Loading...
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/cleancache.h>
27#include <linux/ratelimit.h>
28#include <linux/crc32c.h>
29#include <linux/btrfs.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 "volumes.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 "tests/btrfs-tests.h"
48#include "block-group.h"
49
50#include "qgroup.h"
51#define CREATE_TRACE_POINTS
52#include <trace/events/btrfs.h>
53
54static const struct super_operations btrfs_super_ops;
55
56/*
57 * Types for mounting the default subvolume and a subvolume explicitly
58 * requested by subvol=/path. That way the callchain is straightforward and we
59 * don't have to play tricks with the mount options and recursive calls to
60 * btrfs_mount.
61 *
62 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
63 */
64static struct file_system_type btrfs_fs_type;
65static struct file_system_type btrfs_root_fs_type;
66
67static int btrfs_remount(struct super_block *sb, int *flags, char *data);
68
69const char *btrfs_decode_error(int errno)
70{
71 char *errstr = "unknown";
72
73 switch (errno) {
74 case -EIO:
75 errstr = "IO failure";
76 break;
77 case -ENOMEM:
78 errstr = "Out of memory";
79 break;
80 case -EROFS:
81 errstr = "Readonly filesystem";
82 break;
83 case -EEXIST:
84 errstr = "Object already exists";
85 break;
86 case -ENOSPC:
87 errstr = "No space left";
88 break;
89 case -ENOENT:
90 errstr = "No such entry";
91 break;
92 }
93
94 return errstr;
95}
96
97/*
98 * __btrfs_handle_fs_error decodes expected errors from the caller and
99 * invokes the appropriate error response.
100 */
101__cold
102void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
103 unsigned int line, int errno, const char *fmt, ...)
104{
105 struct super_block *sb = fs_info->sb;
106#ifdef CONFIG_PRINTK
107 const char *errstr;
108#endif
109
110 /*
111 * Special case: if the error is EROFS, and we're already
112 * under SB_RDONLY, then it is safe here.
113 */
114 if (errno == -EROFS && sb_rdonly(sb))
115 return;
116
117#ifdef CONFIG_PRINTK
118 errstr = btrfs_decode_error(errno);
119 if (fmt) {
120 struct va_format vaf;
121 va_list args;
122
123 va_start(args, fmt);
124 vaf.fmt = fmt;
125 vaf.va = &args;
126
127 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
128 sb->s_id, function, line, errno, errstr, &vaf);
129 va_end(args);
130 } else {
131 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
132 sb->s_id, function, line, errno, errstr);
133 }
134#endif
135
136 /*
137 * Today we only save the error info to memory. Long term we'll
138 * also send it down to the disk
139 */
140 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
141
142 /* Don't go through full error handling during mount */
143 if (!(sb->s_flags & SB_BORN))
144 return;
145
146 if (sb_rdonly(sb))
147 return;
148
149 /* btrfs handle error by forcing the filesystem readonly */
150 sb->s_flags |= SB_RDONLY;
151 btrfs_info(fs_info, "forced readonly");
152 /*
153 * Note that a running device replace operation is not canceled here
154 * although there is no way to update the progress. It would add the
155 * risk of a deadlock, therefore the canceling is omitted. The only
156 * penalty is that some I/O remains active until the procedure
157 * completes. The next time when the filesystem is mounted writable
158 * again, the device replace operation continues.
159 */
160}
161
162#ifdef CONFIG_PRINTK
163static const char * const logtypes[] = {
164 "emergency",
165 "alert",
166 "critical",
167 "error",
168 "warning",
169 "notice",
170 "info",
171 "debug",
172};
173
174
175/*
176 * Use one ratelimit state per log level so that a flood of less important
177 * messages doesn't cause more important ones to be dropped.
178 */
179static struct ratelimit_state printk_limits[] = {
180 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
181 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
182 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
183 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
184 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
185 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
186 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
187 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
188};
189
190void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
191{
192 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
193 struct va_format vaf;
194 va_list args;
195 int kern_level;
196 const char *type = logtypes[4];
197 struct ratelimit_state *ratelimit = &printk_limits[4];
198
199 va_start(args, fmt);
200
201 while ((kern_level = printk_get_level(fmt)) != 0) {
202 size_t size = printk_skip_level(fmt) - fmt;
203
204 if (kern_level >= '0' && kern_level <= '7') {
205 memcpy(lvl, fmt, size);
206 lvl[size] = '\0';
207 type = logtypes[kern_level - '0'];
208 ratelimit = &printk_limits[kern_level - '0'];
209 }
210 fmt += size;
211 }
212
213 vaf.fmt = fmt;
214 vaf.va = &args;
215
216 if (__ratelimit(ratelimit))
217 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
218 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
219
220 va_end(args);
221}
222#endif
223
224/*
225 * We only mark the transaction aborted and then set the file system read-only.
226 * This will prevent new transactions from starting or trying to join this
227 * one.
228 *
229 * This means that error recovery at the call site is limited to freeing
230 * any local memory allocations and passing the error code up without
231 * further cleanup. The transaction should complete as it normally would
232 * in the call path but will return -EIO.
233 *
234 * We'll complete the cleanup in btrfs_end_transaction and
235 * btrfs_commit_transaction.
236 */
237__cold
238void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
239 const char *function,
240 unsigned int line, int errno)
241{
242 struct btrfs_fs_info *fs_info = trans->fs_info;
243
244 trans->aborted = errno;
245 /* Nothing used. The other threads that have joined this
246 * transaction may be able to continue. */
247 if (!trans->dirty && list_empty(&trans->new_bgs)) {
248 const char *errstr;
249
250 errstr = btrfs_decode_error(errno);
251 btrfs_warn(fs_info,
252 "%s:%d: Aborting unused transaction(%s).",
253 function, line, errstr);
254 return;
255 }
256 WRITE_ONCE(trans->transaction->aborted, errno);
257 /* Wake up anybody who may be waiting on this transaction */
258 wake_up(&fs_info->transaction_wait);
259 wake_up(&fs_info->transaction_blocked_wait);
260 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
261}
262/*
263 * __btrfs_panic decodes unexpected, fatal errors from the caller,
264 * issues an alert, and either panics or BUGs, depending on mount options.
265 */
266__cold
267void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
268 unsigned int line, int errno, const char *fmt, ...)
269{
270 char *s_id = "<unknown>";
271 const char *errstr;
272 struct va_format vaf = { .fmt = fmt };
273 va_list args;
274
275 if (fs_info)
276 s_id = fs_info->sb->s_id;
277
278 va_start(args, fmt);
279 vaf.va = &args;
280
281 errstr = btrfs_decode_error(errno);
282 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
283 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
284 s_id, function, line, &vaf, errno, errstr);
285
286 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
287 function, line, &vaf, errno, errstr);
288 va_end(args);
289 /* Caller calls BUG() */
290}
291
292static void btrfs_put_super(struct super_block *sb)
293{
294 close_ctree(btrfs_sb(sb));
295}
296
297enum {
298 Opt_acl, Opt_noacl,
299 Opt_clear_cache,
300 Opt_commit_interval,
301 Opt_compress,
302 Opt_compress_force,
303 Opt_compress_force_type,
304 Opt_compress_type,
305 Opt_degraded,
306 Opt_device,
307 Opt_fatal_errors,
308 Opt_flushoncommit, Opt_noflushoncommit,
309 Opt_inode_cache, Opt_noinode_cache,
310 Opt_max_inline,
311 Opt_barrier, Opt_nobarrier,
312 Opt_datacow, Opt_nodatacow,
313 Opt_datasum, Opt_nodatasum,
314 Opt_defrag, Opt_nodefrag,
315 Opt_discard, Opt_nodiscard,
316 Opt_nologreplay,
317 Opt_norecovery,
318 Opt_ratio,
319 Opt_rescan_uuid_tree,
320 Opt_skip_balance,
321 Opt_space_cache, Opt_no_space_cache,
322 Opt_space_cache_version,
323 Opt_ssd, Opt_nossd,
324 Opt_ssd_spread, Opt_nossd_spread,
325 Opt_subvol,
326 Opt_subvol_empty,
327 Opt_subvolid,
328 Opt_thread_pool,
329 Opt_treelog, Opt_notreelog,
330 Opt_usebackuproot,
331 Opt_user_subvol_rm_allowed,
332
333 /* Deprecated options */
334 Opt_alloc_start,
335 Opt_recovery,
336 Opt_subvolrootid,
337
338 /* Debugging options */
339 Opt_check_integrity,
340 Opt_check_integrity_including_extent_data,
341 Opt_check_integrity_print_mask,
342 Opt_enospc_debug, Opt_noenospc_debug,
343#ifdef CONFIG_BTRFS_DEBUG
344 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
345#endif
346#ifdef CONFIG_BTRFS_FS_REF_VERIFY
347 Opt_ref_verify,
348#endif
349 Opt_err,
350};
351
352static const match_table_t tokens = {
353 {Opt_acl, "acl"},
354 {Opt_noacl, "noacl"},
355 {Opt_clear_cache, "clear_cache"},
356 {Opt_commit_interval, "commit=%u"},
357 {Opt_compress, "compress"},
358 {Opt_compress_type, "compress=%s"},
359 {Opt_compress_force, "compress-force"},
360 {Opt_compress_force_type, "compress-force=%s"},
361 {Opt_degraded, "degraded"},
362 {Opt_device, "device=%s"},
363 {Opt_fatal_errors, "fatal_errors=%s"},
364 {Opt_flushoncommit, "flushoncommit"},
365 {Opt_noflushoncommit, "noflushoncommit"},
366 {Opt_inode_cache, "inode_cache"},
367 {Opt_noinode_cache, "noinode_cache"},
368 {Opt_max_inline, "max_inline=%s"},
369 {Opt_barrier, "barrier"},
370 {Opt_nobarrier, "nobarrier"},
371 {Opt_datacow, "datacow"},
372 {Opt_nodatacow, "nodatacow"},
373 {Opt_datasum, "datasum"},
374 {Opt_nodatasum, "nodatasum"},
375 {Opt_defrag, "autodefrag"},
376 {Opt_nodefrag, "noautodefrag"},
377 {Opt_discard, "discard"},
378 {Opt_nodiscard, "nodiscard"},
379 {Opt_nologreplay, "nologreplay"},
380 {Opt_norecovery, "norecovery"},
381 {Opt_ratio, "metadata_ratio=%u"},
382 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
383 {Opt_skip_balance, "skip_balance"},
384 {Opt_space_cache, "space_cache"},
385 {Opt_no_space_cache, "nospace_cache"},
386 {Opt_space_cache_version, "space_cache=%s"},
387 {Opt_ssd, "ssd"},
388 {Opt_nossd, "nossd"},
389 {Opt_ssd_spread, "ssd_spread"},
390 {Opt_nossd_spread, "nossd_spread"},
391 {Opt_subvol, "subvol=%s"},
392 {Opt_subvol_empty, "subvol="},
393 {Opt_subvolid, "subvolid=%s"},
394 {Opt_thread_pool, "thread_pool=%u"},
395 {Opt_treelog, "treelog"},
396 {Opt_notreelog, "notreelog"},
397 {Opt_usebackuproot, "usebackuproot"},
398 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
399
400 /* Deprecated options */
401 {Opt_alloc_start, "alloc_start=%s"},
402 {Opt_recovery, "recovery"},
403 {Opt_subvolrootid, "subvolrootid=%d"},
404
405 /* Debugging options */
406 {Opt_check_integrity, "check_int"},
407 {Opt_check_integrity_including_extent_data, "check_int_data"},
408 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
409 {Opt_enospc_debug, "enospc_debug"},
410 {Opt_noenospc_debug, "noenospc_debug"},
411#ifdef CONFIG_BTRFS_DEBUG
412 {Opt_fragment_data, "fragment=data"},
413 {Opt_fragment_metadata, "fragment=metadata"},
414 {Opt_fragment_all, "fragment=all"},
415#endif
416#ifdef CONFIG_BTRFS_FS_REF_VERIFY
417 {Opt_ref_verify, "ref_verify"},
418#endif
419 {Opt_err, NULL},
420};
421
422/*
423 * Regular mount options parser. Everything that is needed only when
424 * reading in a new superblock is parsed here.
