tree_root);
305 /* FIXME: need to fix VFS to return error? */
306 /* AV: return it _where_? ->put_super() can be triggered by any number
307 * of async events, up to and including delivery of SIGKILL to the
308 * last process that kept it busy. Or segfault in the aforementioned
309 * process... Whom would you report that to?
310 */
311 }
312
313 enum {
314 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
315 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
316 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
317 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
318 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
319 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
320 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
321 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
322 Opt_check_integrity, Opt_check_integrity_including_extent_data,
323 Opt_check_integrity_print_mask, Opt_fatal_errors,
324 Opt_commit_interval,
325 Opt_err,
326 };
327
328 static match_table_t tokens = {
329 {Opt_degraded, "degraded"},
330 {Opt_subvol, "subvol=%s"},
331 {Opt_subvolid, "subvolid=%s"},
332 {Opt_device, "device=%s"},
333 {Opt_nodatasum, "nodatasum"},
334 {Opt_nodatacow, "nodatacow"},
335 {Opt_nobarrier, "nobarrier"},
336 {Opt_max_inline, "max_inline=%s"},
337 {Opt_alloc_start, "alloc_start=%s"},
338 {Opt_thread_pool, "thread_pool=%d"},
339 {Opt_compress, "compress"},
340 {Opt_compress_type, "compress=%s"},
341 {Opt_compress_force, "compress-force"},
342 {Opt_compress_force_type, "compress-force=%s"},
343 {Opt_ssd, "ssd"},
344 {Opt_ssd_spread, "ssd_spread"},
345 {Opt_nossd, "nossd"},
346 {Opt_noacl, "noacl"},
347 {Opt_notreelog, "notreelog"},
348 {Opt_flushoncommit, "flushoncommit"},
349 {Opt_ratio, "metadata_ratio=%d"},
350 {Opt_discard, "discard"},
351 {Opt_space_cache, "space_cache"},
352 {Opt_clear_cache, "clear_cache"},
353 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
354 {Opt_enospc_debug, "enospc_debug"},
355 {Opt_subvolrootid, "subvolrootid=%d"},
356 {Opt_defrag, "autodefrag"},
357 {Opt_inode_cache, "inode_cache"},
358 {Opt_no_space_cache, "nospace_cache"},
359 {Opt_recovery, "recovery"},
360 {Opt_skip_balance, "skip_balance"},
361 {Opt_check_integrity, "check_int"},
362 {Opt_check_integrity_including_extent_data, "check_int_data"},
363 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
364 {Opt_fatal_errors, "fatal_errors=%s"},
365 {Opt_commit_interval, "commit=%d"},
366 {Opt_err, NULL},
367 };
368
369 /*
370 * Regular mount options parser. Everything that is needed only when
371 * reading in a new superblock is parsed here.
372 * XXX JDM: This needs to be cleaned up for remount.
373 */
374 int btrfs_parse_options(struct btrfs_root *root, char *options)
375 {
376 struct btrfs_fs_info *info = root->fs_info;
377 substring_t args[MAX_OPT_ARGS];
378 char *p, *num, *orig = NULL;
379 u64 cache_gen;
380 int intarg;
381 int ret = 0;
382 char *compress_type;
383 bool compress_force = false;
384
385 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
386 if (cache_gen)
387 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
388
389 if (!options)
390 goto out;
391
392 /*
393 * strsep changes the string, duplicate it because parse_options
394 * gets called twice
395 */
396 options = kstrdup(options, GFP_NOFS);
397 if (!options)
398 return -ENOMEM;
399
400 orig = options;
401
402 while ((p = strsep(&options, ",")) != NULL) {
403 int token;
404 if (!*p)
405 continue;
406
407 token = match_token(p, tokens, args);
408 switch (token) {
409 case Opt_degraded:
410 printk(KERN_INFO "btrfs: allowing degraded mounts\n");
411 btrfs_set_opt(info->mount_opt, DEGRADED);
412 break;
413 case Opt_subvol:
414 case Opt_subvolid:
415 case Opt_subvolrootid:
416 case Opt_device:
417 /*
418 * These are parsed by btrfs_parse_early_options
419 * and can be happily ignored here.
420 */
421 break;
422 case Opt_nodatasum:
423 printk(KERN_INFO "btrfs: setting nodatasum\n");
424 btrfs_set_opt(info->mount_opt, NODATASUM);
425 break;
426 case Opt_nodatacow:
427 if (!btrfs_test_opt(root, COMPRESS) ||
428 !btrfs_test_opt(root, FORCE_COMPRESS)) {
429 printk(KERN_INFO "btrfs: setting nodatacow, compression disabled\n");
430 } else {
431 printk(KERN_INFO "btrfs: setting nodatacow\n");
432 }
433 info->compress_type = BTRFS_COMPRESS_NONE;
434 btrfs_clear_opt(info->mount_opt, COMPRESS);
435 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
436 btrfs_set_opt(info->mount_opt, NODATACOW);
437 btrfs_set_opt(info->mount_opt, NODATASUM);
438 break;
439 case Opt_compress_force:
440 case Opt_compress_force_type:
441 compress_force = true;
442 /* Fallthrough */
443 case Opt_compress:
444 case Opt_compress_type:
445 if (token == Opt_compress ||
446 token == Opt_compress_force ||
447 strcmp(args[0].from, "zlib") == 0) {
448 compress_type = "zlib";
449 info->compress_type = BTRFS_COMPRESS_ZLIB;
450 btrfs_set_opt(info->mount_opt, COMPRESS);
451 btrfs_clear_opt(info->mount_opt, NODATACOW);
452 btrfs_clear_opt(info->mount_opt, NODATASUM);
453 } else if (strcmp(args[0].from, "lzo") == 0) {
454 compress_type = "lzo";
455 info->compress_type = BTRFS_COMPRESS_LZO;
456 btrfs_set_opt(info->mount_opt, COMPRESS);
457 btrfs_clear_opt(info->mount_opt, NODATACOW);
458 btrfs_clear_opt(info->mount_opt, NODATASUM);
459 btrfs_set_fs_incompat(info, COMPRESS_LZO);
460 } else if (strncmp(args[0].from, "no", 2) == 0) {
461 compress_type = "no";
462 info->compress_type = BTRFS_COMPRESS_NONE;
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
476 pr_info("btrfs: use %s compression\n",
477 compress_type);
478 break;
479 case Opt_ssd:
480 printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
481 btrfs_set_opt(info->mount_opt, SSD);
482 break;
483 case Opt_ssd_spread:
484 printk(KERN_INFO "btrfs: use spread ssd "
485 "allocation scheme\n");
