?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'regmap-fix-v4.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux/fpc-iii.git] / fs / btrfs / super.c
blob74ed5aae6cea393755ee6968de0a1df20f689036
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 "hash.h"
52 #include "props.h"
53 #include "xattr.h"
54 #include "volumes.h"
55 #include "export.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
60 #include "backref.h"
61 #include "tests/btrfs-tests.h"
62
63 #include "qgroup.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71
72 const char *btrfs_decode_error(int errno)
73 {
74 char *errstr = "unknown";
75
76 switch (errno) {
77 case -EIO:
78 errstr = "IO failure";
79 break;
80 case -ENOMEM:
81 errstr = "Out of memory";
82 break;
83 case -EROFS:
84 errstr = "Readonly filesystem";
85 break;
86 case -EEXIST:
87 errstr = "Object already exists";
88 break;
89 case -ENOSPC:
90 errstr = "No space left";
91 break;
92 case -ENOENT:
93 errstr = "No such entry";
94 break;
95 }
96
97 return errstr;
98 }
99
100 /* btrfs handle error by forcing the filesystem readonly */
101 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
102 {
103 struct super_block *sb = fs_info->sb;
104
105 if (sb->s_flags & MS_RDONLY)
106 return;
107
108 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
109 sb->s_flags |= MS_RDONLY;
110 btrfs_info(fs_info, "forced readonly");
111 /*
112 * Note that a running device replace operation is not
113 * canceled here although there is no way to update
114 * the progress. It would add the risk of a deadlock,
115 * therefore the canceling is omitted. The only penalty
116 * is that some I/O remains active until the procedure
117 * completes. The next time when the filesystem is
118 * mounted writeable again, the device replace
119 * operation continues.
120 */
121 }
122 }
123
124 /*
125 * __btrfs_handle_fs_error decodes expected errors from the caller and
126 * invokes the approciate error response.
127 */
128 __cold
129 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
130 unsigned int line, int errno, const char *fmt, ...)
131 {
132 struct super_block *sb = fs_info->sb;
133 #ifdef CONFIG_PRINTK
134 const char *errstr;
135 #endif
136
137 /*
138 * Special case: if the error is EROFS, and we're already
139 * under MS_RDONLY, then it is safe here.
140 */
141 if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
142 return;
143
144 #ifdef CONFIG_PRINTK
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 pr_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 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
159 sb->s_id, function, line, errno, errstr);
160 }
161 #endif
162
163 /*
164 * Today we only save the error info to memory. Long term we'll
165 * also send it down to the disk
166 */
167 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
168
169 /* Don't go through full error handling during mount */
170 if (sb->s_flags & MS_BORN)
171 btrfs_handle_error(fs_info);
172 }
173
174 #ifdef CONFIG_PRINTK
175 static const char * const logtypes[] = {
176 "emergency",
177 "alert",
178 "critical",
179 "error",
180 "warning",
181 "notice",
182 "info",
183 "debug",
184 };
185
186
187 /*
188 * Use one ratelimit state per log level so that a flood of less important
189 * messages doesn't cause more important ones to be dropped.
190 */
191 static struct ratelimit_state printk_limits[] = {
192 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
193 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
194 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
195 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
196 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
197 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
198 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
199 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
200 };
201
202 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
203 {
204 struct super_block *sb = fs_info->sb;
205 char lvl[4];
206 struct va_format vaf;
207 va_list args;
208 const char *type = logtypes[4];
209 int kern_level;
210 struct ratelimit_state *ratelimit;
211
212 va_start(args, fmt);
213
214 kern_level = printk_get_level(fmt);
215 if (kern_level) {
216 size_t size = printk_skip_level(fmt) - fmt;
217 memcpy(lvl, fmt, size);
218 lvl[size] = '\0';
219 fmt += size;
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
222 } else {
223 *lvl = '\0';
224 /* Default to debug output */
225 ratelimit = &printk_limits[7];
226 }
227
228 vaf.fmt = fmt;
229 vaf.va = &args;
230
231 if (__ratelimit(ratelimit))
232 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
233
234 va_end(args);
235 }
236 #endif
237
238 /*
239 * We only mark the transaction aborted and then set the file system read-only.
240 * This will prevent new transactions from starting or trying to join this
241 * one.
242 *
243 * This means that error recovery at the call site is limited to freeing
244 * any local memory allocations and passing the error code up without
245 * further cleanup. The transaction should complete as it normally would
246 * in the call path but will return -EIO.
247 *
248 * We'll complete the cleanup in btrfs_end_transaction and
249 * btrfs_commit_transaction.
250 */
251 __cold
252 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
253 const char *function,
254 unsigned int line, int errno)
255 {
256 struct btrfs_fs_info *fs_info = trans->fs_info;
257
258 trans->aborted = errno;
259 /* Nothing used. The other threads that have joined this
260 * transaction may be able to continue. */
261 if (!trans->dirty && list_empty(&trans->new_bgs)) {
262 const char *errstr;
263
264 errstr = btrfs_decode_error(errno);
265 btrfs_warn(fs_info,
266 "%s:%d: Aborting unused transaction(%s).",
267 function, line, errstr);
268 return;
269 }
270 ACCESS_ONCE(trans->transaction->aborted) = errno;
271 /* Wake up anybody who may be waiting on this transaction */
272 wake_up(&fs_info->transaction_wait);
273 wake_up(&fs_info->transaction_blocked_wait);
274 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
275 }
276 /*
277 * __btrfs_panic decodes unexpected, fatal errors from the caller,
278 * issues an alert, and either panics or BUGs, depending on mount options.
279 */
280 __cold
281 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
282 unsigned int line, int errno, const char *fmt, ...)
283 {
284 char *s_id = "<unknown>";
285 const char *errstr;
286 struct va_format vaf = { .fmt = fmt };
287 va_list args;
288
289 if (fs_info)
290 s_id = fs_info->sb->s_id;
291
292 va_start(args, fmt);
293 vaf.va = &args;
294
295 errstr = btrfs_decode_error(errno);
296 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
297 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
298 s_id, function, line, &vaf, errno, errstr);
299
300 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
301 function, line, &vaf, errno, errstr);
302 va_end(args);
303 /* Caller calls BUG() */
304 }
305
306 static void btrfs_put_super(struct super_block *sb)
307 {
308 close_ctree(btrfs_sb(sb)->tree_root);
309 }
310
311 enum {
312 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
313 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
314 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
315 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
316 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
317 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
318 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
319 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
320 Opt_skip_balance, Opt_check_integrity,
321 Opt_check_integrity_including_extent_data,
322 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
323 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
324 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
325 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
326 Opt_nologreplay, Opt_norecovery,
327 #ifdef CONFIG_BTRFS_DEBUG
328 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
329 #endif
330 Opt_err,
331 };
332
333 static const match_table_t tokens = {
334 {Opt_degraded, "degraded"},
335 {Opt_subvol, "subvol=%s"},
336 {Opt_subvolid, "subvolid=%s"},
337 {Opt_device, "device=%s"},
338 {Opt_nodatasum, "nodatasum"},
339 {Opt_datasum, "datasum"},
340 {Opt_nodatacow, "nodatacow"},
341 {Opt_datacow, "datacow"},
342 {Opt_nobarrier, "nobarrier"},
343 {Opt_barrier, "barrier"},
344 {Opt_max_inline, "max_inline=%s"},
345 {Opt_alloc_start, "alloc_start=%s"},
346 {Opt_thread_pool, "thread_pool=%d"},
347 {Opt_compress, "compress"},
348 {Opt_compress_type, "compress=%s"},
349 {Opt_compress_force, "compress-force"},
350 {Opt_compress_force_type, "compress-force=%s"},
351 {Opt_ssd, "ssd"},
352 {Opt_ssd_spread, "ssd_spread"},
353 {Opt_nossd, "nossd"},
354 {Opt_acl, "acl"},
355 {Opt_noacl, "noacl"},
356 {Opt_notreelog, "notreelog"},
357 {Opt_treelog, "treelog"},
358 {Opt_nologreplay, "nologreplay"},
359 {Opt_norecovery, "norecovery"},
360 {Opt_flushoncommit, "flushoncommit"},
361 {Opt_noflushoncommit, "noflushoncommit"},
362 {Opt_ratio, "metadata_ratio=%d"},
363 {Opt_discard, "discard"},
364 {Opt_nodiscard, "nodiscard"},
365 {Opt_space_cache, "space_cache"},
366 {Opt_space_cache_version, "space_cache=%s"},
367 {Opt_clear_cache, "clear_cache"},
368 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
369 {Opt_enospc_debug, "enospc_debug"},
370 {Opt_noenospc_debug, "noenospc_debug"},
371 {Opt_subvolrootid, "subvolrootid=%d"},
372 {Opt_defrag, "autodefrag"},
373 {Opt_nodefrag, "noautodefrag"},
374 {Opt_inode_cache, "inode_cache"},
375 {Opt_noinode_cache, "noinode_cache"},
376 {Opt_no_space_cache, "nospace_cache"},
377 {Opt_recovery, "recovery"}, /* deprecated */
378 {Opt_usebackuproot, "usebackuproot"},
379 {Opt_skip_balance, "skip_balance"},
380 {Opt_check_integrity, "check_int"},
381 {Opt_check_integrity_including_extent_data, "check_int_data"},
382 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
383 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
384 {Opt_fatal_errors, "fatal_errors=%s"},
385 {Opt_commit_interval, "commit=%d"},
386 #ifdef CONFIG_BTRFS_DEBUG
387 {Opt_fragment_data, "fragment=data"},
388 {Opt_fragment_metadata, "fragment=metadata"},
389 {Opt_fragment_all, "fragment=all"},
390 #endif
391 {Opt_err, NULL},
392 };
393
394 /*
395 * Regular mount options parser. Everything that is needed only when
396 * reading in a new superblock is parsed here.