425 * XXX JDM: This needs to be cleaned up for remount.
426 */
427int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
428 unsigned long new_flags)
429{
430 substring_t args[MAX_OPT_ARGS];
431 char *p, *num;
432 u64 cache_gen;
433 int intarg;
434 int ret = 0;
435 char *compress_type;
436 bool compress_force = false;
437 enum btrfs_compression_type saved_compress_type;
438 bool saved_compress_force;
439 int no_compress = 0;
440
441 cache_gen = btrfs_super_cache_generation(info->super_copy);
442 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
443 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
444 else if (cache_gen)
445 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
446
447 /*
448 * Even the options are empty, we still need to do extra check
449 * against new flags
450 */
451 if (!options)
452 goto check;
453
454 while ((p = strsep(&options, ",")) != NULL) {
455 int token;
456 if (!*p)
457 continue;
458
459 token = match_token(p, tokens, args);
460 switch (token) {
461 case Opt_degraded:
462 btrfs_info(info, "allowing degraded mounts");
463 btrfs_set_opt(info->mount_opt, DEGRADED);
464 break;
465 case Opt_subvol:
466 case Opt_subvol_empty:
467 case Opt_subvolid:
468 case Opt_subvolrootid:
469 case Opt_device:
470 /*
471 * These are parsed by btrfs_parse_subvol_options or
472 * btrfs_parse_device_options and can be ignored here.
473 */
474 break;
475 case Opt_nodatasum:
476 btrfs_set_and_info(info, NODATASUM,
477 "setting nodatasum");
478 break;
479 case Opt_datasum:
480 if (btrfs_test_opt(info, NODATASUM)) {
481 if (btrfs_test_opt(info, NODATACOW))
482 btrfs_info(info,
483 "setting datasum, datacow enabled");
484 else
485 btrfs_info(info, "setting datasum");
486 }
487 btrfs_clear_opt(info->mount_opt, NODATACOW);
488 btrfs_clear_opt(info->mount_opt, NODATASUM);
489 break;
490 case Opt_nodatacow:
491 if (!btrfs_test_opt(info, NODATACOW)) {
492 if (!btrfs_test_opt(info, COMPRESS) ||
493 !btrfs_test_opt(info, FORCE_COMPRESS)) {
494 btrfs_info(info,
495 "setting nodatacow, compression disabled");
496 } else {
497 btrfs_info(info, "setting nodatacow");
498 }
499 }
500 btrfs_clear_opt(info->mount_opt, COMPRESS);
501 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
502 btrfs_set_opt(info->mount_opt, NODATACOW);
503 btrfs_set_opt(info->mount_opt, NODATASUM);
504 break;
505 case Opt_datacow:
506 btrfs_clear_and_info(info, NODATACOW,
507 "setting datacow");
508 break;
509 case Opt_compress_force:
510 case Opt_compress_force_type:
511 compress_force = true;
512 /* Fallthrough */
513 case Opt_compress:
514 case Opt_compress_type:
515 saved_compress_type = btrfs_test_opt(info,
516 COMPRESS) ?
517 info->compress_type : BTRFS_COMPRESS_NONE;
518 saved_compress_force =
519 btrfs_test_opt(info, FORCE_COMPRESS);
520 if (token == Opt_compress ||
521 token == Opt_compress_force ||
522 strncmp(args[0].from, "zlib", 4) == 0) {
523 compress_type = "zlib";
524
525 info->compress_type = BTRFS_COMPRESS_ZLIB;
526 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
527 /*
528 * args[0] contains uninitialized data since
529 * for these tokens we don't expect any
530 * parameter.
531 */
532 if (token != Opt_compress &&
533 token != Opt_compress_force)
534 info->compress_level =
535 btrfs_compress_str2level(
536 BTRFS_COMPRESS_ZLIB,
537 args[0].from + 4);
538 btrfs_set_opt(info->mount_opt, COMPRESS);
539 btrfs_clear_opt(info->mount_opt, NODATACOW);
540 btrfs_clear_opt(info->mount_opt, NODATASUM);
541 no_compress = 0;
542 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
543 compress_type = "lzo";
544 info->compress_type = BTRFS_COMPRESS_LZO;
545 btrfs_set_opt(info->mount_opt, COMPRESS);
546 btrfs_clear_opt(info->mount_opt, NODATACOW);
547 btrfs_clear_opt(info->mount_opt, NODATASUM);
548 btrfs_set_fs_incompat(info, COMPRESS_LZO);
549 no_compress = 0;
550 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
551 compress_type = "zstd";
552 info->compress_type = BTRFS_COMPRESS_ZSTD;
553 info->compress_level =
554 btrfs_compress_str2level(
555 BTRFS_COMPRESS_ZSTD,
556 args[0].from + 4);
557 btrfs_set_opt(info->mount_opt, COMPRESS);
558 btrfs_clear_opt(info->mount_opt, NODATACOW);
559 btrfs_clear_opt(info->mount_opt, NODATASUM);
560 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
561 no_compress = 0;
562 } else if (strncmp(args[0].from, "no", 2) == 0) {
563 compress_type = "no";
564 btrfs_clear_opt(info->mount_opt, COMPRESS);
565 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
566 compress_force = false;
567 no_compress++;
568 } else {
569 ret = -EINVAL;
570 goto out;
571 }
572
573 if (compress_force) {
574 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
575 } else {
576 /*
577 * If we remount from compress-force=xxx to
578 * compress=xxx, we need clear FORCE_COMPRESS
579 * flag, otherwise, there is no way for users
580 * to disable forcible compression separately.