486 btrfs_set_opt(info->mount_opt, SSD);
487 btrfs_set_opt(info->mount_opt, SSD_SPREAD);
488 break;
489 case Opt_nossd:
490 printk(KERN_INFO "btrfs: not using ssd allocation "
491 "scheme\n");
492 btrfs_set_opt(info->mount_opt, NOSSD);
493 btrfs_clear_opt(info->mount_opt, SSD);
494 btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
495 break;
496 case Opt_nobarrier:
497 printk(KERN_INFO "btrfs: turning off barriers\n");
498 btrfs_set_opt(info->mount_opt, NOBARRIER);
499 break;
500 case Opt_thread_pool:
501 ret = match_int(&args[0], &intarg);
502 if (ret) {
503 goto out;
504 } else if (intarg > 0) {
505 info->thread_pool_size = intarg;
506 } else {
507 ret = -EINVAL;
508 goto out;
509 }
510 break;
511 case Opt_max_inline:
512 num = match_strdup(&args[0]);
513 if (num) {
514 info->max_inline = memparse(num, NULL);
515 kfree(num);
516
517 if (info->max_inline) {
518 info->max_inline = max_t(u64,
519 info->max_inline,
520 root->sectorsize);
521 }
522 printk(KERN_INFO "btrfs: max_inline at %llu\n",
523 (unsigned long long)info->max_inline);
524 } else {
525 ret = -ENOMEM;
526 goto out;
527 }
528 break;
529 case Opt_alloc_start:
530 num = match_strdup(&args[0]);
531 if (num) {
532 mutex_lock(&info->chunk_mutex);
533 info->alloc_start = memparse(num, NULL);
534 mutex_unlock(&info->chunk_mutex);
535 kfree(num);
536 printk(KERN_INFO
537 "btrfs: allocations start at %llu\n",
538 (unsigned long long)info->alloc_start);
539 } else {
540 ret = -ENOMEM;
541 goto out;
542 }
543 break;
544 case Opt_noacl:
545 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
546 break;
547 case Opt_notreelog:
548 printk(KERN_INFO "btrfs: disabling tree log\n");
549 btrfs_set_opt(info->mount_opt, NOTREELOG);
550 break;
551 case Opt_flushoncommit:
552 printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
553 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
554 break;
555 case Opt_ratio:
556 ret = match_int(&args[0], &intarg);
557 if (ret) {
558 goto out;
559 } else if (intarg >= 0) {
560 info->metadata_ratio = intarg;
561 printk(KERN_INFO "btrfs: metadata ratio %d\n",
562 info->metadata_ratio);
563 } else {
564 ret = -EINVAL;
565 goto out;
566 }
567 break;
568 case Opt_discard:
569 btrfs_set_opt(info->mount_opt, DISCARD);
570 break;
571 case Opt_space_cache:
572 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
573 break;
574 case Opt_no_space_cache:
575 printk(KERN_INFO "btrfs: disabling disk space caching\n");
576 btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
577 break;
578 case Opt_inode_cache:
579 printk(KERN_INFO "btrfs: enabling inode map caching\n");
580 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
581 break;
582 case Opt_clear_cache:
583 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
584 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
585 break;
586 case Opt_user_subvol_rm_allowed:
587 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
588 break;
589 case Opt_enospc_debug:
590 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
591 break;
592 case Opt_defrag:
593 printk(KERN_INFO "btrfs: enabling auto defrag\n");
594 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
595 break;
596 case Opt_recovery:
597 printk(KERN_INFO "btrfs: enabling auto recovery\n");
598 btrfs_set_opt(info->mount_opt, RECOVERY);
599 break;
600 case Opt_skip_balance:
601 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
602 break;
603 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
604 case Opt_check_integrity_including_extent_data:
605 printk(KERN_INFO "btrfs: enabling check integrity"
606 " including extent data\n");
607 btrfs_set_opt(info->mount_opt,
608 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
609 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
610 break;
611 case Opt_check_integrity:
612 printk(KERN_INFO "btrfs: enabling check integrity\n");
613 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
614 break;
615 case Opt_check_integrity_print_mask:
616 ret = match_int(&args[0], &intarg);
617 if (ret) {
618 goto out;
619 } else if (intarg >= 0) {
620 info->check_integrity_print_mask = intarg;
621 printk(KERN_INFO "btrfs:"
622 " check_integrity_print_mask 0x%x\n",
623 info->check_integrity_print_mask);
624 } else {
625 ret = -EINVAL;
626 goto out;
627 }
628 break;
629 #else
630 case Opt_check_integrity_including_extent_data:
631 case Opt_check_integrity:
632 case Opt_check_integrity_print_mask:
633 printk(KERN_ERR "btrfs: support for check_integrity*"
634 " not compiled in!\n");
635 ret = -EINVAL;
636 goto out;
637 #endif
638 case Opt_fatal_errors:
639 if (strcmp(args[0].from, "panic") == 0)
640 btrfs_set_opt(info->mount_opt,
641 PANIC_ON_FATAL_ERROR);
642 else if (strcmp(args[0].from, "bug") == 0)
643 btrfs_clear_opt(info->mount_opt,
644 PANIC_ON_FATAL_ERROR);
645 else {
646 ret = -EINVAL;
647 goto out;
648 }
649 break;
650 case Opt_commit_interval:
651 intarg = 0;
652 ret = match_int(&args[0], &intarg);
653 if (ret < 0) {
654 printk(KERN_ERR
655 "btrfs: invalid commit interval\n");
656 ret = -EINVAL;
657 goto out;
658 }
659 if (intarg > 0) {
660 if (intarg > 300) {
661 printk(KERN_WARNING
662 "btrfs: excessive commit interval %d\n",
663 intarg);
664 }
665 info->commit_interval = intarg;
666 } else {
667 printk(KERN_INFO
668 "btrfs: using default commit interval %ds\n",
669 BTRFS_DEFAULT_COMMIT_INTERVAL);
670 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
671 }
672 break;
673 case Opt_err:
674 printk(KERN_INFO "btrfs: unrecognized mount option "
675 "'%s'\n", p);
676 ret = -EINVAL;
677 goto out;
678 default:
679 break;
680 }
681 }
682 out:
683 if (!ret && btrfs_test_opt(root, SPACE_CACHE))
684 printk(KERN_INFO "btrfs: disk space caching is enabled\n");
685 kfree(orig);
686 return ret;
687 }
688
689 /*
690 * Parse mount options that are required early in the mount process.