397 * XXX JDM: This needs to be cleaned up for remount.
398 */
399 int btrfs_parse_options(struct btrfs_root *root, char *options,
400 unsigned long new_flags)
401 {
402 struct btrfs_fs_info *info = root->fs_info;
403 substring_t args[MAX_OPT_ARGS];
404 char *p, *num, *orig = NULL;
405 u64 cache_gen;
406 int intarg;
407 int ret = 0;
408 char *compress_type;
409 bool compress_force = false;
410 enum btrfs_compression_type saved_compress_type;
411 bool saved_compress_force;
412 int no_compress = 0;
413
414 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
415 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE))
416 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
417 else if (cache_gen)
418 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
419
420 /*
421 * Even the options are empty, we still need to do extra check
422 * against new flags
423 */
424 if (!options)
425 goto check;
426
427 /*
428 * strsep changes the string, duplicate it because parse_options
429 * gets called twice
430 */
431 options = kstrdup(options, GFP_NOFS);
432 if (!options)
433 return -ENOMEM;
434
435 orig = options;
436
437 while ((p = strsep(&options, ",")) != NULL) {
438 int token;
439 if (!*p)
440 continue;
441
442 token = match_token(p, tokens, args);
443 switch (token) {
444 case Opt_degraded:
445 btrfs_info(root->fs_info, "allowing degraded mounts");
446 btrfs_set_opt(info->mount_opt, DEGRADED);
447 break;
448 case Opt_subvol:
449 case Opt_subvolid:
450 case Opt_subvolrootid:
451 case Opt_device:
452 /*
453 * These are parsed by btrfs_parse_early_options
454 * and can be happily ignored here.
455 */
456 break;
457 case Opt_nodatasum:
458 btrfs_set_and_info(info, NODATASUM,
459 "setting nodatasum");
460 break;
461 case Opt_datasum:
462 if (btrfs_test_opt(info, NODATASUM)) {
463 if (btrfs_test_opt(info, NODATACOW))
464 btrfs_info(root->fs_info,
465 "setting datasum, datacow enabled");
466 else
467 btrfs_info(root->fs_info,
468 "setting datasum");
469 }
470 btrfs_clear_opt(info->mount_opt, NODATACOW);
471 btrfs_clear_opt(info->mount_opt, NODATASUM);
472 break;
473 case Opt_nodatacow:
474 if (!btrfs_test_opt(info, NODATACOW)) {
475 if (!btrfs_test_opt(info, COMPRESS) ||
476 !btrfs_test_opt(info, FORCE_COMPRESS)) {
477 btrfs_info(root->fs_info,
478 "setting nodatacow, compression disabled");
479 } else {
480 btrfs_info(root->fs_info,
481 "setting nodatacow");
482 }
483 }
484 btrfs_clear_opt(info->mount_opt, COMPRESS);
485 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
486 btrfs_set_opt(info->mount_opt, NODATACOW);
487 btrfs_set_opt(info->mount_opt, NODATASUM);
488 break;
489 case Opt_datacow:
490 btrfs_clear_and_info(info, NODATACOW,
491 "setting datacow");
492 break;
493 case Opt_compress_force:
494 case Opt_compress_force_type:
495 compress_force = true;
496 /* Fallthrough */
497 case Opt_compress:
498 case Opt_compress_type:
499 saved_compress_type = btrfs_test_opt(info,
500 COMPRESS) ?
501 info->compress_type : BTRFS_COMPRESS_NONE;
502 saved_compress_force =
503 btrfs_test_opt(info, FORCE_COMPRESS);
504 if (token == Opt_compress ||
505 token == Opt_compress_force ||
506 strcmp(args[0].from, "zlib") == 0) {
507 compress_type = "zlib";
508 info->compress_type = BTRFS_COMPRESS_ZLIB;
509 btrfs_set_opt(info->mount_opt, COMPRESS);
510 btrfs_clear_opt(info->mount_opt, NODATACOW);
511 btrfs_clear_opt(info->mount_opt, NODATASUM);
512 no_compress = 0;
513 } else if (strcmp(args[0].from, "lzo") == 0) {
514 compress_type = "lzo";
515 info->compress_type = BTRFS_COMPRESS_LZO;
516 btrfs_set_opt(info->mount_opt, COMPRESS);
517 btrfs_clear_opt(info->mount_opt, NODATACOW);
518 btrfs_clear_opt(info->mount_opt, NODATASUM);
519 btrfs_set_fs_incompat(info, COMPRESS_LZO);
520 no_compress = 0;
521 } else if (strncmp(args[0].from, "no", 2) == 0) {
522 compress_type = "no";
523 btrfs_clear_opt(info->mount_opt, COMPRESS);
524 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
525 compress_force = false;
526 no_compress++;
527 } else {
528 ret = -EINVAL;
529 goto out;
530 }
531
532 if (compress_force) {
533 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
534 } else {
535 /*
536 * If we remount from compress-force=xxx to
537 * compress=xxx, we need clear FORCE_COMPRESS
538 * flag, otherwise, there is no way for users
539 * to disable forcible compression separately.
540 */
541 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
542 }
543 if ((btrfs_test_opt(info, COMPRESS) &&
544 (info->compress_type != saved_compress_type ||
545 compress_force != saved_compress_force)) ||
546 (!btrfs_test_opt(info, COMPRESS) &&
547 no_compress == 1)) {
548 btrfs_info(root->fs_info,
549 "%s %s compression",
550 (compress_force) ? "force" : "use",
551 compress_type);
552 }
553 compress_force = false;
554 break;
555 case Opt_ssd:
556 btrfs_set_and_info(info, SSD,
557 "use ssd allocation scheme");
558 break;
559 case Opt_ssd_spread:
560 btrfs_set_and_info(info, SSD_SPREAD,
561 "use spread ssd allocation scheme");
562 btrfs_set_opt(info->mount_opt, SSD);
563 break;
564 case Opt_nossd:
565 btrfs_set_and_info(info, NOSSD,
566 "not using ssd allocation scheme");
567 btrfs_clear_opt(info->mount_opt, SSD);
568 break;
569 case Opt_barrier:
570 btrfs_clear_and_info(info, NOBARRIER,
571 "turning on barriers");
572 break;
573 case Opt_nobarrier:
574 btrfs_set_and_info(info, NOBARRIER,
575 "turning off barriers");
576 break;
577 case Opt_thread_pool:
578 ret = match_int(&args[0], &intarg);
579 if (ret) {
580 goto out;
581 } else if (intarg > 0) {
582 info->thread_pool_size = intarg;
583 } else {
584 ret = -EINVAL;
585 goto out;
586 }
587 break;
588 case Opt_max_inline:
589 num = match_strdup(&args[0]);
590 if (num) {
591 info->max_inline = memparse(num, NULL);
592 kfree(num);
593
594 if (info->max_inline) {
595 info->max_inline = min_t(u64,
596 info->max_inline,
597 root->sectorsize);
598 }
599 btrfs_info(root->fs_info, "max_inline at %llu",
600 info->max_inline);
601 } else {
602 ret = -ENOMEM;
603 goto out;
604 }
605 break;
606 case Opt_alloc_start:
607 num = match_strdup(&args[0]);
608 if (num) {
609 mutex_lock(&info->chunk_mutex);
610 info->alloc_start = memparse(num, NULL);
611 mutex_unlock(&info->chunk_mutex);
612 kfree(num);
613 btrfs_info(root->fs_info,
614 "allocations start at %llu",
615 info->alloc_start);
616 } else {
617 ret = -ENOMEM;
618 goto out;
619 }
620 break;
621 case Opt_acl:
622 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
623 root->fs_info->sb->s_flags |= MS_POSIXACL;
624 break;
625 #else
626 btrfs_err(root->fs_info,
627 "support for ACL not compiled in!");
628 ret = -EINVAL;
629 goto out;
630 #endif
631 case Opt_noacl:
632 root->fs_info->sb->s_flags &= ~MS_POSIXACL;
633 break;
634 case Opt_notreelog:
635 btrfs_set_and_info(info, NOTREELOG,
636 "disabling tree log");
637 break;
638 case Opt_treelog:
639 btrfs_clear_and_info(info, NOTREELOG,
640 "enabling tree log");
641 break;
642 case Opt_norecovery:
643 case Opt_nologreplay:
644 btrfs_set_and_info(info, NOLOGREPLAY,
645 "disabling log replay at mount time");
646 break;
647 case Opt_flushoncommit:
648 btrfs_set_and_info(info, FLUSHONCOMMIT,
649 "turning on flush-on-commit");
650 break;
651 case Opt_noflushoncommit:
652 btrfs_clear_and_info(info, FLUSHONCOMMIT,
653 "turning off flush-on-commit");
654 break;
655 case Opt_ratio:
656 ret = match_int(&args[0], &intarg);
657 if (ret) {
658 goto out;
659 } else if (intarg >= 0) {
660 info->metadata_ratio = intarg;
661 btrfs_info(root->fs_info, "metadata ratio %d",
662 info->metadata_ratio);
663 } else {
664 ret = -EINVAL;
665 goto out;
666 }
667 break;
668 case Opt_discard:
669 btrfs_set_and_info(info, DISCARD,
670 "turning on discard");
671 break;
672 case Opt_nodiscard:
673 btrfs_clear_and_info(info, DISCARD,
674 "turning off discard");
675 break;
676 case Opt_space_cache:
677 case Opt_space_cache_version:
678 if (token == Opt_space_cache ||
679 strcmp(args[0].from, "v1") == 0) {
680 btrfs_clear_opt(root->fs_info->mount_opt,
681 FREE_SPACE_TREE);
682 btrfs_set_and_info(info, SPACE_CACHE,
683 "enabling disk space caching");
684 } else if (strcmp(args[0].from, "v2") == 0) {