581 */
582 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
583 }
584 if ((btrfs_test_opt(info, COMPRESS) &&
585 (info->compress_type != saved_compress_type ||
586 compress_force != saved_compress_force)) ||
587 (!btrfs_test_opt(info, COMPRESS) &&
588 no_compress == 1)) {
589 btrfs_info(info, "%s %s compression, level %d",
590 (compress_force) ? "force" : "use",
591 compress_type, info->compress_level);
592 }
593 compress_force = false;
594 break;
595 case Opt_ssd:
596 btrfs_set_and_info(info, SSD,
597 "enabling ssd optimizations");
598 btrfs_clear_opt(info->mount_opt, NOSSD);
599 break;
600 case Opt_ssd_spread:
601 btrfs_set_and_info(info, SSD,
602 "enabling ssd optimizations");
603 btrfs_set_and_info(info, SSD_SPREAD,
604 "using spread ssd allocation scheme");
605 btrfs_clear_opt(info->mount_opt, NOSSD);
606 break;
607 case Opt_nossd:
608 btrfs_set_opt(info->mount_opt, NOSSD);
609 btrfs_clear_and_info(info, SSD,
610 "not using ssd optimizations");
611 /* Fallthrough */
612 case Opt_nossd_spread:
613 btrfs_clear_and_info(info, SSD_SPREAD,
614 "not using spread ssd allocation scheme");
615 break;
616 case Opt_barrier:
617 btrfs_clear_and_info(info, NOBARRIER,
618 "turning on barriers");
619 break;
620 case Opt_nobarrier:
621 btrfs_set_and_info(info, NOBARRIER,
622 "turning off barriers");
623 break;
624 case Opt_thread_pool:
625 ret = match_int(&args[0], &intarg);
626 if (ret) {
627 goto out;
628 } else if (intarg == 0) {
629 ret = -EINVAL;
630 goto out;
631 }
632 info->thread_pool_size = intarg;
633 break;
634 case Opt_max_inline:
635 num = match_strdup(&args[0]);
636 if (num) {
637 info->max_inline = memparse(num, NULL);
638 kfree(num);
639
640 if (info->max_inline) {
641 info->max_inline = min_t(u64,
642 info->max_inline,
643 info->sectorsize);
644 }
645 btrfs_info(info, "max_inline at %llu",
646 info->max_inline);
647 } else {
648 ret = -ENOMEM;
649 goto out;
650 }
651 break;
652 case Opt_alloc_start:
653 btrfs_info(info,
654 "option alloc_start is obsolete, ignored");
655 break;
656 case Opt_acl:
657#ifdef CONFIG_BTRFS_FS_POSIX_ACL
658 info->sb->s_flags |= SB_POSIXACL;
659 break;
660#else
661 btrfs_err(info, "support for ACL not compiled in!");
662 ret = -EINVAL;
663 goto out;
664#endif
665 case Opt_noacl:
666 info->sb->s_flags &= ~SB_POSIXACL;
667 break;
668 case Opt_notreelog:
669 btrfs_set_and_info(info, NOTREELOG,
670 "disabling tree log");
671 break;
672 case Opt_treelog:
673 btrfs_clear_and_info(info, NOTREELOG,
674 "enabling tree log");
675 break;
676 case Opt_norecovery:
677 case Opt_nologreplay:
678 btrfs_set_and_info(info, NOLOGREPLAY,
679 "disabling log replay at mount time");
680 break;
681 case Opt_flushoncommit:
682 btrfs_set_and_info(info, FLUSHONCOMMIT,
683 "turning on flush-on-commit");
684 break;
685 case Opt_noflushoncommit:
686 btrfs_clear_and_info(info, FLUSHONCOMMIT,
687 "turning off flush-on-commit");
688 break;
689 case Opt_ratio:
690 ret = match_int(&args[0], &intarg);
691 if (ret)
692 goto out;
693 info->metadata_ratio = intarg;
694 btrfs_info(info, "metadata ratio %u",
695 info->metadata_ratio);
696 break;
697 case Opt_discard:
698 btrfs_set_and_info(info, DISCARD,
699 "turning on discard");
700 break;
701 case Opt_nodiscard:
702 btrfs_clear_and_info(info, DISCARD,
703 "turning off discard");
704 break;
705 case Opt_space_cache:
706 case Opt_space_cache_version:
707 if (token == Opt_space_cache ||
708 strcmp(args[0].from, "v1") == 0) {
709 btrfs_clear_opt(info->mount_opt,
710 FREE_SPACE_TREE);
711 btrfs_set_and_info(info, SPACE_CACHE,
712 "enabling disk space caching");
713 } else if (strcmp(args[0].from, "v2") == 0) {
714 btrfs_clear_opt(info->mount_opt,
715 SPACE_CACHE);
716 btrfs_set_and_info(info, FREE_SPACE_TREE,
717 "enabling free space tree");
718 } else {
719 ret = -EINVAL;
720 goto out;
721 }
722 break;
723 case Opt_rescan_uuid_tree:
724 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
725 break;
726 case Opt_no_space_cache:
727 if (btrfs_test_opt(info, SPACE_CACHE)) {
728 btrfs_clear_and_info(info, SPACE_CACHE,
729 "disabling disk space caching");
730 }
731 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
732 btrfs_clear_and_info(info, FREE_SPACE_TREE,
733 "disabling free space tree");
734 }
735 break;
736 case Opt_inode_cache:
737 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
738 "enabling inode map caching");
739 break;
740 case Opt_noinode_cache:
741 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
742 "disabling inode map caching");
743 break;
744 case Opt_clear_cache:
745 btrfs_set_and_info(info, CLEAR_CACHE,
746 "force clearing of disk cache");
747 break;
748 case Opt_user_subvol_rm_allowed:
749 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
750 break;
751 case Opt_enospc_debug:
752 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
753 break;
754 case Opt_noenospc_debug:
755 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
756 break;
757 case Opt_defrag:
758 btrfs_set_and_info(info, AUTO_DEFRAG,
759 "enabling auto defrag");
760 break;
761 case Opt_nodefrag:
762 btrfs_clear_and_info(info, AUTO_DEFRAG,
763 "disabling auto defrag");
764 break;
765 case Opt_recovery:
766 btrfs_warn(info,
767 "'recovery' is deprecated, use 'usebackuproot' instead");
768 /* fall through */
769 case Opt_usebackuproot:
770 btrfs_info(info,
771 "trying to use backup root at mount time");
772 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
773 break;
774 case Opt_skip_balance:
775 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
776 break;
777#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
778 case Opt_check_integrity_including_extent_data:
779 btrfs_info(info,
780 "enabling check integrity including extent data");
781 btrfs_set_opt(info->mount_opt,
782 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
783 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
784 break;
785 case Opt_check_integrity:
786 btrfs_info(info, "enabling check integrity");
787 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
788 break;
789 case Opt_check_integrity_print_mask:
790 ret = match_int(&args[0], &intarg);
791 if (ret)
792 goto out;
793 info->check_integrity_print_mask = intarg;
794 btrfs_info(info, "check_integrity_print_mask 0x%x",
795 info->check_integrity_print_mask);
796 break;
797#else
798 case Opt_check_integrity_including_extent_data:
799 case Opt_check_integrity:
800 case Opt_check_integrity_print_mask:
801 btrfs_err(info,
802 "support for check_integrity* not compiled in!");
803 ret = -EINVAL;
804 goto out;
805#endif
806 case Opt_fatal_errors:
807 if (strcmp(args[0].from, "panic") == 0)
808 btrfs_set_opt(info->mount_opt,
809 PANIC_ON_FATAL_ERROR);
810 else if (strcmp(args[0].from, "bug") == 0)
811 btrfs_clear_opt(info->mount_opt,
812 PANIC_ON_FATAL_ERROR);
813 else {
814 ret = -EINVAL;
815 goto out;
816 }
817 break;
818 case Opt_commit_interval:
819 intarg = 0;
820 ret = match_int(&args[0], &intarg);
821 if (ret)
822 goto out;
823 if (intarg == 0) {
824 btrfs_info(info,
825 "using default commit interval %us",
826 BTRFS_DEFAULT_COMMIT_INTERVAL);
827 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
828 } else if (intarg > 300) {
829 btrfs_warn(info, "excessive commit interval %d",
830 intarg);
831 }
832 info->commit_interval = intarg;
833 break;
834#ifdef CONFIG_BTRFS_DEBUG
835 case Opt_fragment_all:
836 btrfs_info(info, "fragmenting all space");
837 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
838 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
839 break;
840 case Opt_fragment_metadata:
841 btrfs_info(info, "fragmenting metadata");
842 btrfs_set_opt(info->mount_opt,
843 FRAGMENT_METADATA);
844 break;
845 case Opt_fragment_data:
846 btrfs_info(info, "fragmenting data");
847 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
848 break;
849#endif
850#ifdef CONFIG_BTRFS_FS_REF_VERIFY
851 case Opt_ref_verify:
852 btrfs_info(info, "doing ref verification");
853 btrfs_set_opt(info->mount_opt, REF_VERIFY);
854 break;
855#endif
856 case Opt_err:
857 btrfs_info(info, "unrecognized mount option '%s'", p);
858 ret = -EINVAL;
859 goto out;
860 default:
861 break;
862 }
863 }
864check:
865 /*
866 * Extra check for current option against current flag
867 */
868 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
869 btrfs_err(info,
870 "nologreplay must be used with ro mount option");
871 ret = -EINVAL;
872 }
873out:
874 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
875 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
876 !btrfs_test_opt(info, CLEAR_CACHE)) {
877 btrfs_err(info, "cannot disable free space tree");
878 ret = -EINVAL;
879
880 }
881 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
882 btrfs_info(info, "disk space caching is enabled");
883 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
884 btrfs_info(info, "using free space tree");
885 return ret;
886}
887
888/*
889 * Parse mount options that are required early in the mount process.
890 *
891 * All other options will be parsed on much later in the mount process and
892 * only when we need to allocate a new super block.
