Loading...
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
67static const struct super_operations btrfs_super_ops;
68static struct file_system_type btrfs_fs_type;
69
70static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71
72const 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 */
101static 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
129void __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
175static 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 */
191static 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
202void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
203{
204 struct super_block *sb = fs_info->sb;
205 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
206 struct va_format vaf;
207 va_list args;
208 int kern_level;
209 const char *type = logtypes[4];
210 struct ratelimit_state *ratelimit = &printk_limits[4];
211
212 va_start(args, fmt);
213
214 while ((kern_level = printk_get_level(fmt)) != 0) {
215 size_t size = printk_skip_level(fmt) - fmt;
216
217 if (kern_level >= '0' && kern_level <= '7') {
218 memcpy(lvl, fmt, size);
219 lvl[size] = '\0';
220 type = logtypes[kern_level - '0'];
221 ratelimit = &printk_limits[kern_level - '0'];
222 }
223 fmt += size;
224 }
225
226 vaf.fmt = fmt;
227 vaf.va = &args;
228
229 if (__ratelimit(ratelimit))
230 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
231
232 va_end(args);
233}
234#endif
235
236/*
237 * We only mark the transaction aborted and then set the file system read-only.
238 * This will prevent new transactions from starting or trying to join this
239 * one.
240 *
241 * This means that error recovery at the call site is limited to freeing
242 * any local memory allocations and passing the error code up without
243 * further cleanup. The transaction should complete as it normally would
244 * in the call path but will return -EIO.
245 *
246 * We'll complete the cleanup in btrfs_end_transaction and
247 * btrfs_commit_transaction.
248 */
249__cold
250void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
251 const char *function,
252 unsigned int line, int errno)
253{
254 struct btrfs_fs_info *fs_info = trans->fs_info;
255
256 trans->aborted = errno;
257 /* Nothing used. The other threads that have joined this
258 * transaction may be able to continue. */
259 if (!trans->dirty && list_empty(&trans->new_bgs)) {
260 const char *errstr;
261
262 errstr = btrfs_decode_error(errno);
263 btrfs_warn(fs_info,
264 "%s:%d: Aborting unused transaction(%s).",
265 function, line, errstr);
266 return;
267 }
268 ACCESS_ONCE(trans->transaction->aborted) = errno;
269 /* Wake up anybody who may be waiting on this transaction */
270 wake_up(&fs_info->transaction_wait);
271 wake_up(&fs_info->transaction_blocked_wait);
272 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
273}
274/*
275 * __btrfs_panic decodes unexpected, fatal errors from the caller,
276 * issues an alert, and either panics or BUGs, depending on mount options.
277 */
278__cold
279void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
280 unsigned int line, int errno, const char *fmt, ...)
281{
282 char *s_id = "<unknown>";
283 const char *errstr;
284 struct va_format vaf = { .fmt = fmt };
285 va_list args;
286
287 if (fs_info)
288 s_id = fs_info->sb->s_id;
289
290 va_start(args, fmt);
291 vaf.va = &args;
292
293 errstr = btrfs_decode_error(errno);
294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
296 s_id, function, line, &vaf, errno, errstr);
297
298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
299 function, line, &vaf, errno, errstr);
300 va_end(args);
301 /* Caller calls BUG() */
302}
303
304static void btrfs_put_super(struct super_block *sb)
305{
306 close_ctree(btrfs_sb(sb));
307}
308
309enum {
310 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
311 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
312 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
313 Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
314 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
315 Opt_space_cache, Opt_space_cache_version, Opt_clear_cache,
316 Opt_user_subvol_rm_allowed, Opt_enospc_debug, Opt_subvolrootid,
317 Opt_defrag, Opt_inode_cache, Opt_no_space_cache, Opt_recovery,
318 Opt_skip_balance, Opt_check_integrity,
319 Opt_check_integrity_including_extent_data,
320 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
321 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
322 Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
323 Opt_datasum, Opt_treelog, Opt_noinode_cache, Opt_usebackuproot,
324 Opt_nologreplay, Opt_norecovery,
325#ifdef CONFIG_BTRFS_DEBUG
326 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
327#endif
328 Opt_err,
329};
330
331static const match_table_t tokens = {
332 {Opt_degraded, "degraded"},
333 {Opt_subvol, "subvol=%s"},
334 {Opt_subvolid, "subvolid=%s"},
335 {Opt_device, "device=%s"},
336 {Opt_nodatasum, "nodatasum"},
337 {Opt_datasum, "datasum"},
338 {Opt_nodatacow, "nodatacow"},
339 {Opt_datacow, "datacow"},
340 {Opt_nobarrier, "nobarrier"},
341 {Opt_barrier, "barrier"},
342 {Opt_max_inline, "max_inline=%s"},
343 {Opt_alloc_start, "alloc_start=%s"},
344 {Opt_thread_pool, "thread_pool=%d"},
345 {Opt_compress, "compress"},
346 {Opt_compress_type, "compress=%s"},
347 {Opt_compress_force, "compress-force"},
348 {Opt_compress_force_type, "compress-force=%s"},
349 {Opt_ssd, "ssd"},
350 {Opt_ssd_spread, "ssd_spread"},
351 {Opt_nossd, "nossd"},
352 {Opt_acl, "acl"},
353 {Opt_noacl, "noacl"},
354 {Opt_notreelog, "notreelog"},
355 {Opt_treelog, "treelog"},
356 {Opt_nologreplay, "nologreplay"},
357 {Opt_norecovery, "norecovery"},
358 {Opt_flushoncommit, "flushoncommit"},
359 {Opt_noflushoncommit, "noflushoncommit"},
360 {Opt_ratio, "metadata_ratio=%d"},
361 {Opt_discard, "discard"},
362 {Opt_nodiscard, "nodiscard"},
363 {Opt_space_cache, "space_cache"},
364 {Opt_space_cache_version, "space_cache=%s"},
365 {Opt_clear_cache, "clear_cache"},
366 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
367 {Opt_enospc_debug, "enospc_debug"},
368 {Opt_noenospc_debug, "noenospc_debug"},
369 {Opt_subvolrootid, "subvolrootid=%d"},
370 {Opt_defrag, "autodefrag"},
371 {Opt_nodefrag, "noautodefrag"},
372 {Opt_inode_cache, "inode_cache"},
373 {Opt_noinode_cache, "noinode_cache"},
374 {Opt_no_space_cache, "nospace_cache"},
375 {Opt_recovery, "recovery"}, /* deprecated */
376 {Opt_usebackuproot, "usebackuproot"},
377 {Opt_skip_balance, "skip_balance"},
378 {Opt_check_integrity, "check_int"},
379 {Opt_check_integrity_including_extent_data, "check_int_data"},
380 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
381 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
382 {Opt_fatal_errors, "fatal_errors=%s"},
383 {Opt_commit_interval, "commit=%d"},
384#ifdef CONFIG_BTRFS_DEBUG
385 {Opt_fragment_data, "fragment=data"},
386 {Opt_fragment_metadata, "fragment=metadata"},
387 {Opt_fragment_all, "fragment=all"},
388#endif
389 {Opt_err, NULL},
390};
391
392/*
393 * Regular mount options parser. Everything that is needed only when
394 * reading in a new superblock is parsed here.
395 * XXX JDM: This needs to be cleaned up for remount.
396 */
397int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
398 unsigned long new_flags)
399{
400 substring_t args[MAX_OPT_ARGS];
401 char *p, *num, *orig = NULL;
402 u64 cache_gen;
403 int intarg;
404 int ret = 0;
405 char *compress_type;
406 bool compress_force = false;
407 enum btrfs_compression_type saved_compress_type;
408 bool saved_compress_force;
409 int no_compress = 0;
410
411 cache_gen = btrfs_super_cache_generation(info->super_copy);
412 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
413 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
414 else if (cache_gen)
415 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
416
417 /*
418 * Even the options are empty, we still need to do extra check
419 * against new flags
420 */
421 if (!options)
422 goto check;
423
424 /*
425 * strsep changes the string, duplicate it because parse_options
426 * gets called twice
427 */
428 options = kstrdup(options, GFP_NOFS);
429 if (!options)
430 return -ENOMEM;
431
432 orig = options;
433
434 while ((p = strsep(&options, ",")) != NULL) {
435 int token;
436 if (!*p)
437 continue;
438
439 token = match_token(p, tokens, args);
440 switch (token) {
441 case Opt_degraded:
442 btrfs_info(info, "allowing degraded mounts");
443 btrfs_set_opt(info->mount_opt, DEGRADED);
444 break;
445 case Opt_subvol:
446 case Opt_subvolid:
447 case Opt_subvolrootid:
448 case Opt_device:
449 /*
450 * These are parsed by btrfs_parse_early_options
451 * and can be happily ignored here.
452 */
453 break;
454 case Opt_nodatasum:
455 btrfs_set_and_info(info, NODATASUM,
456 "setting nodatasum");
457 break;
458 case Opt_datasum:
459 if (btrfs_test_opt(info, NODATASUM)) {
460 if (btrfs_test_opt(info, NODATACOW))
461 btrfs_info(info,
462 "setting datasum, datacow enabled");
463 else
464 btrfs_info(info, "setting datasum");
465 }
466 btrfs_clear_opt(info->mount_opt, NODATACOW);
467 btrfs_clear_opt(info->mount_opt, NODATASUM);
468 break;
469 case Opt_nodatacow:
470 if (!btrfs_test_opt(info, NODATACOW)) {
471 if (!btrfs_test_opt(info, COMPRESS) ||
472 !btrfs_test_opt(info, FORCE_COMPRESS)) {
473 btrfs_info(info,
474 "setting nodatacow, compression disabled");
475 } else {
476 btrfs_info(info, "setting nodatacow");
477 }
478 }
479 btrfs_clear_opt(info->mount_opt, COMPRESS);
480 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
481 btrfs_set_opt(info->mount_opt, NODATACOW);
482 btrfs_set_opt(info->mount_opt, NODATASUM);
483 break;
484 case Opt_datacow:
485 btrfs_clear_and_info(info, NODATACOW,
486 "setting datacow");
487 break;
488 case Opt_compress_force:
489 case Opt_compress_force_type:
490 compress_force = true;
491 /* Fallthrough */
492 case Opt_compress:
493 case Opt_compress_type:
494 saved_compress_type = btrfs_test_opt(info,
495 COMPRESS) ?
496 info->compress_type : BTRFS_COMPRESS_NONE;
497 saved_compress_force =
498 btrfs_test_opt(info, FORCE_COMPRESS);
499 if (token == Opt_compress ||
500 token == Opt_compress_force ||
501 strcmp(args[0].from, "zlib") == 0) {
502 compress_type = "zlib";
503 info->compress_type = BTRFS_COMPRESS_ZLIB;
504 btrfs_set_opt(info->mount_opt, COMPRESS);
505 btrfs_clear_opt(info->mount_opt, NODATACOW);
506 btrfs_clear_opt(info->mount_opt, NODATASUM);
507 no_compress = 0;
508 } else if (strcmp(args[0].from, "lzo") == 0) {
509 compress_type = "lzo";
510 info->compress_type = BTRFS_COMPRESS_LZO;
511 btrfs_set_opt(info->mount_opt, COMPRESS);
512 btrfs_clear_opt(info->mount_opt, NODATACOW);
513 btrfs_clear_opt(info->mount_opt, NODATASUM);
514 btrfs_set_fs_incompat(info, COMPRESS_LZO);
515 no_compress = 0;
516 } else if (strncmp(args[0].from, "no", 2) == 0) {
517 compress_type = "no";
518 btrfs_clear_opt(info->mount_opt, COMPRESS);
519 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
520 compress_force = false;
521 no_compress++;
522 } else {
523 ret = -EINVAL;
524 goto out;
525 }
526
527 if (compress_force) {
528 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
529 } else {
530 /*
531 * If we remount from compress-force=xxx to
532 * compress=xxx, we need clear FORCE_COMPRESS
533 * flag, otherwise, there is no way for users
534 * to disable forcible compression separately.
535 */
536 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
537 }
538 if ((btrfs_test_opt(info, COMPRESS) &&
539 (info->compress_type != saved_compress_type ||
540 compress_force != saved_compress_force)) ||
541 (!btrfs_test_opt(info, COMPRESS) &&
542 no_compress == 1)) {
543 btrfs_info(info, "%s %s compression",
544 (compress_force) ? "force" : "use",
545 compress_type);
546 }
547 compress_force = false;
548 break;
549 case Opt_ssd:
550 btrfs_set_and_info(info, SSD,
551 "use ssd allocation scheme");
552 break;
553 case Opt_ssd_spread:
554 btrfs_set_and_info(info, SSD_SPREAD,
555 "use spread ssd allocation scheme");
556 btrfs_set_opt(info->mount_opt, SSD);
557 break;
558 case Opt_nossd:
559 btrfs_set_and_info(info, NOSSD,
560 "not using ssd allocation scheme");
561 btrfs_clear_opt(info->mount_opt, SSD);
562 break;
563 case Opt_barrier:
564 btrfs_clear_and_info(info, NOBARRIER,
565 "turning on barriers");
566 break;
567 case Opt_nobarrier:
568 btrfs_set_and_info(info, NOBARRIER,
569 "turning off barriers");
570 break;
571 case Opt_thread_pool:
572 ret = match_int(&args[0], &intarg);
573 if (ret) {
574 goto out;
575 } else if (intarg > 0) {
576 info->thread_pool_size = intarg;
577 } else {
578 ret = -EINVAL;
579 goto out;
580 }
581 break;
582 case Opt_max_inline:
583 num = match_strdup(&args[0]);
584 if (num) {
585 info->max_inline = memparse(num, NULL);
586 kfree(num);
587
588 if (info->max_inline) {
589 info->max_inline = min_t(u64,
590 info->max_inline,
591 info->sectorsize);
592 }
593 btrfs_info(info, "max_inline at %llu",
594 info->max_inline);
595 } else {
596 ret = -ENOMEM;
597 goto out;
598 }
599 break;
600 case Opt_alloc_start:
601 num = match_strdup(&args[0]);
602 if (num) {
603 mutex_lock(&info->chunk_mutex);
604 info->alloc_start = memparse(num, NULL);
605 mutex_unlock(&info->chunk_mutex);
606 kfree(num);
607 btrfs_info(info, "allocations start at %llu",
608 info->alloc_start);
609 } else {
610 ret = -ENOMEM;
611 goto out;
612 }
613 break;
614 case Opt_acl:
615#ifdef CONFIG_BTRFS_FS_POSIX_ACL
616 info->sb->s_flags |= MS_POSIXACL;
617 break;
618#else
619 btrfs_err(info, "support for ACL not compiled in!");
620 ret = -EINVAL;
621 goto out;
622#endif
623 case Opt_noacl:
624 info->sb->s_flags &= ~MS_POSIXACL;
625 break;
626 case Opt_notreelog:
627 btrfs_set_and_info(info, NOTREELOG,
628 "disabling tree log");
629 break;
630 case Opt_treelog:
631 btrfs_clear_and_info(info, NOTREELOG,
632 "enabling tree log");
633 break;
634 case Opt_norecovery:
635 case Opt_nologreplay:
636 btrfs_set_and_info(info, NOLOGREPLAY,
637 "disabling log replay at mount time");
638 break;
639 case Opt_flushoncommit:
640 btrfs_set_and_info(info, FLUSHONCOMMIT,
641 "turning on flush-on-commit");
642 break;
643 case Opt_noflushoncommit:
644 btrfs_clear_and_info(info, FLUSHONCOMMIT,
645 "turning off flush-on-commit");
646 break;
647 case Opt_ratio:
648 ret = match_int(&args[0], &intarg);
649 if (ret) {
650 goto out;
651 } else if (intarg >= 0) {
652 info->metadata_ratio = intarg;
653 btrfs_info(info, "metadata ratio %d",
654 info->metadata_ratio);
655 } else {
656 ret = -EINVAL;
657 goto out;
658 }
659 break;
660 case Opt_discard:
661 btrfs_set_and_info(info, DISCARD,
662 "turning on discard");
663 break;
664 case Opt_nodiscard:
665 btrfs_clear_and_info(info, DISCARD,
666 "turning off discard");
667 break;
668 case Opt_space_cache:
669 case Opt_space_cache_version:
670 if (token == Opt_space_cache ||
671 strcmp(args[0].from, "v1") == 0) {
672 btrfs_clear_opt(info->mount_opt,
673 FREE_SPACE_TREE);
674 btrfs_set_and_info(info, SPACE_CACHE,
675 "enabling disk space caching");
676 } else if (strcmp(args[0].from, "v2") == 0) {
677 btrfs_clear_opt(info->mount_opt,
678 SPACE_CACHE);
679 btrfs_set_and_info(info, FREE_SPACE_TREE,
680 "enabling free space tree");
681 } else {
682 ret = -EINVAL;
683 goto out;
684 }
685 break;
686 case Opt_rescan_uuid_tree:
687 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
688 break;
689 case Opt_no_space_cache:
690 if (btrfs_test_opt(info, SPACE_CACHE)) {
691 btrfs_clear_and_info(info, SPACE_CACHE,
692 "disabling disk space caching");
693 }
694 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
695 btrfs_clear_and_info(info, FREE_SPACE_TREE,
696 "disabling free space tree");
697 }
698 break;
699 case Opt_inode_cache:
700 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
701 "enabling inode map caching");
702 break;
703 case Opt_noinode_cache:
704 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
705 "disabling inode map caching");
706 break;
707 case Opt_clear_cache:
708 btrfs_set_and_info(info, CLEAR_CACHE,
709 "force clearing of disk cache");
710 break;
711 case Opt_user_subvol_rm_allowed:
712 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
713 break;
714 case Opt_enospc_debug:
715 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
716 break;
717 case Opt_noenospc_debug:
718 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
719 break;
720 case Opt_defrag:
721 btrfs_set_and_info(info, AUTO_DEFRAG,
722 "enabling auto defrag");
723 break;
724 case Opt_nodefrag:
725 btrfs_clear_and_info(info, AUTO_DEFRAG,
726 "disabling auto defrag");
727 break;
728 case Opt_recovery:
729 btrfs_warn(info,
730 "'recovery' is deprecated, use 'usebackuproot' instead");
731 case Opt_usebackuproot:
732 btrfs_info(info,
733 "trying to use backup root at mount time");
734 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
735 break;
736 case Opt_skip_balance:
737 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
738 break;
739#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
740 case Opt_check_integrity_including_extent_data:
741 btrfs_info(info,
742 "enabling check integrity including extent data");
743 btrfs_set_opt(info->mount_opt,
744 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
745 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
746 break;
747 case Opt_check_integrity:
748 btrfs_info(info, "enabling check integrity");
749 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
750 break;
751 case Opt_check_integrity_print_mask:
752 ret = match_int(&args[0], &intarg);
753 if (ret) {
754 goto out;
755 } else if (intarg >= 0) {
756 info->check_integrity_print_mask = intarg;
757 btrfs_info(info,
758 "check_integrity_print_mask 0x%x",
759 info->check_integrity_print_mask);
760 } else {
761 ret = -EINVAL;
762 goto out;
763 }
764 break;
765#else
766 case Opt_check_integrity_including_extent_data:
767 case Opt_check_integrity:
768 case Opt_check_integrity_print_mask:
769 btrfs_err(info,
770 "support for check_integrity* not compiled in!");
771 ret = -EINVAL;
772 goto out;
773#endif
774 case Opt_fatal_errors:
775 if (strcmp(args[0].from, "panic") == 0)
776 btrfs_set_opt(info->mount_opt,
777 PANIC_ON_FATAL_ERROR);
778 else if (strcmp(args[0].from, "bug") == 0)
779 btrfs_clear_opt(info->mount_opt,
780 PANIC_ON_FATAL_ERROR);
781 else {
782 ret = -EINVAL;
783 goto out;
784 }
785 break;
786 case Opt_commit_interval:
787 intarg = 0;
788 ret = match_int(&args[0], &intarg);
789 if (ret < 0) {
790 btrfs_err(info, "invalid commit interval");
791 ret = -EINVAL;
792 goto out;
793 }
794 if (intarg > 0) {
795 if (intarg > 300) {
796 btrfs_warn(info,
797 "excessive commit interval %d",
798 intarg);
799 }
800 info->commit_interval = intarg;
801 } else {
802 btrfs_info(info,
803 "using default commit interval %ds",
804 BTRFS_DEFAULT_COMMIT_INTERVAL);
805 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
806 }
807 break;
808#ifdef CONFIG_BTRFS_DEBUG
809 case Opt_fragment_all:
810 btrfs_info(info, "fragmenting all space");
811 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
812 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
813 break;
814 case Opt_fragment_metadata:
815 btrfs_info(info, "fragmenting metadata");
816 btrfs_set_opt(info->mount_opt,
817 FRAGMENT_METADATA);
818 break;
819 case Opt_fragment_data:
820 btrfs_info(info, "fragmenting data");
821 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
822 break;
823#endif
824 case Opt_err:
825 btrfs_info(info, "unrecognized mount option '%s'", p);
826 ret = -EINVAL;
827 goto out;
828 default:
829 break;
830 }
831 }
832check:
833 /*
834 * Extra check for current option against current flag
835 */
836 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & MS_RDONLY)) {
837 btrfs_err(info,
838 "nologreplay must be used with ro mount option");
839 ret = -EINVAL;
840 }
841out:
842 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
843 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
844 !btrfs_test_opt(info, CLEAR_CACHE)) {
845 btrfs_err(info, "cannot disable free space tree");
846 ret = -EINVAL;
847
848 }
849 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
850 btrfs_info(info, "disk space caching is enabled");
851 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
852 btrfs_info(info, "using free space tree");
853 kfree(orig);
854 return ret;
855}
856
857/*
858 * Parse mount options that are required early in the mount process.