685 btrfs_clear_opt(root->fs_info->mount_opt,
686 SPACE_CACHE);
687 btrfs_set_and_info(info,
688 FREE_SPACE_TREE,
689 "enabling free space tree");
690 } else {
691 ret = -EINVAL;
692 goto out;
693 }
694 break;
695 case Opt_rescan_uuid_tree:
696 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
697 break;
698 case Opt_no_space_cache:
699 if (btrfs_test_opt(info, SPACE_CACHE)) {
700 btrfs_clear_and_info(info,
701 SPACE_CACHE,
702 "disabling disk space caching");
703 }
704 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
705 btrfs_clear_and_info(info,
706 FREE_SPACE_TREE,
707 "disabling free space tree");
708 }
709 break;
710 case Opt_inode_cache:
711 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
712 "enabling inode map caching");
713 break;
714 case Opt_noinode_cache:
715 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
716 "disabling inode map caching");
717 break;
718 case Opt_clear_cache:
719 btrfs_set_and_info(info, CLEAR_CACHE,
720 "force clearing of disk cache");
721 break;
722 case Opt_user_subvol_rm_allowed:
723 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
724 break;
725 case Opt_enospc_debug:
726 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
727 break;
728 case Opt_noenospc_debug:
729 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
730 break;
731 case Opt_defrag:
732 btrfs_set_and_info(info, AUTO_DEFRAG,
733 "enabling auto defrag");
734 break;
735 case Opt_nodefrag:
736 btrfs_clear_and_info(info, AUTO_DEFRAG,
737 "disabling auto defrag");
738 break;
739 case Opt_recovery:
740 btrfs_warn(root->fs_info,
741 "'recovery' is deprecated, use 'usebackuproot' instead");
742 case Opt_usebackuproot:
743 btrfs_info(root->fs_info,
744 "trying to use backup root at mount time");
745 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
746 break;
747 case Opt_skip_balance:
748 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
749 break;
750 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
751 case Opt_check_integrity_including_extent_data:
752 btrfs_info(root->fs_info,
753 "enabling check integrity including extent data");
754 btrfs_set_opt(info->mount_opt,
755 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
756 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
757 break;
758 case Opt_check_integrity:
759 btrfs_info(root->fs_info, "enabling check integrity");
760 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
761 break;
762 case Opt_check_integrity_print_mask:
763 ret = match_int(&args[0], &intarg);
764 if (ret) {
765 goto out;
766 } else if (intarg >= 0) {
767 info->check_integrity_print_mask = intarg;
768 btrfs_info(root->fs_info,
769 "check_integrity_print_mask 0x%x",
770 info->check_integrity_print_mask);
771 } else {
772 ret = -EINVAL;
773 goto out;
774 }
775 break;
776 #else
777 case Opt_check_integrity_including_extent_data:
778 case Opt_check_integrity:
779 case Opt_check_integrity_print_mask:
780 btrfs_err(root->fs_info,
781 "support for check_integrity* not compiled in!");
782 ret = -EINVAL;
783 goto out;
784 #endif
785 case Opt_fatal_errors:
786 if (strcmp(args[0].from, "panic") == 0)
787 btrfs_set_opt(info->mount_opt,
788 PANIC_ON_FATAL_ERROR);
789 else if (strcmp(args[0].from, "bug") == 0)
790 btrfs_clear_opt(info->mount_opt,
791 PANIC_ON_FATAL_ERROR);
792 else {
793 ret = -EINVAL;
794 goto out;
795 }
796 break;
797 case Opt_commit_interval:
798 intarg = 0;
799 ret = match_int(&args[0], &intarg);
800 if (ret < 0) {
801 btrfs_err(root->fs_info,
802 "invalid commit interval");
803 ret = -EINVAL;
804 goto out;
805 }
806 if (intarg > 0) {
807 if (intarg > 300) {
808 btrfs_warn(root->fs_info,
809 "excessive commit interval %d",
810 intarg);
811 }
812 info->commit_interval = intarg;
813 } else {
814 btrfs_info(root->fs_info,
815 "using default commit interval %ds",
816 BTRFS_DEFAULT_COMMIT_INTERVAL);
817 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
818 }
819 break;
820 #ifdef CONFIG_BTRFS_DEBUG
821 case Opt_fragment_all:
822 btrfs_info(root->fs_info, "fragmenting all space");
823 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
824 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
825 break;
826 case Opt_fragment_metadata:
827 btrfs_info(root->fs_info, "fragmenting metadata");
828 btrfs_set_opt(info->mount_opt,
829 FRAGMENT_METADATA);
830 break;
831 case Opt_fragment_data:
832 btrfs_info(root->fs_info, "fragmenting data");
833 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
834 break;
835 #endif
836 case Opt_err:
837 btrfs_info(root->fs_info,
838 "unrecognized mount option '%s'", p);
839 ret = -EINVAL;
840 goto out;
841 default:
842 break;
843 }
844 }
845 check:
846 /*
847 * Extra check for current option against current flag
848 */
849 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
850 btrfs_err(root->fs_info,
851 "nologreplay must be used with ro mount option");
852 ret = -EINVAL;
853 }
854 out:
855 if (btrfs_fs_compat_ro(root->fs_info, FREE_SPACE_TREE) &&
856 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
857 !btrfs_test_opt(info, CLEAR_CACHE)) {
858 btrfs_err(root->fs_info, "cannot disable free space tree");
859 ret = -EINVAL;
860
861 }
862 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
863 btrfs_info(root->fs_info, "disk space caching is enabled");
864 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
865 btrfs_info(root->fs_info, "using free space tree");
866 kfree(orig);
867 return ret;
868 }
869
870 /*
871 * Parse mount options that are required early in the mount process.
872 *
873 * All other options will be parsed on much later in the mount process and
874 * only when we need to allocate a new super block.
875 */
876 static int btrfs_parse_early_options(const char *options, fmode_t flags,
877 void *holder, char **subvol_name, u64 *subvol_objectid,
878 struct btrfs_fs_devices **fs_devices)
879 {
880 substring_t args[MAX_OPT_ARGS];
881 char *device_name, *opts, *orig, *p;
882 char *num = NULL;
883 int error = 0;
884
885 if (!options)
886 return 0;
887
888 /*
889 * strsep changes the string, duplicate it because parse_options
890 * gets called twice
891 */
892 opts = kstrdup(options, GFP_KERNEL);
893 if (!opts)
894 return -ENOMEM;
895 orig = opts;
896
897 while ((p = strsep(&opts, ",")) != NULL) {
898 int token;
899 if (!*p)
900 continue;
901
902 token = match_token(p, tokens, args);
903 switch (token) {
904 case Opt_subvol:
905 kfree(*subvol_name);
906 *subvol_name = match_strdup(&args[0]);
907 if (!*subvol_name) {
908 error = -ENOMEM;
909 goto out;
910 }
911 break;
912 case Opt_subvolid:
913 num = match_strdup(&args[0]);
914 if (num) {
915 *subvol_objectid = memparse(num, NULL);
916 kfree(num);
917 /* we want the original fs_tree */
918 if (!*subvol_objectid)
919 *subvol_objectid =
920 BTRFS_FS_TREE_OBJECTID;
921 } else {
922 error = -EINVAL;
923 goto out;
924 }
925 break;
926 case Opt_subvolrootid:
927 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
928 break;
929 case Opt_device:
930 device_name = match_strdup(&args[0]);
931 if (!device_name) {
932 error = -ENOMEM;
933 goto out;
934 }
935 error = btrfs_scan_one_device(device_name,
936 flags, holder, fs_devices);
937 kfree(device_name);
938 if (error)
939 goto out;
940 break;
941 default:
942 break;
943 }
944 }
945
946 out:
947 kfree(orig);
948 return error;
949 }
950
951 static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
952 u64 subvol_objectid)
953 {
954 struct btrfs_root *root = fs_info->tree_root;
955 struct btrfs_root *fs_root;
956 struct btrfs_root_ref *root_ref;
957 struct btrfs_inode_ref *inode_ref;
958 struct btrfs_key key;
959 struct btrfs_path *path = NULL;
960 char *name = NULL, *ptr;
961 u64 dirid;
962 int len;
963 int ret;
964
965 path = btrfs_alloc_path();
966 if (!path) {
967 ret = -ENOMEM;
968 goto err;
969 }
970 path->leave_spinning = 1;
971
972 name = kmalloc(PATH_MAX, GFP_NOFS);
973 if (!name) {
974 ret = -ENOMEM;
975 goto err;
976 }
977 ptr = name + PATH_MAX - 1;
978 ptr[0] = '\0';
979
980 /*
981 * Walk up the subvolume trees in the tree of tree roots by root
982 * backrefs until we hit the top-level subvolume.