893 */
894static int btrfs_parse_device_options(const char *options, fmode_t flags,
895 void *holder)
896{
897 substring_t args[MAX_OPT_ARGS];
898 char *device_name, *opts, *orig, *p;
899 struct btrfs_device *device = NULL;
900 int error = 0;
901
902 lockdep_assert_held(&uuid_mutex);
903
904 if (!options)
905 return 0;
906
907 /*
908 * strsep changes the string, duplicate it because btrfs_parse_options
909 * gets called later
910 */
911 opts = kstrdup(options, GFP_KERNEL);
912 if (!opts)
913 return -ENOMEM;
914 orig = opts;
915
916 while ((p = strsep(&opts, ",")) != NULL) {
917 int token;
918
919 if (!*p)
920 continue;
921
922 token = match_token(p, tokens, args);
923 if (token == Opt_device) {
924 device_name = match_strdup(&args[0]);
925 if (!device_name) {
926 error = -ENOMEM;
927 goto out;
928 }
929 device = btrfs_scan_one_device(device_name, flags,
930 holder);
931 kfree(device_name);
932 if (IS_ERR(device)) {
933 error = PTR_ERR(device);
934 goto out;
935 }
936 }
937 }
938
939out:
940 kfree(orig);
941 return error;
942}
943
944/*
945 * Parse mount options that are related to subvolume id
946 *
947 * The value is later passed to mount_subvol()
948 */
949static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
950 u64 *subvol_objectid)
951{
952 substring_t args[MAX_OPT_ARGS];
953 char *opts, *orig, *p;
954 int error = 0;
955 u64 subvolid;
956
957 if (!options)
958 return 0;
959
960 /*
961 * strsep changes the string, duplicate it because
962 * btrfs_parse_device_options gets called later
963 */
964 opts = kstrdup(options, GFP_KERNEL);
965 if (!opts)
966 return -ENOMEM;
967 orig = opts;
968
969 while ((p = strsep(&opts, ",")) != NULL) {
970 int token;
971 if (!*p)
972 continue;
973
974 token = match_token(p, tokens, args);
975 switch (token) {
976 case Opt_subvol:
977 kfree(*subvol_name);
978 *subvol_name = match_strdup(&args[0]);
979 if (!*subvol_name) {
980 error = -ENOMEM;
981 goto out;
982 }
983 break;
984 case Opt_subvolid:
985 error = match_u64(&args[0], &subvolid);
986 if (error)
987 goto out;
988
989 /* we want the original fs_tree */
990 if (subvolid == 0)
991 subvolid = BTRFS_FS_TREE_OBJECTID;
992
993 *subvol_objectid = subvolid;
994 break;
995 case Opt_subvolrootid:
996 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
997 break;
998 default:
999 break;
1000 }
1001 }
1002
1003out:
1004 kfree(orig);
1005 return error;
1006}
1007
1008static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1009 u64 subvol_objectid)
1010{
1011 struct btrfs_root *root = fs_info->tree_root;
1012 struct btrfs_root *fs_root;
1013 struct btrfs_root_ref *root_ref;
1014 struct btrfs_inode_ref *inode_ref;
1015 struct btrfs_key key;
1016 struct btrfs_path *path = NULL;
1017 char *name = NULL, *ptr;
1018 u64 dirid;
1019 int len;
1020 int ret;
1021
1022 path = btrfs_alloc_path();
1023 if (!path) {
1024 ret = -ENOMEM;
1025 goto err;
1026 }
1027 path->leave_spinning = 1;
1028
1029 name = kmalloc(PATH_MAX, GFP_KERNEL);
1030 if (!name) {
1031 ret = -ENOMEM;
1032 goto err;
1033 }
1034 ptr = name + PATH_MAX - 1;
1035 ptr[0] = '\0';
1036
1037 /*
1038 * Walk up the subvolume trees in the tree of tree roots by root
1039 * backrefs until we hit the top-level subvolume.
1040 */
1041 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1042 key.objectid = subvol_objectid;
1043 key.type = BTRFS_ROOT_BACKREF_KEY;
1044 key.offset = (u64)-1;
1045
1046 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1047 if (ret < 0) {
1048 goto err;
1049 } else if (ret > 0) {
1050 ret = btrfs_previous_item(root, path, subvol_objectid,
1051 BTRFS_ROOT_BACKREF_KEY);
1052 if (ret < 0) {
1053 goto err;
1054 } else if (ret > 0) {
1055 ret = -ENOENT;
1056 goto err;
1057 }
1058 }
1059
1060 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1061 subvol_objectid = key.offset;
1062
1063 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1064 struct btrfs_root_ref);
1065 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1066 ptr -= len + 1;
1067 if (ptr < name) {
1068 ret = -ENAMETOOLONG;
1069 goto err;
1070 }
1071 read_extent_buffer(path->nodes[0], ptr + 1,
1072 (unsigned long)(root_ref + 1), len);
1073 ptr[0] = '/';
1074 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1075 btrfs_release_path(path);
1076
1077 key.objectid = subvol_objectid;
1078 key.type = BTRFS_ROOT_ITEM_KEY;
1079 key.offset = (u64)-1;
1080 fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1081 if (IS_ERR(fs_root)) {
1082 ret = PTR_ERR(fs_root);
1083 goto err;
1084 }
1085
1086 /*
1087 * Walk up the filesystem tree by inode refs until we hit the
1088 * root directory.
1089 */
1090 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1091 key.objectid = dirid;
1092 key.type = BTRFS_INODE_REF_KEY;
1093 key.offset = (u64)-1;
1094
1095 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1096 if (ret < 0) {
1097 goto err;
1098 } else if (ret > 0) {
1099 ret = btrfs_previous_item(fs_root, path, dirid,
1100 BTRFS_INODE_REF_KEY);
1101 if (ret < 0) {
1102 goto err;
1103 } else if (ret > 0) {
1104 ret = -ENOENT;
1105 goto err;
1106 }
1107 }
1108
1109 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1110 dirid = key.offset;
1111
1112 inode_ref = btrfs_item_ptr(path->nodes[0],
1113 path->slots[0],
1114 struct btrfs_inode_ref);
1115 len = btrfs_inode_ref_name_len(path->nodes[0],
1116 inode_ref);
1117 ptr -= len + 1;
1118 if (ptr < name) {
1119 ret = -ENAMETOOLONG;
1120 goto err;
1121 }
1122 read_extent_buffer(path->nodes[0], ptr + 1,
1123 (unsigned long)(inode_ref + 1), len);
1124 ptr[0] = '/';
1125 btrfs_release_path(path);
1126 }
1127 }
1128
1129 btrfs_free_path(path);
1130 if (ptr == name + PATH_MAX - 1) {
1131 name[0] = '/';
1132 name[1] = '\0';
1133 } else {
1134 memmove(name, ptr, name + PATH_MAX - ptr);
1135 }
1136 return name;
1137
1138err:
1139 btrfs_free_path(path);
1140 kfree(name);
1141 return ERR_PTR(ret);
1142}
1143
1144static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1145{
1146 struct btrfs_root *root = fs_info->tree_root;
1147 struct btrfs_dir_item *di;
1148 struct btrfs_path *path;
1149 struct btrfs_key location;
1150 u64 dir_id;
1151
1152 path = btrfs_alloc_path();
1153 if (!path)
1154 return -ENOMEM;
1155 path->leave_spinning = 1;
1156
1157 /*
1158 * Find the "default" dir item which points to the root item that we
1159 * will mount by default if we haven't been given a specific subvolume
1160 * to mount.
1161 */
1162 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1163 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1164 if (IS_ERR(di)) {
1165 btrfs_free_path(path);
1166 return PTR_ERR(di);
1167 }
1168 if (!di) {
1169 /*
1170 * Ok the default dir item isn't there. This is weird since
1171 * it's always been there, but don't freak out, just try and
1172 * mount the top-level subvolume.
1173 */
1174 btrfs_free_path(path);
1175 *objectid = BTRFS_FS_TREE_OBJECTID;
1176 return 0;
1177 }
1178
1179 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1180 btrfs_free_path(path);
1181 *objectid = location.objectid;
1182 return 0;
1183}
1184
1185static int btrfs_fill_super(struct super_block *sb,
1186 struct btrfs_fs_devices *fs_devices,
1187 void *data)
1188{
1189 struct inode *inode;
1190 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1191 struct btrfs_key key;
1192 int err;
1193
1194 sb->s_maxbytes = MAX_LFS_FILESIZE;
1195 sb->s_magic = BTRFS_SUPER_MAGIC;
1196 sb->s_op = &btrfs_super_ops;
1197 sb->s_d_op = &btrfs_dentry_operations;
1198 sb->s_export_op = &btrfs_export_ops;
1199 sb->s_xattr = btrfs_xattr_handlers;
1200 sb->s_time_gran = 1;
1201#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1202 sb->s_flags |= SB_POSIXACL;
1203#endif
1204 sb->s_flags |= SB_I_VERSION;
1205 sb->s_iflags |= SB_I_CGROUPWB;
1206
1207 err = super_setup_bdi(sb);
1208 if (err) {
1209 btrfs_err(fs_info, "super_setup_bdi failed");
1210 return err;
1211 }
1212
1213 err = open_ctree(sb, fs_devices, (char *)data);
1214 if (err) {
1215 btrfs_err(fs_info, "open_ctree failed");
1216 return err;
1217 }
1218
1219 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1220 key.type = BTRFS_INODE_ITEM_KEY;
1221 key.offset = 0;
1222 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1223 if (IS_ERR(inode)) {
1224 err = PTR_ERR(inode);
1225 goto fail_close;
1226 }
1227
1228 sb->s_root = d_make_root(inode);
1229 if (!sb->s_root) {
1230 err = -ENOMEM;
1231 goto fail_close;
1232 }
1233
1234 cleancache_init_fs(sb);
1235 sb->s_flags |= SB_ACTIVE;
1236 return 0;
1237
1238fail_close:
1239 close_ctree(fs_info);
1240 return err;
1241}
1242
1243int btrfs_sync_fs(struct super_block *sb, int wait)
1244{
1245 struct btrfs_trans_handle *trans;
1246 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1247 struct btrfs_root *root = fs_info->tree_root;
1248
1249 trace_btrfs_sync_fs(fs_info, wait);
1250
1251 if (!wait) {
1252 filemap_flush(fs_info->btree_inode->i_mapping);
1253 return 0;
1254 }
1255
1256 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1257
1258 trans = btrfs_attach_transaction_barrier(root);
1259 if (IS_ERR(trans)) {
1260 /* no transaction, don't bother */
1261 if (PTR_ERR(trans) == -ENOENT) {
1262 /*
1263 * Exit unless we have some pending changes
1264 * that need to go through commit
1265 */
1266 if (fs_info->pending_changes == 0)
1267 return 0;
1268 /*
1269 * A non-blocking test if the fs is frozen. We must not
1270 * start a new transaction here otherwise a deadlock
1271 * happens. The pending operations are delayed to the
1272 * next commit after thawing.