691 *
692 * All other options will be parsed on much later in the mount process and
693 * only when we need to allocate a new super block.
694 */
695 static int btrfs_parse_early_options(const char *options, fmode_t flags,
696 void *holder, char **subvol_name, u64 *subvol_objectid,
697 struct btrfs_fs_devices **fs_devices)
698 {
699 substring_t args[MAX_OPT_ARGS];
700 char *device_name, *opts, *orig, *p;
701 char *num = NULL;
702 int error = 0;
703
704 if (!options)
705 return 0;
706
707 /*
708 * strsep changes the string, duplicate it because parse_options
709 * gets called twice
710 */
711 opts = kstrdup(options, GFP_KERNEL);
712 if (!opts)
713 return -ENOMEM;
714 orig = opts;
715
716 while ((p = strsep(&opts, ",")) != NULL) {
717 int token;
718 if (!*p)
719 continue;
720
721 token = match_token(p, tokens, args);
722 switch (token) {
723 case Opt_subvol:
724 kfree(*subvol_name);
725 *subvol_name = match_strdup(&args[0]);
726 if (!*subvol_name) {
727 error = -ENOMEM;
728 goto out;
729 }
730 break;
731 case Opt_subvolid:
732 num = match_strdup(&args[0]);
733 if (num) {
734 *subvol_objectid = memparse(num, NULL);
735 kfree(num);
736 /* we want the original fs_tree */
737 if (!*subvol_objectid)
738 *subvol_objectid =
739 BTRFS_FS_TREE_OBJECTID;
740 } else {
741 error = -EINVAL;
742 goto out;
743 }
744 break;
745 case Opt_subvolrootid:
746 printk(KERN_WARNING
747 "btrfs: 'subvolrootid' mount option is deprecated and has no effect\n");
748 break;
749 case Opt_device:
750 device_name = match_strdup(&args[0]);
751 if (!device_name) {
752 error = -ENOMEM;
753 goto out;
754 }
755 error = btrfs_scan_one_device(device_name,
756 flags, holder, fs_devices);
757 kfree(device_name);
758 if (error)
759 goto out;
760 break;
761 default:
762 break;
763 }
764 }
765
766 out:
767 kfree(orig);
768 return error;
769 }
770
771 static struct dentry *get_default_root(struct super_block *sb,
772 u64 subvol_objectid)
773 {
774 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
775 struct btrfs_root *root = fs_info->tree_root;
776 struct btrfs_root *new_root;
777 struct btrfs_dir_item *di;
778 struct btrfs_path *path;
779 struct btrfs_key location;
780 struct inode *inode;
781 u64 dir_id;
782 int new = 0;
783
784 /*
785 * We have a specific subvol we want to mount, just setup location and
786 * go look up the root.
787 */
788 if (subvol_objectid) {
789 location.objectid = subvol_objectid;
790 location.type = BTRFS_ROOT_ITEM_KEY;
791 location.offset = (u64)-1;
792 goto find_root;
793 }
794
795 path = btrfs_alloc_path();
796 if (!path)
797 return ERR_PTR(-ENOMEM);
798 path->leave_spinning = 1;
799
800 /*
801 * Find the "default" dir item which points to the root item that we
802 * will mount by default if we haven't been given a specific subvolume
803 * to mount.
804 */
805 dir_id = btrfs_super_root_dir(fs_info->super_copy);
806 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
807 if (IS_ERR(di)) {
808 btrfs_free_path(path);
809 return ERR_CAST(di);
810 }
811 if (!di) {
812 /*
813 * Ok the default dir item isn't there. This is weird since
814 * it's always been there, but don't freak out, just try and
815 * mount to root most subvolume.
816 */
817 btrfs_free_path(path);
818 dir_id = BTRFS_FIRST_FREE_OBJECTID;
819 new_root = fs_info->fs_root;
820 goto setup_root;
821 }
822
823 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
824 btrfs_free_path(path);
825
826 find_root:
827 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
828 if (IS_ERR(new_root))
829 return ERR_CAST(new_root);
830
831 dir_id = btrfs_root_dirid(&new_root->root_item);
832 setup_root:
833 location.objectid = dir_id;
834 location.type = BTRFS_INODE_ITEM_KEY;
835 location.offset = 0;
836
837 inode = btrfs_iget(sb, &location, new_root, &new);
838 if (IS_ERR(inode))
839 return ERR_CAST(inode);
840
841 /*
842 * If we're just mounting the root most subvol put the inode and return
843 * a reference to the dentry. We will have already gotten a reference
844 * to the inode in btrfs_fill_super so we're good to go.