859 *
860 * All other options will be parsed on much later in the mount process and
861 * only when we need to allocate a new super block.
862 */
863static int btrfs_parse_early_options(const char *options, fmode_t flags,
864 void *holder, char **subvol_name, u64 *subvol_objectid,
865 struct btrfs_fs_devices **fs_devices)
866{
867 substring_t args[MAX_OPT_ARGS];
868 char *device_name, *opts, *orig, *p;
869 char *num = NULL;
870 int error = 0;
871
872 if (!options)
873 return 0;
874
875 /*
876 * strsep changes the string, duplicate it because parse_options
877 * gets called twice
878 */
879 opts = kstrdup(options, GFP_KERNEL);
880 if (!opts)
881 return -ENOMEM;
882 orig = opts;
883
884 while ((p = strsep(&opts, ",")) != NULL) {
885 int token;
886 if (!*p)
887 continue;
888
889 token = match_token(p, tokens, args);
890 switch (token) {
891 case Opt_subvol:
892 kfree(*subvol_name);
893 *subvol_name = match_strdup(&args[0]);
894 if (!*subvol_name) {
895 error = -ENOMEM;
896 goto out;
897 }
898 break;
899 case Opt_subvolid:
900 num = match_strdup(&args[0]);
901 if (num) {
902 *subvol_objectid = memparse(num, NULL);
903 kfree(num);
904 /* we want the original fs_tree */
905 if (!*subvol_objectid)
906 *subvol_objectid =
907 BTRFS_FS_TREE_OBJECTID;
908 } else {
909 error = -EINVAL;
910 goto out;
911 }
912 break;
913 case Opt_subvolrootid:
914 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
915 break;
916 case Opt_device:
917 device_name = match_strdup(&args[0]);
918 if (!device_name) {
919 error = -ENOMEM;
920 goto out;
921 }
922 error = btrfs_scan_one_device(device_name,
923 flags, holder, fs_devices);
924 kfree(device_name);
925 if (error)
926 goto out;
927 break;
928 default:
929 break;
930 }
931 }
932
933out:
934 kfree(orig);
935 return error;
936}
937
938static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
939 u64 subvol_objectid)
940{
941 struct btrfs_root *root = fs_info->tree_root;
942 struct btrfs_root *fs_root;
943 struct btrfs_root_ref *root_ref;
944 struct btrfs_inode_ref *inode_ref;
945 struct btrfs_key key;
946 struct btrfs_path *path = NULL;
947 char *name = NULL, *ptr;
948 u64 dirid;
949 int len;
950 int ret;
951
952 path = btrfs_alloc_path();
953 if (!path) {
954 ret = -ENOMEM;
955 goto err;
956 }
957 path->leave_spinning = 1;
958
959 name = kmalloc(PATH_MAX, GFP_NOFS);
960 if (!name) {
961 ret = -ENOMEM;
962 goto err;
963 }
964 ptr = name + PATH_MAX - 1;
965 ptr[0] = '\0';
966
967 /*
968 * Walk up the subvolume trees in the tree of tree roots by root
969 * backrefs until we hit the top-level subvolume.
970 */
971 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
972 key.objectid = subvol_objectid;
973 key.type = BTRFS_ROOT_BACKREF_KEY;
974 key.offset = (u64)-1;
975
976 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
977 if (ret < 0) {
978 goto err;
979 } else if (ret > 0) {
980 ret = btrfs_previous_item(root, path, subvol_objectid,
981 BTRFS_ROOT_BACKREF_KEY);
982 if (ret < 0) {
983 goto err;
984 } else if (ret > 0) {
985 ret = -ENOENT;
986 goto err;
987 }
988 }
989
990 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
991 subvol_objectid = key.offset;
992
993 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
994 struct btrfs_root_ref);
995 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
996 ptr -= len + 1;
997 if (ptr < name) {
998 ret = -ENAMETOOLONG;
999 goto err;
1000 }
1001 read_extent_buffer(path->nodes[0], ptr + 1,
1002 (unsigned long)(root_ref + 1), len);
1003 ptr[0] = '/';
1004 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1005 btrfs_release_path(path);
1006
1007 key.objectid = subvol_objectid;
1008 key.type = BTRFS_ROOT_ITEM_KEY;
1009 key.offset = (u64)-1;
1010 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1011 if (IS_ERR(fs_root)) {
1012 ret = PTR_ERR(fs_root);
1013 goto err;
1014 }
1015
1016 /*
1017 * Walk up the filesystem tree by inode refs until we hit the
1018 * root directory.
1019 */
1020 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1021 key.objectid = dirid;
1022 key.type = BTRFS_INODE_REF_KEY;
1023 key.offset = (u64)-1;
1024
1025 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1026 if (ret < 0) {
1027 goto err;
1028 } else if (ret > 0) {
1029 ret = btrfs_previous_item(fs_root, path, dirid,
1030 BTRFS_INODE_REF_KEY);
1031 if (ret < 0) {
1032 goto err;
1033 } else if (ret > 0) {
1034 ret = -ENOENT;
1035 goto err;
1036 }
1037 }
1038
1039 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1040 dirid = key.offset;
1041
1042 inode_ref = btrfs_item_ptr(path->nodes[0],
1043 path->slots[0],
1044 struct btrfs_inode_ref);
1045 len = btrfs_inode_ref_name_len(path->nodes[0],
1046 inode_ref);
1047 ptr -= len + 1;
1048 if (ptr < name) {
1049 ret = -ENAMETOOLONG;
1050 goto err;
1051 }
1052 read_extent_buffer(path->nodes[0], ptr + 1,
1053 (unsigned long)(inode_ref + 1), len);
1054 ptr[0] = '/';
1055 btrfs_release_path(path);
1056 }
1057 }
1058
1059 btrfs_free_path(path);
1060 if (ptr == name + PATH_MAX - 1) {
1061 name[0] = '/';
1062 name[1] = '\0';
1063 } else {
1064 memmove(name, ptr, name + PATH_MAX - ptr);
1065 }
1066 return name;
1067
1068err:
1069 btrfs_free_path(path);
1070 kfree(name);
1071 return ERR_PTR(ret);
1072}
1073
1074static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1075{
1076 struct btrfs_root *root = fs_info->tree_root;
1077 struct btrfs_dir_item *di;
1078 struct btrfs_path *path;
1079 struct btrfs_key location;
1080 u64 dir_id;
1081
1082 path = btrfs_alloc_path();
1083 if (!path)
1084 return -ENOMEM;
1085 path->leave_spinning = 1;
1086
1087 /*
1088 * Find the "default" dir item which points to the root item that we
1089 * will mount by default if we haven't been given a specific subvolume
1090 * to mount.
1091 */
1092 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1093 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1094 if (IS_ERR(di)) {
1095 btrfs_free_path(path);
1096 return PTR_ERR(di);
1097 }
1098 if (!di) {
1099 /*
1100 * Ok the default dir item isn't there. This is weird since
1101 * it's always been there, but don't freak out, just try and
1102 * mount the top-level subvolume.
1103 */
1104 btrfs_free_path(path);
1105 *objectid = BTRFS_FS_TREE_OBJECTID;
1106 return 0;
1107 }
1108
1109 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1110 btrfs_free_path(path);
1111 *objectid = location.objectid;
1112 return 0;
1113}
1114
1115static int btrfs_fill_super(struct super_block *sb,
1116 struct btrfs_fs_devices *fs_devices,
1117 void *data, int silent)
1118{
1119 struct inode *inode;
1120 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1121 struct btrfs_key key;
1122 int err;
1123
1124 sb->s_maxbytes = MAX_LFS_FILESIZE;
1125 sb->s_magic = BTRFS_SUPER_MAGIC;
1126 sb->s_op = &btrfs_super_ops;
1127 sb->s_d_op = &btrfs_dentry_operations;
1128 sb->s_export_op = &btrfs_export_ops;
1129 sb->s_xattr = btrfs_xattr_handlers;
1130 sb->s_time_gran = 1;
1131#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1132 sb->s_flags |= MS_POSIXACL;
1133#endif
1134 sb->s_flags |= MS_I_VERSION;
1135 sb->s_iflags |= SB_I_CGROUPWB;
1136 err = open_ctree(sb, fs_devices, (char *)data);
1137 if (err) {
1138 btrfs_err(fs_info, "open_ctree failed");
1139 return err;
1140 }
1141
1142 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1143 key.type = BTRFS_INODE_ITEM_KEY;
1144 key.offset = 0;
1145 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1146 if (IS_ERR(inode)) {
1147 err = PTR_ERR(inode);
1148 goto fail_close;
1149 }
1150
1151 sb->s_root = d_make_root(inode);
1152 if (!sb->s_root) {
1153 err = -ENOMEM;
1154 goto fail_close;
1155 }
1156
1157 save_mount_options(sb, data);
1158 cleancache_init_fs(sb);
1159 sb->s_flags |= MS_ACTIVE;
1160 return 0;
1161
1162fail_close:
1163 close_ctree(fs_info);
1164 return err;
1165}
1166
1167int btrfs_sync_fs(struct super_block *sb, int wait)
1168{
1169 struct btrfs_trans_handle *trans;
1170 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1171 struct btrfs_root *root = fs_info->tree_root;
1172
1173 trace_btrfs_sync_fs(fs_info, wait);
1174
1175 if (!wait) {
1176 filemap_flush(fs_info->btree_inode->i_mapping);
1177 return 0;
1178 }
1179
1180 btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
1181
1182 trans = btrfs_attach_transaction_barrier(root);
1183 if (IS_ERR(trans)) {
1184 /* no transaction, don't bother */
1185 if (PTR_ERR(trans) == -ENOENT) {
1186 /*
1187 * Exit unless we have some pending changes
1188 * that need to go through commit
1189 */
1190 if (fs_info->pending_changes == 0)
1191 return 0;
1192 /*
1193 * A non-blocking test if the fs is frozen. We must not
1194 * start a new transaction here otherwise a deadlock
1195 * happens. The pending operations are delayed to the
1196 * next commit after thawing.
1197 */
1198 if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
1199 __sb_end_write(sb, SB_FREEZE_WRITE);
1200 else
1201 return 0;
1202 trans = btrfs_start_transaction(root, 0);
1203 }
1204 if (IS_ERR(trans))
1205 return PTR_ERR(trans);
1206 }
1207 return btrfs_commit_transaction(trans);
1208}
1209
1210static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1211{
1212 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1213 char *compress_type;
1214
1215 if (btrfs_test_opt(info, DEGRADED))
1216 seq_puts(seq, ",degraded");
1217 if (btrfs_test_opt(info, NODATASUM))
1218 seq_puts(seq, ",nodatasum");
1219 if (btrfs_test_opt(info, NODATACOW))
1220 seq_puts(seq, ",nodatacow");
1221 if (btrfs_test_opt(info, NOBARRIER))
1222 seq_puts(seq, ",nobarrier");
1223 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1224 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1225 if (info->alloc_start != 0)
1226 seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1227 if (info->thread_pool_size != min_t(unsigned long,
1228 num_online_cpus() + 2, 8))
1229 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1230 if (btrfs_test_opt(info, COMPRESS)) {
1231 if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1232 compress_type = "zlib";
1233 else
1234 compress_type = "lzo";
1235 if (btrfs_test_opt(info, FORCE_COMPRESS))
1236 seq_printf(seq, ",compress-force=%s", compress_type);
1237 else
1238 seq_printf(seq, ",compress=%s", compress_type);
1239 }
1240 if (btrfs_test_opt(info, NOSSD))
1241 seq_puts(seq, ",nossd");
1242 if (btrfs_test_opt(info, SSD_SPREAD))
1243 seq_puts(seq, ",ssd_spread");
1244 else if (btrfs_test_opt(info, SSD))
1245 seq_puts(seq, ",ssd");
1246 if (btrfs_test_opt(info, NOTREELOG))
1247 seq_puts(seq, ",notreelog");
1248 if (btrfs_test_opt(info, NOLOGREPLAY))
1249 seq_puts(seq, ",nologreplay");
1250 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1251 seq_puts(seq, ",flushoncommit");
1252 if (btrfs_test_opt(info, DISCARD))
1253 seq_puts(seq, ",discard");
1254 if (!(info->sb->s_flags & MS_POSIXACL))
1255 seq_puts(seq, ",noacl");
1256 if (btrfs_test_opt(info, SPACE_CACHE))
1257 seq_puts(seq, ",space_cache");
1258 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1259 seq_puts(seq, ",space_cache=v2");
1260 else
1261 seq_puts(seq, ",nospace_cache");
1262 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1263 seq_puts(seq, ",rescan_uuid_tree");
1264 if (btrfs_test_opt(info, CLEAR_CACHE))
1265 seq_puts(seq, ",clear_cache");
1266 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1267 seq_puts(seq, ",user_subvol_rm_allowed");
1268 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1269 seq_puts(seq, ",enospc_debug");
1270 if (btrfs_test_opt(info, AUTO_DEFRAG))
1271 seq_puts(seq, ",autodefrag");
1272 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1273 seq_puts(seq, ",inode_cache");
1274 if (btrfs_test_opt(info, SKIP_BALANCE))
1275 seq_puts(seq, ",skip_balance");
1276#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1277 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1278 seq_puts(seq, ",check_int_data");
1279 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1280 seq_puts(seq, ",check_int");
1281 if (info->check_integrity_print_mask)
1282 seq_printf(seq, ",check_int_print_mask=%d",
1283 info->check_integrity_print_mask);
1284#endif
1285 if (info->metadata_ratio)
1286 seq_printf(seq, ",metadata_ratio=%d",
1287 info->metadata_ratio);
1288 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1289 seq_puts(seq, ",fatal_errors=panic");
1290 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1291 seq_printf(seq, ",commit=%d", info->commit_interval);
1292#ifdef CONFIG_BTRFS_DEBUG
1293 if (btrfs_test_opt(info, FRAGMENT_DATA))
1294 seq_puts(seq, ",fragment=data");
1295 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1296 seq_puts(seq, ",fragment=metadata");
1297#endif
1298 seq_printf(seq, ",subvolid=%llu",
1299 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1300 seq_puts(seq, ",subvol=");
1301 seq_dentry(seq, dentry, " \t\n\\");
1302 return 0;
1303}
1304
1305static int btrfs_test_super(struct super_block *s, void *data)
1306{
1307 struct btrfs_fs_info *p = data;
1308 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1309
1310 return fs_info->fs_devices == p->fs_devices;
1311}
1312
1313static int btrfs_set_super(struct super_block *s, void *data)
1314{
1315 int err = set_anon_super(s, data);
1316 if (!err)
1317 s->s_fs_info = data;
1318 return err;
1319}
1320
1321/*
1322 * subvolumes are identified by ino 256
1323 */
1324static inline int is_subvolume_inode(struct inode *inode)
1325{
1326 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1327 return 1;
1328 return 0;
1329}
1330
1331/*
1332 * This will add subvolid=0 to the argument string while removing any subvol=
1333 * and subvolid= arguments to make sure we get the top-level root for path
1334 * walking to the subvol we want.