1273 */
1274 if (sb_start_write_trylock(sb))
1275 sb_end_write(sb);
1276 else
1277 return 0;
1278 trans = btrfs_start_transaction(root, 0);
1279 }
1280 if (IS_ERR(trans))
1281 return PTR_ERR(trans);
1282 }
1283 return btrfs_commit_transaction(trans);
1284}
1285
1286static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1287{
1288 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1289 const char *compress_type;
1290
1291 if (btrfs_test_opt(info, DEGRADED))
1292 seq_puts(seq, ",degraded");
1293 if (btrfs_test_opt(info, NODATASUM))
1294 seq_puts(seq, ",nodatasum");
1295 if (btrfs_test_opt(info, NODATACOW))
1296 seq_puts(seq, ",nodatacow");
1297 if (btrfs_test_opt(info, NOBARRIER))
1298 seq_puts(seq, ",nobarrier");
1299 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1300 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1301 if (info->thread_pool_size != min_t(unsigned long,
1302 num_online_cpus() + 2, 8))
1303 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1304 if (btrfs_test_opt(info, COMPRESS)) {
1305 compress_type = btrfs_compress_type2str(info->compress_type);
1306 if (btrfs_test_opt(info, FORCE_COMPRESS))
1307 seq_printf(seq, ",compress-force=%s", compress_type);
1308 else
1309 seq_printf(seq, ",compress=%s", compress_type);
1310 if (info->compress_level)
1311 seq_printf(seq, ":%d", info->compress_level);
1312 }
1313 if (btrfs_test_opt(info, NOSSD))
1314 seq_puts(seq, ",nossd");
1315 if (btrfs_test_opt(info, SSD_SPREAD))
1316 seq_puts(seq, ",ssd_spread");
1317 else if (btrfs_test_opt(info, SSD))
1318 seq_puts(seq, ",ssd");
1319 if (btrfs_test_opt(info, NOTREELOG))
1320 seq_puts(seq, ",notreelog");
1321 if (btrfs_test_opt(info, NOLOGREPLAY))
1322 seq_puts(seq, ",nologreplay");
1323 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1324 seq_puts(seq, ",flushoncommit");
1325 if (btrfs_test_opt(info, DISCARD))
1326 seq_puts(seq, ",discard");
1327 if (!(info->sb->s_flags & SB_POSIXACL))
1328 seq_puts(seq, ",noacl");
1329 if (btrfs_test_opt(info, SPACE_CACHE))
1330 seq_puts(seq, ",space_cache");
1331 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1332 seq_puts(seq, ",space_cache=v2");
1333 else
1334 seq_puts(seq, ",nospace_cache");
1335 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1336 seq_puts(seq, ",rescan_uuid_tree");
1337 if (btrfs_test_opt(info, CLEAR_CACHE))
1338 seq_puts(seq, ",clear_cache");
1339 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1340 seq_puts(seq, ",user_subvol_rm_allowed");
1341 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1342 seq_puts(seq, ",enospc_debug");
1343 if (btrfs_test_opt(info, AUTO_DEFRAG))
1344 seq_puts(seq, ",autodefrag");
1345 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1346 seq_puts(seq, ",inode_cache");
1347 if (btrfs_test_opt(info, SKIP_BALANCE))
1348 seq_puts(seq, ",skip_balance");
1349#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1350 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1351 seq_puts(seq, ",check_int_data");
1352 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1353 seq_puts(seq, ",check_int");
1354 if (info->check_integrity_print_mask)
1355 seq_printf(seq, ",check_int_print_mask=%d",
1356 info->check_integrity_print_mask);
1357#endif
1358 if (info->metadata_ratio)
1359 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1360 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1361 seq_puts(seq, ",fatal_errors=panic");
1362 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1363 seq_printf(seq, ",commit=%u", info->commit_interval);
1364#ifdef CONFIG_BTRFS_DEBUG
1365 if (btrfs_test_opt(info, FRAGMENT_DATA))
1366 seq_puts(seq, ",fragment=data");
1367 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1368 seq_puts(seq, ",fragment=metadata");
1369#endif
1370 if (btrfs_test_opt(info, REF_VERIFY))
1371 seq_puts(seq, ",ref_verify");
1372 seq_printf(seq, ",subvolid=%llu",
1373 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1374 seq_puts(seq, ",subvol=");
1375 seq_dentry(seq, dentry, " \t\n\\");
1376 return 0;
1377}
1378
1379static int btrfs_test_super(struct super_block *s, void *data)
1380{
1381 struct btrfs_fs_info *p = data;
1382 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1383
1384 return fs_info->fs_devices == p->fs_devices;
1385}
1386
1387static int btrfs_set_super(struct super_block *s, void *data)
1388{
1389 int err = set_anon_super(s, data);
1390 if (!err)
1391 s->s_fs_info = data;
1392 return err;
1393}
1394
1395/*
1396 * subvolumes are identified by ino 256
1397 */
1398static inline int is_subvolume_inode(struct inode *inode)
1399{
1400 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1401 return 1;
1402 return 0;
1403}
1404
1405static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1406 struct vfsmount *mnt)
1407{
1408 struct dentry *root;
1409 int ret;
1410
1411 if (!subvol_name) {
1412 if (!subvol_objectid) {
1413 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1414 &subvol_objectid);
1415 if (ret) {
1416 root = ERR_PTR(ret);
1417 goto out;
1418 }
1419 }
1420 subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1421 subvol_objectid);
1422 if (IS_ERR(subvol_name)) {
1423 root = ERR_CAST(subvol_name);
1424 subvol_name = NULL;
1425 goto out;
1426 }
1427
1428 }
1429
1430 root = mount_subtree(mnt, subvol_name);
1431 /* mount_subtree() drops our reference on the vfsmount. */
1432 mnt = NULL;
1433
1434 if (!IS_ERR(root)) {
1435 struct super_block *s = root->d_sb;
1436 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1437 struct inode *root_inode = d_inode(root);
1438 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1439
1440 ret = 0;
1441 if (!is_subvolume_inode(root_inode)) {
1442 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1443 subvol_name);
1444 ret = -EINVAL;
1445 }
1446 if (subvol_objectid && root_objectid != subvol_objectid) {
1447 /*
1448 * This will also catch a race condition where a
1449 * subvolume which was passed by ID is renamed and
1450 * another subvolume is renamed over the old location.
1451 */
1452 btrfs_err(fs_info,
1453 "subvol '%s' does not match subvolid %llu",
1454 subvol_name, subvol_objectid);
1455 ret = -EINVAL;
1456 }
1457 if (ret) {
1458 dput(root);
1459 root = ERR_PTR(ret);
1460 deactivate_locked_super(s);
1461 }
1462 }
1463
1464out:
1465 mntput(mnt);
1466 kfree(subvol_name);
1467 return root;
1468}
1469
1470/*
1471 * Find a superblock for the given device / mount point.
1472 *
1473 * Note: This is based on mount_bdev from fs/super.c with a few additions
1474 * for multiple device setup. Make sure to keep it in sync.
1475 */
1476static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1477 int flags, const char *device_name, void *data)
1478{
1479 struct block_device *bdev = NULL;
1480 struct super_block *s;
1481 struct btrfs_device *device = NULL;
1482 struct btrfs_fs_devices *fs_devices = NULL;
1483 struct btrfs_fs_info *fs_info = NULL;
1484 void *new_sec_opts = NULL;
1485 fmode_t mode = FMODE_READ;
1486 int error = 0;
1487
1488 if (!(flags & SB_RDONLY))
1489 mode |= FMODE_WRITE;
1490
1491 if (data) {
1492 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1493 if (error)
1494 return ERR_PTR(error);
1495 }
1496
1497 /*
1498 * Setup a dummy root and fs_info for test/set super. This is because
1499 * we don't actually fill this stuff out until open_ctree, but we need
1500 * it for searching for existing supers, so this lets us do that and
1501 * then open_ctree will properly initialize everything later.
1502 */
1503 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1504 if (!fs_info) {
1505 error = -ENOMEM;
1506 goto error_sec_opts;
1507 }
1508
1509 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1510 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1511 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1512 error = -ENOMEM;
1513 goto error_fs_info;
1514 }
1515
1516 mutex_lock(&uuid_mutex);
1517 error = btrfs_parse_device_options(data, mode, fs_type);
1518 if (error) {
1519 mutex_unlock(&uuid_mutex);
1520 goto error_fs_info;
1521 }
1522
1523 device = btrfs_scan_one_device(device_name, mode, fs_type);
1524 if (IS_ERR(device)) {
1525 mutex_unlock(&uuid_mutex);
1526 error = PTR_ERR(device);
1527 goto error_fs_info;
1528 }
1529
1530 fs_devices = device->fs_devices;
1531 fs_info->fs_devices = fs_devices;
1532
1533 error = btrfs_open_devices(fs_devices, mode, fs_type);
1534 mutex_unlock(&uuid_mutex);
1535 if (error)
1536 goto error_fs_info;
1537
1538 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1539 error = -EACCES;
1540 goto error_close_devices;
1541 }
1542
1543 bdev = fs_devices->latest_bdev;
1544 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1545 fs_info);
1546 if (IS_ERR(s)) {
1547 error = PTR_ERR(s);
1548 goto error_close_devices;
1549 }
1550
1551 if (s->s_root) {
1552 btrfs_close_devices(fs_devices);
1553 free_fs_info(fs_info);
1554 if ((flags ^ s->s_flags) & SB_RDONLY)
1555 error = -EBUSY;
1556 } else {
1557 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1558 btrfs_sb(s)->bdev_holder = fs_type;
1559 if (!strstr(crc32c_impl(), "generic"))
1560 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1561 error = btrfs_fill_super(s, fs_devices, data);
1562 }
1563 if (!error)
1564 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1565 security_free_mnt_opts(&new_sec_opts);
1566 if (error) {
1567 deactivate_locked_super(s);
1568 return ERR_PTR(error);
1569 }
1570
1571 return dget(s->s_root);
1572
1573error_close_devices:
1574 btrfs_close_devices(fs_devices);
1575error_fs_info:
1576 free_fs_info(fs_info);
1577error_sec_opts:
1578 security_free_mnt_opts(&new_sec_opts);
1579 return ERR_PTR(error);
1580}
1581
1582/*
1583 * Mount function which is called by VFS layer.