845 */
846 if (!new && sb->s_root->d_inode == inode) {
847 iput(inode);
848 return dget(sb->s_root);
849 }
850
851 return d_obtain_alias(inode);
852 }
853
854 static int btrfs_fill_super(struct super_block *sb,
855 struct btrfs_fs_devices *fs_devices,
856 void *data, int silent)
857 {
858 struct inode *inode;
859 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
860 struct btrfs_key key;
861 int err;
862
863 sb->s_maxbytes = MAX_LFS_FILESIZE;
864 sb->s_magic = BTRFS_SUPER_MAGIC;
865 sb->s_op = &btrfs_super_ops;
866 sb->s_d_op = &btrfs_dentry_operations;
867 sb->s_export_op = &btrfs_export_ops;
868 sb->s_xattr = btrfs_xattr_handlers;
869 sb->s_time_gran = 1;
870 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
871 sb->s_flags |= MS_POSIXACL;
872 #endif
873 sb->s_flags |= MS_I_VERSION;
874 err = open_ctree(sb, fs_devices, (char *)data);
875 if (err) {
876 printk("btrfs: open_ctree failed\n");
877 return err;
878 }
879
880 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
881 key.type = BTRFS_INODE_ITEM_KEY;
882 key.offset = 0;
883 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
884 if (IS_ERR(inode)) {
885 err = PTR_ERR(inode);
886 goto fail_close;
887 }
888
889 sb->s_root = d_make_root(inode);
890 if (!sb->s_root) {
891 err = -ENOMEM;
892 goto fail_close;
893 }
894
895 save_mount_options(sb, data);
896 cleancache_init_fs(sb);
897 sb->s_flags |= MS_ACTIVE;
898 return 0;
899
900 fail_close:
901 close_ctree(fs_info->tree_root);
902 return err;
903 }
904
905 int btrfs_sync_fs(struct super_block *sb, int wait)
906 {
907 struct btrfs_trans_handle *trans;
908 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
909 struct btrfs_root *root = fs_info->tree_root;
910
911 trace_btrfs_sync_fs(wait);
912
913 if (!wait) {
914 filemap_flush(fs_info->btree_inode->i_mapping);
915 return 0;
916 }
917
918 btrfs_wait_all_ordered_extents(fs_info, 1);
919
920 trans = btrfs_attach_transaction_barrier(root);
921 if (IS_ERR(trans)) {
922 /* no transaction, don't bother */
923 if (PTR_ERR(trans) == -ENOENT)
924 return 0;
925 return PTR_ERR(trans);
926 }
927 return btrfs_commit_transaction(trans, root);
928 }
929
930 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
931 {
932 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
933 struct btrfs_root *root = info->tree_root;
934 char *compress_type;
935
936 if (btrfs_test_opt(root, DEGRADED))
937 seq_puts(seq, ",degraded");
938 if (btrfs_test_opt(root, NODATASUM))
939 seq_puts(seq, ",nodatasum");
940 if (btrfs_test_opt(root, NODATACOW))
941 seq_puts(seq, ",nodatacow");
942 if (btrfs_test_opt(root, NOBARRIER))
943 seq_puts(seq, ",nobarrier");
944 if (info->max_inline != 8192 * 1024)
945 seq_printf(seq, ",max_inline=%llu",
946 (unsigned long long)info->max_inline);
947 if (info->alloc_start != 0)
948 seq_printf(seq, ",alloc_start=%llu",
949 (unsigned long long)info->alloc_start);
950 if (info->thread_pool_size != min_t(unsigned long,
951 num_online_cpus() + 2, 8))
952 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
953 if (btrfs_test_opt(root, COMPRESS)) {
954 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
955 compress_type = "zlib";
956 else
957 compress_type = "lzo";
958 if (btrfs_test_opt(root, FORCE_COMPRESS))
959 seq_printf(seq, ",compress-force=%s", compress_type);
960 else
961 seq_printf(seq, ",compress=%s", compress_type);
962 }
963 if (btrfs_test_opt(root, NOSSD))
964 seq_puts(seq, ",nossd");
965 if (btrfs_test_opt(root, SSD_SPREAD))
966 seq_puts(seq, ",ssd_spread");
967 else if (btrfs_test_opt(root, SSD))
968 seq_puts(seq, ",ssd");
969 if (btrfs_test_opt(root, NOTREELOG))
970 seq_puts(seq, ",notreelog");
971 if (btrfs_test_opt(root, FLUSHONCOMMIT))
972 seq_puts(seq, ",flushoncommit");
973 if (btrfs_test_opt(root, DISCARD))
974 seq_puts(seq, ",discard");
975 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
976 seq_puts(seq, ",noacl");
977 if (btrfs_test_opt(root, SPACE_CACHE))
978 seq_puts(seq, ",space_cache");
979 else
980 seq_puts(seq, ",nospace_cache");
981 if (btrfs_test_opt(root, CLEAR_CACHE))
982 seq_puts(seq, ",clear_cache");
983 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
984 seq_puts(seq, ",user_subvol_rm_allowed");
985 if (btrfs_test_opt(root, ENOSPC_DEBUG))
986 seq_puts(seq, ",enospc_debug");
987 if (btrfs_test_opt(root, AUTO_DEFRAG))
988 seq_puts(seq, ",autodefrag");
989 if (btrfs_test_opt(root, INODE_MAP_CACHE))
990 seq_puts(seq, ",inode_cache");
991 if (btrfs_test_opt(root, SKIP_BALANCE))
992 seq_puts(seq, ",skip_balance");
993 if (btrfs_test_opt(root, RECOVERY))
994 seq_puts(seq, ",recovery");
995 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
996 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
997 seq_puts(seq, ",check_int_data");
998 else if (btrfs_test_opt(root, CHECK_INTEGRITY))
999 seq_puts(seq, ",check_int");
1000 if (info->check_integrity_print_mask)
1001 seq_printf(seq, ",check_int_print_mask=%d",
1002 info->check_integrity_print_mask);
1003 #endif
1004 if (info->metadata_ratio)
1005 seq_printf(seq, ",metadata_ratio=%d",
1006 info->metadata_ratio);
1007 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1008 seq_puts(seq, ",fatal_errors=panic");
1009 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1010 seq_printf(seq, ",commit=%d", info->commit_interval);
1011 return 0;
1012 }
1013
1014 static int btrfs_test_super(struct super_block *s, void *data)
1015 {
1016 struct btrfs_fs_info *p = data;
1017 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1018
1019 return fs_info->fs_devices == p->fs_devices;
1020 }
1021
1022 static int btrfs_set_super(struct super_block *s, void *data)
1023 {
1024 int err = set_anon_super(s, data);
1025 if (!err)
1026 s->s_fs_info = data;
1027 return err;
1028 }
1029
1030 /*
1031 * subvolumes are identified by ino 256
1032 */
1033 static inline int is_subvolume_inode(struct inode *inode)
1034 {
1035 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1036 return 1;
1037 return 0;
1038 }
1039
1040 /*
1041 * This will strip out the subvol=%s argument for an argument string and add
1042 * subvolid=0 to make sure we get the actual tree root for path walking to the
1043 * subvol we want.
1044 */
1045 static char *setup_root_args(char *args)
1046 {
1047 unsigned len = strlen(args) + 2 + 1;
1048 char *src, *dst, *buf;
1049
1050 /*
1051 * We need the same args as before, but with this substitution:
1052 * s!subvol=[^,]+!subvolid=0!