983 */
984 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
985 key.objectid = subvol_objectid;
986 key.type = BTRFS_ROOT_BACKREF_KEY;
987 key.offset = (u64)-1;
988
989 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
990 if (ret < 0) {
991 goto err;
992 } else if (ret > 0) {
993 ret = btrfs_previous_item(root, path, subvol_objectid,
994 BTRFS_ROOT_BACKREF_KEY);
995 if (ret < 0) {
996 goto err;
997 } else if (ret > 0) {
998 ret = -ENOENT;
999 goto err;
1000 }
1001 }
1002
1003 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1004 subvol_objectid = key.offset;
1005
1006 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1007 struct btrfs_root_ref);
1008 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1009 ptr -= len + 1;
1010 if (ptr < name) {
1011 ret = -ENAMETOOLONG;
1012 goto err;
1013 }
1014 read_extent_buffer(path->nodes[0], ptr + 1,
1015 (unsigned long)(root_ref + 1), len);
1016 ptr[0] = '/';
1017 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1018 btrfs_release_path(path);
1019
1020 key.objectid = subvol_objectid;
1021 key.type = BTRFS_ROOT_ITEM_KEY;
1022 key.offset = (u64)-1;
1023 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1024 if (IS_ERR(fs_root)) {
1025 ret = PTR_ERR(fs_root);
1026 goto err;
1027 }
1028
1029 /*
1030 * Walk up the filesystem tree by inode refs until we hit the
1031 * root directory.
1032 */
1033 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1034 key.objectid = dirid;
1035 key.type = BTRFS_INODE_REF_KEY;
1036 key.offset = (u64)-1;
1037
1038 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1039 if (ret < 0) {
1040 goto err;
1041 } else if (ret > 0) {
1042 ret = btrfs_previous_item(fs_root, path, dirid,
1043 BTRFS_INODE_REF_KEY);
1044 if (ret < 0) {
1045 goto err;
1046 } else if (ret > 0) {
1047 ret = -ENOENT;
1048 goto err;
1049 }
1050 }
1051
1052 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1053 dirid = key.offset;
1054
1055 inode_ref = btrfs_item_ptr(path->nodes[0],
1056 path->slots[0],
1057 struct btrfs_inode_ref);
1058 len = btrfs_inode_ref_name_len(path->nodes[0],
1059 inode_ref);
1060 ptr -= len + 1;
1061 if (ptr < name) {
1062 ret = -ENAMETOOLONG;
1063 goto err;
1064 }
1065 read_extent_buffer(path->nodes[0], ptr + 1,
1066 (unsigned long)(inode_ref + 1), len);
1067 ptr[0] = '/';
1068 btrfs_release_path(path);
1069 }
1070 }
1071
1072 btrfs_free_path(path);
1073 if (ptr == name + PATH_MAX - 1) {
1074 name[0] = '/';
1075 name[1] = '\0';
1076 } else {
1077 memmove(name, ptr, name + PATH_MAX - ptr);
1078 }
1079 return name;
1080
1081 err:
1082 btrfs_free_path(path);
1083 kfree(name);
1084 return ERR_PTR(ret);
1085 }
1086
1087 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1088 {
1089 struct btrfs_root *root = fs_info->tree_root;
1090 struct btrfs_dir_item *di;
1091 struct btrfs_path *path;
1092 struct btrfs_key location;
1093 u64 dir_id;
1094
1095 path = btrfs_alloc_path();
1096 if (!path)
1097 return -ENOMEM;
1098 path->leave_spinning = 1;
1099
1100 /*
1101 * Find the "default" dir item which points to the root item that we
1102 * will mount by default if we haven't been given a specific subvolume
1103 * to mount.
1104 */
1105 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1106 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1107 if (IS_ERR(di)) {
1108 btrfs_free_path(path);
1109 return PTR_ERR(di);
1110 }
1111 if (!di) {
1112 /*
1113 * Ok the default dir item isn't there. This is weird since
1114 * it's always been there, but don't freak out, just try and
1115 * mount the top-level subvolume.
1116 */
1117 btrfs_free_path(path);
1118 *objectid = BTRFS_FS_TREE_OBJECTID;
1119 return 0;
1120 }
1121
1122 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1123 btrfs_free_path(path);
1124 *objectid = location.objectid;
1125 return 0;
1126 }
1127
1128 static int btrfs_fill_super(struct super_block *sb,
1129 struct btrfs_fs_devices *fs_devices,
1130 void *data, int silent)
1131 {
1132 struct inode *inode;
1133 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1134 struct btrfs_key key;
1135 int err;
1136
1137 sb->s_maxbytes = MAX_LFS_FILESIZE;
1138 sb->s_magic = BTRFS_SUPER_MAGIC;
1139 sb->s_op = &btrfs_super_ops;
1140 sb->s_d_op = &btrfs_dentry_operations;
1141 sb->s_export_op = &btrfs_export_ops;
1142 sb->s_xattr = btrfs_xattr_handlers;
1143 sb->s_time_gran = 1;
1144 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1145 sb->s_flags |= MS_POSIXACL;
1146 #endif
1147 sb->s_flags |= MS_I_VERSION;
1148 sb->s_iflags |= SB_I_CGROUPWB;
1149 err = open_ctree(sb, fs_devices, (char *)data);
1150 if (err) {
1151 btrfs_err(fs_info, "open_ctree failed");
1152 return err;
1153 }
1154
1155 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1156 key.type = BTRFS_INODE_ITEM_KEY;
1157 key.offset = 0;
1158 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1159 if (IS_ERR(inode)) {
1160 err = PTR_ERR(inode);
1161 goto fail_close;
1162 }
1163
1164 sb->s_root = d_make_root(inode);
1165 if (!sb->s_root) {
1166 err = -ENOMEM;
1167 goto fail_close;
1168 }
1169
1170 save_mount_options(sb, data);
1171 cleancache_init_fs(sb);
1172 sb->s_flags |= MS_ACTIVE;
1173 return 0;
1174
1175 fail_close:
1176 close_ctree(fs_info->tree_root);
1177 return err;
1178 }
1179
1180 int btrfs_sync_fs(struct super_block *sb, int wait)
1181 {
1182 struct btrfs_trans_handle *trans;
1183 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1184 struct btrfs_root *root = fs_info->tree_root;
1185
1186 trace_btrfs_sync_fs(fs_info, wait);
1187
1188 if (!wait) {
1189 filemap_flush(fs_info->btree_inode->i_mapping);
1190 return 0;
1191 }
1192
1193 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1194
1195 trans = btrfs_attach_transaction_barrier(root);
1196 if (IS_ERR(trans)) {
1197 /* no transaction, don't bother */
1198 if (PTR_ERR(trans) == -ENOENT) {
1199 /*
1200 * Exit unless we have some pending changes
1201 * that need to go through commit
1202 */
1203 if (fs_info->pending_changes == 0)
1204 return 0;
1205 /*
1206 * A non-blocking test if the fs is frozen. We must not
1207 * start a new transaction here otherwise a deadlock
1208 * happens. The pending operations are delayed to the
1209 * next commit after thawing.