1584 *
1585 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1586 * which needs vfsmount* of device's root (/). This means device's root has to
1587 * be mounted internally in any case.
1588 *
1589 * Operation flow:
1590 * 1. Parse subvol id related options for later use in mount_subvol().
1591 *
1592 * 2. Mount device's root (/) by calling vfs_kern_mount().
1593 *
1594 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1595 * first place. In order to avoid calling btrfs_mount() again, we use
1596 * different file_system_type which is not registered to VFS by
1597 * register_filesystem() (btrfs_root_fs_type). As a result,
1598 * btrfs_mount_root() is called. The return value will be used by
1599 * mount_subtree() in mount_subvol().
1600 *
1601 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1602 * "btrfs subvolume set-default", mount_subvol() is called always.
1603 */
1604static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1605 const char *device_name, void *data)
1606{
1607 struct vfsmount *mnt_root;
1608 struct dentry *root;
1609 char *subvol_name = NULL;
1610 u64 subvol_objectid = 0;
1611 int error = 0;
1612
1613 error = btrfs_parse_subvol_options(data, &subvol_name,
1614 &subvol_objectid);
1615 if (error) {
1616 kfree(subvol_name);
1617 return ERR_PTR(error);
1618 }
1619
1620 /* mount device's root (/) */
1621 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1622 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1623 if (flags & SB_RDONLY) {
1624 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1625 flags & ~SB_RDONLY, device_name, data);
1626 } else {
1627 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1628 flags | SB_RDONLY, device_name, data);
1629 if (IS_ERR(mnt_root)) {
1630 root = ERR_CAST(mnt_root);
1631 kfree(subvol_name);
1632 goto out;
1633 }
1634
1635 down_write(&mnt_root->mnt_sb->s_umount);
1636 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1637 up_write(&mnt_root->mnt_sb->s_umount);
1638 if (error < 0) {
1639 root = ERR_PTR(error);
1640 mntput(mnt_root);
1641 kfree(subvol_name);
1642 goto out;
1643 }
1644 }
1645 }
1646 if (IS_ERR(mnt_root)) {
1647 root = ERR_CAST(mnt_root);
1648 kfree(subvol_name);
1649 goto out;
1650 }
1651
1652 /* mount_subvol() will free subvol_name and mnt_root */
1653 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1654
1655out:
1656 return root;
1657}
1658
1659static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1660 u32 new_pool_size, u32 old_pool_size)
1661{
1662 if (new_pool_size == old_pool_size)
1663 return;
1664
1665 fs_info->thread_pool_size = new_pool_size;
1666
1667 btrfs_info(fs_info, "resize thread pool %d -> %d",
1668 old_pool_size, new_pool_size);
1669
1670 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1671 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1672 btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1673 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1674 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1675 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1676 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1677 new_pool_size);
1678 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1679 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1680 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1681 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1682 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1683 new_pool_size);
1684}
1685
1686static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1687{
1688 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1689}
1690
1691static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1692 unsigned long old_opts, int flags)
1693{
1694 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1695 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1696 (flags & SB_RDONLY))) {
1697 /* wait for any defraggers to finish */
1698 wait_event(fs_info->transaction_wait,
1699 (atomic_read(&fs_info->defrag_running) == 0));
1700 if (flags & SB_RDONLY)
1701 sync_filesystem(fs_info->sb);
1702 }
1703}
1704
1705static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1706 unsigned long old_opts)
1707{
1708 /*
1709 * We need to cleanup all defragable inodes if the autodefragment is
1710 * close or the filesystem is read only.
1711 */
1712 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1713 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1714 btrfs_cleanup_defrag_inodes(fs_info);
1715 }
1716
1717 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1718}
1719
1720static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1721{
1722 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1723 struct btrfs_root *root = fs_info->tree_root;
1724 unsigned old_flags = sb->s_flags;
1725 unsigned long old_opts = fs_info->mount_opt;
1726 unsigned long old_compress_type = fs_info->compress_type;
1727 u64 old_max_inline = fs_info->max_inline;
1728 u32 old_thread_pool_size = fs_info->thread_pool_size;
1729 u32 old_metadata_ratio = fs_info->metadata_ratio;
1730 int ret;
1731
1732 sync_filesystem(sb);
1733 btrfs_remount_prepare(fs_info);
1734
1735 if (data) {
1736 void *new_sec_opts = NULL;
1737
1738 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1739 if (!ret)
1740 ret = security_sb_remount(sb, new_sec_opts);
1741 security_free_mnt_opts(&new_sec_opts);
1742 if (ret)
1743 goto restore;
1744 }
1745
1746 ret = btrfs_parse_options(fs_info, data, *flags);
1747 if (ret)
1748 goto restore;
1749
1750 btrfs_remount_begin(fs_info, old_opts, *flags);
1751 btrfs_resize_thread_pool(fs_info,
1752 fs_info->thread_pool_size, old_thread_pool_size);
1753
1754 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1755 goto out;
1756
1757 if (*flags & SB_RDONLY) {
1758 /*
1759 * this also happens on 'umount -rf' or on shutdown, when
1760 * the filesystem is busy.
1761 */
1762 cancel_work_sync(&fs_info->async_reclaim_work);
1763
1764 /* wait for the uuid_scan task to finish */
1765 down(&fs_info->uuid_tree_rescan_sem);
1766 /* avoid complains from lockdep et al. */
1767 up(&fs_info->uuid_tree_rescan_sem);
1768
1769 sb->s_flags |= SB_RDONLY;
1770
1771 /*
1772 * Setting SB_RDONLY will put the cleaner thread to
1773 * sleep at the next loop if it's already active.
1774 * If it's already asleep, we'll leave unused block
1775 * groups on disk until we're mounted read-write again
1776 * unless we clean them up here.
1777 */
1778 btrfs_delete_unused_bgs(fs_info);
1779
1780 btrfs_dev_replace_suspend_for_unmount(fs_info);
1781 btrfs_scrub_cancel(fs_info);
1782 btrfs_pause_balance(fs_info);
1783
1784 ret = btrfs_commit_super(fs_info);
1785 if (ret)
1786 goto restore;
1787 } else {
1788 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1789 btrfs_err(fs_info,
1790 "Remounting read-write after error is not allowed");
1791 ret = -EINVAL;
1792 goto restore;
1793 }
1794 if (fs_info->fs_devices->rw_devices == 0) {
1795 ret = -EACCES;
1796 goto restore;
1797 }
1798
1799 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1800 btrfs_warn(fs_info,
1801 "too many missing devices, writable remount is not allowed");
1802 ret = -EACCES;
1803 goto restore;
1804 }
1805
1806 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1807 ret = -EINVAL;
1808 goto restore;
1809 }
1810
1811 ret = btrfs_cleanup_fs_roots(fs_info);
1812 if (ret)
1813 goto restore;
1814
1815 /* recover relocation */
1816 mutex_lock(&fs_info->cleaner_mutex);
1817 ret = btrfs_recover_relocation(root);
1818 mutex_unlock(&fs_info->cleaner_mutex);
1819 if (ret)
1820 goto restore;
1821
1822 ret = btrfs_resume_balance_async(fs_info);
1823 if (ret)
1824 goto restore;
1825
1826 ret = btrfs_resume_dev_replace_async(fs_info);
1827 if (ret) {
1828 btrfs_warn(fs_info, "failed to resume dev_replace");
1829 goto restore;
1830 }
1831
1832 btrfs_qgroup_rescan_resume(fs_info);
1833
1834 if (!fs_info->uuid_root) {
1835 btrfs_info(fs_info, "creating UUID tree");
1836 ret = btrfs_create_uuid_tree(fs_info);
1837 if (ret) {
1838 btrfs_warn(fs_info,
1839 "failed to create the UUID tree %d",
1840 ret);
1841 goto restore;
1842 }
1843 }
1844 sb->s_flags &= ~SB_RDONLY;
1845
1846 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1847 }
1848out:
1849 wake_up_process(fs_info->transaction_kthread);
1850 btrfs_remount_cleanup(fs_info, old_opts);
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 sb->s_flags = old_flags;
1858 fs_info->mount_opt = old_opts;
1859 fs_info->compress_type = old_compress_type;
1860 fs_info->max_inline = old_max_inline;
1861 btrfs_resize_thread_pool(fs_info,
1862 old_thread_pool_size, fs_info->thread_pool_size);
1863 fs_info->metadata_ratio = old_metadata_ratio;
1864 btrfs_remount_cleanup(fs_info, old_opts);
1865 return ret;
1866}
1867
1868/* Used to sort the devices by max_avail(descending sort) */
1869static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
1870 const void *dev_info2)
1871{
1872 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1873 ((struct btrfs_device_info *)dev_info2)->max_avail)
1874 return -1;
1875 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1876 ((struct btrfs_device_info *)dev_info2)->max_avail)
1877 return 1;
1878 else
1879 return 0;
1880}
1881
1882/*
1883 * sort the devices by max_avail, in which max free extent size of each device
1884 * is stored.(Descending Sort)
1885 */
1886static inline void btrfs_descending_sort_devices(
1887 struct btrfs_device_info *devices,
1888 size_t nr_devices)
1889{
1890 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1891 btrfs_cmp_device_free_bytes, NULL);
1892}
1893
1894/*
1895 * The helper to calc the free space on the devices that can be used to store
1896 * file data.
1897 */
1898static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1899 u64 *free_bytes)
1900{
1901 struct btrfs_device_info *devices_info;
1902 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1903 struct btrfs_device *device;
1904 u64 type;
1905 u64 avail_space;
1906 u64 min_stripe_size;
1907 int num_stripes = 1;
1908 int i = 0, nr_devices;
1909 const struct btrfs_raid_attr *rattr;
1910
1911 /*
1912 * We aren't under the device list lock, so this is racy-ish, but good
1913 * enough for our purposes.