1053 *
1054 * Since the replacement string is up to 2 bytes longer than the
1055 * original, allocate strlen(args) + 2 + 1 bytes.
1056 */
1057
1058 src = strstr(args, "subvol=");
1059 /* This shouldn't happen, but just in case.. */
1060 if (!src)
1061 return NULL;
1062
1063 buf = dst = kmalloc(len, GFP_NOFS);
1064 if (!buf)
1065 return NULL;
1066
1067 /*
1068 * If the subvol= arg is not at the start of the string,
1069 * copy whatever precedes it into buf.
1070 */
1071 if (src != args) {
1072 *src++ = '\0';
1073 strcpy(buf, args);
1074 dst += strlen(args);
1075 }
1076
1077 strcpy(dst, "subvolid=0");
1078 dst += strlen("subvolid=0");
1079
1080 /*
1081 * If there is a "," after the original subvol=... string,
1082 * copy that suffix into our buffer. Otherwise, we're done.
1083 */
1084 src = strchr(src, ',');
1085 if (src)
1086 strcpy(dst, src);
1087
1088 return buf;
1089 }
1090
1091 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1092 const char *device_name, char *data)
1093 {
1094 struct dentry *root;
1095 struct vfsmount *mnt;
1096 char *newargs;
1097
1098 newargs = setup_root_args(data);
1099 if (!newargs)
1100 return ERR_PTR(-ENOMEM);
1101 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1102 newargs);
1103 kfree(newargs);
1104 if (IS_ERR(mnt))
1105 return ERR_CAST(mnt);
1106
1107 root = mount_subtree(mnt, subvol_name);
1108
1109 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1110 struct super_block *s = root->d_sb;
1111 dput(root);
1112 root = ERR_PTR(-EINVAL);
1113 deactivate_locked_super(s);
1114 printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1115 subvol_name);
1116 }
1117
1118 return root;
1119 }
1120
1121 /*
1122 * Find a superblock for the given device / mount point.
1123 *
1124 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1125 * for multiple device setup. Make sure to keep it in sync.
1126 */
1127 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1128 const char *device_name, void *data)
1129 {
1130 struct block_device *bdev = NULL;
1131 struct super_block *s;
1132 struct dentry *root;
1133 struct btrfs_fs_devices *fs_devices = NULL;
1134 struct btrfs_fs_info *fs_info = NULL;
1135 fmode_t mode = FMODE_READ;
1136 char *subvol_name = NULL;
1137 u64 subvol_objectid = 0;
1138 int error = 0;
1139
1140 if (!(flags & MS_RDONLY))
1141 mode |= FMODE_WRITE;
1142
1143 error = btrfs_parse_early_options(data, mode, fs_type,
1144 &subvol_name, &subvol_objectid,
1145 &fs_devices);
1146 if (error) {
1147 kfree(subvol_name);
1148 return ERR_PTR(error);
1149 }
1150
1151 if (subvol_name) {
1152 root = mount_subvol(subvol_name, flags, device_name, data);
1153 kfree(subvol_name);
1154 return root;
1155 }
1156
1157 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1158 if (error)
1159 return ERR_PTR(error);
1160
1161 /*
1162 * Setup a dummy root and fs_info for test/set super. This is because
1163 * we don't actually fill this stuff out until open_ctree, but we need
1164 * it for searching for existing supers, so this lets us do that and
1165 * then open_ctree will properly initialize everything later.
1166 */
1167 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1168 if (!fs_info)
1169 return ERR_PTR(-ENOMEM);
1170
1171 fs_info->fs_devices = fs_devices;
1172
1173 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1174 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1175 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1176 error = -ENOMEM;
1177 goto error_fs_info;
1178 }
1179
1180 error = btrfs_open_devices(fs_devices, mode, fs_type);
1181 if (error)
1182 goto error_fs_info;
1183
1184 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1185 error = -EACCES;
1186 goto error_close_devices;
1187 }
1188
1189 bdev = fs_devices->latest_bdev;
1190 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1191 fs_info);
1192 if (IS_ERR(s)) {
1193 error = PTR_ERR(s);
1194 goto error_close_devices;
1195 }
1196
1197 if (s->s_root) {
1198 btrfs_close_devices(fs_devices);
1199 free_fs_info(fs_info);
1200 if ((flags ^ s->s_flags) & MS_RDONLY)
1201 error = -EBUSY;
1202 } else {
1203 char b[BDEVNAME_SIZE];
1204
1205 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1206 btrfs_sb(s)->bdev_holder = fs_type;
1207 error = btrfs_fill_super(s, fs_devices, data,
1208 flags & MS_SILENT ? 1 : 0);
1209 }
1210
1211 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1212 if (IS_ERR(root))
1213 deactivate_locked_super(s);
1214
1215 return root;
1216
1217 error_close_devices:
1218 btrfs_close_devices(fs_devices);
1219 error_fs_info:
1220 free_fs_info(fs_info);
1221 return ERR_PTR(error);
1222 }
1223
1224 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1225 {
1226 spin_lock_irq(&workers->lock);
1227 workers->max_workers = new_limit;
1228 spin_unlock_irq(&workers->lock);
1229 }
1230
1231 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1232 int new_pool_size, int old_pool_size)
1233 {
1234 if (new_pool_size == old_pool_size)
1235 return;
1236
1237 fs_info->thread_pool_size = new_pool_size;
1238
1239 printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1240 old_pool_size, new_pool_size);
1241
1242 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1243 btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1244 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1245 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1246 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1247 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1248 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1249 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1250 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1251 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1252 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1253 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1254 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1255 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers,
1256 new_pool_size);
1257 }
1258
1259 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1260 {
1261 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1262 }
1263
1264 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1265 unsigned long old_opts, int flags)
1266 {
1267 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1268 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1269 (flags & MS_RDONLY))) {
1270 /* wait for any defraggers to finish */
1271 wait_event(fs_info->transaction_wait,
1272 (atomic_read(&fs_info->defrag_running) == 0));
1273 if (flags & MS_RDONLY)
1274 sync_filesystem(fs_info->sb);
1275 }
1276 }
1277
1278 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1279 unsigned long old_opts)
1280 {
1281 /*
1282 * We need cleanup all defragable inodes if the autodefragment is
1283 * close or the fs is R/O.