1210 */
1211 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1212 __sb_end_write(sb, SB_FREEZE_WRITE);
1213 else
1214 return 0;
1215 trans = btrfs_start_transaction(root, 0);
1216 }
1217 if (IS_ERR(trans))
1218 return PTR_ERR(trans);
1219 }
1220 return btrfs_commit_transaction(trans, root);
1221 }
1222
1223 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1224 {
1225 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1226 struct btrfs_root *root = info->tree_root;
1227 char *compress_type;
1228
1229 if (btrfs_test_opt(info, DEGRADED))
1230 seq_puts(seq, ",degraded");
1231 if (btrfs_test_opt(info, NODATASUM))
1232 seq_puts(seq, ",nodatasum");
1233 if (btrfs_test_opt(info, NODATACOW))
1234 seq_puts(seq, ",nodatacow");
1235 if (btrfs_test_opt(info, NOBARRIER))
1236 seq_puts(seq, ",nobarrier");
1237 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1238 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1239 if (info->alloc_start != 0)
1240 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1241 if (info->thread_pool_size != min_t(unsigned long,
1242 num_online_cpus() + 2, 8))
1243 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1244 if (btrfs_test_opt(info, COMPRESS)) {
1245 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1246 compress_type = "zlib";
1247 else
1248 compress_type = "lzo";
1249 if (btrfs_test_opt(info, FORCE_COMPRESS))
1250 seq_printf(seq, ",compress-force=%s", compress_type);
1251 else
1252 seq_printf(seq, ",compress=%s", compress_type);
1253 }
1254 if (btrfs_test_opt(info, NOSSD))
1255 seq_puts(seq, ",nossd");
1256 if (btrfs_test_opt(info, SSD_SPREAD))
1257 seq_puts(seq, ",ssd_spread");
1258 else if (btrfs_test_opt(info, SSD))
1259 seq_puts(seq, ",ssd");
1260 if (btrfs_test_opt(info, NOTREELOG))
1261 seq_puts(seq, ",notreelog");
1262 if (btrfs_test_opt(info, NOLOGREPLAY))
1263 seq_puts(seq, ",nologreplay");
1264 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1265 seq_puts(seq, ",flushoncommit");
1266 if (btrfs_test_opt(info, DISCARD))
1267 seq_puts(seq, ",discard");
1268 if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1269 seq_puts(seq, ",noacl");
1270 if (btrfs_test_opt(info, SPACE_CACHE))
1271 seq_puts(seq, ",space_cache");
1272 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1273 seq_puts(seq, ",space_cache=v2");
1274 else
1275 seq_puts(seq, ",nospace_cache");
1276 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1277 seq_puts(seq, ",rescan_uuid_tree");
1278 if (btrfs_test_opt(info, CLEAR_CACHE))
1279 seq_puts(seq, ",clear_cache");
1280 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1281 seq_puts(seq, ",user_subvol_rm_allowed");
1282 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1283 seq_puts(seq, ",enospc_debug");
1284 if (btrfs_test_opt(info, AUTO_DEFRAG))
1285 seq_puts(seq, ",autodefrag");
1286 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1287 seq_puts(seq, ",inode_cache");
1288 if (btrfs_test_opt(info, SKIP_BALANCE))
1289 seq_puts(seq, ",skip_balance");
1290 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1291 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1292 seq_puts(seq, ",check_int_data");
1293 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1294 seq_puts(seq, ",check_int");
1295 if (info->check_integrity_print_mask)
1296 seq_printf(seq, ",check_int_print_mask=%d",
1297 info->check_integrity_print_mask);
1298 #endif
1299 if (info->metadata_ratio)
1300 seq_printf(seq, ",metadata_ratio=%d",
1301 info->metadata_ratio);
1302 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1303 seq_puts(seq, ",fatal_errors=panic");
1304 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1305 seq_printf(seq, ",commit=%d", info->commit_interval);
1306 #ifdef CONFIG_BTRFS_DEBUG
1307 if (btrfs_test_opt(info, FRAGMENT_DATA))
1308 seq_puts(seq, ",fragment=data");
1309 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1310 seq_puts(seq, ",fragment=metadata");
1311 #endif
1312 seq_printf(seq, ",subvolid=%llu",
1313 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1314 seq_puts(seq, ",subvol=");
1315 seq_dentry(seq, dentry, " \t\n\\");
1316 return 0;
1317 }
1318
1319 static int btrfs_test_super(struct super_block *s, void *data)
1320 {
1321 struct btrfs_fs_info *p = data;
1322 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1323
1324 return fs_info->fs_devices == p->fs_devices;
1325 }
1326
1327 static int btrfs_set_super(struct super_block *s, void *data)
1328 {
1329 int err = set_anon_super(s, data);
1330 if (!err)
1331 s->s_fs_info = data;
1332 return err;
1333 }
1334
1335 /*
1336 * subvolumes are identified by ino 256
1337 */
1338 static inline int is_subvolume_inode(struct inode *inode)
1339 {
1340 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1341 return 1;
1342 return 0;
1343 }
1344
1345 /*
1346 * This will add subvolid=0 to the argument string while removing any subvol=
1347 * and subvolid= arguments to make sure we get the top-level root for path
1348 * walking to the subvol we want.
1349 */
1350 static char *setup_root_args(char *args)
1351 {
1352 char *buf, *dst, *sep;
1353
1354 if (!args)
1355 return kstrdup("subvolid=0", GFP_NOFS);
1356
1357 /* The worst case is that we add ",subvolid=0" to the end. */
1358 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1359 if (!buf)
1360 return NULL;
1361
1362 while (1) {
1363 sep = strchrnul(args, ',');
1364 if (!strstarts(args, "subvol=") &&
1365 !strstarts(args, "subvolid=")) {
1366 memcpy(dst, args, sep - args);
1367 dst += sep - args;
1368 *dst++ = ',';
1369 }
1370 if (*sep)
1371 args = sep + 1;
1372 else
1373 break;
1374 }
1375 strcpy(dst, "subvolid=0");
1376
1377 return buf;
1378 }
1379
1380 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1381 int flags, const char *device_name,
1382 char *data)
1383 {
1384 struct dentry *root;
1385 struct vfsmount *mnt = NULL;
1386 char *newargs;
1387 int ret;
1388
1389 newargs = setup_root_args(data);
1390 if (!newargs) {
1391 root = ERR_PTR(-ENOMEM);
1392 goto out;
1393 }
1394
1395 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1396 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1397 if (flags & MS_RDONLY) {
1398 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1399 device_name, newargs);
1400 } else {
1401 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1402 device_name, newargs);
1403 if (IS_ERR(mnt)) {
1404 root = ERR_CAST(mnt);
1405 mnt = NULL;
1406 goto out;
1407 }
1408
1409 down_write(&mnt->mnt_sb->s_umount);
1410 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1411 up_write(&mnt->mnt_sb->s_umount);
1412 if (ret < 0) {
1413 root = ERR_PTR(ret);
1414 goto out;
1415 }
1416 }
1417 }
1418 if (IS_ERR(mnt)) {
1419 root = ERR_CAST(mnt);
1420 mnt = NULL;
1421 goto out;
1422 }
1423
1424 if (!subvol_name) {
1425 if (!subvol_objectid) {
1426 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1427 &subvol_objectid);
1428 if (ret) {
1429 root = ERR_PTR(ret);
1430 goto out;
1431 }
1432 }
1433 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1434 subvol_objectid);
1435 if (IS_ERR(subvol_name)) {
1436 root = ERR_CAST(subvol_name);
1437 subvol_name = NULL;
1438 goto out;
1439 }
1440
1441 }
1442
1443 root = mount_subtree(mnt, subvol_name);
1444 /* mount_subtree() drops our reference on the vfsmount. */
1445 mnt = NULL;
1446
1447 if (!IS_ERR(root)) {
1448 struct super_block *s = root->d_sb;
1449 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1450 struct inode *root_inode = d_inode(root);
1451 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1452
1453 ret = 0;
1454 if (!is_subvolume_inode(root_inode)) {
1455 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1456 subvol_name);
1457 ret = -EINVAL;
1458 }
1459 if (subvol_objectid && root_objectid != subvol_objectid) {
1460 /*
1461 * This will also catch a race condition where a
1462 * subvolume which was passed by ID is renamed and
1463 * another subvolume is renamed over the old location.
1464 */
1465 btrfs_err(fs_info,
1466 "subvol '%s' does not match subvolid %llu",
1467 subvol_name, subvol_objectid);
1468 ret = -EINVAL;
1469 }
1470 if (ret) {
1471 dput(root);
1472 root = ERR_PTR(ret);
1473 deactivate_locked_super(s);
1474 }
1475 }
1476
1477 out:
1478 mntput(mnt);
1479 kfree(newargs);
1480 kfree(subvol_name);
1481 return root;
1482 }
1483
1484 static int parse_security_options(char *orig_opts,
1485 struct security_mnt_opts *sec_opts)
1486 {
1487 char *secdata = NULL;
1488 int ret = 0;
1489
1490 secdata = alloc_secdata();
1491 if (!secdata)
1492 return -ENOMEM;
1493 ret = security_sb_copy_data(orig_opts, secdata);
1494 if (ret) {
1495 free_secdata(secdata);
1496 return ret;
1497 }
1498 ret = security_sb_parse_opts_str(secdata, sec_opts);
1499 free_secdata(secdata);
1500 return ret;
1501 }
1502
1503 static int setup_security_options(struct btrfs_fs_info *fs_info,
1504 struct super_block *sb,
1505 struct security_mnt_opts *sec_opts)
1506 {
1507 int ret = 0;
1508
1509 /*
1510 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1511 * is valid.
1512 */
1513 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1514 if (ret)
1515 return ret;
1516
1517 #ifdef CONFIG_SECURITY
1518 if (!fs_info->security_opts.num_mnt_opts) {
1519 /* first time security setup, copy sec_opts to fs_info */
1520 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1521 } else {
1522 /*
1523 * Since SELinux (the only one supporting security_mnt_opts)
1524 * does NOT support changing context during remount/mount of
1525 * the same sb, this must be the same or part of the same
1526 * security options, just free it.
1527 */
1528 security_free_mnt_opts(sec_opts);
1529 }
1530 #endif
1531 return ret;
1532 }
1533
1534 /*
1535 * Find a superblock for the given device / mount point.
1536 *
1537 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1538 * for multiple device setup. Make sure to keep it in sync.
1539 */
1540 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1541 const char *device_name, void *data)
1542 {
1543 struct block_device *bdev = NULL;
1544 struct super_block *s;
1545 struct btrfs_fs_devices *fs_devices = NULL;
1546 struct btrfs_fs_info *fs_info = NULL;
1547 struct security_mnt_opts new_sec_opts;
1548 fmode_t mode = FMODE_READ;
1549 char *subvol_name = NULL;
1550 u64 subvol_objectid = 0;
1551 int error = 0;
1552
1553 if (!(flags & MS_RDONLY))
1554 mode |= FMODE_WRITE;
1555
1556 error = btrfs_parse_early_options(data, mode, fs_type,
1557 &subvol_name, &subvol_objectid,
1558 &fs_devices);
1559 if (error) {
1560 kfree(subvol_name);
1561 return ERR_PTR(error);
1562 }
1563
1564 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1565 /* mount_subvol() will free subvol_name. */
1566 return mount_subvol(subvol_name, subvol_objectid, flags,
1567 device_name, data);
1568 }
1569
1570 security_init_mnt_opts(&new_sec_opts);
1571 if (data) {
1572 error = parse_security_options(data, &new_sec_opts);
1573 if (error)
1574 return ERR_PTR(error);
1575 }
1576
1577 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1578 if (error)
1579 goto error_sec_opts;
1580
1581 /*
1582 * Setup a dummy root and fs_info for test/set super. This is because
1583 * we don't actually fill this stuff out until open_ctree, but we need
1584 * it for searching for existing supers, so this lets us do that and
1585 * then open_ctree will properly initialize everything later.