1914 */
1915 nr_devices = fs_info->fs_devices->open_devices;
1916 if (!nr_devices) {
1917 smp_mb();
1918 nr_devices = fs_info->fs_devices->open_devices;
1919 ASSERT(nr_devices);
1920 if (!nr_devices) {
1921 *free_bytes = 0;
1922 return 0;
1923 }
1924 }
1925
1926 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1927 GFP_KERNEL);
1928 if (!devices_info)
1929 return -ENOMEM;
1930
1931 /* calc min stripe number for data space allocation */
1932 type = btrfs_data_alloc_profile(fs_info);
1933 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1934
1935 if (type & BTRFS_BLOCK_GROUP_RAID0)
1936 num_stripes = nr_devices;
1937 else if (type & BTRFS_BLOCK_GROUP_RAID1)
1938 num_stripes = 2;
1939 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1940 num_stripes = 4;
1941
1942 /* Adjust for more than 1 stripe per device */
1943 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1944
1945 rcu_read_lock();
1946 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1947 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1948 &device->dev_state) ||
1949 !device->bdev ||
1950 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1951 continue;
1952
1953 if (i >= nr_devices)
1954 break;
1955
1956 avail_space = device->total_bytes - device->bytes_used;
1957
1958 /* align with stripe_len */
1959 avail_space = rounddown(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 * This ensures we have at least min_stripe_size free space
1967 * after excluding 1MB.
1968 */
1969 if (avail_space <= SZ_1M + min_stripe_size)
1970 continue;
1971
1972 avail_space -= SZ_1M;
1973
1974 devices_info[i].dev = device;
1975 devices_info[i].max_avail = avail_space;
1976
1977 i++;
1978 }
1979 rcu_read_unlock();
1980
1981 nr_devices = i;
1982
1983 btrfs_descending_sort_devices(devices_info, nr_devices);
1984
1985 i = nr_devices - 1;
1986 avail_space = 0;
1987 while (nr_devices >= rattr->devs_min) {
1988 num_stripes = min(num_stripes, nr_devices);
1989
1990 if (devices_info[i].max_avail >= min_stripe_size) {
1991 int j;
1992 u64 alloc_size;
1993
1994 avail_space += devices_info[i].max_avail * num_stripes;
1995 alloc_size = devices_info[i].max_avail;
1996 for (j = i + 1 - num_stripes; j <= i; j++)
1997 devices_info[j].max_avail -= alloc_size;
1998 }
1999 i--;
2000 nr_devices--;
2001 }
2002
2003 kfree(devices_info);
2004 *free_bytes = avail_space;
2005 return 0;
2006}
2007
2008/*
2009 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2010 *
2011 * If there's a redundant raid level at DATA block groups, use the respective
2012 * multiplier to scale the sizes.
2013 *
2014 * Unused device space usage is based on simulating the chunk allocator
2015 * algorithm that respects the device sizes and order of allocations. This is
2016 * a close approximation of the actual use but there are other factors that may
2017 * change the result (like a new metadata chunk).
2018 *
2019 * If metadata is exhausted, f_bavail will be 0.
2020 */
2021static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2022{
2023 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2024 struct btrfs_super_block *disk_super = fs_info->super_copy;
2025 struct list_head *head = &fs_info->space_info;
2026 struct btrfs_space_info *found;
2027 u64 total_used = 0;
2028 u64 total_free_data = 0;
2029 u64 total_free_meta = 0;
2030 int bits = dentry->d_sb->s_blocksize_bits;
2031 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2032 unsigned factor = 1;
2033 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2034 int ret;
2035 u64 thresh = 0;
2036 int mixed = 0;
2037
2038 rcu_read_lock();
2039 list_for_each_entry_rcu(found, head, list) {
2040 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2041 int i;
2042
2043 total_free_data += found->disk_total - found->disk_used;
2044 total_free_data -=
2045 btrfs_account_ro_block_groups_free_space(found);
2046
2047 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2048 if (!list_empty(&found->block_groups[i]))
2049 factor = btrfs_bg_type_to_factor(
2050 btrfs_raid_array[i].bg_flag);
2051 }
2052 }
2053
2054 /*
2055 * Metadata in mixed block goup profiles are accounted in data
2056 */
2057 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2058 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2059 mixed = 1;
2060 else
2061 total_free_meta += found->disk_total -
2062 found->disk_used;
2063 }
2064
2065 total_used += found->disk_used;
2066 }
2067
2068 rcu_read_unlock();
2069
2070 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2071 buf->f_blocks >>= bits;
2072 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2073
2074 /* Account global block reserve as used, it's in logical size already */
2075 spin_lock(&block_rsv->lock);
2076 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2077 if (buf->f_bfree >= block_rsv->size >> bits)
2078 buf->f_bfree -= block_rsv->size >> bits;
2079 else
2080 buf->f_bfree = 0;
2081 spin_unlock(&block_rsv->lock);
2082
2083 buf->f_bavail = div_u64(total_free_data, factor);
2084 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2085 if (ret)
2086 return ret;
2087 buf->f_bavail += div_u64(total_free_data, factor);
2088 buf->f_bavail = buf->f_bavail >> bits;
2089
2090 /*
2091 * We calculate the remaining metadata space minus global reserve. If
2092 * this is (supposedly) smaller than zero, there's no space. But this
2093 * does not hold in practice, the exhausted state happens where's still
2094 * some positive delta. So we apply some guesswork and compare the
2095 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2096 *
2097 * We probably cannot calculate the exact threshold value because this
2098 * depends on the internal reservations requested by various
2099 * operations, so some operations that consume a few metadata will
2100 * succeed even if the Avail is zero. But this is better than the other
2101 * way around.
2102 */
2103 thresh = SZ_4M;
2104
2105 if (!mixed && total_free_meta - thresh < block_rsv->size)
2106 buf->f_bavail = 0;
2107
2108 buf->f_type = BTRFS_SUPER_MAGIC;
2109 buf->f_bsize = dentry->d_sb->s_blocksize;
2110 buf->f_namelen = BTRFS_NAME_LEN;
2111
2112 /* We treat it as constant endianness (it doesn't matter _which_)
2113 because we want the fsid to come out the same whether mounted
2114 on a big-endian or little-endian host */
2115 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2116 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2117 /* Mask in the root object ID too, to disambiguate subvols */
2118 buf->f_fsid.val[0] ^=
2119 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2120 buf->f_fsid.val[1] ^=
2121 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2122
2123 return 0;
2124}
2125
2126static void btrfs_kill_super(struct super_block *sb)
2127{
2128 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2129 kill_anon_super(sb);
2130 free_fs_info(fs_info);
2131}
2132
2133static struct file_system_type btrfs_fs_type = {
2134 .owner = THIS_MODULE,
2135 .name = "btrfs",
2136 .mount = btrfs_mount,
2137 .kill_sb = btrfs_kill_super,
2138 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2139};
2140
2141static struct file_system_type btrfs_root_fs_type = {
2142 .owner = THIS_MODULE,
2143 .name = "btrfs",
2144 .mount = btrfs_mount_root,
2145 .kill_sb = btrfs_kill_super,
2146 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2147};
2148
2149MODULE_ALIAS_FS("btrfs");
2150
2151static int btrfs_control_open(struct inode *inode, struct file *file)
2152{
2153 /*
2154 * The control file's private_data is used to hold the
2155 * transaction when it is started and is used to keep
2156 * track of whether a transaction is already in progress.
2157 */
2158 file->private_data = NULL;
2159 return 0;
2160}
2161
2162/*
2163 * used by btrfsctl to scan devices when no FS is mounted
2164 */
2165static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2166 unsigned long arg)
2167{
2168 struct btrfs_ioctl_vol_args *vol;
2169 struct btrfs_device *device = NULL;
2170 int ret = -ENOTTY;
2171
2172 if (!capable(CAP_SYS_ADMIN))
2173 return -EPERM;
2174
2175 vol = memdup_user((void __user *)arg, sizeof(*vol));
2176 if (IS_ERR(vol))
2177 return PTR_ERR(vol);
2178 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2179
2180 switch (cmd) {
2181 case BTRFS_IOC_SCAN_DEV:
2182 mutex_lock(&uuid_mutex);
2183 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2184 &btrfs_root_fs_type);
2185 ret = PTR_ERR_OR_ZERO(device);
2186 mutex_unlock(&uuid_mutex);
2187 break;
2188 case BTRFS_IOC_FORGET_DEV:
2189 ret = btrfs_forget_devices(vol->name);
2190 break;
2191 case BTRFS_IOC_DEVICES_READY:
2192 mutex_lock(&uuid_mutex);
2193 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2194 &btrfs_root_fs_type);
2195 if (IS_ERR(device)) {
2196 mutex_unlock(&uuid_mutex);
2197 ret = PTR_ERR(device);
2198 break;
2199 }
2200 ret = !(device->fs_devices->num_devices ==
2201 device->fs_devices->total_devices);
2202 mutex_unlock(&uuid_mutex);
2203 break;
2204 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2205 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2206 break;
2207 }
2208
2209 kfree(vol);
2210 return ret;
2211}
2212
2213static int btrfs_freeze(struct super_block *sb)
2214{
2215 struct btrfs_trans_handle *trans;
2216 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2217 struct btrfs_root *root = fs_info->tree_root;
2218
2219 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2220 /*
2221 * We don't need a barrier here, we'll wait for any transaction that
2222 * could be in progress on other threads (and do delayed iputs that
2223 * we want to avoid on a frozen filesystem), or do the commit
2224 * ourselves.