1335 */
1336static char *setup_root_args(char *args)
1337{
1338 char *buf, *dst, *sep;
1339
1340 if (!args)
1341 return kstrdup("subvolid=0", GFP_NOFS);
1342
1343 /* The worst case is that we add ",subvolid=0" to the end. */
1344 buf = dst = kmalloc(strlen(args) + strlen(",subvolid=0") + 1, GFP_NOFS);
1345 if (!buf)
1346 return NULL;
1347
1348 while (1) {
1349 sep = strchrnul(args, ',');
1350 if (!strstarts(args, "subvol=") &&
1351 !strstarts(args, "subvolid=")) {
1352 memcpy(dst, args, sep - args);
1353 dst += sep - args;
1354 *dst++ = ',';
1355 }
1356 if (*sep)
1357 args = sep + 1;
1358 else
1359 break;
1360 }
1361 strcpy(dst, "subvolid=0");
1362
1363 return buf;
1364}
1365
1366static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1367 int flags, const char *device_name,
1368 char *data)
1369{
1370 struct dentry *root;
1371 struct vfsmount *mnt = NULL;
1372 char *newargs;
1373 int ret;
1374
1375 newargs = setup_root_args(data);
1376 if (!newargs) {
1377 root = ERR_PTR(-ENOMEM);
1378 goto out;
1379 }
1380
1381 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, newargs);
1382 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY) {
1383 if (flags & MS_RDONLY) {
1384 mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY,
1385 device_name, newargs);
1386 } else {
1387 mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY,
1388 device_name, newargs);
1389 if (IS_ERR(mnt)) {
1390 root = ERR_CAST(mnt);
1391 mnt = NULL;
1392 goto out;
1393 }
1394
1395 down_write(&mnt->mnt_sb->s_umount);
1396 ret = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1397 up_write(&mnt->mnt_sb->s_umount);
1398 if (ret < 0) {
1399 root = ERR_PTR(ret);
1400 goto out;
1401 }
1402 }
1403 }
1404 if (IS_ERR(mnt)) {
1405 root = ERR_CAST(mnt);
1406 mnt = NULL;
1407 goto out;
1408 }
1409
1410 if (!subvol_name) {
1411 if (!subvol_objectid) {
1412 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1413 &subvol_objectid);
1414 if (ret) {
1415 root = ERR_PTR(ret);
1416 goto out;
1417 }
1418 }
1419 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1420 subvol_objectid);
1421 if (IS_ERR(subvol_name)) {
1422 root = ERR_CAST(subvol_name);
1423 subvol_name = NULL;
1424 goto out;
1425 }
1426
1427 }
1428
1429 root = mount_subtree(mnt, subvol_name);
1430 /* mount_subtree() drops our reference on the vfsmount. */
1431 mnt = NULL;
1432
1433 if (!IS_ERR(root)) {
1434 struct super_block *s = root->d_sb;
1435 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1436 struct inode *root_inode = d_inode(root);
1437 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1438
1439 ret = 0;
1440 if (!is_subvolume_inode(root_inode)) {
1441 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1442 subvol_name);
1443 ret = -EINVAL;
1444 }
1445 if (subvol_objectid && root_objectid != subvol_objectid) {
1446 /*
1447 * This will also catch a race condition where a
1448 * subvolume which was passed by ID is renamed and
1449 * another subvolume is renamed over the old location.
1450 */
1451 btrfs_err(fs_info,
1452 "subvol '%s' does not match subvolid %llu",
1453 subvol_name, subvol_objectid);
1454 ret = -EINVAL;
1455 }
1456 if (ret) {
1457 dput(root);
1458 root = ERR_PTR(ret);
1459 deactivate_locked_super(s);
1460 }
1461 }
1462
1463out:
1464 mntput(mnt);
1465 kfree(newargs);
1466 kfree(subvol_name);
1467 return root;
1468}
1469
1470static int parse_security_options(char *orig_opts,
1471 struct security_mnt_opts *sec_opts)
1472{
1473 char *secdata = NULL;
1474 int ret = 0;
1475
1476 secdata = alloc_secdata();
1477 if (!secdata)
1478 return -ENOMEM;
1479 ret = security_sb_copy_data(orig_opts, secdata);
1480 if (ret) {
1481 free_secdata(secdata);
1482 return ret;
1483 }
1484 ret = security_sb_parse_opts_str(secdata, sec_opts);
1485 free_secdata(secdata);
1486 return ret;
1487}
1488
1489static int setup_security_options(struct btrfs_fs_info *fs_info,
1490 struct super_block *sb,
1491 struct security_mnt_opts *sec_opts)
1492{
1493 int ret = 0;
1494
1495 /*
1496 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1497 * is valid.
1498 */
1499 ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1500 if (ret)
1501 return ret;
1502
1503#ifdef CONFIG_SECURITY
1504 if (!fs_info->security_opts.num_mnt_opts) {
1505 /* first time security setup, copy sec_opts to fs_info */
1506 memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1507 } else {
1508 /*
1509 * Since SELinux (the only one supporting security_mnt_opts)
1510 * does NOT support changing context during remount/mount of
1511 * the same sb, this must be the same or part of the same
1512 * security options, just free it.
1513 */
1514 security_free_mnt_opts(sec_opts);
1515 }
1516#endif
1517 return ret;
1518}
1519
1520/*
1521 * Find a superblock for the given device / mount point.
1522 *
1523 * Note: This is based on get_sb_bdev from fs/super.c with a few additions
1524 * for multiple device setup. Make sure to keep it in sync.
1525 */
1526static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1527 const char *device_name, void *data)
1528{
1529 struct block_device *bdev = NULL;
1530 struct super_block *s;
1531 struct btrfs_fs_devices *fs_devices = NULL;
1532 struct btrfs_fs_info *fs_info = NULL;
1533 struct security_mnt_opts new_sec_opts;
1534 fmode_t mode = FMODE_READ;
1535 char *subvol_name = NULL;
1536 u64 subvol_objectid = 0;
1537 int error = 0;
1538
1539 if (!(flags & MS_RDONLY))
1540 mode |= FMODE_WRITE;
1541
1542 error = btrfs_parse_early_options(data, mode, fs_type,
1543 &subvol_name, &subvol_objectid,
1544 &fs_devices);
1545 if (error) {
1546 kfree(subvol_name);
1547 return ERR_PTR(error);
1548 }
1549
1550 if (subvol_name || subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1551 /* mount_subvol() will free subvol_name. */
1552 return mount_subvol(subvol_name, subvol_objectid, flags,
1553 device_name, data);
1554 }
1555
1556 security_init_mnt_opts(&new_sec_opts);
1557 if (data) {
1558 error = parse_security_options(data, &new_sec_opts);
1559 if (error)
1560 return ERR_PTR(error);
1561 }
1562
1563 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1564 if (error)
1565 goto error_sec_opts;
1566
1567 /*
1568 * Setup a dummy root and fs_info for test/set super. This is because
1569 * we don't actually fill this stuff out until open_ctree, but we need
1570 * it for searching for existing supers, so this lets us do that and
1571 * then open_ctree will properly initialize everything later.
1572 */
1573 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1574 if (!fs_info) {
1575 error = -ENOMEM;
1576 goto error_sec_opts;
1577 }
1578
1579 fs_info->fs_devices = fs_devices;
1580
1581 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1582 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1583 security_init_mnt_opts(&fs_info->security_opts);
1584 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1585 error = -ENOMEM;
1586 goto error_fs_info;
1587 }
1588
1589 error = btrfs_open_devices(fs_devices, mode, fs_type);
1590 if (error)
1591 goto error_fs_info;
1592
1593 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1594 error = -EACCES;
1595 goto error_close_devices;
1596 }
1597
1598 bdev = fs_devices->latest_bdev;
1599 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1600 fs_info);
1601 if (IS_ERR(s)) {
1602 error = PTR_ERR(s);
1603 goto error_close_devices;
1604 }
1605
1606 if (s->s_root) {
1607 btrfs_close_devices(fs_devices);
1608 free_fs_info(fs_info);
1609 if ((flags ^ s->s_flags) & MS_RDONLY)
1610 error = -EBUSY;
1611 } else {
1612 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1613 btrfs_sb(s)->bdev_holder = fs_type;
1614 error = btrfs_fill_super(s, fs_devices, data,
1615 flags & MS_SILENT ? 1 : 0);
1616 }
1617 if (error) {
1618 deactivate_locked_super(s);
1619 goto error_sec_opts;
1620 }
1621
1622 fs_info = btrfs_sb(s);
1623 error = setup_security_options(fs_info, s, &new_sec_opts);
1624 if (error) {
1625 deactivate_locked_super(s);
1626 goto error_sec_opts;
1627 }
1628
1629 return dget(s->s_root);
1630
1631error_close_devices:
1632 btrfs_close_devices(fs_devices);
1633error_fs_info:
1634 free_fs_info(fs_info);
1635error_sec_opts:
1636 security_free_mnt_opts(&new_sec_opts);
1637 return ERR_PTR(error);
1638}
1639
1640static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1641 int new_pool_size, int old_pool_size)
1642{
1643 if (new_pool_size == old_pool_size)
1644 return;
1645
1646 fs_info->thread_pool_size = new_pool_size;
1647
1648 btrfs_info(fs_info, "resize thread pool %d -> %d",
1649 old_pool_size, new_pool_size);
1650
1651 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1652 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1653 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1654 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1655 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1656 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1657 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1658 new_pool_size);
1659 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1660 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1661 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1662 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1663 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1664 new_pool_size);
1665}
1666
1667static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1668{
1669 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1670}
1671
1672static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1673 unsigned long old_opts, int flags)
1674{
1675 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1676 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1677 (flags & MS_RDONLY))) {
1678 /* wait for any defraggers to finish */
1679 wait_event(fs_info->transaction_wait,
1680 (atomic_read(&fs_info->defrag_running) == 0));
1681 if (flags & MS_RDONLY)
1682 sync_filesystem(fs_info->sb);
1683 }
1684}
1685
1686static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1687 unsigned long old_opts)
1688{
1689 /*
1690 * We need to cleanup all defragable inodes if the autodefragment is
1691 * close or the filesystem is read only.
1692 */
1693 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1694 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1695 (fs_info->sb->s_flags & MS_RDONLY))) {
1696 btrfs_cleanup_defrag_inodes(fs_info);
1697 }
1698
1699 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1700}
1701
1702static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1703{
1704 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1705 struct btrfs_root *root = fs_info->tree_root;
1706 unsigned old_flags = sb->s_flags;
1707 unsigned long old_opts = fs_info->mount_opt;
1708 unsigned long old_compress_type = fs_info->compress_type;
1709 u64 old_max_inline = fs_info->max_inline;
1710 u64 old_alloc_start = fs_info->alloc_start;
1711 int old_thread_pool_size = fs_info->thread_pool_size;
1712 unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1713 int ret;
1714
1715 sync_filesystem(sb);
1716 btrfs_remount_prepare(fs_info);
1717
1718 if (data) {
1719 struct security_mnt_opts new_sec_opts;
1720
1721 security_init_mnt_opts(&new_sec_opts);
1722 ret = parse_security_options(data, &new_sec_opts);
1723 if (ret)
1724 goto restore;
1725 ret = setup_security_options(fs_info, sb,
1726 &new_sec_opts);
1727 if (ret) {
1728 security_free_mnt_opts(&new_sec_opts);
1729 goto restore;
1730 }
1731 }
1732
1733 ret = btrfs_parse_options(fs_info, data, *flags);
1734 if (ret) {
1735 ret = -EINVAL;
1736 goto restore;
1737 }
1738
1739 btrfs_remount_begin(fs_info, old_opts, *flags);
1740 btrfs_resize_thread_pool(fs_info,
1741 fs_info->thread_pool_size, old_thread_pool_size);
1742
1743 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1744 goto out;
1745
1746 if (*flags & MS_RDONLY) {
1747 /*
1748 * this also happens on 'umount -rf' or on shutdown, when
1749 * the filesystem is busy.
1750 */
1751 cancel_work_sync(&fs_info->async_reclaim_work);
1752
1753 /* wait for the uuid_scan task to finish */
1754 down(&fs_info->uuid_tree_rescan_sem);
1755 /* avoid complains from lockdep et al. */
1756 up(&fs_info->uuid_tree_rescan_sem);
1757
1758 sb->s_flags |= MS_RDONLY;
1759
1760 /*
1761 * Setting MS_RDONLY will put the cleaner thread to
1762 * sleep at the next loop if it's already active.
1763 * If it's already asleep, we'll leave unused block
1764 * groups on disk until we're mounted read-write again
1765 * unless we clean them up here.
1766 */
1767 btrfs_delete_unused_bgs(fs_info);
1768
1769 btrfs_dev_replace_suspend_for_unmount(fs_info);
1770 btrfs_scrub_cancel(fs_info);
1771 btrfs_pause_balance(fs_info);
1772
1773 ret = btrfs_commit_super(fs_info);
1774 if (ret)
1775 goto restore;
1776 } else {
1777 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1778 btrfs_err(fs_info,
1779 "Remounting read-write after error is not allowed");
1780 ret = -EINVAL;
1781 goto restore;
1782 }
1783 if (fs_info->fs_devices->rw_devices == 0) {
1784 ret = -EACCES;
1785 goto restore;
1786 }
1787
1788 if (fs_info->fs_devices->missing_devices >
1789 fs_info->num_tolerated_disk_barrier_failures &&
1790 !(*flags & MS_RDONLY)) {
1791 btrfs_warn(fs_info,
1792 "too many missing devices, writeable remount is not allowed");
1793 ret = -EACCES;
1794 goto restore;
1795 }
1796
1797 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1798 ret = -EINVAL;
1799 goto restore;
1800 }
1801
1802 ret = btrfs_cleanup_fs_roots(fs_info);
1803 if (ret)
1804 goto restore;
1805
1806 /* recover relocation */
1807 mutex_lock(&fs_info->cleaner_mutex);
1808 ret = btrfs_recover_relocation(root);
1809 mutex_unlock(&fs_info->cleaner_mutex);
1810 if (ret)
1811 goto restore;
1812
1813 ret = btrfs_resume_balance_async(fs_info);
1814 if (ret)
1815 goto restore;
1816
1817 ret = btrfs_resume_dev_replace_async(fs_info);
1818 if (ret) {
1819 btrfs_warn(fs_info, "failed to resume dev_replace");
1820 goto restore;
1821 }
1822
1823 if (!fs_info->uuid_root) {
1824 btrfs_info(fs_info, "creating UUID tree");
1825 ret = btrfs_create_uuid_tree(fs_info);
1826 if (ret) {
1827 btrfs_warn(fs_info,
1828 "failed to create the UUID tree %d",
1829 ret);
1830 goto restore;
1831 }
1832 }
1833 sb->s_flags &= ~MS_RDONLY;
1834
1835 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1836 }
1837out:
1838 wake_up_process(fs_info->transaction_kthread);
1839 btrfs_remount_cleanup(fs_info, old_opts);
1840 return 0;
1841
1842restore:
1843 /* We've hit an error - don't reset MS_RDONLY */
1844 if (sb->s_flags & MS_RDONLY)
1845 old_flags |= MS_RDONLY;
1846 sb->s_flags = old_flags;
1847 fs_info->mount_opt = old_opts;
1848 fs_info->compress_type = old_compress_type;
1849 fs_info->max_inline = old_max_inline;
1850 mutex_lock(&fs_info->chunk_mutex);
1851 fs_info->alloc_start = old_alloc_start;
1852 mutex_unlock(&fs_info->chunk_mutex);
1853 btrfs_resize_thread_pool(fs_info,
1854 old_thread_pool_size, fs_info->thread_pool_size);
1855 fs_info->metadata_ratio = old_metadata_ratio;
1856 btrfs_remount_cleanup(fs_info, old_opts);
1857 return ret;
1858}
1859
1860/* Used to sort the devices by max_avail(descending sort) */
1861static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1862 const void *dev_info2)
1863{
1864 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1865 ((struct btrfs_device_info *)dev_info2)->max_avail)
1866 return -1;
1867 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1868 ((struct btrfs_device_info *)dev_info2)->max_avail)
1869 return 1;
1870 else
1871 return 0;
1872}
1873
1874/*
1875 * sort the devices by max_avail, in which max free extent size of each device
1876 * is stored.(Descending Sort)
1877 */
1878static inline void btrfs_descending_sort_devices(
1879 struct btrfs_device_info *devices,
1880 size_t nr_devices)
1881{
1882 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1883 btrfs_cmp_device_free_bytes, NULL);
1884}
1885
1886/*
1887 * The helper to calc the free space on the devices that can be used to store
1888 * file data.
1889 */
1890static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1891 u64 *free_bytes)
1892{
1893 struct btrfs_root *root = fs_info->tree_root;
1894 struct btrfs_device_info *devices_info;
1895 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1896 struct btrfs_device *device;
1897 u64 skip_space;
1898 u64 type;
1899 u64 avail_space;
1900 u64 used_space;
1901 u64 min_stripe_size;
1902 int min_stripes = 1, num_stripes = 1;
1903 int i = 0, nr_devices;
1904 int ret;
1905
1906 /*
1907 * We aren't under the device list lock, so this is racy-ish, but good
1908 * enough for our purposes.
1909 */
1910 nr_devices = fs_info->fs_devices->open_devices;
1911 if (!nr_devices) {
1912 smp_mb();
1913 nr_devices = fs_info->fs_devices->open_devices;
1914 ASSERT(nr_devices);
1915 if (!nr_devices) {
1916 *free_bytes = 0;
1917 return 0;
1918 }
1919 }
1920
1921 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1922 GFP_NOFS);
1923 if (!devices_info)
1924 return -ENOMEM;
1925
1926 /* calc min stripe number for data space allocation */
1927 type = btrfs_get_alloc_profile(root, 1);
1928 if (type & BTRFS_BLOCK_GROUP_RAID0) {
1929 min_stripes = 2;
1930 num_stripes = nr_devices;
1931 } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1932 min_stripes = 2;
1933 num_stripes = 2;
1934 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1935 min_stripes = 4;
1936 num_stripes = 4;
1937 }
1938
1939 if (type & BTRFS_BLOCK_GROUP_DUP)
1940 min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1941 else
1942 min_stripe_size = BTRFS_STRIPE_LEN;
1943
1944 if (fs_info->alloc_start)
1945 mutex_lock(&fs_devices->device_list_mutex);
1946 rcu_read_lock();
1947 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1948 if (!device->in_fs_metadata || !device->bdev ||
1949 device->is_tgtdev_for_dev_replace)
1950 continue;
1951
1952 if (i >= nr_devices)
1953 break;
1954
1955 avail_space = device->total_bytes - device->bytes_used;
1956
1957 /* align with stripe_len */
1958 avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
1959 avail_space *= BTRFS_STRIPE_LEN;
1960
1961 /*
1962 * In order to avoid overwriting the superblock on the drive,
1963 * btrfs starts at an offset of at least 1MB when doing chunk
1964 * allocation.
1965 */
1966 skip_space = SZ_1M;
1967
1968 /* user can set the offset in fs_info->alloc_start. */
1969 if (fs_info->alloc_start &&
1970 fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1971 device->total_bytes) {
1972 rcu_read_unlock();
1973 skip_space = max(fs_info->alloc_start, skip_space);
1974
1975 /*
1976 * btrfs can not use the free space in
1977 * [0, skip_space - 1], we must subtract it from the
1978 * total. In order to implement it, we account the used
1979 * space in this range first.
1980 */
1981 ret = btrfs_account_dev_extents_size(device, 0,
1982 skip_space - 1,
1983 &used_space);
1984 if (ret) {
1985 kfree(devices_info);
1986 mutex_unlock(&fs_devices->device_list_mutex);
1987 return ret;
1988 }
1989
1990 rcu_read_lock();
1991
1992 /* calc the free space in [0, skip_space - 1] */
1993 skip_space -= used_space;
1994 }
1995
1996 /*
1997 * we can use the free space in [0, skip_space - 1], subtract
1998 * it from the total.