1586 */
1587 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1588 if (!fs_info) {
1589 error = -ENOMEM;
1590 goto error_sec_opts;
1591 }
1592
1593 fs_info->fs_devices = fs_devices;
1594
1595 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1596 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1597 security_init_mnt_opts(&fs_info->security_opts);
1598 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1599 error = -ENOMEM;
1600 goto error_fs_info;
1601 }
1602
1603 error = btrfs_open_devices(fs_devices, mode, fs_type);
1604 if (error)
1605 goto error_fs_info;
1606
1607 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1608 error = -EACCES;
1609 goto error_close_devices;
1610 }
1611
1612 bdev = fs_devices->latest_bdev;
1613 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1614 fs_info);
1615 if (IS_ERR(s)) {
1616 error = PTR_ERR(s);
1617 goto error_close_devices;
1618 }
1619
1620 if (s->s_root) {
1621 btrfs_close_devices(fs_devices);
1622 free_fs_info(fs_info);
1623 if ((flags ^ s->s_flags) & MS_RDONLY)
1624 error = -EBUSY;
1625 } else {
1626 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1627 btrfs_sb(s)->bdev_holder = fs_type;
1628 error = btrfs_fill_super(s, fs_devices, data,
1629 flags & MS_SILENT ? 1 : 0);
1630 }
1631 if (error) {
1632 deactivate_locked_super(s);
1633 goto error_sec_opts;
1634 }
1635
1636 fs_info = btrfs_sb(s);
1637 error = setup_security_options(fs_info, s, &new_sec_opts);
1638 if (error) {
1639 deactivate_locked_super(s);
1640 goto error_sec_opts;
1641 }
1642
1643 return dget(s->s_root);
1644
1645 error_close_devices:
1646 btrfs_close_devices(fs_devices);
1647 error_fs_info:
1648 free_fs_info(fs_info);
1649 error_sec_opts:
1650 security_free_mnt_opts(&new_sec_opts);
1651 return ERR_PTR(error);
1652 }
1653
1654 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1655 int new_pool_size, int old_pool_size)
1656 {
1657 if (new_pool_size == old_pool_size)
1658 return;
1659
1660 fs_info->thread_pool_size = new_pool_size;
1661
1662 btrfs_info(fs_info, "resize thread pool %d -> %d",
1663 old_pool_size, new_pool_size);
1664
1665 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1666 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1667 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1668 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1669 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1670 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1671 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1672 new_pool_size);
1673 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1674 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1675 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1676 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1677 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1678 new_pool_size);
1679 }
1680
1681 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1682 {
1683 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1684 }
1685
1686 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1687 unsigned long old_opts, int flags)
1688 {
1689 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1690 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1691 (flags & MS_RDONLY))) {
1692 /* wait for any defraggers to finish */
1693 wait_event(fs_info->transaction_wait,
1694 (atomic_read(&fs_info->defrag_running) == 0));
1695 if (flags & MS_RDONLY)
1696 sync_filesystem(fs_info->sb);
1697 }
1698 }
1699
1700 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1701 unsigned long old_opts)
1702 {
1703 /*
1704 * We need to cleanup all defragable inodes if the autodefragment is
1705 * close or the filesystem is read only.
1706 */
1707 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1708 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1709 (fs_info->sb->s_flags & MS_RDONLY))) {
1710 btrfs_cleanup_defrag_inodes(fs_info);
1711 }
1712
1713 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1714 }
1715
1716 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1717 {
1718 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1719 struct btrfs_root *root = fs_info->tree_root;
1720 unsigned old_flags = sb->s_flags;
1721 unsigned long old_opts = fs_info->mount_opt;
1722 unsigned long old_compress_type = fs_info->compress_type;
1723 u64 old_max_inline = fs_info->max_inline;
1724 u64 old_alloc_start = fs_info->alloc_start;
1725 int old_thread_pool_size = fs_info->thread_pool_size;
1726 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1727 int ret;
1728
1729 sync_filesystem(sb);
1730 btrfs_remount_prepare(fs_info);
1731
1732 if (data) {
1733 struct security_mnt_opts new_sec_opts;
1734
1735 security_init_mnt_opts(&new_sec_opts);
1736 ret = parse_security_options(data, &new_sec_opts);
1737 if (ret)
1738 goto restore;
1739 ret = setup_security_options(fs_info, sb,
1740 &new_sec_opts);
1741 if (ret) {
1742 security_free_mnt_opts(&new_sec_opts);
1743 goto restore;
1744 }
1745 }
1746
1747 ret = btrfs_parse_options(root, data, *flags);
1748 if (ret) {
1749 ret = -EINVAL;
1750 goto restore;
1751 }
1752
1753 btrfs_remount_begin(fs_info, old_opts, *flags);
1754 btrfs_resize_thread_pool(fs_info,
1755 fs_info->thread_pool_size, old_thread_pool_size);
1756
1757 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1758 goto out;
1759
1760 if (*flags & MS_RDONLY) {
1761 /*
1762 * this also happens on 'umount -rf' or on shutdown, when
1763 * the filesystem is busy.
1764 */
1765 cancel_work_sync(&fs_info->async_reclaim_work);
1766
1767 /* wait for the uuid_scan task to finish */
1768 down(&fs_info->uuid_tree_rescan_sem);
1769 /* avoid complains from lockdep et al. */
1770 up(&fs_info->uuid_tree_rescan_sem);
1771
1772 sb->s_flags |= MS_RDONLY;
1773
1774 /*
1775 * Setting MS_RDONLY will put the cleaner thread to
1776 * sleep at the next loop if it's already active.
1777 * If it's already asleep, we'll leave unused block
1778 * groups on disk until we're mounted read-write again
1779 * unless we clean them up here.
1780 */
1781 btrfs_delete_unused_bgs(fs_info);
1782
1783 btrfs_dev_replace_suspend_for_unmount(fs_info);
1784 btrfs_scrub_cancel(fs_info);
1785 btrfs_pause_balance(fs_info);
1786
1787 ret = btrfs_commit_super(root);
1788 if (ret)
1789 goto restore;
1790 } else {
1791 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1792 btrfs_err(fs_info,
1793 "Remounting read-write after error is not allowed");
1794 ret = -EINVAL;
1795 goto restore;
1796 }
1797 if (fs_info->fs_devices->rw_devices == 0) {
1798 ret = -EACCES;
1799 goto restore;
1800 }
1801
1802 if (fs_info->fs_devices->missing_devices >
1803 fs_info->num_tolerated_disk_barrier_failures &&
1804 !(*flags & MS_RDONLY)) {
1805 btrfs_warn(fs_info,
1806 "too many missing devices, writeable remount is not allowed");
1807 ret = -EACCES;
1808 goto restore;
1809 }
1810
1811 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1812 ret = -EINVAL;
1813 goto restore;
1814 }
1815
1816 ret = btrfs_cleanup_fs_roots(fs_info);
1817 if (ret)
1818 goto restore;
1819
1820 /* recover relocation */
1821 mutex_lock(&fs_info->cleaner_mutex);
1822 ret = btrfs_recover_relocation(root);
1823 mutex_unlock(&fs_info->cleaner_mutex);
1824 if (ret)
1825 goto restore;
1826
1827 ret = btrfs_resume_balance_async(fs_info);
1828 if (ret)
1829 goto restore;
1830
1831 ret = btrfs_resume_dev_replace_async(fs_info);
1832 if (ret) {
1833 btrfs_warn(fs_info, "failed to resume dev_replace");
1834 goto restore;
1835 }
1836
1837 if (!fs_info->uuid_root) {
1838 btrfs_info(fs_info, "creating UUID tree");
1839 ret = btrfs_create_uuid_tree(fs_info);
1840 if (ret) {
1841 btrfs_warn(fs_info,
1842 "failed to create the UUID tree %d",
1843 ret);
1844 goto restore;
1845 }
1846 }
1847 sb->s_flags &= ~MS_RDONLY;
1848
1849 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1850 }
1851 out:
1852 wake_up_process(fs_info->transaction_kthread);
1853 btrfs_remount_cleanup(fs_info, old_opts);
1854 return 0;
1855
1856 restore:
1857 /* We've hit an error - don't reset MS_RDONLY */
1858 if (sb->s_flags & MS_RDONLY)
1859 old_flags |= MS_RDONLY;
1860 sb->s_flags = old_flags;
1861 fs_info->mount_opt = old_opts;
1862 fs_info->compress_type = old_compress_type;
1863 fs_info->max_inline = old_max_inline;
1864 mutex_lock(&fs_info->chunk_mutex);
1865 fs_info->alloc_start = old_alloc_start;
1866 mutex_unlock(&fs_info->chunk_mutex);
1867 btrfs_resize_thread_pool(fs_info,
1868 old_thread_pool_size, fs_info->thread_pool_size);
1869 fs_info->metadata_ratio = old_metadata_ratio;
1870 btrfs_remount_cleanup(fs_info, old_opts);
1871 return ret;
1872 }
1873
1874 /* Used to sort the devices by max_avail(descending sort) */
1875 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1876 const void *dev_info2)
1877 {
1878 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1879 ((struct btrfs_device_info *)dev_info2)->max_avail)
1880 return -1;
1881 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1882 ((struct btrfs_device_info *)dev_info2)->max_avail)
1883 return 1;
1884 else
1885 return 0;
1886 }
1887
1888 /*
1889 * sort the devices by max_avail, in which max free extent size of each device
1890 * is stored.(Descending Sort)
1891 */
1892 static inline void btrfs_descending_sort_devices(
1893 struct btrfs_device_info *devices,
1894 size_t nr_devices)
1895 {
1896 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1897 btrfs_cmp_device_free_bytes, NULL);
1898 }
1899
1900 /*
1901 * The helper to calc the free space on the devices that can be used to store
1902 * file data.