1284 */
1285 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1286 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1287 (fs_info->sb->s_flags & MS_RDONLY))) {
1288 btrfs_cleanup_defrag_inodes(fs_info);
1289 }
1290
1291 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1292 }
1293
1294 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1295 {
1296 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1297 struct btrfs_root *root = fs_info->tree_root;
1298 unsigned old_flags = sb->s_flags;
1299 unsigned long old_opts = fs_info->mount_opt;
1300 unsigned long old_compress_type = fs_info->compress_type;
1301 u64 old_max_inline = fs_info->max_inline;
1302 u64 old_alloc_start = fs_info->alloc_start;
1303 int old_thread_pool_size = fs_info->thread_pool_size;
1304 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1305 int ret;
1306
1307 btrfs_remount_prepare(fs_info);
1308
1309 ret = btrfs_parse_options(root, data);
1310 if (ret) {
1311 ret = -EINVAL;
1312 goto restore;
1313 }
1314
1315 btrfs_remount_begin(fs_info, old_opts, *flags);
1316 btrfs_resize_thread_pool(fs_info,
1317 fs_info->thread_pool_size, old_thread_pool_size);
1318
1319 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1320 goto out;
1321
1322 if (*flags & MS_RDONLY) {
1323 /*
1324 * this also happens on 'umount -rf' or on shutdown, when
1325 * the filesystem is busy.
1326 */
1327 sb->s_flags |= MS_RDONLY;
1328
1329 btrfs_dev_replace_suspend_for_unmount(fs_info);
1330 btrfs_scrub_cancel(fs_info);
1331 btrfs_pause_balance(fs_info);
1332
1333 ret = btrfs_commit_super(root);
1334 if (ret)
1335 goto restore;
1336 } else {
1337 if (fs_info->fs_devices->rw_devices == 0) {
1338 ret = -EACCES;
1339 goto restore;
1340 }
1341
1342 if (fs_info->fs_devices->missing_devices >
1343 fs_info->num_tolerated_disk_barrier_failures &&
1344 !(*flags & MS_RDONLY)) {
1345 printk(KERN_WARNING
1346 "Btrfs: too many missing devices, writeable remount is not allowed\n");
1347 ret = -EACCES;
1348 goto restore;
1349 }
1350
1351 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1352 ret = -EINVAL;
1353 goto restore;
1354 }
1355
1356 ret = btrfs_cleanup_fs_roots(fs_info);
1357 if (ret)
1358 goto restore;
1359
1360 /* recover relocation */
1361 ret = btrfs_recover_relocation(root);
1362 if (ret)
1363 goto restore;
1364
1365 ret = btrfs_resume_balance_async(fs_info);
1366 if (ret)
1367 goto restore;
1368
1369 ret = btrfs_resume_dev_replace_async(fs_info);
1370 if (ret) {
1371 pr_warn("btrfs: failed to resume dev_replace\n");
1372 goto restore;
1373 }
1374 sb->s_flags &= ~MS_RDONLY;
1375 }
1376 out:
1377 btrfs_remount_cleanup(fs_info, old_opts);
1378 return 0;
1379
1380 restore:
1381 /* We've hit an error - don't reset MS_RDONLY */
1382 if (sb->s_flags & MS_RDONLY)
1383 old_flags |= MS_RDONLY;
1384 sb->s_flags = old_flags;
1385 fs_info->mount_opt = old_opts;
1386 fs_info->compress_type = old_compress_type;
1387 fs_info->max_inline = old_max_inline;
1388 mutex_lock(&fs_info->chunk_mutex);
1389 fs_info->alloc_start = old_alloc_start;
1390 mutex_unlock(&fs_info->chunk_mutex);
1391 btrfs_resize_thread_pool(fs_info,
1392 old_thread_pool_size, fs_info->thread_pool_size);
1393 fs_info->metadata_ratio = old_metadata_ratio;
1394 btrfs_remount_cleanup(fs_info, old_opts);
1395 return ret;
1396 }
1397
1398 /* Used to sort the devices by max_avail(descending sort) */
1399 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1400 const void *dev_info2)
1401 {
1402 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1403 ((struct btrfs_device_info *)dev_info2)->max_avail)
1404 return -1;
1405 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1406 ((struct btrfs_device_info *)dev_info2)->max_avail)
1407 return 1;
1408 else
1409 return 0;
1410 }
1411
1412 /*
1413 * sort the devices by max_avail, in which max free extent size of each device
1414 * is stored.(Descending Sort)
1415 */
1416 static inline void btrfs_descending_sort_devices(
1417 struct btrfs_device_info *devices,
1418 size_t nr_devices)
1419 {
1420 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1421 btrfs_cmp_device_free_bytes, NULL);
1422 }
1423
1424 /*
1425 * The helper to calc the free space on the devices that can be used to store
1426 * file data.
1427 */
1428 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1429 {
1430 struct btrfs_fs_info *fs_info = root->fs_info;
1431 struct btrfs_device_info *devices_info;
1432 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1433 struct btrfs_device *device;
1434 u64 skip_space;
1435 u64 type;
1436 u64 avail_space;
1437 u64 used_space;
1438 u64 min_stripe_size;
1439 int min_stripes = 1, num_stripes = 1;
1440 int i = 0, nr_devices;
1441 int ret;
1442
1443 nr_devices = fs_info->fs_devices->open_devices;
1444 BUG_ON(!nr_devices);
1445
1446 devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1447 GFP_NOFS);
1448 if (!devices_info)
1449 return -ENOMEM;
1450
1451 /* calc min stripe number for data space alloction */
1452 type = btrfs_get_alloc_profile(root, 1);
1453 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1454 min_stripes = 2;
1455 num_stripes = nr_devices;
1456 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1457 min_stripes = 2;
1458 num_stripes = 2;
1459 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1460 min_stripes = 4;
1461 num_stripes = 4;
1462 }
1463
1464 if (type & BTRFS_BLOCK_GROUP_DUP)
1465 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1466 else
1467 min_stripe_size = BTRFS_STRIPE_LEN;
1468
1469 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1470 if (!device->in_fs_metadata || !device->bdev ||
1471 device->is_tgtdev_for_dev_replace)
1472 continue;
1473
1474 avail_space = device->total_bytes - device->bytes_used;
1475
1476 /* align with stripe_len */
1477 do_div(avail_space, BTRFS_STRIPE_LEN);
1478 avail_space *= BTRFS_STRIPE_LEN;
1479
1480 /*
1481 * In order to avoid overwritting the superblock on the drive,
1482 * btrfs starts at an offset of at least 1MB when doing chunk
1483 * allocation.