1999 */
2000 if (avail_space && avail_space >= skip_space)
2001 avail_space -= skip_space;
2002 else
2003 avail_space = 0;
2004
2005 if (avail_space < min_stripe_size)
2006 continue;
2007
2008 devices_info[i].dev = device;
2009 devices_info[i].max_avail = avail_space;
2010
2011 i++;
2012 }
2013 rcu_read_unlock();
2014 if (fs_info->alloc_start)
2015 mutex_unlock(&fs_devices->device_list_mutex);
2016
2017 nr_devices = i;
2018
2019 btrfs_descending_sort_devices(devices_info, nr_devices);
2020
2021 i = nr_devices - 1;
2022 avail_space = 0;
2023 while (nr_devices >= min_stripes) {
2024 if (num_stripes > nr_devices)
2025 num_stripes = nr_devices;
2026
2027 if (devices_info[i].max_avail >= min_stripe_size) {
2028 int j;
2029 u64 alloc_size;
2030
2031 avail_space += devices_info[i].max_avail * num_stripes;
2032 alloc_size = devices_info[i].max_avail;
2033 for (j = i + 1 - num_stripes; j <= i; j++)
2034 devices_info[j].max_avail -= alloc_size;
2035 }
2036 i--;
2037 nr_devices--;
2038 }
2039
2040 kfree(devices_info);
2041 *free_bytes = avail_space;
2042 return 0;
2043}
2044
2045/*
2046 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2047 *
2048 * If there's a redundant raid level at DATA block groups, use the respective
2049 * multiplier to scale the sizes.
2050 *
2051 * Unused device space usage is based on simulating the chunk allocator
2052 * algorithm that respects the device sizes, order of allocations and the
2053 * 'alloc_start' value, this is a close approximation of the actual use but
2054 * there are other factors that may change the result (like a new metadata
2055 * chunk).
2056 *
2057 * If metadata is exhausted, f_bavail will be 0.
2058 */
2059static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2060{
2061 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2062 struct btrfs_super_block *disk_super = fs_info->super_copy;
2063 struct list_head *head = &fs_info->space_info;
2064 struct btrfs_space_info *found;
2065 u64 total_used = 0;
2066 u64 total_free_data = 0;
2067 u64 total_free_meta = 0;
2068 int bits = dentry->d_sb->s_blocksize_bits;
2069 __be32 *fsid = (__be32 *)fs_info->fsid;
2070 unsigned factor = 1;
2071 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2072 int ret;
2073 u64 thresh = 0;
2074 int mixed = 0;
2075
2076 rcu_read_lock();
2077 list_for_each_entry_rcu(found, head, list) {
2078 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2079 int i;
2080
2081 total_free_data += found->disk_total - found->disk_used;
2082 total_free_data -=
2083 btrfs_account_ro_block_groups_free_space(found);
2084
2085 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2086 if (!list_empty(&found->block_groups[i])) {
2087 switch (i) {
2088 case BTRFS_RAID_DUP:
2089 case BTRFS_RAID_RAID1:
2090 case BTRFS_RAID_RAID10:
2091 factor = 2;
2092 }
2093 }
2094 }
2095 }
2096
2097 /*
2098 * Metadata in mixed block goup profiles are accounted in data
2099 */
2100 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2101 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2102 mixed = 1;
2103 else
2104 total_free_meta += found->disk_total -
2105 found->disk_used;
2106 }
2107
2108 total_used += found->disk_used;
2109 }
2110
2111 rcu_read_unlock();
2112
2113 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2114 buf->f_blocks >>= bits;
2115 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2116
2117 /* Account global block reserve as used, it's in logical size already */
2118 spin_lock(&block_rsv->lock);
2119 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2120 if (buf->f_bfree >= block_rsv->size >> bits)
2121 buf->f_bfree -= block_rsv->size >> bits;
2122 else
2123 buf->f_bfree = 0;
2124 spin_unlock(&block_rsv->lock);
2125
2126 buf->f_bavail = div_u64(total_free_data, factor);
2127 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2128 if (ret)
2129 return ret;
2130 buf->f_bavail += div_u64(total_free_data, factor);
2131 buf->f_bavail = buf->f_bavail >> bits;
2132
2133 /*
2134 * We calculate the remaining metadata space minus global reserve. If
2135 * this is (supposedly) smaller than zero, there's no space. But this
2136 * does not hold in practice, the exhausted state happens where's still
2137 * some positive delta. So we apply some guesswork and compare the
2138 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2139 *
2140 * We probably cannot calculate the exact threshold value because this
2141 * depends on the internal reservations requested by various
2142 * operations, so some operations that consume a few metadata will
2143 * succeed even if the Avail is zero. But this is better than the other
2144 * way around.
2145 */
2146 thresh = 4 * 1024 * 1024;
2147
2148 if (!mixed && total_free_meta - thresh < block_rsv->size)
2149 buf->f_bavail = 0;
2150
2151 buf->f_type = BTRFS_SUPER_MAGIC;
2152 buf->f_bsize = dentry->d_sb->s_blocksize;
2153 buf->f_namelen = BTRFS_NAME_LEN;
2154
2155 /* We treat it as constant endianness (it doesn't matter _which_)
2156 because we want the fsid to come out the same whether mounted
2157 on a big-endian or little-endian host */
2158 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2159 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2160 /* Mask in the root object ID too, to disambiguate subvols */
2161 buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2162 buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
2163
2164 return 0;
2165}
2166
2167static void btrfs_kill_super(struct super_block *sb)
2168{
2169 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2170 kill_anon_super(sb);
2171 free_fs_info(fs_info);
2172}
2173
2174static struct file_system_type btrfs_fs_type = {
2175 .owner = THIS_MODULE,
2176 .name = "btrfs",
2177 .mount = btrfs_mount,
2178 .kill_sb = btrfs_kill_super,
2179 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2180};
2181MODULE_ALIAS_FS("btrfs");
2182
2183static int btrfs_control_open(struct inode *inode, struct file *file)
2184{
2185 /*
2186 * The control file's private_data is used to hold the
2187 * transaction when it is started and is used to keep
2188 * track of whether a transaction is already in progress.
2189 */
2190 file->private_data = NULL;
2191 return 0;
2192}
2193
2194/*
2195 * used by btrfsctl to scan devices when no FS is mounted
2196 */
2197static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2198 unsigned long arg)
2199{
2200 struct btrfs_ioctl_vol_args *vol;
2201 struct btrfs_fs_devices *fs_devices;
2202 int ret = -ENOTTY;
2203
2204 if (!capable(CAP_SYS_ADMIN))
2205 return -EPERM;
2206
2207 vol = memdup_user((void __user *)arg, sizeof(*vol));
2208 if (IS_ERR(vol))
2209 return PTR_ERR(vol);
2210
2211 switch (cmd) {
2212 case BTRFS_IOC_SCAN_DEV:
2213 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2214 &btrfs_fs_type, &fs_devices);
2215 break;
2216 case BTRFS_IOC_DEVICES_READY:
2217 ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2218 &btrfs_fs_type, &fs_devices);
2219 if (ret)
2220 break;
2221 ret = !(fs_devices->num_devices == fs_devices->total_devices);
2222 break;
2223 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2224 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2225 break;
2226 }
2227
2228 kfree(vol);
2229 return ret;
2230}
2231
2232static int btrfs_freeze(struct super_block *sb)
2233{
2234 struct btrfs_trans_handle *trans;
2235 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2236 struct btrfs_root *root = fs_info->tree_root;
2237
2238 fs_info->fs_frozen = 1;
2239 /*
2240 * We don't need a barrier here, we'll wait for any transaction that
2241 * could be in progress on other threads (and do delayed iputs that
2242 * we want to avoid on a frozen filesystem), or do the commit
2243 * ourselves.
2244 */
2245 trans = btrfs_attach_transaction_barrier(root);
2246 if (IS_ERR(trans)) {
2247 /* no transaction, don't bother */
2248 if (PTR_ERR(trans) == -ENOENT)
2249 return 0;
2250 return PTR_ERR(trans);
2251 }
2252 return btrfs_commit_transaction(trans);
2253}
2254
2255static int btrfs_unfreeze(struct super_block *sb)
2256{
2257 btrfs_sb(sb)->fs_frozen = 0;
2258 return 0;
2259}
2260
2261static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2262{
2263 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2264 struct btrfs_fs_devices *cur_devices;
2265 struct btrfs_device *dev, *first_dev = NULL;
2266 struct list_head *head;
2267 struct rcu_string *name;
2268
2269 mutex_lock(&fs_info->fs_devices->device_list_mutex);
2270 cur_devices = fs_info->fs_devices;
2271 while (cur_devices) {
2272 head = &cur_devices->devices;
2273 list_for_each_entry(dev, head, dev_list) {
2274 if (dev->missing)
2275 continue;
2276 if (!dev->name)
2277 continue;
2278 if (!first_dev || dev->devid < first_dev->devid)
2279 first_dev = dev;
2280 }
2281 cur_devices = cur_devices->seed;
2282 }
2283
2284 if (first_dev) {
2285 rcu_read_lock();
2286 name = rcu_dereference(first_dev->name);
2287 seq_escape(m, name->str, " \t\n\\");
2288 rcu_read_unlock();
2289 } else {
2290 WARN_ON(1);
2291 }
2292 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2293 return 0;
2294}
2295
2296static const struct super_operations btrfs_super_ops = {
2297 .drop_inode = btrfs_drop_inode,
2298 .evict_inode = btrfs_evict_inode,
2299 .put_super = btrfs_put_super,
2300 .sync_fs = btrfs_sync_fs,
2301 .show_options = btrfs_show_options,
2302 .show_devname = btrfs_show_devname,
2303 .write_inode = btrfs_write_inode,
2304 .alloc_inode = btrfs_alloc_inode,
2305 .destroy_inode = btrfs_destroy_inode,
2306 .statfs = btrfs_statfs,
2307 .remount_fs = btrfs_remount,
2308 .freeze_fs = btrfs_freeze,
2309 .unfreeze_fs = btrfs_unfreeze,
2310};
2311
2312static const struct file_operations btrfs_ctl_fops = {
2313 .open = btrfs_control_open,
2314 .unlocked_ioctl = btrfs_control_ioctl,
2315 .compat_ioctl = btrfs_control_ioctl,
2316 .owner = THIS_MODULE,
2317 .llseek = noop_llseek,
2318};
2319
2320static struct miscdevice btrfs_misc = {
2321 .minor = BTRFS_MINOR,
2322 .name = "btrfs-control",
2323 .fops = &btrfs_ctl_fops
2324};
2325
2326MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2327MODULE_ALIAS("devname:btrfs-control");
2328
2329static int btrfs_interface_init(void)
2330{
2331 return misc_register(&btrfs_misc);
2332}
2333
2334static void btrfs_interface_exit(void)
2335{
2336 misc_deregister(&btrfs_misc);
2337}
2338
2339static void btrfs_print_mod_info(void)
2340{
2341 pr_info("Btrfs loaded, crc32c=%s"
2342#ifdef CONFIG_BTRFS_DEBUG
2343 ", debug=on"
2344#endif
2345#ifdef CONFIG_BTRFS_ASSERT
2346 ", assert=on"
2347#endif
2348#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2349 ", integrity-checker=on"
2350#endif
2351 "\n",
2352 btrfs_crc32c_impl());
2353}
2354
2355static int __init init_btrfs_fs(void)
2356{
2357 int err;
2358
2359 err = btrfs_hash_init();
2360 if (err)
2361 return err;
2362
2363 btrfs_props_init();
2364
2365 err = btrfs_init_sysfs();
2366 if (err)
2367 goto free_hash;
2368
2369 btrfs_init_compress();
2370
2371 err = btrfs_init_cachep();
2372 if (err)
2373 goto free_compress;
2374
2375 err = extent_io_init();
2376 if (err)
2377 goto free_cachep;
2378
2379 err = extent_map_init();
2380 if (err)
2381 goto free_extent_io;
2382
2383 err = ordered_data_init();
2384 if (err)
2385 goto free_extent_map;
2386
2387 err = btrfs_delayed_inode_init();
2388 if (err)
2389 goto free_ordered_data;
2390
2391 err = btrfs_auto_defrag_init();
2392 if (err)
2393 goto free_delayed_inode;
2394
2395 err = btrfs_delayed_ref_init();
2396 if (err)
2397 goto free_auto_defrag;
2398
2399 err = btrfs_prelim_ref_init();
2400 if (err)
2401 goto free_delayed_ref;
2402
2403 err = btrfs_end_io_wq_init();
2404 if (err)
2405 goto free_prelim_ref;
2406
2407 err = btrfs_interface_init();
2408 if (err)
2409 goto free_end_io_wq;
2410
2411 btrfs_init_lockdep();
2412
2413 btrfs_print_mod_info();
2414
2415 err = btrfs_run_sanity_tests();
2416 if (err)
2417 goto unregister_ioctl;
2418
2419 err = register_filesystem(&btrfs_fs_type);
2420 if (err)
2421 goto unregister_ioctl;
2422
2423 return 0;
2424
2425unregister_ioctl:
2426 btrfs_interface_exit();
2427free_end_io_wq:
2428 btrfs_end_io_wq_exit();
2429free_prelim_ref:
2430 btrfs_prelim_ref_exit();
2431free_delayed_ref:
2432 btrfs_delayed_ref_exit();
2433free_auto_defrag:
2434 btrfs_auto_defrag_exit();
2435free_delayed_inode:
2436 btrfs_delayed_inode_exit();
2437free_ordered_data:
2438 ordered_data_exit();
2439free_extent_map:
2440 extent_map_exit();
2441free_extent_io:
2442 extent_io_exit();
2443free_cachep:
2444 btrfs_destroy_cachep();
2445free_compress:
2446 btrfs_exit_compress();
2447 btrfs_exit_sysfs();
2448free_hash:
2449 btrfs_hash_exit();
2450 return err;
2451}
2452
2453static void __exit exit_btrfs_fs(void)
2454{
2455 btrfs_destroy_cachep();
2456 btrfs_delayed_ref_exit();
2457 btrfs_auto_defrag_exit();
2458 btrfs_delayed_inode_exit();
2459 btrfs_prelim_ref_exit();
2460 ordered_data_exit();
2461 extent_map_exit();
2462 extent_io_exit();
2463 btrfs_interface_exit();
2464 btrfs_end_io_wq_exit();
2465 unregister_filesystem(&btrfs_fs_type);
2466 btrfs_exit_sysfs();
2467 btrfs_cleanup_fs_uuids();
2468 btrfs_exit_compress();
2469 btrfs_hash_exit();
2470}
2471
2472late_initcall(init_btrfs_fs);
2473module_exit(exit_btrfs_fs)
2474
2475MODULE_LICENSE("GPL");
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/module.h>
8#include <linux/fs.h>
9#include <linux/pagemap.h>
10#include <linux/highmem.h>
11#include <linux/time.h>
12#include <linux/init.h>
13#include <linux/seq_file.h>
14#include <linux/string.h>
15#include <linux/backing-dev.h>
16#include <linux/mount.h>
17#include <linux/writeback.h>
18#include <linux/statfs.h>
19#include <linux/compat.h>
20#include <linux/parser.h>
21#include <linux/ctype.h>
22#include <linux/namei.h>
23#include <linux/miscdevice.h>
24#include <linux/magic.h>
25#include <linux/slab.h>
26#include <linux/ratelimit.h>
27#include <linux/crc32c.h>
28#include <linux/btrfs.h>
29#include "messages.h"
30#include "delayed-inode.h"
31#include "ctree.h"
32#include "disk-io.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "props.h"
37#include "xattr.h"
38#include "bio.h"
39#include "export.h"
40#include "compression.h"
41#include "rcu-string.h"
42#include "dev-replace.h"
43#include "free-space-cache.h"
44#include "backref.h"
45#include "space-info.h"
46#include "sysfs.h"
47#include "zoned.h"
48#include "tests/btrfs-tests.h"
49#include "block-group.h"
50#include "discard.h"
51#include "qgroup.h"
52#include "raid56.h"
53#include "fs.h"
54#include "accessors.h"
55#include "defrag.h"
56#include "dir-item.h"
57#include "ioctl.h"
58#include "scrub.h"
59#include "verity.h"
60#include "super.h"
61#define CREATE_TRACE_POINTS
62#include <trace/events/btrfs.h>
63
64static const struct super_operations btrfs_super_ops;
65
66/*
67 * Types for mounting the default subvolume and a subvolume explicitly
68 * requested by subvol=/path. That way the callchain is straightforward and we
69 * don't have to play tricks with the mount options and recursive calls to
70 * btrfs_mount.