1903 */
1904 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1905 {
1906 struct btrfs_fs_info *fs_info = root->fs_info;
1907 struct btrfs_device_info *devices_info;
1908 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1909 struct btrfs_device *device;
1910 u64 skip_space;
1911 u64 type;
1912 u64 avail_space;
1913 u64 used_space;
1914 u64 min_stripe_size;
1915 int min_stripes = 1, num_stripes = 1;
1916 int i = 0, nr_devices;
1917 int ret;
1918
1919 /*
1920 * We aren't under the device list lock, so this is racy-ish, but good
1921 * enough for our purposes.
1922 */
1923 nr_devices = fs_info->fs_devices->open_devices;
1924 if (!nr_devices) {
1925 smp_mb();
1926 nr_devices = fs_info->fs_devices->open_devices;
1927 ASSERT(nr_devices);
1928 if (!nr_devices) {
1929 *free_bytes = 0;
1930 return 0;
1931 }
1932 }
1933
1934 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1935 GFP_NOFS);
1936 if (!devices_info)
1937 return -ENOMEM;
1938
1939 /* calc min stripe number for data space allocation */
1940 type = btrfs_get_alloc_profile(root, 1);
1941 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1942 min_stripes = 2;
1943 num_stripes = nr_devices;
1944 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1945 min_stripes = 2;
1946 num_stripes = 2;
1947 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1948 min_stripes = 4;
1949 num_stripes = 4;
1950 }
1951
1952 if (type & BTRFS_BLOCK_GROUP_DUP)
1953 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1954 else
1955 min_stripe_size = BTRFS_STRIPE_LEN;
1956
1957 if (fs_info->alloc_start)
1958 mutex_lock(&fs_devices->device_list_mutex);
1959 rcu_read_lock();
1960 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1961 if (!device->in_fs_metadata || !device->bdev ||
1962 device->is_tgtdev_for_dev_replace)
1963 continue;
1964
1965 if (i >= nr_devices)
1966 break;
1967
1968 avail_space = device->total_bytes - device->bytes_used;
1969
1970 /* align with stripe_len */
1971 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1972 avail_space *= BTRFS_STRIPE_LEN;
1973
1974 /*
1975 * In order to avoid overwriting the superblock on the drive,
1976 * btrfs starts at an offset of at least 1MB when doing chunk
1977 * allocation.
1978 */
1979 skip_space = SZ_1M;
1980
1981 /* user can set the offset in fs_info->alloc_start. */
1982 if (fs_info->alloc_start &&
1983 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1984 device->total_bytes) {
1985 rcu_read_unlock();
1986 skip_space = max(fs_info->alloc_start, skip_space);
1987
1988 /*
1989 * btrfs can not use the free space in
1990 * [0, skip_space - 1], we must subtract it from the
1991 * total. In order to implement it, we account the used
1992 * space in this range first.
1993 */
1994 ret = btrfs_account_dev_extents_size(device, 0,
1995 skip_space - 1,
1996 &used_space);
1997 if (ret) {
1998 kfree(devices_info);
1999 mutex_unlock(&fs_devices->device_list_mutex);
2000 return ret;
2001 }
2002
2003 rcu_read_lock();
2004
2005 /* calc the free space in [0, skip_space - 1] */
2006 skip_space -= used_space;
2007 }
2008
2009 /*
2010 * we can use the free space in [0, skip_space - 1], subtract
2011 * it from the total.
2012 */
2013 if (avail_space && avail_space >= skip_space)
2014 avail_space -= skip_space;
2015 else
2016 avail_space = 0;
2017
2018 if (avail_space < min_stripe_size)
2019 continue;
2020
2021 devices_info[i].dev = device;
2022 devices_info[i].max_avail = avail_space;
2023
2024 i++;
2025 }
2026 rcu_read_unlock();
2027 if (fs_info->alloc_start)
2028 mutex_unlock(&fs_devices->device_list_mutex);
2029
2030 nr_devices = i;
2031
2032 btrfs_descending_sort_devices(devices_info, nr_devices);
2033
2034 i = nr_devices - 1;
2035 avail_space = 0;
2036 while (nr_devices >= min_stripes) {
2037 if (num_stripes > nr_devices)
2038 num_stripes = nr_devices;
2039
2040 if (devices_info[i].max_avail >= min_stripe_size) {
2041 int j;
2042 u64 alloc_size;
2043
2044 avail_space += devices_info[i].max_avail * num_stripes;
2045 alloc_size = devices_info[i].max_avail;
2046 for (j = i + 1 - num_stripes; j <= i; j++)
2047 devices_info[j].max_avail -= alloc_size;
2048 }
2049 i--;
2050 nr_devices--;
2051 }
2052
2053 kfree(devices_info);
2054 *free_bytes = avail_space;
2055 return 0;
2056 }
2057
2058 /*
2059 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2060 *
2061 * If there's a redundant raid level at DATA block groups, use the respective
2062 * multiplier to scale the sizes.
2063 *
2064 * Unused device space usage is based on simulating the chunk allocator
2065 * algorithm that respects the device sizes, order of allocations and the
2066 * 'alloc_start' value, this is a close approximation of the actual use but
2067 * there are other factors that may change the result (like a new metadata
2068 * chunk).
2069 *
2070 * If metadata is exhausted, f_bavail will be 0.
2071 */
2072 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2073 {
2074 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2075 struct btrfs_super_block *disk_super = fs_info->super_copy;
2076 struct list_head *head = &fs_info->space_info;
2077 struct btrfs_space_info *found;
2078 u64 total_used = 0;
2079 u64 total_free_data = 0;
2080 u64 total_free_meta = 0;
2081 int bits = dentry->d_sb->s_blocksize_bits;
2082 __be32 *fsid = (__be32 *)fs_info->fsid;
2083 unsigned factor = 1;
2084 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2085 int ret;
2086 u64 thresh = 0;
2087 int mixed = 0;
2088
2089 /*
2090 * holding chunk_mutex to avoid allocating new chunks, holding
2091 * device_list_mutex to avoid the device being removed
2092 */
2093 rcu_read_lock();
2094 list_for_each_entry_rcu(found, head, list) {
2095 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2096 int i;
2097
2098 total_free_data += found->disk_total - found->disk_used;
2099 total_free_data -=
2100 btrfs_account_ro_block_groups_free_space(found);
2101
2102 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2103 if (!list_empty(&found->block_groups[i])) {
2104 switch (i) {
2105 case BTRFS_RAID_DUP:
2106 case BTRFS_RAID_RAID1:
2107 case BTRFS_RAID_RAID10:
2108 factor = 2;
2109 }
2110 }
2111 }
2112 }
2113
2114 /*
2115 * Metadata in mixed block goup profiles are accounted in data
2116 */
2117 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2118 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2119 mixed = 1;
2120 else
2121 total_free_meta += found->disk_total -
2122 found->disk_used;
2123 }
2124
2125 total_used += found->disk_used;
2126 }
2127
2128 rcu_read_unlock();
2129
2130 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2131 buf->f_blocks >>= bits;
2132 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2133
2134 /* Account global block reserve as used, it's in logical size already */
2135 spin_lock(&block_rsv->lock);
2136 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2137 if (buf->f_bfree >= block_rsv->size >> bits)
2138 buf->f_bfree -= block_rsv->size >> bits;
2139 else
2140 buf->f_bfree = 0;
2141 spin_unlock(&block_rsv->lock);
2142
2143 buf->f_bavail = div_u64(total_free_data, factor);
2144 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
2145 if (ret)
2146 return ret;
2147 buf->f_bavail += div_u64(total_free_data, factor);
2148 buf->f_bavail = buf->f_bavail >> bits;
2149
2150 /*
2151 * We calculate the remaining metadata space minus global reserve. If
2152 * this is (supposedly) smaller than zero, there's no space. But this
2153 * does not hold in practice, the exhausted state happens where's still
2154 * some positive delta. So we apply some guesswork and compare the
2155 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2156 *
2157 * We probably cannot calculate the exact threshold value because this
2158 * depends on the internal reservations requested by various
2159 * operations, so some operations that consume a few metadata will
2160 * succeed even if the Avail is zero. But this is better than the other
2161 * way around.
2162 */
2163 thresh = 4 * 1024 * 1024;
2164
2165 if (!mixed && total_free_meta - thresh < block_rsv->size)
2166 buf->f_bavail = 0;
2167
2168 buf->f_type = BTRFS_SUPER_MAGIC;
2169 buf->f_bsize = dentry->d_sb->s_blocksize;
2170 buf->f_namelen = BTRFS_NAME_LEN;
2171
2172 /* We treat it as constant endianness (it doesn't matter _which_)
2173 because we want the fsid to come out the same whether mounted
2174 on a big-endian or little-endian host */
2175 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2176 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2177 /* Mask in the root object ID too, to disambiguate subvols */
2178 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2179 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2180
2181 return 0;
2182 }
2183
2184 static void btrfs_kill_super(struct super_block *sb)
2185 {
2186 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2187 kill_anon_super(sb);
2188 free_fs_info(fs_info);
2189 }
2190
2191 static struct file_system_type btrfs_fs_type = {
2192 .owner = THIS_MODULE,
2193 .name = "btrfs",
2194 .mount = btrfs_mount,
2195 .kill_sb = btrfs_kill_super,
2196 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2197 };
2198 MODULE_ALIAS_FS("btrfs");
2199
2200 static int btrfs_control_open(struct inode *inode, struct file *file)
2201 {
2202 /*
2203 * The control file's private_data is used to hold the
2204 * transaction when it is started and is used to keep
2205 * track of whether a transaction is already in progress.