1484 */
1485 skip_space = 1024 * 1024;
1486
1487 /* user can set the offset in fs_info->alloc_start. */
1488 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1489 device->total_bytes)
1490 skip_space = max(fs_info->alloc_start, skip_space);
1491
1492 /*
1493 * btrfs can not use the free space in [0, skip_space - 1],
1494 * we must subtract it from the total. In order to implement
1495 * it, we account the used space in this range first.
1496 */
1497 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1498 &used_space);
1499 if (ret) {
1500 kfree(devices_info);
1501 return ret;
1502 }
1503
1504 /* calc the free space in [0, skip_space - 1] */
1505 skip_space -= used_space;
1506
1507 /*
1508 * we can use the free space in [0, skip_space - 1], subtract
1509 * it from the total.
1510 */
1511 if (avail_space && avail_space >= skip_space)
1512 avail_space -= skip_space;
1513 else
1514 avail_space = 0;
1515
1516 if (avail_space < min_stripe_size)
1517 continue;
1518
1519 devices_info[i].dev = device;
1520 devices_info[i].max_avail = avail_space;
1521
1522 i++;
1523 }
1524
1525 nr_devices = i;
1526
1527 btrfs_descending_sort_devices(devices_info, nr_devices);
1528
1529 i = nr_devices - 1;
1530 avail_space = 0;
1531 while (nr_devices >= min_stripes) {
1532 if (num_stripes > nr_devices)
1533 num_stripes = nr_devices;
1534
1535 if (devices_info[i].max_avail >= min_stripe_size) {
1536 int j;
1537 u64 alloc_size;
1538
1539 avail_space += devices_info[i].max_avail * num_stripes;
1540 alloc_size = devices_info[i].max_avail;
1541 for (j = i + 1 - num_stripes; j <= i; j++)
1542 devices_info[j].max_avail -= alloc_size;
1543 }
1544 i--;
1545 nr_devices--;
1546 }
1547
1548 kfree(devices_info);
1549 *free_bytes = avail_space;
1550 return 0;
1551 }
1552
1553 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1554 {
1555 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1556 struct btrfs_super_block *disk_super = fs_info->super_copy;
1557 struct list_head *head = &fs_info->space_info;
1558 struct btrfs_space_info *found;
1559 u64 total_used = 0;
1560 u64 total_free_data = 0;
1561 int bits = dentry->d_sb->s_blocksize_bits;
1562 __be32 *fsid = (__be32 *)fs_info->fsid;
1563 int ret;
1564
1565 /* holding chunk_muext to avoid allocating new chunks */
1566 mutex_lock(&fs_info->chunk_mutex);
1567 rcu_read_lock();
1568 list_for_each_entry_rcu(found, head, list) {
1569 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1570 total_free_data += found->disk_total - found->disk_used;
1571 total_free_data -=
1572 btrfs_account_ro_block_groups_free_space(found);
1573 }
1574
1575 total_used += found->disk_used;
1576 }
1577 rcu_read_unlock();
1578
1579 buf->f_namelen = BTRFS_NAME_LEN;
1580 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1581 buf->f_bfree = buf->f_blocks - (total_used >> bits);
1582 buf->f_bsize = dentry->d_sb->s_blocksize;
1583 buf->f_type = BTRFS_SUPER_MAGIC;
1584 buf->f_bavail = total_free_data;
1585 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1586 if (ret) {
1587 mutex_unlock(&fs_info->chunk_mutex);
1588 return ret;
1589 }
1590 buf->f_bavail += total_free_data;
1591 buf->f_bavail = buf->f_bavail >> bits;
1592 mutex_unlock(&fs_info->chunk_mutex);
1593
1594 /* We treat it as constant endianness (it doesn't matter _which_)
1595 because we want the fsid to come out the same whether mounted
1596 on a big-endian or little-endian host */
1597 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1598 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1599 /* Mask in the root object ID too, to disambiguate subvols */
1600 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1601 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1602
1603 return 0;
1604 }
1605
1606 static void btrfs_kill_super(struct super_block *sb)
1607 {
1608 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1609 kill_anon_super(sb);
1610 free_fs_info(fs_info);
1611 }
1612
1613 static struct file_system_type btrfs_fs_type = {
1614 .owner = THIS_MODULE,
1615 .name = "btrfs",
1616 .mount = btrfs_mount,
1617 .kill_sb = btrfs_kill_super,
1618 .fs_flags = FS_REQUIRES_DEV,
1619 };
1620 MODULE_ALIAS_FS("btrfs");
1621
1622 /*
1623 * used by btrfsctl to scan devices when no FS is mounted
1624 */
1625 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1626 unsigned long arg)
1627 {
1628 struct btrfs_ioctl_vol_args *vol;
1629 struct btrfs_fs_devices *fs_devices;
1630 int ret = -ENOTTY;
1631
1632 if (!capable(CAP_SYS_ADMIN))
1633 return -EPERM;
1634
1635 vol = memdup_user((void __user *)arg, sizeof(*vol));
1636 if (IS_ERR(vol))
1637 return PTR_ERR(vol);
1638
1639 switch (cmd) {
1640 case BTRFS_IOC_SCAN_DEV:
1641 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1642 &btrfs_fs_type, &fs_devices);
1643 break;
1644 case BTRFS_IOC_DEVICES_READY:
1645 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1646 &btrfs_fs_type, &fs_devices);
1647 if (ret)
1648 break;
1649 ret = !(fs_devices->num_devices == fs_devices->total_devices);
1650 break;
1651 }
1652
1653 kfree(vol);
1654 return ret;
1655 }
1656
1657 static int btrfs_freeze(struct super_block *sb)