71 *
72 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
73 */
74static struct file_system_type btrfs_fs_type;
75static struct file_system_type btrfs_root_fs_type;
76
77static int btrfs_remount(struct super_block *sb, int *flags, char *data);
78
79static void btrfs_put_super(struct super_block *sb)
80{
81 close_ctree(btrfs_sb(sb));
82}
83
84enum {
85 Opt_acl, Opt_noacl,
86 Opt_clear_cache,
87 Opt_commit_interval,
88 Opt_compress,
89 Opt_compress_force,
90 Opt_compress_force_type,
91 Opt_compress_type,
92 Opt_degraded,
93 Opt_device,
94 Opt_fatal_errors,
95 Opt_flushoncommit, Opt_noflushoncommit,
96 Opt_max_inline,
97 Opt_barrier, Opt_nobarrier,
98 Opt_datacow, Opt_nodatacow,
99 Opt_datasum, Opt_nodatasum,
100 Opt_defrag, Opt_nodefrag,
101 Opt_discard, Opt_nodiscard,
102 Opt_discard_mode,
103 Opt_norecovery,
104 Opt_ratio,
105 Opt_rescan_uuid_tree,
106 Opt_skip_balance,
107 Opt_space_cache, Opt_no_space_cache,
108 Opt_space_cache_version,
109 Opt_ssd, Opt_nossd,
110 Opt_ssd_spread, Opt_nossd_spread,
111 Opt_subvol,
112 Opt_subvol_empty,
113 Opt_subvolid,
114 Opt_thread_pool,
115 Opt_treelog, Opt_notreelog,
116 Opt_user_subvol_rm_allowed,
117
118 /* Rescue options */
119 Opt_rescue,
120 Opt_usebackuproot,
121 Opt_nologreplay,
122 Opt_ignorebadroots,
123 Opt_ignoredatacsums,
124 Opt_rescue_all,
125
126 /* Deprecated options */
127 Opt_recovery,
128 Opt_inode_cache, Opt_noinode_cache,
129
130 /* Debugging options */
131 Opt_check_integrity,
132 Opt_check_integrity_including_extent_data,
133 Opt_check_integrity_print_mask,
134 Opt_enospc_debug, Opt_noenospc_debug,
135#ifdef CONFIG_BTRFS_DEBUG
136 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
137#endif
138#ifdef CONFIG_BTRFS_FS_REF_VERIFY
139 Opt_ref_verify,
140#endif
141 Opt_err,
142};
143
144static const match_table_t tokens = {
145 {Opt_acl, "acl"},
146 {Opt_noacl, "noacl"},
147 {Opt_clear_cache, "clear_cache"},
148 {Opt_commit_interval, "commit=%u"},
149 {Opt_compress, "compress"},
150 {Opt_compress_type, "compress=%s"},
151 {Opt_compress_force, "compress-force"},
152 {Opt_compress_force_type, "compress-force=%s"},
153 {Opt_degraded, "degraded"},
154 {Opt_device, "device=%s"},
155 {Opt_fatal_errors, "fatal_errors=%s"},
156 {Opt_flushoncommit, "flushoncommit"},
157 {Opt_noflushoncommit, "noflushoncommit"},
158 {Opt_inode_cache, "inode_cache"},
159 {Opt_noinode_cache, "noinode_cache"},
160 {Opt_max_inline, "max_inline=%s"},
161 {Opt_barrier, "barrier"},
162 {Opt_nobarrier, "nobarrier"},
163 {Opt_datacow, "datacow"},
164 {Opt_nodatacow, "nodatacow"},
165 {Opt_datasum, "datasum"},
166 {Opt_nodatasum, "nodatasum"},
167 {Opt_defrag, "autodefrag"},
168 {Opt_nodefrag, "noautodefrag"},
169 {Opt_discard, "discard"},
170 {Opt_discard_mode, "discard=%s"},
171 {Opt_nodiscard, "nodiscard"},
172 {Opt_norecovery, "norecovery"},
173 {Opt_ratio, "metadata_ratio=%u"},
174 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
175 {Opt_skip_balance, "skip_balance"},
176 {Opt_space_cache, "space_cache"},
177 {Opt_no_space_cache, "nospace_cache"},
178 {Opt_space_cache_version, "space_cache=%s"},
179 {Opt_ssd, "ssd"},
180 {Opt_nossd, "nossd"},
181 {Opt_ssd_spread, "ssd_spread"},
182 {Opt_nossd_spread, "nossd_spread"},
183 {Opt_subvol, "subvol=%s"},
184 {Opt_subvol_empty, "subvol="},
185 {Opt_subvolid, "subvolid=%s"},
186 {Opt_thread_pool, "thread_pool=%u"},
187 {Opt_treelog, "treelog"},
188 {Opt_notreelog, "notreelog"},
189 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
190
191 /* Rescue options */
192 {Opt_rescue, "rescue=%s"},
193 /* Deprecated, with alias rescue=nologreplay */
194 {Opt_nologreplay, "nologreplay"},
195 /* Deprecated, with alias rescue=usebackuproot */
196 {Opt_usebackuproot, "usebackuproot"},
197
198 /* Deprecated options */
199 {Opt_recovery, "recovery"},
200
201 /* Debugging options */
202 {Opt_check_integrity, "check_int"},
203 {Opt_check_integrity_including_extent_data, "check_int_data"},
204 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
205 {Opt_enospc_debug, "enospc_debug"},
206 {Opt_noenospc_debug, "noenospc_debug"},
207#ifdef CONFIG_BTRFS_DEBUG
208 {Opt_fragment_data, "fragment=data"},
209 {Opt_fragment_metadata, "fragment=metadata"},
210 {Opt_fragment_all, "fragment=all"},
211#endif
212#ifdef CONFIG_BTRFS_FS_REF_VERIFY
213 {Opt_ref_verify, "ref_verify"},
214#endif
215 {Opt_err, NULL},
216};
217
218static const match_table_t rescue_tokens = {
219 {Opt_usebackuproot, "usebackuproot"},
220 {Opt_nologreplay, "nologreplay"},
221 {Opt_ignorebadroots, "ignorebadroots"},
222 {Opt_ignorebadroots, "ibadroots"},
223 {Opt_ignoredatacsums, "ignoredatacsums"},
224 {Opt_ignoredatacsums, "idatacsums"},
225 {Opt_rescue_all, "all"},
226 {Opt_err, NULL},
227};
228
229static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
230 const char *opt_name)
231{
232 if (fs_info->mount_opt & opt) {
233 btrfs_err(fs_info, "%s must be used with ro mount option",
234 opt_name);
235 return true;
236 }
237 return false;
238}
239
240static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
241{
242 char *opts;
243 char *orig;
244 char *p;
245 substring_t args[MAX_OPT_ARGS];
246 int ret = 0;
247
248 opts = kstrdup(options, GFP_KERNEL);
249 if (!opts)
250 return -ENOMEM;
251 orig = opts;
252
253 while ((p = strsep(&opts, ":")) != NULL) {
254 int token;
255
256 if (!*p)
257 continue;
258 token = match_token(p, rescue_tokens, args);
259 switch (token){
260 case Opt_usebackuproot:
261 btrfs_info(info,
262 "trying to use backup root at mount time");
263 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
264 break;
265 case Opt_nologreplay:
266 btrfs_set_and_info(info, NOLOGREPLAY,
267 "disabling log replay at mount time");
268 break;
269 case Opt_ignorebadroots:
270 btrfs_set_and_info(info, IGNOREBADROOTS,
271 "ignoring bad roots");
272 break;
273 case Opt_ignoredatacsums:
274 btrfs_set_and_info(info, IGNOREDATACSUMS,
275 "ignoring data csums");
276 break;
277 case Opt_rescue_all:
278 btrfs_info(info, "enabling all of the rescue options");
279 btrfs_set_and_info(info, IGNOREDATACSUMS,
280 "ignoring data csums");
281 btrfs_set_and_info(info, IGNOREBADROOTS,
282 "ignoring bad roots");
283 btrfs_set_and_info(info, NOLOGREPLAY,
284 "disabling log replay at mount time");
285 break;
286 case Opt_err:
287 btrfs_info(info, "unrecognized rescue option '%s'", p);
288 ret = -EINVAL;
289 goto out;
290 default:
291 break;
292 }
293
294 }
295out:
296 kfree(orig);
297 return ret;
298}
299
300/*
301 * Regular mount options parser. Everything that is needed only when
302 * reading in a new superblock is parsed here.
303 * XXX JDM: This needs to be cleaned up for remount.
304 */
305int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
306 unsigned long new_flags)
307{
308 substring_t args[MAX_OPT_ARGS];
309 char *p, *num;
310 int intarg;
311 int ret = 0;
312 char *compress_type;
313 bool compress_force = false;
314 enum btrfs_compression_type saved_compress_type;
315 int saved_compress_level;
316 bool saved_compress_force;
317 int no_compress = 0;
318 const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
319
320 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
321 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
322 else if (btrfs_free_space_cache_v1_active(info)) {
323 if (btrfs_is_zoned(info)) {
324 btrfs_info(info,
325 "zoned: clearing existing space cache");
326 btrfs_set_super_cache_generation(info->super_copy, 0);
327 } else {
328 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
329 }
330 }
331
332 /*
333 * Even the options are empty, we still need to do extra check
334 * against new flags
335 */
336 if (!options)
337 goto check;
338
339 while ((p = strsep(&options, ",")) != NULL) {
340 int token;
341 if (!*p)
342 continue;
343
344 token = match_token(p, tokens, args);
345 switch (token) {
346 case Opt_degraded:
347 btrfs_info(info, "allowing degraded mounts");
348 btrfs_set_opt(info->mount_opt, DEGRADED);
349 break;
350 case Opt_subvol:
351 case Opt_subvol_empty:
352 case Opt_subvolid:
353 case Opt_device:
354 /*
355 * These are parsed by btrfs_parse_subvol_options or
356 * btrfs_parse_device_options and can be ignored here.
357 */
358 break;
359 case Opt_nodatasum:
360 btrfs_set_and_info(info, NODATASUM,
361 "setting nodatasum");
362 break;
363 case Opt_datasum:
364 if (btrfs_test_opt(info, NODATASUM)) {
365 if (btrfs_test_opt(info, NODATACOW))
366 btrfs_info(info,
367 "setting datasum, datacow enabled");
368 else
369 btrfs_info(info, "setting datasum");
370 }
371 btrfs_clear_opt(info->mount_opt, NODATACOW);
372 btrfs_clear_opt(info->mount_opt, NODATASUM);
373 break;
374 case Opt_nodatacow:
375 if (!btrfs_test_opt(info, NODATACOW)) {
376 if (!btrfs_test_opt(info, COMPRESS) ||
377 !btrfs_test_opt(info, FORCE_COMPRESS)) {
378 btrfs_info(info,
379 "setting nodatacow, compression disabled");
380 } else {
381 btrfs_info(info, "setting nodatacow");
382 }
383 }
384 btrfs_clear_opt(info->mount_opt, COMPRESS);
385 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
386 btrfs_set_opt(info->mount_opt, NODATACOW);
387 btrfs_set_opt(info->mount_opt, NODATASUM);
388 break;
389 case Opt_datacow:
390 btrfs_clear_and_info(info, NODATACOW,
391 "setting datacow");
392 break;
393 case Opt_compress_force:
394 case Opt_compress_force_type:
395 compress_force = true;
396 fallthrough;
397 case Opt_compress:
398 case Opt_compress_type:
399 saved_compress_type = btrfs_test_opt(info,
400 COMPRESS) ?
401 info->compress_type : BTRFS_COMPRESS_NONE;
402 saved_compress_force =
403 btrfs_test_opt(info, FORCE_COMPRESS);
404 saved_compress_level = info->compress_level;
405 if (token == Opt_compress ||
406 token == Opt_compress_force ||
407 strncmp(args[0].from, "zlib", 4) == 0) {
408 compress_type = "zlib";
409
410 info->compress_type = BTRFS_COMPRESS_ZLIB;
411 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
412 /*
413 * args[0] contains uninitialized data since
414 * for these tokens we don't expect any
415 * parameter.
416 */
417 if (token != Opt_compress &&
418 token != Opt_compress_force)
419 info->compress_level =
420 btrfs_compress_str2level(
421 BTRFS_COMPRESS_ZLIB,
422 args[0].from + 4);
423 btrfs_set_opt(info->mount_opt, COMPRESS);
424 btrfs_clear_opt(info->mount_opt, NODATACOW);
425 btrfs_clear_opt(info->mount_opt, NODATASUM);
426 no_compress = 0;
427 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
428 compress_type = "lzo";
429 info->compress_type = BTRFS_COMPRESS_LZO;
430 info->compress_level = 0;
431 btrfs_set_opt(info->mount_opt, COMPRESS);
432 btrfs_clear_opt(info->mount_opt, NODATACOW);
433 btrfs_clear_opt(info->mount_opt, NODATASUM);
434 btrfs_set_fs_incompat(info, COMPRESS_LZO);
435 no_compress = 0;
436 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
437 compress_type = "zstd";
438 info->compress_type = BTRFS_COMPRESS_ZSTD;
439 info->compress_level =
440 btrfs_compress_str2level(
441 BTRFS_COMPRESS_ZSTD,
442 args[0].from + 4);
443 btrfs_set_opt(info->mount_opt, COMPRESS);
444 btrfs_clear_opt(info->mount_opt, NODATACOW);
445 btrfs_clear_opt(info->mount_opt, NODATASUM);
446 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
447 no_compress = 0;
448 } else if (strncmp(args[0].from, "no", 2) == 0) {
449 compress_type = "no";
450 info->compress_level = 0;
451 info->compress_type = 0;
452 btrfs_clear_opt(info->mount_opt, COMPRESS);
453 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
454 compress_force = false;
455 no_compress++;
456 } else {
457 btrfs_err(info, "unrecognized compression value %s",
458 args[0].from);
459 ret = -EINVAL;
460 goto out;
461 }
462
463 if (compress_force) {
464 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
465 } else {
466 /*
467 * If we remount from compress-force=xxx to
468 * compress=xxx, we need clear FORCE_COMPRESS
469 * flag, otherwise, there is no way for users
470 * to disable forcible compression separately.
471 */
472 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
473 }
474 if (no_compress == 1) {
475 btrfs_info(info, "use no compression");
476 } else if ((info->compress_type != saved_compress_type) ||
477 (compress_force != saved_compress_force) ||
478 (info->compress_level != saved_compress_level)) {
479 btrfs_info(info, "%s %s compression, level %d",
480 (compress_force) ? "force" : "use",
481 compress_type, info->compress_level);
482 }
483 compress_force = false;
484 break;
485 case Opt_ssd:
486 btrfs_set_and_info(info, SSD,
487 "enabling ssd optimizations");
488 btrfs_clear_opt(info->mount_opt, NOSSD);
489 break;
490 case Opt_ssd_spread:
491 btrfs_set_and_info(info, SSD,
492 "enabling ssd optimizations");
493 btrfs_set_and_info(info, SSD_SPREAD,
494 "using spread ssd allocation scheme");
495 btrfs_clear_opt(info->mount_opt, NOSSD);
496 break;
497 case Opt_nossd:
498 btrfs_set_opt(info->mount_opt, NOSSD);
499 btrfs_clear_and_info(info, SSD,
500 "not using ssd optimizations");
501 fallthrough;
502 case Opt_nossd_spread:
503 btrfs_clear_and_info(info, SSD_SPREAD,
504 "not using spread ssd allocation scheme");
505 break;
506 case Opt_barrier:
507 btrfs_clear_and_info(info, NOBARRIER,
508 "turning on barriers");
509 break;
510 case Opt_nobarrier:
511 btrfs_set_and_info(info, NOBARRIER,
512 "turning off barriers");
513 break;
514 case Opt_thread_pool:
515 ret = match_int(&args[0], &intarg);
516 if (ret) {
517 btrfs_err(info, "unrecognized thread_pool value %s",
518 args[0].from);
519 goto out;
520 } else if (intarg == 0) {
521 btrfs_err(info, "invalid value 0 for thread_pool");
522 ret = -EINVAL;
523 goto out;
524 }
525 info->thread_pool_size = intarg;
526 break;
527 case Opt_max_inline:
528 num = match_strdup(&args[0]);
529 if (num) {
530 info->max_inline = memparse(num, NULL);
531 kfree(num);
532
533 if (info->max_inline) {
534 info->max_inline = min_t(u64,
535 info->max_inline,
536 info->sectorsize);
537 }
538 btrfs_info(info, "max_inline at %llu",
539 info->max_inline);
540 } else {
541 ret = -ENOMEM;
542 goto out;
543 }
544 break;
545 case Opt_acl:
546#ifdef CONFIG_BTRFS_FS_POSIX_ACL
547 info->sb->s_flags |= SB_POSIXACL;
548 break;
549#else
550 btrfs_err(info, "support for ACL not compiled in!");
551 ret = -EINVAL;
552 goto out;
553#endif
554 case Opt_noacl:
555 info->sb->s_flags &= ~SB_POSIXACL;
556 break;
557 case Opt_notreelog:
558 btrfs_set_and_info(info, NOTREELOG,
559 "disabling tree log");
560 break;
561 case Opt_treelog:
562 btrfs_clear_and_info(info, NOTREELOG,
563 "enabling tree log");
564 break;
565 case Opt_norecovery:
566 case Opt_nologreplay:
567 btrfs_warn(info,
568 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
569 btrfs_set_and_info(info, NOLOGREPLAY,
570 "disabling log replay at mount time");
571 break;
572 case Opt_flushoncommit:
573 btrfs_set_and_info(info, FLUSHONCOMMIT,
574 "turning on flush-on-commit");
575 break;
576 case Opt_noflushoncommit:
577 btrfs_clear_and_info(info, FLUSHONCOMMIT,
578 "turning off flush-on-commit");
579 break;
580 case Opt_ratio:
581 ret = match_int(&args[0], &intarg);
582 if (ret) {
583 btrfs_err(info, "unrecognized metadata_ratio value %s",
584 args[0].from);
585 goto out;
586 }
587 info->metadata_ratio = intarg;
588 btrfs_info(info, "metadata ratio %u",
589 info->metadata_ratio);
590 break;
591 case Opt_discard:
592 case Opt_discard_mode:
593 if (token == Opt_discard ||
594 strcmp(args[0].from, "sync") == 0) {
595 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
596 btrfs_set_and_info(info, DISCARD_SYNC,
597 "turning on sync discard");
598 } else if (strcmp(args[0].from, "async") == 0) {
599 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
600 btrfs_set_and_info(info, DISCARD_ASYNC,
601 "turning on async discard");
602 } else {
603 btrfs_err(info, "unrecognized discard mode value %s",
604 args[0].from);
605 ret = -EINVAL;
606 goto out;
607 }
608 btrfs_clear_opt(info->mount_opt, NODISCARD);
609 break;
610 case Opt_nodiscard:
611 btrfs_clear_and_info(info, DISCARD_SYNC,
612 "turning off discard");
613 btrfs_clear_and_info(info, DISCARD_ASYNC,
614 "turning off async discard");
615 btrfs_set_opt(info->mount_opt, NODISCARD);
616 break;
617 case Opt_space_cache:
618 case Opt_space_cache_version:
619 /*
620 * We already set FREE_SPACE_TREE above because we have
621 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
622 * to allow v1 to be set for extent tree v2, simply
623 * ignore this setting if we're extent tree v2.
624 */
625 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
626 break;
627 if (token == Opt_space_cache ||
628 strcmp(args[0].from, "v1") == 0) {
629 btrfs_clear_opt(info->mount_opt,
630 FREE_SPACE_TREE);
631 btrfs_set_and_info(info, SPACE_CACHE,
632 "enabling disk space caching");
633 } else if (strcmp(args[0].from, "v2") == 0) {
634 btrfs_clear_opt(info->mount_opt,
635 SPACE_CACHE);
636 btrfs_set_and_info(info, FREE_SPACE_TREE,
637 "enabling free space tree");
638 } else {
639 btrfs_err(info, "unrecognized space_cache value %s",
640 args[0].from);
641 ret = -EINVAL;
642 goto out;
643 }
644 break;
645 case Opt_rescan_uuid_tree:
646 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
647 break;
648 case Opt_no_space_cache:
649 /*
650 * We cannot operate without the free space tree with
651 * extent tree v2, ignore this option.
652 */
653 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
654 break;
655 if (btrfs_test_opt(info, SPACE_CACHE)) {
656 btrfs_clear_and_info(info, SPACE_CACHE,
657 "disabling disk space caching");
658 }
659 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
660 btrfs_clear_and_info(info, FREE_SPACE_TREE,
661 "disabling free space tree");
662 }
663 break;
664 case Opt_inode_cache:
665 case Opt_noinode_cache:
666 btrfs_warn(info,
667 "the 'inode_cache' option is deprecated and has no effect since 5.11");
668 break;
669 case Opt_clear_cache:
670 /*
671 * We cannot clear the free space tree with extent tree
672 * v2, ignore this option.