2206 */
2207 file->private_data = NULL;
2208 return 0;
2209 }
2210
2211 /*
2212 * used by btrfsctl to scan devices when no FS is mounted
2213 */
2214 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2215 unsigned long arg)
2216 {
2217 struct btrfs_ioctl_vol_args *vol;
2218 struct btrfs_fs_devices *fs_devices;
2219 int ret = -ENOTTY;
2220
2221 if (!capable(CAP_SYS_ADMIN))
2222 return -EPERM;
2223
2224 vol = memdup_user((void __user *)arg, sizeof(*vol));
2225 if (IS_ERR(vol))
2226 return PTR_ERR(vol);
2227
2228 switch (cmd) {
2229 case BTRFS_IOC_SCAN_DEV:
2230 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2231 &btrfs_fs_type, &fs_devices);
2232 break;
2233 case BTRFS_IOC_DEVICES_READY:
2234 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2235 &btrfs_fs_type, &fs_devices);
2236 if (ret)
2237 break;
2238 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2239 break;
2240 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2241 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2242 break;
2243 }
2244
2245 kfree(vol);
2246 return ret;
2247 }
2248
2249 static int btrfs_freeze(struct super_block *sb)
2250 {
2251 struct btrfs_trans_handle *trans;
2252 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2253
2254 root->fs_info->fs_frozen = 1;
2255 /*
2256 * We don't need a barrier here, we'll wait for any transaction that
2257 * could be in progress on other threads (and do delayed iputs that
2258 * we want to avoid on a frozen filesystem), or do the commit
2259 * ourselves.
2260 */
2261 trans = btrfs_attach_transaction_barrier(root);
2262 if (IS_ERR(trans)) {
2263 /* no transaction, don't bother */
2264 if (PTR_ERR(trans) == -ENOENT)
2265 return 0;
2266 return PTR_ERR(trans);
2267 }
2268 return btrfs_commit_transaction(trans, root);
2269 }
2270
2271 static int btrfs_unfreeze(struct super_block *sb)
2272 {
2273 struct btrfs_root *root = btrfs_sb(sb)->tree_root;
2274
2275 root->fs_info->fs_frozen = 0;
2276 return 0;
2277 }
2278
2279 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2280 {
2281 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2282 struct btrfs_fs_devices *cur_devices;
2283 struct btrfs_device *dev, *first_dev = NULL;
2284 struct list_head *head;
2285 struct rcu_string *name;
2286
2287 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2288 cur_devices = fs_info->fs_devices;
2289 while (cur_devices) {
2290 head = &cur_devices->devices;
2291 list_for_each_entry(dev, head, dev_list) {
2292 if (dev->missing)
2293 continue;
2294 if (!dev->name)
2295 continue;
2296 if (!first_dev || dev->devid < first_dev->devid)
2297 first_dev = dev;
2298 }
2299 cur_devices = cur_devices->seed;
2300 }
2301
2302 if (first_dev) {
2303 rcu_read_lock();
2304 name = rcu_dereference(first_dev->name);
2305 seq_escape(m, name->str, " \t\n\\");
2306 rcu_read_unlock();
2307 } else {
2308 WARN_ON(1);
2309 }
2310 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2311 return 0;
2312 }
2313
2314 static const struct super_operations btrfs_super_ops = {
2315 .drop_inode = btrfs_drop_inode,
2316 .evict_inode = btrfs_evict_inode,
2317 .put_super = btrfs_put_super,
2318 .sync_fs = btrfs_sync_fs,
2319 .show_options = btrfs_show_options,
2320 .show_devname = btrfs_show_devname,
2321 .write_inode = btrfs_write_inode,
2322 .alloc_inode = btrfs_alloc_inode,
2323 .destroy_inode = btrfs_destroy_inode,
2324 .statfs = btrfs_statfs,
2325 .remount_fs = btrfs_remount,
2326 .freeze_fs = btrfs_freeze,
2327 .unfreeze_fs = btrfs_unfreeze,
2328 };
2329
2330 static const struct file_operations btrfs_ctl_fops = {
2331 .open = btrfs_control_open,
2332 .unlocked_ioctl = btrfs_control_ioctl,
2333 .compat_ioctl = btrfs_control_ioctl,
2334 .owner = THIS_MODULE,
2335 .llseek = noop_llseek,
2336 };
2337
2338 static struct miscdevice btrfs_misc = {
2339 .minor = BTRFS_MINOR,
2340 .name = "btrfs-control",
2341 .fops = &btrfs_ctl_fops
2342 };
2343
2344 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2345 MODULE_ALIAS("devname:btrfs-control");
2346
2347 static int btrfs_interface_init(void)
2348 {
2349 return misc_register(&btrfs_misc);
2350 }
2351
2352 static void btrfs_interface_exit(void)
2353 {
2354 misc_deregister(&btrfs_misc);
2355 }
2356
2357 static void btrfs_print_mod_info(void)
2358 {
2359 pr_info("Btrfs loaded, crc32c=%s"
2360 #ifdef CONFIG_BTRFS_DEBUG
2361 ", debug=on"
2362 #endif
2363 #ifdef CONFIG_BTRFS_ASSERT
2364 ", assert=on"
2365 #endif
2366 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2367 ", integrity-checker=on"
2368 #endif
2369 "\n",
2370 btrfs_crc32c_impl());
2371 }
2372
2373 static int __init init_btrfs_fs(void)
2374 {
2375 int err;
2376
2377 err = btrfs_hash_init();
2378 if (err)
2379 return err;
2380
2381 btrfs_props_init();
2382
2383 err = btrfs_init_sysfs();
2384 if (err)
2385 goto free_hash;
2386
2387 btrfs_init_compress();
2388
2389 err = btrfs_init_cachep();
2390 if (err)
2391 goto free_compress;
2392
2393 err = extent_io_init();
2394 if (err)
2395 goto free_cachep;
2396
2397 err = extent_map_init();
2398 if (err)
2399 goto free_extent_io;
2400
2401 err = ordered_data_init();
2402 if (err)
2403 goto free_extent_map;
2404
2405 err = btrfs_delayed_inode_init();
2406 if (err)
2407 goto free_ordered_data;
2408
2409 err = btrfs_auto_defrag_init();
2410 if (err)
2411 goto free_delayed_inode;
2412
2413 err = btrfs_delayed_ref_init();
2414 if (err)
2415 goto free_auto_defrag;
2416
2417 err = btrfs_prelim_ref_init();
2418 if (err)
2419 goto free_delayed_ref;
2420
2421 err = btrfs_end_io_wq_init();
2422 if (err)
2423 goto free_prelim_ref;
2424
2425 err = btrfs_interface_init();
2426 if (err)
2427 goto free_end_io_wq;
2428
2429 btrfs_init_lockdep();
2430
2431 btrfs_print_mod_info();
2432
2433 err = btrfs_run_sanity_tests();
2434 if (err)
2435 goto unregister_ioctl;
2436
2437 err = register_filesystem(&btrfs_fs_type);
2438 if (err)
2439 goto unregister_ioctl;
2440
2441 return 0;
2442
2443 unregister_ioctl:
2444 btrfs_interface_exit();
2445 free_end_io_wq:
2446 btrfs_end_io_wq_exit();
2447 free_prelim_ref:
2448 btrfs_prelim_ref_exit();
2449 free_delayed_ref:
2450 btrfs_delayed_ref_exit();
2451 free_auto_defrag:
2452 btrfs_auto_defrag_exit();
2453 free_delayed_inode:
2454 btrfs_delayed_inode_exit();
2455 free_ordered_data:
2456 ordered_data_exit();
2457 free_extent_map:
2458 extent_map_exit();
2459 free_extent_io:
2460 extent_io_exit();
2461 free_cachep:
2462 btrfs_destroy_cachep();
2463 free_compress:
2464 btrfs_exit_compress();
2465 btrfs_exit_sysfs();
2466 free_hash:
2467 btrfs_hash_exit();
2468 return err;
2469 }
2470
2471 static void __exit exit_btrfs_fs(void)
2472 {
2473 btrfs_destroy_cachep();
2474 btrfs_delayed_ref_exit();
2475 btrfs_auto_defrag_exit();
2476 btrfs_delayed_inode_exit();
2477 btrfs_prelim_ref_exit();
2478 ordered_data_exit();
2479 extent_map_exit();
2480 extent_io_exit();
2481 btrfs_interface_exit();
2482 btrfs_end_io_wq_exit();
2483 unregister_filesystem(&btrfs_fs_type);
2484 btrfs_exit_sysfs();
2485 btrfs_cleanup_fs_uuids();
2486 btrfs_exit_compress();
2487 btrfs_hash_exit();
2488 }
2489
2490 late_initcall(init_btrfs_fs);
2491 module_exit(exit_btrfs_fs)
2492
2493 MODULE_LICENSE("GPL");
Linux with added FPC-III support