1658 {
1659 struct btrfs_trans_handle *trans;
1660 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1661
1662 trans = btrfs_attach_transaction_barrier(root);
1663 if (IS_ERR(trans)) {
1664 /* no transaction, don't bother */
1665 if (PTR_ERR(trans) == -ENOENT)
1666 return 0;
1667 return PTR_ERR(trans);
1668 }
1669 return btrfs_commit_transaction(trans, root);
1670 }
1671
1672 static int btrfs_unfreeze(struct super_block *sb)
1673 {
1674 return 0;
1675 }
1676
1677 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1678 {
1679 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1680 struct btrfs_fs_devices *cur_devices;
1681 struct btrfs_device *dev, *first_dev = NULL;
1682 struct list_head *head;
1683 struct rcu_string *name;
1684
1685 mutex_lock(&fs_info->fs_devices->device_list_mutex);
1686 cur_devices = fs_info->fs_devices;
1687 while (cur_devices) {
1688 head = &cur_devices->devices;
1689 list_for_each_entry(dev, head, dev_list) {
1690 if (dev->missing)
1691 continue;
1692 if (!first_dev || dev->devid < first_dev->devid)
1693 first_dev = dev;
1694 }
1695 cur_devices = cur_devices->seed;
1696 }
1697
1698 if (first_dev) {
1699 rcu_read_lock();
1700 name = rcu_dereference(first_dev->name);
1701 seq_escape(m, name->str, " \t\n\\");
1702 rcu_read_unlock();
1703 } else {
1704 WARN_ON(1);
1705 }
1706 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1707 return 0;
1708 }
1709
1710 static const struct super_operations btrfs_super_ops = {
1711 .drop_inode = btrfs_drop_inode,
1712 .evict_inode = btrfs_evict_inode,
1713 .put_super = btrfs_put_super,
1714 .sync_fs = btrfs_sync_fs,
1715 .show_options = btrfs_show_options,
1716 .show_devname = btrfs_show_devname,
1717 .write_inode = btrfs_write_inode,
1718 .alloc_inode = btrfs_alloc_inode,
1719 .destroy_inode = btrfs_destroy_inode,
1720 .statfs = btrfs_statfs,
1721 .remount_fs = btrfs_remount,
1722 .freeze_fs = btrfs_freeze,
1723 .unfreeze_fs = btrfs_unfreeze,
1724 };
1725
1726 static const struct file_operations btrfs_ctl_fops = {
1727 .unlocked_ioctl = btrfs_control_ioctl,
1728 .compat_ioctl = btrfs_control_ioctl,
1729 .owner = THIS_MODULE,
1730 .llseek = noop_llseek,
1731 };
1732
1733 static struct miscdevice btrfs_misc = {
1734 .minor = BTRFS_MINOR,
1735 .name = "btrfs-control",
1736 .fops = &btrfs_ctl_fops
1737 };
1738
1739 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1740 MODULE_ALIAS("devname:btrfs-control");
1741
1742 static int btrfs_interface_init(void)
1743 {
1744 return misc_register(&btrfs_misc);
1745 }
1746
1747 static void btrfs_interface_exit(void)
1748 {
1749 if (misc_deregister(&btrfs_misc) < 0)
1750 printk(KERN_INFO "btrfs: misc_deregister failed for control device\n");
1751 }
1752
1753 static void btrfs_print_info(void)
1754 {
1755 printk(KERN_INFO "Btrfs loaded"
1756 #ifdef CONFIG_BTRFS_DEBUG
1757 ", debug=on"
1758 #endif
1759 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1760 ", integrity-checker=on"
1761 #endif
1762 "\n");
1763 }
1764
1765 static int __init init_btrfs_fs(void)
1766 {
1767 int err;
1768
1769 err = btrfs_init_sysfs();
1770 if (err)
1771 return err;
1772
1773 btrfs_init_compress();
1774
1775 err = btrfs_init_cachep();
1776 if (err)
1777 goto free_compress;
1778
1779 err = extent_io_init();
1780 if (err)
1781 goto free_cachep;
1782
1783 err = extent_map_init();
1784 if (err)
1785 goto free_extent_io;
1786
1787 err = ordered_data_init();
1788 if (err)
1789 goto free_extent_map;
1790
1791 err = btrfs_delayed_inode_init();
1792 if (err)
1793 goto free_ordered_data;
1794
1795 err = btrfs_auto_defrag_init();
1796 if (err)
1797 goto free_delayed_inode;
1798
1799 err = btrfs_delayed_ref_init();
1800 if (err)
1801 goto free_auto_defrag;
1802
1803 err = btrfs_interface_init();
1804 if (err)
1805 goto free_delayed_ref;
1806
1807 err = register_filesystem(&btrfs_fs_type);
1808 if (err)
1809 goto unregister_ioctl;
1810
1811 btrfs_init_lockdep();
1812
1813 btrfs_print_info();
1814 btrfs_test_free_space_cache();
1815
1816 return 0;
1817
1818 unregister_ioctl:
1819 btrfs_interface_exit();
1820 free_delayed_ref:
1821 btrfs_delayed_ref_exit();
1822 free_auto_defrag:
1823 btrfs_auto_defrag_exit();
1824 free_delayed_inode:
1825 btrfs_delayed_inode_exit();
1826 free_ordered_data:
1827 ordered_data_exit();
1828 free_extent_map:
1829 extent_map_exit();
1830 free_extent_io:
1831 extent_io_exit();
1832 free_cachep:
1833 btrfs_destroy_cachep();
1834 free_compress:
1835 btrfs_exit_compress();
1836 btrfs_exit_sysfs();
1837 return err;
1838 }
1839
1840 static void __exit exit_btrfs_fs(void)
1841 {
1842 btrfs_destroy_cachep();
1843 btrfs_delayed_ref_exit();
1844 btrfs_auto_defrag_exit();
1845 btrfs_delayed_inode_exit();
1846 ordered_data_exit();
1847 extent_map_exit();
1848 extent_io_exit();
1849 btrfs_interface_exit();
1850 unregister_filesystem(&btrfs_fs_type);
1851 btrfs_exit_sysfs();
1852 btrfs_cleanup_fs_uuids();
1853 btrfs_exit_compress();
1854 }
1855
1856 module_init(init_btrfs_fs)
1857 module_exit(exit_btrfs_fs)
1858
1859 MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.110168 seconds and 5 git commands to generate.