673 */
674 if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
675 break;
676 btrfs_set_and_info(info, CLEAR_CACHE,
677 "force clearing of disk cache");
678 break;
679 case Opt_user_subvol_rm_allowed:
680 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
681 break;
682 case Opt_enospc_debug:
683 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
684 break;
685 case Opt_noenospc_debug:
686 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
687 break;
688 case Opt_defrag:
689 btrfs_set_and_info(info, AUTO_DEFRAG,
690 "enabling auto defrag");
691 break;
692 case Opt_nodefrag:
693 btrfs_clear_and_info(info, AUTO_DEFRAG,
694 "disabling auto defrag");
695 break;
696 case Opt_recovery:
697 case Opt_usebackuproot:
698 btrfs_warn(info,
699 "'%s' is deprecated, use 'rescue=usebackuproot' instead",
700 token == Opt_recovery ? "recovery" :
701 "usebackuproot");
702 btrfs_info(info,
703 "trying to use backup root at mount time");
704 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
705 break;
706 case Opt_skip_balance:
707 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
708 break;
709#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
710 case Opt_check_integrity_including_extent_data:
711 btrfs_info(info,
712 "enabling check integrity including extent data");
713 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
714 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
715 break;
716 case Opt_check_integrity:
717 btrfs_info(info, "enabling check integrity");
718 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
719 break;
720 case Opt_check_integrity_print_mask:
721 ret = match_int(&args[0], &intarg);
722 if (ret) {
723 btrfs_err(info,
724 "unrecognized check_integrity_print_mask value %s",
725 args[0].from);
726 goto out;
727 }
728 info->check_integrity_print_mask = intarg;
729 btrfs_info(info, "check_integrity_print_mask 0x%x",
730 info->check_integrity_print_mask);
731 break;
732#else
733 case Opt_check_integrity_including_extent_data:
734 case Opt_check_integrity:
735 case Opt_check_integrity_print_mask:
736 btrfs_err(info,
737 "support for check_integrity* not compiled in!");
738 ret = -EINVAL;
739 goto out;
740#endif
741 case Opt_fatal_errors:
742 if (strcmp(args[0].from, "panic") == 0) {
743 btrfs_set_opt(info->mount_opt,
744 PANIC_ON_FATAL_ERROR);
745 } else if (strcmp(args[0].from, "bug") == 0) {
746 btrfs_clear_opt(info->mount_opt,
747 PANIC_ON_FATAL_ERROR);
748 } else {
749 btrfs_err(info, "unrecognized fatal_errors value %s",
750 args[0].from);
751 ret = -EINVAL;
752 goto out;
753 }
754 break;
755 case Opt_commit_interval:
756 intarg = 0;
757 ret = match_int(&args[0], &intarg);
758 if (ret) {
759 btrfs_err(info, "unrecognized commit_interval value %s",
760 args[0].from);
761 ret = -EINVAL;
762 goto out;
763 }
764 if (intarg == 0) {
765 btrfs_info(info,
766 "using default commit interval %us",
767 BTRFS_DEFAULT_COMMIT_INTERVAL);
768 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
769 } else if (intarg > 300) {
770 btrfs_warn(info, "excessive commit interval %d",
771 intarg);
772 }
773 info->commit_interval = intarg;
774 break;
775 case Opt_rescue:
776 ret = parse_rescue_options(info, args[0].from);
777 if (ret < 0) {
778 btrfs_err(info, "unrecognized rescue value %s",
779 args[0].from);
780 goto out;
781 }
782 break;
783#ifdef CONFIG_BTRFS_DEBUG
784 case Opt_fragment_all:
785 btrfs_info(info, "fragmenting all space");
786 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
787 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
788 break;
789 case Opt_fragment_metadata:
790 btrfs_info(info, "fragmenting metadata");
791 btrfs_set_opt(info->mount_opt,
792 FRAGMENT_METADATA);
793 break;
794 case Opt_fragment_data:
795 btrfs_info(info, "fragmenting data");
796 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
797 break;
798#endif
799#ifdef CONFIG_BTRFS_FS_REF_VERIFY
800 case Opt_ref_verify:
801 btrfs_info(info, "doing ref verification");
802 btrfs_set_opt(info->mount_opt, REF_VERIFY);
803 break;
804#endif
805 case Opt_err:
806 btrfs_err(info, "unrecognized mount option '%s'", p);
807 ret = -EINVAL;
808 goto out;
809 default:
810 break;
811 }
812 }
813check:
814 /* We're read-only, don't have to check. */
815 if (new_flags & SB_RDONLY)
816 goto out;
817
818 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
819 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
820 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
821 ret = -EINVAL;
822out:
823 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
824 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
825 !btrfs_test_opt(info, CLEAR_CACHE)) {
826 btrfs_err(info, "cannot disable free space tree");
827 ret = -EINVAL;
828
829 }
830 if (!ret)
831 ret = btrfs_check_mountopts_zoned(info);
832 if (!ret && !remounting) {
833 if (btrfs_test_opt(info, SPACE_CACHE))
834 btrfs_info(info, "disk space caching is enabled");
835 if (btrfs_test_opt(info, FREE_SPACE_TREE))
836 btrfs_info(info, "using free space tree");
837 }
838 return ret;
839}
840
841/*
842 * Parse mount options that are required early in the mount process.
843 *
844 * All other options will be parsed on much later in the mount process and
845 * only when we need to allocate a new super block.
846 */
847static int btrfs_parse_device_options(const char *options, fmode_t flags,
848 void *holder)
849{
850 substring_t args[MAX_OPT_ARGS];
851 char *device_name, *opts, *orig, *p;
852 struct btrfs_device *device = NULL;
853 int error = 0;
854
855 lockdep_assert_held(&uuid_mutex);
856
857 if (!options)
858 return 0;
859
860 /*
861 * strsep changes the string, duplicate it because btrfs_parse_options
862 * gets called later
863 */
864 opts = kstrdup(options, GFP_KERNEL);
865 if (!opts)
866 return -ENOMEM;
867 orig = opts;
868
869 while ((p = strsep(&opts, ",")) != NULL) {
870 int token;
871
872 if (!*p)
873 continue;
874
875 token = match_token(p, tokens, args);
876 if (token == Opt_device) {
877 device_name = match_strdup(&args[0]);
878 if (!device_name) {
879 error = -ENOMEM;
880 goto out;
881 }
882 device = btrfs_scan_one_device(device_name, flags,
883 holder);
884 kfree(device_name);
885 if (IS_ERR(device)) {
886 error = PTR_ERR(device);
887 goto out;
888 }
889 }
890 }
891
892out:
893 kfree(orig);
894 return error;
895}
896
897/*
898 * Parse mount options that are related to subvolume id
899 *
900 * The value is later passed to mount_subvol()
901 */
902static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
903 u64 *subvol_objectid)
904{
905 substring_t args[MAX_OPT_ARGS];
906 char *opts, *orig, *p;
907 int error = 0;
908 u64 subvolid;
909
910 if (!options)
911 return 0;
912
913 /*
914 * strsep changes the string, duplicate it because
915 * btrfs_parse_device_options gets called later
916 */
917 opts = kstrdup(options, GFP_KERNEL);
918 if (!opts)
919 return -ENOMEM;
920 orig = opts;
921
922 while ((p = strsep(&opts, ",")) != NULL) {
923 int token;
924 if (!*p)
925 continue;
926
927 token = match_token(p, tokens, args);
928 switch (token) {
929 case Opt_subvol:
930 kfree(*subvol_name);
931 *subvol_name = match_strdup(&args[0]);
932 if (!*subvol_name) {
933 error = -ENOMEM;
934 goto out;
935 }
936 break;
937 case Opt_subvolid:
938 error = match_u64(&args[0], &subvolid);
939 if (error)
940 goto out;
941
942 /* we want the original fs_tree */
943 if (subvolid == 0)
944 subvolid = BTRFS_FS_TREE_OBJECTID;
945
946 *subvol_objectid = subvolid;
947 break;
948 default:
949 break;
950 }
951 }
952
953out:
954 kfree(orig);
955 return error;
956}
957
958char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
959 u64 subvol_objectid)
960{
961 struct btrfs_root *root = fs_info->tree_root;
962 struct btrfs_root *fs_root = NULL;
963 struct btrfs_root_ref *root_ref;
964 struct btrfs_inode_ref *inode_ref;
965 struct btrfs_key key;
966 struct btrfs_path *path = NULL;
967 char *name = NULL, *ptr;
968 u64 dirid;
969 int len;
970 int ret;
971
972 path = btrfs_alloc_path();
973 if (!path) {
974 ret = -ENOMEM;
975 goto err;
976 }
977
978 name = kmalloc(PATH_MAX, GFP_KERNEL);
979 if (!name) {
980 ret = -ENOMEM;
981 goto err;
982 }
983 ptr = name + PATH_MAX - 1;
984 ptr[0] = '\0';
985
986 /*
987 * Walk up the subvolume trees in the tree of tree roots by root
988 * backrefs until we hit the top-level subvolume.
989 */
990 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
991 key.objectid = subvol_objectid;
992 key.type = BTRFS_ROOT_BACKREF_KEY;
993 key.offset = (u64)-1;
994
995 ret = btrfs_search_backwards(root, &key, path);
996 if (ret < 0) {
997 goto err;
998 } else if (ret > 0) {
999 ret = -ENOENT;
1000 goto err;
1001 }
1002
1003 subvol_objectid = key.offset;
1004
1005 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1006 struct btrfs_root_ref);
1007 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1008 ptr -= len + 1;
1009 if (ptr < name) {
1010 ret = -ENAMETOOLONG;
1011 goto err;
1012 }
1013 read_extent_buffer(path->nodes[0], ptr + 1,
1014 (unsigned long)(root_ref + 1), len);
1015 ptr[0] = '/';
1016 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1017 btrfs_release_path(path);
1018
1019 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1020 if (IS_ERR(fs_root)) {
1021 ret = PTR_ERR(fs_root);
1022 fs_root = NULL;
1023 goto err;
1024 }
1025
1026 /*
1027 * Walk up the filesystem tree by inode refs until we hit the
1028 * root directory.
1029 */
1030 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1031 key.objectid = dirid;
1032 key.type = BTRFS_INODE_REF_KEY;
1033 key.offset = (u64)-1;
1034
1035 ret = btrfs_search_backwards(fs_root, &key, path);
1036 if (ret < 0) {
1037 goto err;
1038 } else if (ret > 0) {
1039 ret = -ENOENT;
1040 goto err;
1041 }
1042
1043 dirid = key.offset;
1044
1045 inode_ref = btrfs_item_ptr(path->nodes[0],
1046 path->slots[0],
1047 struct btrfs_inode_ref);
1048 len = btrfs_inode_ref_name_len(path->nodes[0],
1049 inode_ref);
1050 ptr -= len + 1;
1051 if (ptr < name) {
1052 ret = -ENAMETOOLONG;
1053 goto err;
1054 }
1055 read_extent_buffer(path->nodes[0], ptr + 1,
1056 (unsigned long)(inode_ref + 1), len);
1057 ptr[0] = '/';
1058 btrfs_release_path(path);
1059 }
1060 btrfs_put_root(fs_root);
1061 fs_root = NULL;
1062 }
1063
1064 btrfs_free_path(path);
1065 if (ptr == name + PATH_MAX - 1) {
1066 name[0] = '/';
1067 name[1] = '\0';
1068 } else {
1069 memmove(name, ptr, name + PATH_MAX - ptr);
1070 }
1071 return name;
1072
1073err:
1074 btrfs_put_root(fs_root);
1075 btrfs_free_path(path);
1076 kfree(name);
1077 return ERR_PTR(ret);
1078}
1079
1080static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1081{
1082 struct btrfs_root *root = fs_info->tree_root;
1083 struct btrfs_dir_item *di;
1084 struct btrfs_path *path;
1085 struct btrfs_key location;
1086 struct fscrypt_str name = FSTR_INIT("default", 7);
1087 u64 dir_id;
1088
1089 path = btrfs_alloc_path();
1090 if (!path)
1091 return -ENOMEM;
1092
1093 /*
1094 * Find the "default" dir item which points to the root item that we
1095 * will mount by default if we haven't been given a specific subvolume
1096 * to mount.
1097 */
1098 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1099 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1100 if (IS_ERR(di)) {
1101 btrfs_free_path(path);
1102 return PTR_ERR(di);
1103 }
1104 if (!di) {
1105 /*
1106 * Ok the default dir item isn't there. This is weird since
1107 * it's always been there, but don't freak out, just try and
1108 * mount the top-level subvolume.
1109 */
1110 btrfs_free_path(path);
1111 *objectid = BTRFS_FS_TREE_OBJECTID;
1112 return 0;
1113 }
1114
1115 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1116 btrfs_free_path(path);
1117 *objectid = location.objectid;
1118 return 0;
1119}
1120
1121static int btrfs_fill_super(struct super_block *sb,
1122 struct btrfs_fs_devices *fs_devices,
1123 void *data)
1124{
1125 struct inode *inode;
1126 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1127 int err;
1128
1129 sb->s_maxbytes = MAX_LFS_FILESIZE;
1130 sb->s_magic = BTRFS_SUPER_MAGIC;
1131 sb->s_op = &btrfs_super_ops;
1132 sb->s_d_op = &btrfs_dentry_operations;
1133 sb->s_export_op = &btrfs_export_ops;
1134#ifdef CONFIG_FS_VERITY
1135 sb->s_vop = &btrfs_verityops;
1136#endif
1137 sb->s_xattr = btrfs_xattr_handlers;
1138 sb->s_time_gran = 1;
1139#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1140 sb->s_flags |= SB_POSIXACL;
1141#endif
1142 sb->s_flags |= SB_I_VERSION;
1143 sb->s_iflags |= SB_I_CGROUPWB;
1144
1145 err = super_setup_bdi(sb);
1146 if (err) {
1147 btrfs_err(fs_info, "super_setup_bdi failed");
1148 return err;
1149 }
1150
1151 err = open_ctree(sb, fs_devices, (char *)data);
1152 if (err) {
1153 btrfs_err(fs_info, "open_ctree failed");
1154 return err;
1155 }
1156
1157 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1158 if (IS_ERR(inode)) {
1159 err = PTR_ERR(inode);
1160 goto fail_close;
1161 }
1162
1163 sb->s_root = d_make_root(inode);
1164 if (!sb->s_root) {
1165 err = -ENOMEM;
1166 goto fail_close;
1167 }
1168
1169 sb->s_flags |= SB_ACTIVE;
1170 return 0;
1171
1172fail_close:
1173 close_ctree(fs_info);
1174 return err;
1175}
1176
1177int btrfs_sync_fs(struct super_block *sb, int wait)
1178{
1179 struct btrfs_trans_handle *trans;
1180 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1181 struct btrfs_root *root = fs_info->tree_root;
1182
1183 trace_btrfs_sync_fs(fs_info, wait);
1184
1185 if (!wait) {
1186 filemap_flush(fs_info->btree_inode->i_mapping);
1187 return 0;
1188 }
1189
1190 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1191
1192 trans = btrfs_attach_transaction_barrier(root);
1193 if (IS_ERR(trans)) {
1194 /* no transaction, don't bother */
1195 if (PTR_ERR(trans) == -ENOENT) {
1196 /*
1197 * Exit unless we have some pending changes
1198 * that need to go through commit
1199 */
1200 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1201 &fs_info->flags))
1202 return 0;
1203 /*
1204 * A non-blocking test if the fs is frozen. We must not
1205 * start a new transaction here otherwise a deadlock
1206 * happens. The pending operations are delayed to the
1207 * next commit after thawing.
1208 */
1209 if (sb_start_write_trylock(sb))
1210 sb_end_write(sb);
1211 else
1212 return 0;
1213 trans = btrfs_start_transaction(root, 0);
1214 }
1215 if (IS_ERR(trans))
1216 return PTR_ERR(trans);
1217 }
1218 return btrfs_commit_transaction(trans);
1219}
1220
1221static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1222{
1223 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1224 *printed = true;
1225}
1226
1227static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1228{
1229 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1230 const char *compress_type;
1231 const char *subvol_name;
1232 bool printed = false;
1233
1234 if (btrfs_test_opt(info, DEGRADED))
1235 seq_puts(seq, ",degraded");
1236 if (btrfs_test_opt(info, NODATASUM))
1237 seq_puts(seq, ",nodatasum");
1238 if (btrfs_test_opt(info, NODATACOW))
1239 seq_puts(seq, ",nodatacow");
1240 if (btrfs_test_opt(info, NOBARRIER))
1241 seq_puts(seq, ",nobarrier");
1242 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1243 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1244 if (info->thread_pool_size != min_t(unsigned long,
1245 num_online_cpus() + 2, 8))
1246 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1247 if (btrfs_test_opt(info, COMPRESS)) {
1248 compress_type = btrfs_compress_type2str(info->compress_type);
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 if (info->compress_level)
1254 seq_printf(seq, ":%d", info->compress_level);
1255 }
1256 if (btrfs_test_opt(info, NOSSD))
1257 seq_puts(seq, ",nossd");
1258 if (btrfs_test_opt(info, SSD_SPREAD))
1259 seq_puts(seq, ",ssd_spread");
1260 else if (btrfs_test_opt(info, SSD))
1261 seq_puts(seq, ",ssd");
1262 if (btrfs_test_opt(info, NOTREELOG))
1263 seq_puts(seq, ",notreelog");
1264 if (btrfs_test_opt(info, NOLOGREPLAY))
1265 print_rescue_option(seq, "nologreplay", &printed);
1266 if (btrfs_test_opt(info, USEBACKUPROOT))
1267 print_rescue_option(seq, "usebackuproot", &printed);
1268 if (btrfs_test_opt(info, IGNOREBADROOTS))
1269 print_rescue_option(seq, "ignorebadroots", &printed);
1270 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1271 print_rescue_option(seq, "ignoredatacsums", &printed);
1272 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1273 seq_puts(seq, ",flushoncommit");
1274 if (btrfs_test_opt(info, DISCARD_SYNC))
1275 seq_puts(seq, ",discard");
1276 if (btrfs_test_opt(info, DISCARD_ASYNC))
1277 seq_puts(seq, ",discard=async");
1278 if (!(info->sb->s_flags & SB_POSIXACL))
1279 seq_puts(seq, ",noacl");
1280 if (btrfs_free_space_cache_v1_active(info))
1281 seq_puts(seq, ",space_cache");
1282 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1283 seq_puts(seq, ",space_cache=v2");
1284 else
1285 seq_puts(seq, ",nospace_cache");
1286 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1287 seq_puts(seq, ",rescan_uuid_tree");
1288 if (btrfs_test_opt(info, CLEAR_CACHE))
1289 seq_puts(seq, ",clear_cache");
1290 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1291 seq_puts(seq, ",user_subvol_rm_allowed");
1292 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1293 seq_puts(seq, ",enospc_debug");
1294 if (btrfs_test_opt(info, AUTO_DEFRAG))
1295 seq_puts(seq, ",autodefrag");
1296 if (btrfs_test_opt(info, SKIP_BALANCE))
1297 seq_puts(seq, ",skip_balance");
1298#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1299 if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1300 seq_puts(seq, ",check_int_data");
1301 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1302 seq_puts(seq, ",check_int");
1303 if (info->check_integrity_print_mask)
1304 seq_printf(seq, ",check_int_print_mask=%d",
1305 info->check_integrity_print_mask);
1306#endif
1307 if (info->metadata_ratio)
1308 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1309 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1310 seq_puts(seq, ",fatal_errors=panic");
1311 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1312 seq_printf(seq, ",commit=%u", info->commit_interval);
1313#ifdef CONFIG_BTRFS_DEBUG
1314 if (btrfs_test_opt(info, FRAGMENT_DATA))
1315 seq_puts(seq, ",fragment=data");
1316 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1317 seq_puts(seq, ",fragment=metadata");
1318#endif
1319 if (btrfs_test_opt(info, REF_VERIFY))
1320 seq_puts(seq, ",ref_verify");
1321 seq_printf(seq, ",subvolid=%llu",
1322 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1323 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1324 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1325 if (!IS_ERR(subvol_name)) {
1326 seq_puts(seq, ",subvol=");
1327 seq_escape(seq, subvol_name, " \t\n\\");
1328 kfree(subvol_name);
1329 }
1330 return 0;
1331}
1332
1333static int btrfs_test_super(struct super_block *s, void *data)
1334{
1335 struct btrfs_fs_info *p = data;
1336 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1337
1338 return fs_info->fs_devices == p->fs_devices;
1339}
1340
1341static int btrfs_set_super(struct super_block *s, void *data)
1342{
1343 int err = set_anon_super(s, data);
1344 if (!err)
1345 s->s_fs_info = data;
1346 return err;
1347}
1348
1349/*
1350 * subvolumes are identified by ino 256
1351 */
1352static inline int is_subvolume_inode(struct inode *inode)
1353{
1354 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1355 return 1;
1356 return 0;
1357}
1358
1359static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1360 struct vfsmount *mnt)
1361{
1362 struct dentry *root;
1363 int ret;
1364
1365 if (!subvol_name) {
1366 if (!subvol_objectid) {
1367 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1368 &subvol_objectid);
1369 if (ret) {
1370 root = ERR_PTR(ret);
1371 goto out;
1372 }
1373 }
1374 subvol_name = btrfs_get_subvol_name_from_objectid(
1375 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1376 if (IS_ERR(subvol_name)) {
1377 root = ERR_CAST(subvol_name);
1378 subvol_name = NULL;
1379 goto out;
1380 }
1381
1382 }
1383
1384 root = mount_subtree(mnt, subvol_name);
1385 /* mount_subtree() drops our reference on the vfsmount. */
1386 mnt = NULL;
1387
1388 if (!IS_ERR(root)) {
1389 struct super_block *s = root->d_sb;
1390 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1391 struct inode *root_inode = d_inode(root);
1392 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1393
1394 ret = 0;
1395 if (!is_subvolume_inode(root_inode)) {
1396 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1397 subvol_name);
1398 ret = -EINVAL;
1399 }
1400 if (subvol_objectid && root_objectid != subvol_objectid) {
1401 /*
1402 * This will also catch a race condition where a
1403 * subvolume which was passed by ID is renamed and
1404 * another subvolume is renamed over the old location.
1405 */
1406 btrfs_err(fs_info,
1407 "subvol '%s' does not match subvolid %llu",
1408 subvol_name, subvol_objectid);
1409 ret = -EINVAL;
1410 }
1411 if (ret) {
1412 dput(root);
1413 root = ERR_PTR(ret);
1414 deactivate_locked_super(s);
1415 }
1416 }
1417
1418out:
1419 mntput(mnt);
1420 kfree(subvol_name);
1421 return root;
1422}
1423
1424/*
1425 * Find a superblock for the given device / mount point.
1426 *
1427 * Note: This is based on mount_bdev from fs/super.c with a few additions
1428 * for multiple device setup. Make sure to keep it in sync.
1429 */
1430static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1431 int flags, const char *device_name, void *data)
1432{
1433 struct block_device *bdev = NULL;
1434 struct super_block *s;
1435 struct btrfs_device *device = NULL;
1436 struct btrfs_fs_devices *fs_devices = NULL;
1437 struct btrfs_fs_info *fs_info = NULL;
1438 void *new_sec_opts = NULL;
1439 fmode_t mode = FMODE_READ;
1440 int error = 0;
1441
1442 if (!(flags & SB_RDONLY))
1443 mode |= FMODE_WRITE;
1444
1445 if (data) {
1446 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1447 if (error)
1448 return ERR_PTR(error);
1449 }
1450
1451 /*
1452 * Setup a dummy root and fs_info for test/set super. This is because
1453 * we don't actually fill this stuff out until open_ctree, but we need
1454 * then open_ctree will properly initialize the file system specific
1455 * settings later. btrfs_init_fs_info initializes the static elements
1456 * of the fs_info (locks and such) to make cleanup easier if we find a
1457 * superblock with our given fs_devices later on at sget() time.
1458 */
1459 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1460 if (!fs_info) {
1461 error = -ENOMEM;
1462 goto error_sec_opts;
1463 }
1464 btrfs_init_fs_info(fs_info);
1465
1466 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1467 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1468 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1469 error = -ENOMEM;
1470 goto error_fs_info;
1471 }
1472
1473 mutex_lock(&uuid_mutex);
1474 error = btrfs_parse_device_options(data, mode, fs_type);
1475 if (error) {
1476 mutex_unlock(&uuid_mutex);
1477 goto error_fs_info;
1478 }
1479
1480 device = btrfs_scan_one_device(device_name, mode, fs_type);
1481 if (IS_ERR(device)) {
1482 mutex_unlock(&uuid_mutex);
1483 error = PTR_ERR(device);
1484 goto error_fs_info;
1485 }
1486
1487 fs_devices = device->fs_devices;
1488 fs_info->fs_devices = fs_devices;
1489
1490 error = btrfs_open_devices(fs_devices, mode, fs_type);
1491 mutex_unlock(&uuid_mutex);
1492 if (error)
1493 goto error_fs_info;
1494
1495 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1496 error = -EACCES;
1497 goto error_close_devices;
1498 }
1499
1500 bdev = fs_devices->latest_dev->bdev;
1501 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1502 fs_info);
1503 if (IS_ERR(s)) {
1504 error = PTR_ERR(s);
1505 goto error_close_devices;
1506 }
1507
1508 if (s->s_root) {
1509 btrfs_close_devices(fs_devices);
1510 btrfs_free_fs_info(fs_info);
1511 if ((flags ^ s->s_flags) & SB_RDONLY)
1512 error = -EBUSY;
1513 } else {
1514 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1515 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1516 s->s_id);
1517 btrfs_sb(s)->bdev_holder = fs_type;
1518 if (!strstr(crc32c_impl(), "generic"))
1519 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1520 error = btrfs_fill_super(s, fs_devices, data);
1521 }
1522 if (!error)
1523 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1524 security_free_mnt_opts(&new_sec_opts);
1525 if (error) {
1526 deactivate_locked_super(s);
1527 return ERR_PTR(error);
1528 }
1529
1530 return dget(s->s_root);
1531
1532error_close_devices:
1533 btrfs_close_devices(fs_devices);
1534error_fs_info:
1535 btrfs_free_fs_info(fs_info);
1536error_sec_opts:
1537 security_free_mnt_opts(&new_sec_opts);
1538 return ERR_PTR(error);
1539}
1540
1541/*
1542 * Mount function which is called by VFS layer.
1543 *
1544 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1545 * which needs vfsmount* of device's root (/). This means device's root has to
1546 * be mounted internally in any case.
1547 *
1548 * Operation flow:
1549 * 1. Parse subvol id related options for later use in mount_subvol().
1550 *
1551 * 2. Mount device's root (/) by calling vfs_kern_mount().
1552 *
1553 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1554 * first place. In order to avoid calling btrfs_mount() again, we use
1555 * different file_system_type which is not registered to VFS by
1556 * register_filesystem() (btrfs_root_fs_type). As a result,
1557 * btrfs_mount_root() is called. The return value will be used by
1558 * mount_subtree() in mount_subvol().
1559 *
1560 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1561 * "btrfs subvolume set-default", mount_subvol() is called always.
1562 */
1563static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1564 const char *device_name, void *data)
1565{
1566 struct vfsmount *mnt_root;
1567 struct dentry *root;
1568 char *subvol_name = NULL;
1569 u64 subvol_objectid = 0;
1570 int error = 0;
1571
1572 error = btrfs_parse_subvol_options(data, &subvol_name,
1573 &subvol_objectid);
1574 if (error) {
1575 kfree(subvol_name);
1576 return ERR_PTR(error);
1577 }
1578
1579 /* mount device's root (/) */
1580 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1581 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1582 if (flags & SB_RDONLY) {
1583 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1584 flags & ~SB_RDONLY, device_name, data);
1585 } else {
1586 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1587 flags | SB_RDONLY, device_name, data);
1588 if (IS_ERR(mnt_root)) {
1589 root = ERR_CAST(mnt_root);
1590 kfree(subvol_name);
1591 goto out;
1592 }
1593
1594 down_write(&mnt_root->mnt_sb->s_umount);
1595 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1596 up_write(&mnt_root->mnt_sb->s_umount);
1597 if (error < 0) {
1598 root = ERR_PTR(error);
1599 mntput(mnt_root);
1600 kfree(subvol_name);
1601 goto out;
1602 }
1603 }
1604 }
1605 if (IS_ERR(mnt_root)) {
1606 root = ERR_CAST(mnt_root);
1607 kfree(subvol_name);
1608 goto out;
1609 }
1610
1611 /* mount_subvol() will free subvol_name and mnt_root */
1612 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1613
1614out:
1615 return root;
1616}
1617
1618static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1619 u32 new_pool_size, u32 old_pool_size)
1620{
1621 if (new_pool_size == old_pool_size)
1622 return;
1623
1624 fs_info->thread_pool_size = new_pool_size;
1625
1626 btrfs_info(fs_info, "resize thread pool %d -> %d",
1627 old_pool_size, new_pool_size);
1628
1629 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1630 btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1631 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1632 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1633 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1634 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1635 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1636}
1637
1638static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1639 unsigned long old_opts, int flags)
1640{
1641 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1642 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1643 (flags & SB_RDONLY))) {
1644 /* wait for any defraggers to finish */
1645 wait_event(fs_info->transaction_wait,
1646 (atomic_read(&fs_info->defrag_running) == 0));
1647 if (flags & SB_RDONLY)
1648 sync_filesystem(fs_info->sb);
1649 }
1650}
1651
1652static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1653 unsigned long old_opts)
1654{
1655 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1656
1657 /*
1658 * We need to cleanup all defragable inodes if the autodefragment is
1659 * close or the filesystem is read only.
1660 */
1661 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1662 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1663 btrfs_cleanup_defrag_inodes(fs_info);
1664 }
1665
1666 /* If we toggled discard async */
1667 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1668 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1669 btrfs_discard_resume(fs_info);
1670 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1671 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1672 btrfs_discard_cleanup(fs_info);
1673
1674 /* If we toggled space cache */
1675 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1676 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1677}
1678
1679static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1680{
1681 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1682 unsigned old_flags = sb->s_flags;
1683 unsigned long old_opts = fs_info->mount_opt;
1684 unsigned long old_compress_type = fs_info->compress_type;
1685 u64 old_max_inline = fs_info->max_inline;
1686 u32 old_thread_pool_size = fs_info->thread_pool_size;
1687 u32 old_metadata_ratio = fs_info->metadata_ratio;
1688 int ret;
1689
1690 sync_filesystem(sb);
1691 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1692
1693 if (data) {
1694 void *new_sec_opts = NULL;
1695
1696 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1697 if (!ret)
1698 ret = security_sb_remount(sb, new_sec_opts);
1699 security_free_mnt_opts(&new_sec_opts);
1700 if (ret)
1701 goto restore;
1702 }
1703
1704 ret = btrfs_parse_options(fs_info, data, *flags);
1705 if (ret)
1706 goto restore;
1707
1708 ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
1709 if (ret < 0)
1710 goto restore;
1711
1712 btrfs_remount_begin(fs_info, old_opts, *flags);
1713 btrfs_resize_thread_pool(fs_info,
1714 fs_info->thread_pool_size, old_thread_pool_size);
1715
1716 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1717 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1718 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1719 btrfs_warn(fs_info,
1720 "remount supports changing free space tree only from ro to rw");
1721 /* Make sure free space cache options match the state on disk */
1722 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1723 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1724 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1725 }
1726 if (btrfs_free_space_cache_v1_active(fs_info)) {
1727 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1728 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1729 }
1730 }
1731
1732 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1733 goto out;
1734
1735 if (*flags & SB_RDONLY) {
1736 /*
1737 * this also happens on 'umount -rf' or on shutdown, when
1738 * the filesystem is busy.
1739 */
1740 cancel_work_sync(&fs_info->async_reclaim_work);
1741 cancel_work_sync(&fs_info->async_data_reclaim_work);
1742
1743 btrfs_discard_cleanup(fs_info);
1744
1745 /* wait for the uuid_scan task to finish */
1746 down(&fs_info->uuid_tree_rescan_sem);
1747 /* avoid complains from lockdep et al. */
1748 up(&fs_info->uuid_tree_rescan_sem);
1749
1750 btrfs_set_sb_rdonly(sb);
1751
1752 /*
1753 * Setting SB_RDONLY will put the cleaner thread to
1754 * sleep at the next loop if it's already active.
1755 * If it's already asleep, we'll leave unused block
1756 * groups on disk until we're mounted read-write again
1757 * unless we clean them up here.
1758 */
1759 btrfs_delete_unused_bgs(fs_info);
1760
1761 /*
1762 * The cleaner task could be already running before we set the
1763 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1764 * We must make sure that after we finish the remount, i.e. after
1765 * we call btrfs_commit_super(), the cleaner can no longer start
1766 * a transaction - either because it was dropping a dead root,
1767 * running delayed iputs or deleting an unused block group (the
1768 * cleaner picked a block group from the list of unused block
1769 * groups before we were able to in the previous call to
1770 * btrfs_delete_unused_bgs()).
1771 */
1772 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1773 TASK_UNINTERRUPTIBLE);
1774
1775 /*
1776 * We've set the superblock to RO mode, so we might have made
1777 * the cleaner task sleep without running all pending delayed
1778 * iputs. Go through all the delayed iputs here, so that if an
1779 * unmount happens without remounting RW we don't end up at
1780 * finishing close_ctree() with a non-empty list of delayed
1781 * iputs.
1782 */
1783 btrfs_run_delayed_iputs(fs_info);
1784
1785 btrfs_dev_replace_suspend_for_unmount(fs_info);
1786 btrfs_scrub_cancel(fs_info);
1787 btrfs_pause_balance(fs_info);
1788
1789 /*
1790 * Pause the qgroup rescan worker if it is running. We don't want
1791 * it to be still running after we are in RO mode, as after that,
1792 * by the time we unmount, it might have left a transaction open,
1793 * so we would leak the transaction and/or crash.
1794 */
1795 btrfs_qgroup_wait_for_completion(fs_info, false);
1796
1797 ret = btrfs_commit_super(fs_info);
1798 if (ret)
1799 goto restore;
1800 } else {
1801 if (BTRFS_FS_ERROR(fs_info)) {
1802 btrfs_err(fs_info,
1803 "Remounting read-write after error is not allowed");
1804 ret = -EINVAL;
1805 goto restore;
1806 }
1807 if (fs_info->fs_devices->rw_devices == 0) {
1808 ret = -EACCES;
1809 goto restore;
1810 }
1811
1812 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1813 btrfs_warn(fs_info,
1814 "too many missing devices, writable remount is not allowed");
1815 ret = -EACCES;
1816 goto restore;
1817 }
1818
1819 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1820 btrfs_warn(fs_info,
1821 "mount required to replay tree-log, cannot remount read-write");
1822 ret = -EINVAL;
1823 goto restore;
1824 }
1825
1826 /*
1827 * NOTE: when remounting with a change that does writes, don't
1828 * put it anywhere above this point, as we are not sure to be
1829 * safe to write until we pass the above checks.
1830 */
1831 ret = btrfs_start_pre_rw_mount(fs_info);
1832 if (ret)
1833 goto restore;
1834
1835 btrfs_clear_sb_rdonly(sb);
1836
1837 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1838 }
1839out:
1840 /*
1841 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1842 * since the absence of the flag means it can be toggled off by remount.
1843 */
1844 *flags |= SB_I_VERSION;
1845
1846 wake_up_process(fs_info->transaction_kthread);
1847 btrfs_remount_cleanup(fs_info, old_opts);
1848 btrfs_clear_oneshot_options(fs_info);
1849 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1850
1851 return 0;
1852
1853restore:
1854 /* We've hit an error - don't reset SB_RDONLY */
1855 if (sb_rdonly(sb))
1856 old_flags |= SB_RDONLY;
1857 if (!(old_flags & SB_RDONLY))
1858 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1859 sb->s_flags = old_flags;
1860 fs_info->mount_opt = old_opts;
1861 fs_info->compress_type = old_compress_type;
1862 fs_info->max_inline = old_max_inline;
1863 btrfs_resize_thread_pool(fs_info,
1864 old_thread_pool_size, fs_info->thread_pool_size);
1865 fs_info->metadata_ratio = old_metadata_ratio;
1866 btrfs_remount_cleanup(fs_info, old_opts);
1867 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1868
1869 return ret;
1870}
1871
1872/* Used to sort the devices by max_avail(descending sort) */
1873static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1874{
1875 const struct btrfs_device_info *dev_info1 = a;
1876 const struct btrfs_device_info *dev_info2 = b;
1877
1878 if (dev_info1->max_avail > dev_info2->max_avail)
1879 return -1;
1880 else if (dev_info1->max_avail < dev_info2->max_avail)
1881 return 1;
1882 return 0;
1883}
1884
1885/*
1886 * sort the devices by max_avail, in which max free extent size of each device
1887 * is stored.(Descending Sort)
1888 */
1889static inline void btrfs_descending_sort_devices(
1890 struct btrfs_device_info *devices,
1891 size_t nr_devices)
1892{
1893 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1894 btrfs_cmp_device_free_bytes, NULL);
1895}
1896
1897/*
1898 * The helper to calc the free space on the devices that can be used to store
1899 * file data.
1900 */
1901static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1902 u64 *free_bytes)
1903{
1904 struct btrfs_device_info *devices_info;
1905 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1906 struct btrfs_device *device;
1907 u64 type;
1908 u64 avail_space;
1909 u64 min_stripe_size;
1910 int num_stripes = 1;
1911 int i = 0, nr_devices;
1912 const struct btrfs_raid_attr *rattr;
1913
1914 /*
1915 * We aren't under the device list lock, so this is racy-ish, but good
1916 * enough for our purposes.
1917 */
1918 nr_devices = fs_info->fs_devices->open_devices;
1919 if (!nr_devices) {
1920 smp_mb();
1921 nr_devices = fs_info->fs_devices->open_devices;
1922 ASSERT(nr_devices);
1923 if (!nr_devices) {
1924 *free_bytes = 0;
1925 return 0;
1926 }
1927 }
1928
1929 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1930 GFP_KERNEL);
1931 if (!devices_info)
1932 return -ENOMEM;
1933
1934 /* calc min stripe number for data space allocation */
1935 type = btrfs_data_alloc_profile(fs_info);
1936 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1937
1938 if (type & BTRFS_BLOCK_GROUP_RAID0)
1939 num_stripes = nr_devices;
1940 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1941 num_stripes = rattr->ncopies;
1942 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1943 num_stripes = 4;
1944
1945 /* Adjust for more than 1 stripe per device */
1946 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1947
1948 rcu_read_lock();
1949 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1950 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1951 &device->dev_state) ||
1952 !device->bdev ||
1953 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1954 continue;
1955
1956 if (i >= nr_devices)
1957 break;
1958
1959 avail_space = device->total_bytes - device->bytes_used;
1960
1961 /* align with stripe_len */
1962 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1963
1964 /*
1965 * Ensure we have at least min_stripe_size on top of the
1966 * reserved space on the device.
1967 */
1968 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1969 continue;
1970
1971 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1972
1973 devices_info[i].dev = device;
1974 devices_info[i].max_avail = avail_space;
1975
1976 i++;
1977 }
1978 rcu_read_unlock();
1979
1980 nr_devices = i;
1981
1982 btrfs_descending_sort_devices(devices_info, nr_devices);
1983
1984 i = nr_devices - 1;
1985 avail_space = 0;
1986 while (nr_devices >= rattr->devs_min) {
1987 num_stripes = min(num_stripes, nr_devices);
1988
1989 if (devices_info[i].max_avail >= min_stripe_size) {
1990 int j;
1991 u64 alloc_size;
1992
1993 avail_space += devices_info[i].max_avail * num_stripes;
1994 alloc_size = devices_info[i].max_avail;
1995 for (j = i + 1 - num_stripes; j <= i; j++)
1996 devices_info[j].max_avail -= alloc_size;
1997 }
1998 i--;
1999 nr_devices--;
2000 }
2001
2002 kfree(devices_info);
2003 *free_bytes = avail_space;
2004 return 0;
2005}
2006
2007/*
2008 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2009 *
2010 * If there's a redundant raid level at DATA block groups, use the respective
2011 * multiplier to scale the sizes.
2012 *
2013 * Unused device space usage is based on simulating the chunk allocator
2014 * algorithm that respects the device sizes and order of allocations. This is
2015 * a close approximation of the actual use but there are other factors that may
2016 * change the result (like a new metadata chunk).
2017 *
2018 * If metadata is exhausted, f_bavail will be 0.
2019 */
2020static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2021{
2022 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2023 struct btrfs_super_block *disk_super = fs_info->super_copy;
2024 struct btrfs_space_info *found;
2025 u64 total_used = 0;
2026 u64 total_free_data = 0;
2027 u64 total_free_meta = 0;
2028 u32 bits = fs_info->sectorsize_bits;
2029 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2030 unsigned factor = 1;
2031 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2032 int ret;
2033 u64 thresh = 0;
2034 int mixed = 0;
2035
2036 list_for_each_entry(found, &fs_info->space_info, list) {
2037 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2038 int i;
2039
2040 total_free_data += found->disk_total - found->disk_used;
2041 total_free_data -=
2042 btrfs_account_ro_block_groups_free_space(found);
2043
2044 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2045 if (!list_empty(&found->block_groups[i]))
2046 factor = btrfs_bg_type_to_factor(
2047 btrfs_raid_array[i].bg_flag);
2048 }
2049 }
2050
2051 /*
2052 * Metadata in mixed block goup profiles are accounted in data
2053 */
2054 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2055 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2056 mixed = 1;
2057 else
2058 total_free_meta += found->disk_total -
2059 found->disk_used;
2060 }
2061
2062 total_used += found->disk_used;
2063 }
2064
2065 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2066 buf->f_blocks >>= bits;
2067 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2068
2069 /* Account global block reserve as used, it's in logical size already */
2070 spin_lock(&block_rsv->lock);
2071 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2072 if (buf->f_bfree >= block_rsv->size >> bits)
2073 buf->f_bfree -= block_rsv->size >> bits;
2074 else
2075 buf->f_bfree = 0;
2076 spin_unlock(&block_rsv->lock);
2077
2078 buf->f_bavail = div_u64(total_free_data, factor);
2079 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2080 if (ret)
2081 return ret;
2082 buf->f_bavail += div_u64(total_free_data, factor);
2083 buf->f_bavail = buf->f_bavail >> bits;
2084
2085 /*
2086 * We calculate the remaining metadata space minus global reserve. If
2087 * this is (supposedly) smaller than zero, there's no space. But this
2088 * does not hold in practice, the exhausted state happens where's still
2089 * some positive delta. So we apply some guesswork and compare the
2090 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2091 *
2092 * We probably cannot calculate the exact threshold value because this
2093 * depends on the internal reservations requested by various
2094 * operations, so some operations that consume a few metadata will
2095 * succeed even if the Avail is zero. But this is better than the other
2096 * way around.
2097 */
2098 thresh = SZ_4M;
2099
2100 /*
2101 * We only want to claim there's no available space if we can no longer
2102 * allocate chunks for our metadata profile and our global reserve will
2103 * not fit in the free metadata space. If we aren't ->full then we
2104 * still can allocate chunks and thus are fine using the currently
2105 * calculated f_bavail.
2106 */
2107 if (!mixed && block_rsv->space_info->full &&
2108 total_free_meta - thresh < block_rsv->size)
2109 buf->f_bavail = 0;
2110
2111 buf->f_type = BTRFS_SUPER_MAGIC;
2112 buf->f_bsize = dentry->d_sb->s_blocksize;
2113 buf->f_namelen = BTRFS_NAME_LEN;
2114
2115 /* We treat it as constant endianness (it doesn't matter _which_)
2116 because we want the fsid to come out the same whether mounted
2117 on a big-endian or little-endian host */
2118 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2119 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2120 /* Mask in the root object ID too, to disambiguate subvols */
2121 buf->f_fsid.val[0] ^=
2122 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2123 buf->f_fsid.val[1] ^=
2124 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2125
2126 return 0;
2127}
2128
2129static void btrfs_kill_super(struct super_block *sb)
2130{
2131 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2132 kill_anon_super(sb);
2133 btrfs_free_fs_info(fs_info);
2134}
2135
2136static struct file_system_type btrfs_fs_type = {
2137 .owner = THIS_MODULE,
2138 .name = "btrfs",
2139 .mount = btrfs_mount,
2140 .kill_sb = btrfs_kill_super,
2141 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2142};
2143
2144static struct file_system_type btrfs_root_fs_type = {
2145 .owner = THIS_MODULE,
2146 .name = "btrfs",
2147 .mount = btrfs_mount_root,
2148 .kill_sb = btrfs_kill_super,
2149 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2150};
2151
2152MODULE_ALIAS_FS("btrfs");
2153
2154static int btrfs_control_open(struct inode *inode, struct file *file)
2155{
2156 /*
2157 * The control file's private_data is used to hold the
2158 * transaction when it is started and is used to keep
2159 * track of whether a transaction is already in progress.
2160 */
2161 file->private_data = NULL;
2162 return 0;
2163}
2164
2165/*
2166 * Used by /dev/btrfs-control for devices ioctls.
2167 */
2168static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2169 unsigned long arg)
2170{
2171 struct btrfs_ioctl_vol_args *vol;
2172 struct btrfs_device *device = NULL;
2173 dev_t devt = 0;
2174 int ret = -ENOTTY;
2175
2176 if (!capable(CAP_SYS_ADMIN))
2177 return -EPERM;
2178
2179 vol = memdup_user((void __user *)arg, sizeof(*vol));
2180 if (IS_ERR(vol))
2181 return PTR_ERR(vol);
2182 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2183
2184 switch (cmd) {
2185 case BTRFS_IOC_SCAN_DEV:
2186 mutex_lock(&uuid_mutex);
2187 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2188 &btrfs_root_fs_type);
2189 ret = PTR_ERR_OR_ZERO(device);
2190 mutex_unlock(&uuid_mutex);
2191 break;
2192 case BTRFS_IOC_FORGET_DEV:
2193 if (vol->name[0] != 0) {
2194 ret = lookup_bdev(vol->name, &devt);
2195 if (ret)
2196 break;
2197 }
2198 ret = btrfs_forget_devices(devt);
2199 break;
2200 case BTRFS_IOC_DEVICES_READY:
2201 mutex_lock(&uuid_mutex);
2202 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2203 &btrfs_root_fs_type);
2204 if (IS_ERR(device)) {
2205 mutex_unlock(&uuid_mutex);
2206 ret = PTR_ERR(device);
2207 break;
2208 }
2209 ret = !(device->fs_devices->num_devices ==
2210 device->fs_devices->total_devices);
2211 mutex_unlock(&uuid_mutex);
2212 break;
2213 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2214 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2215 break;
2216 }
2217
2218 kfree(vol);
2219 return ret;
2220}
2221
2222static int btrfs_freeze(struct super_block *sb)
2223{
2224 struct btrfs_trans_handle *trans;
2225 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2226 struct btrfs_root *root = fs_info->tree_root;
2227
2228 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2229 /*
2230 * We don't need a barrier here, we'll wait for any transaction that
2231 * could be in progress on other threads (and do delayed iputs that
2232 * we want to avoid on a frozen filesystem), or do the commit
2233 * ourselves.
2234 */
2235 trans = btrfs_attach_transaction_barrier(root);
2236 if (IS_ERR(trans)) {
2237 /* no transaction, don't bother */
2238 if (PTR_ERR(trans) == -ENOENT)
2239 return 0;
2240 return PTR_ERR(trans);
2241 }
2242 return btrfs_commit_transaction(trans);
2243}
2244
2245static int check_dev_super(struct btrfs_device *dev)
2246{
2247 struct btrfs_fs_info *fs_info = dev->fs_info;
2248 struct btrfs_super_block *sb;
2249 u16 csum_type;
2250 int ret = 0;
2251
2252 /* This should be called with fs still frozen. */
2253 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2254
2255 /* Missing dev, no need to check. */
2256 if (!dev->bdev)
2257 return 0;
2258
2259 /* Only need to check the primary super block. */
2260 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2261 if (IS_ERR(sb))
2262 return PTR_ERR(sb);
2263
2264 /* Verify the checksum. */
2265 csum_type = btrfs_super_csum_type(sb);
2266 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2267 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2268 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2269 ret = -EUCLEAN;
2270 goto out;
2271 }
2272
2273 if (btrfs_check_super_csum(fs_info, sb)) {
2274 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2275 ret = -EUCLEAN;
2276 goto out;
2277 }
2278
2279 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2280 ret = btrfs_validate_super(fs_info, sb, 0);
2281 if (ret < 0)
2282 goto out;
2283
2284 if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2285 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2286 btrfs_super_generation(sb),
2287 fs_info->last_trans_committed);
2288 ret = -EUCLEAN;
2289 goto out;
2290 }
2291out:
2292 btrfs_release_disk_super(sb);
2293 return ret;
2294}
2295
2296static int btrfs_unfreeze(struct super_block *sb)
2297{
2298 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2299 struct btrfs_device *device;
2300 int ret = 0;
2301
2302 /*
2303 * Make sure the fs is not changed by accident (like hibernation then
2304 * modified by other OS).
2305 * If we found anything wrong, we mark the fs error immediately.
2306 *
2307 * And since the fs is frozen, no one can modify the fs yet, thus
2308 * we don't need to hold device_list_mutex.
2309 */
2310 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2311 ret = check_dev_super(device);
2312 if (ret < 0) {
2313 btrfs_handle_fs_error(fs_info, ret,
2314 "super block on devid %llu got modified unexpectedly",
2315 device->devid);
2316 break;
2317 }
2318 }
2319 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2320
2321 /*
2322 * We still return 0, to allow VFS layer to unfreeze the fs even the
2323 * above checks failed. Since the fs is either fine or read-only, we're
2324 * safe to continue, without causing further damage.
2325 */
2326 return 0;
2327}
2328
2329static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2330{
2331 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2332
2333 /*
2334 * There should be always a valid pointer in latest_dev, it may be stale
2335 * for a short moment in case it's being deleted but still valid until
2336 * the end of RCU grace period.
2337 */
2338 rcu_read_lock();
2339 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2340 rcu_read_unlock();
2341
2342 return 0;
2343}
2344
2345static const struct super_operations btrfs_super_ops = {
2346 .drop_inode = btrfs_drop_inode,
2347 .evict_inode = btrfs_evict_inode,
2348 .put_super = btrfs_put_super,
2349 .sync_fs = btrfs_sync_fs,
2350 .show_options = btrfs_show_options,
2351 .show_devname = btrfs_show_devname,
2352 .alloc_inode = btrfs_alloc_inode,
2353 .destroy_inode = btrfs_destroy_inode,
2354 .free_inode = btrfs_free_inode,
2355 .statfs = btrfs_statfs,
2356 .remount_fs = btrfs_remount,
2357 .freeze_fs = btrfs_freeze,
2358 .unfreeze_fs = btrfs_unfreeze,
2359};
2360
2361static const struct file_operations btrfs_ctl_fops = {
2362 .open = btrfs_control_open,
2363 .unlocked_ioctl = btrfs_control_ioctl,
2364 .compat_ioctl = compat_ptr_ioctl,
2365 .owner = THIS_MODULE,
2366 .llseek = noop_llseek,
2367};
2368
2369static struct miscdevice btrfs_misc = {
2370 .minor = BTRFS_MINOR,
2371 .name = "btrfs-control",
2372 .fops = &btrfs_ctl_fops
2373};
2374
2375MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2376MODULE_ALIAS("devname:btrfs-control");
2377
2378static int __init btrfs_interface_init(void)
2379{
2380 return misc_register(&btrfs_misc);
2381}
2382
2383static __cold void btrfs_interface_exit(void)
2384{
2385 misc_deregister(&btrfs_misc);
2386}
2387
2388static int __init btrfs_print_mod_info(void)
2389{
2390 static const char options[] = ""
2391#ifdef CONFIG_BTRFS_DEBUG
2392 ", debug=on"
2393#endif
2394#ifdef CONFIG_BTRFS_ASSERT
2395 ", assert=on"
2396#endif
2397#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2398 ", integrity-checker=on"
2399#endif
2400#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2401 ", ref-verify=on"
2402#endif
2403#ifdef CONFIG_BLK_DEV_ZONED
2404 ", zoned=yes"
2405#else
2406 ", zoned=no"
2407#endif
2408#ifdef CONFIG_FS_VERITY
2409 ", fsverity=yes"
2410#else
2411 ", fsverity=no"
2412#endif
2413 ;
2414 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2415 return 0;
2416}
2417
2418static int register_btrfs(void)
2419{
2420 return register_filesystem(&btrfs_fs_type);
2421}
2422
2423static void unregister_btrfs(void)
2424{
2425 unregister_filesystem(&btrfs_fs_type);
2426}
2427
2428/* Helper structure for long init/exit functions. */
2429struct init_sequence {
2430 int (*init_func)(void);
2431 /* Can be NULL if the init_func doesn't need cleanup. */
2432 void (*exit_func)(void);
2433};
2434
2435static const struct init_sequence mod_init_seq[] = {
2436 {
2437 .init_func = btrfs_props_init,
2438 .exit_func = NULL,
2439 }, {
2440 .init_func = btrfs_init_sysfs,
2441 .exit_func = btrfs_exit_sysfs,
2442 }, {
2443 .init_func = btrfs_init_compress,
2444 .exit_func = btrfs_exit_compress,
2445 }, {
2446 .init_func = btrfs_init_cachep,
2447 .exit_func = btrfs_destroy_cachep,
2448 }, {
2449 .init_func = btrfs_transaction_init,
2450 .exit_func = btrfs_transaction_exit,
2451 }, {
2452 .init_func = btrfs_ctree_init,
2453 .exit_func = btrfs_ctree_exit,
2454 }, {
2455 .init_func = btrfs_free_space_init,
2456 .exit_func = btrfs_free_space_exit,
2457 }, {
2458 .init_func = extent_state_init_cachep,
2459 .exit_func = extent_state_free_cachep,
2460 }, {
2461 .init_func = extent_buffer_init_cachep,
2462 .exit_func = extent_buffer_free_cachep,
2463 }, {
2464 .init_func = btrfs_bioset_init,
2465 .exit_func = btrfs_bioset_exit,
2466 }, {
2467 .init_func = extent_map_init,
2468 .exit_func = extent_map_exit,
2469 }, {
2470 .init_func = ordered_data_init,
2471 .exit_func = ordered_data_exit,
2472 }, {
2473 .init_func = btrfs_delayed_inode_init,
2474 .exit_func = btrfs_delayed_inode_exit,
2475 }, {
2476 .init_func = btrfs_auto_defrag_init,
2477 .exit_func = btrfs_auto_defrag_exit,
2478 }, {
2479 .init_func = btrfs_delayed_ref_init,
2480 .exit_func = btrfs_delayed_ref_exit,
2481 }, {
2482 .init_func = btrfs_prelim_ref_init,
2483 .exit_func = btrfs_prelim_ref_exit,
2484 }, {
2485 .init_func = btrfs_interface_init,
2486 .exit_func = btrfs_interface_exit,
2487 }, {
2488 .init_func = btrfs_print_mod_info,
2489 .exit_func = NULL,
2490 }, {
2491 .init_func = btrfs_run_sanity_tests,
2492 .exit_func = NULL,
2493 }, {
2494 .init_func = register_btrfs,
2495 .exit_func = unregister_btrfs,
2496 }
2497};
2498
2499static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2500
2501static __always_inline void btrfs_exit_btrfs_fs(void)
2502{
2503 int i;
2504
2505 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2506 if (!mod_init_result[i])
2507 continue;
2508 if (mod_init_seq[i].exit_func)
2509 mod_init_seq[i].exit_func();
2510 mod_init_result[i] = false;
2511 }
2512}
2513
2514static void __exit exit_btrfs_fs(void)
2515{
2516 btrfs_exit_btrfs_fs();
2517 btrfs_cleanup_fs_uuids();
2518}
2519
2520static int __init init_btrfs_fs(void)
2521{
2522 int ret;
2523 int i;
2524
2525 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2526 ASSERT(!mod_init_result[i]);
2527 ret = mod_init_seq[i].init_func();
2528 if (ret < 0) {
2529 btrfs_exit_btrfs_fs();
2530 return ret;
2531 }
2532 mod_init_result[i] = true;
2533 }
2534 return 0;
2535}
2536
2537late_initcall(init_btrfs_fs);
2538module_exit(exit_btrfs_fs)
2539
2540MODULE_LICENSE("GPL");
2541MODULE_SOFTDEP("pre: crc32c");
2542MODULE_SOFTDEP("pre: xxhash64");
2543MODULE_SOFTDEP("pre: sha256");
2544MODULE_SOFTDEP("pre: blake2b-256");