2225 */
2226 trans = btrfs_attach_transaction_barrier(root);
2227 if (IS_ERR(trans)) {
2228 /* no transaction, don't bother */
2229 if (PTR_ERR(trans) == -ENOENT)
2230 return 0;
2231 return PTR_ERR(trans);
2232 }
2233 return btrfs_commit_transaction(trans);
2234}
2235
2236static int btrfs_unfreeze(struct super_block *sb)
2237{
2238 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2239
2240 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2241 return 0;
2242}
2243
2244static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2245{
2246 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2247 struct btrfs_fs_devices *cur_devices;
2248 struct btrfs_device *dev, *first_dev = NULL;
2249 struct list_head *head;
2250
2251 /*
2252 * Lightweight locking of the devices. We should not need
2253 * device_list_mutex here as we only read the device data and the list
2254 * is protected by RCU. Even if a device is deleted during the list
2255 * traversals, we'll get valid data, the freeing callback will wait at
2256 * least until the rcu_read_unlock.
2257 */
2258 rcu_read_lock();
2259 cur_devices = fs_info->fs_devices;
2260 while (cur_devices) {
2261 head = &cur_devices->devices;
2262 list_for_each_entry_rcu(dev, head, dev_list) {
2263 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2264 continue;
2265 if (!dev->name)
2266 continue;
2267 if (!first_dev || dev->devid < first_dev->devid)
2268 first_dev = dev;
2269 }
2270 cur_devices = cur_devices->seed;
2271 }
2272
2273 if (first_dev)
2274 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2275 else
2276 WARN_ON(1);
2277 rcu_read_unlock();
2278 return 0;
2279}
2280
2281static const struct super_operations btrfs_super_ops = {
2282 .drop_inode = btrfs_drop_inode,
2283 .evict_inode = btrfs_evict_inode,
2284 .put_super = btrfs_put_super,
2285 .sync_fs = btrfs_sync_fs,
2286 .show_options = btrfs_show_options,
2287 .show_devname = btrfs_show_devname,
2288 .alloc_inode = btrfs_alloc_inode,
2289 .destroy_inode = btrfs_destroy_inode,
2290 .free_inode = btrfs_free_inode,
2291 .statfs = btrfs_statfs,
2292 .remount_fs = btrfs_remount,
2293 .freeze_fs = btrfs_freeze,
2294 .unfreeze_fs = btrfs_unfreeze,
2295};
2296
2297static const struct file_operations btrfs_ctl_fops = {
2298 .open = btrfs_control_open,
2299 .unlocked_ioctl = btrfs_control_ioctl,
2300 .compat_ioctl = btrfs_control_ioctl,
2301 .owner = THIS_MODULE,
2302 .llseek = noop_llseek,
2303};
2304
2305static struct miscdevice btrfs_misc = {
2306 .minor = BTRFS_MINOR,
2307 .name = "btrfs-control",
2308 .fops = &btrfs_ctl_fops
2309};
2310
2311MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2312MODULE_ALIAS("devname:btrfs-control");
2313
2314static int __init btrfs_interface_init(void)
2315{
2316 return misc_register(&btrfs_misc);
2317}
2318
2319static __cold void btrfs_interface_exit(void)
2320{
2321 misc_deregister(&btrfs_misc);
2322}
2323
2324static void __init btrfs_print_mod_info(void)
2325{
2326 static const char options[] = ""
2327#ifdef CONFIG_BTRFS_DEBUG
2328 ", debug=on"
2329#endif
2330#ifdef CONFIG_BTRFS_ASSERT
2331 ", assert=on"
2332#endif
2333#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2334 ", integrity-checker=on"
2335#endif
2336#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2337 ", ref-verify=on"
2338#endif
2339 ;
2340 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2341}
2342
2343static int __init init_btrfs_fs(void)
2344{
2345 int err;
2346
2347 btrfs_props_init();
2348
2349 err = btrfs_init_sysfs();
2350 if (err)
2351 return err;
2352
2353 btrfs_init_compress();
2354
2355 err = btrfs_init_cachep();
2356 if (err)
2357 goto free_compress;
2358
2359 err = extent_io_init();
2360 if (err)
2361 goto free_cachep;
2362
2363 err = extent_map_init();
2364 if (err)
2365 goto free_extent_io;
2366
2367 err = ordered_data_init();
2368 if (err)
2369 goto free_extent_map;
2370
2371 err = btrfs_delayed_inode_init();
2372 if (err)
2373 goto free_ordered_data;
2374
2375 err = btrfs_auto_defrag_init();
2376 if (err)
2377 goto free_delayed_inode;
2378
2379 err = btrfs_delayed_ref_init();
2380 if (err)
2381 goto free_auto_defrag;
2382
2383 err = btrfs_prelim_ref_init();
2384 if (err)
2385 goto free_delayed_ref;
2386
2387 err = btrfs_end_io_wq_init();
2388 if (err)
2389 goto free_prelim_ref;
2390
2391 err = btrfs_interface_init();
2392 if (err)
2393 goto free_end_io_wq;
2394
2395 btrfs_init_lockdep();
2396
2397 btrfs_print_mod_info();
2398
2399 err = btrfs_run_sanity_tests();
2400 if (err)
2401 goto unregister_ioctl;
2402
2403 err = register_filesystem(&btrfs_fs_type);
2404 if (err)
2405 goto unregister_ioctl;
2406
2407 return 0;
2408
2409unregister_ioctl:
2410 btrfs_interface_exit();
2411free_end_io_wq:
2412 btrfs_end_io_wq_exit();
2413free_prelim_ref:
2414 btrfs_prelim_ref_exit();
2415free_delayed_ref:
2416 btrfs_delayed_ref_exit();
2417free_auto_defrag:
2418 btrfs_auto_defrag_exit();
2419free_delayed_inode:
2420 btrfs_delayed_inode_exit();
2421free_ordered_data:
2422 ordered_data_exit();
2423free_extent_map:
2424 extent_map_exit();
2425free_extent_io:
2426 extent_io_exit();
2427free_cachep:
2428 btrfs_destroy_cachep();
2429free_compress:
2430 btrfs_exit_compress();
2431 btrfs_exit_sysfs();
2432
2433 return err;
2434}
2435
2436static void __exit exit_btrfs_fs(void)
2437{
2438 btrfs_destroy_cachep();
2439 btrfs_delayed_ref_exit();
2440 btrfs_auto_defrag_exit();
2441 btrfs_delayed_inode_exit();
2442 btrfs_prelim_ref_exit();
2443 ordered_data_exit();
2444 extent_map_exit();
2445 extent_io_exit();
2446 btrfs_interface_exit();
2447 btrfs_end_io_wq_exit();
2448 unregister_filesystem(&btrfs_fs_type);
2449 btrfs_exit_sysfs();
2450 btrfs_cleanup_fs_uuids();
2451 btrfs_exit_compress();
2452}
2453
2454late_initcall(init_btrfs_fs);
2455module_exit(exit_btrfs_fs)
2456
2457MODULE_LICENSE("GPL");
2458MODULE_SOFTDEP("pre: crc32c");
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 <linux/security.h>
30#include <linux/fs_parser.h>
31#include "messages.h"
32#include "delayed-inode.h"
33#include "ctree.h"
34#include "disk-io.h"
35#include "transaction.h"
36#include "btrfs_inode.h"
37#include "props.h"
38#include "xattr.h"
39#include "bio.h"
40#include "export.h"
41#include "compression.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#include "extent-tree.h"
62#define CREATE_TRACE_POINTS
63#include <trace/events/btrfs.h>
64
65static const struct super_operations btrfs_super_ops;
66static struct file_system_type btrfs_fs_type;
67
68static void btrfs_put_super(struct super_block *sb)
69{
70 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
71
72 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
73 close_ctree(fs_info);
74}
75
76/* Store the mount options related information. */
77struct btrfs_fs_context {
78 char *subvol_name;
79 u64 subvol_objectid;
80 u64 max_inline;
81 u32 commit_interval;
82 u32 metadata_ratio;
83 u32 thread_pool_size;
84 unsigned long mount_opt;
85 unsigned long compress_type:4;
86 unsigned int compress_level;
87 refcount_t refs;
88};
89
90enum {
91 Opt_acl,
92 Opt_clear_cache,
93 Opt_commit_interval,
94 Opt_compress,
95 Opt_compress_force,
96 Opt_compress_force_type,
97 Opt_compress_type,
98 Opt_degraded,
99 Opt_device,
100 Opt_fatal_errors,
101 Opt_flushoncommit,
102 Opt_max_inline,
103 Opt_barrier,
104 Opt_datacow,
105 Opt_datasum,
106 Opt_defrag,
107 Opt_discard,
108 Opt_discard_mode,
109 Opt_ratio,
110 Opt_rescan_uuid_tree,
111 Opt_skip_balance,
112 Opt_space_cache,
113 Opt_space_cache_version,
114 Opt_ssd,
115 Opt_ssd_spread,
116 Opt_subvol,
117 Opt_subvol_empty,
118 Opt_subvolid,
119 Opt_thread_pool,
120 Opt_treelog,
121 Opt_user_subvol_rm_allowed,
122
123 /* Rescue options */
124 Opt_rescue,
125 Opt_usebackuproot,
126 Opt_nologreplay,
127 Opt_ignorebadroots,
128 Opt_ignoredatacsums,
129 Opt_rescue_all,
130
131 /* Debugging options */
132 Opt_enospc_debug,
133#ifdef CONFIG_BTRFS_DEBUG
134 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
135#endif
136#ifdef CONFIG_BTRFS_FS_REF_VERIFY
137 Opt_ref_verify,
138#endif
139 Opt_err,
140};
141
142enum {
143 Opt_fatal_errors_panic,
144 Opt_fatal_errors_bug,
145};
146
147static const struct constant_table btrfs_parameter_fatal_errors[] = {
148 { "panic", Opt_fatal_errors_panic },
149 { "bug", Opt_fatal_errors_bug },
150 {}
151};
152
153enum {
154 Opt_discard_sync,
155 Opt_discard_async,
156};
157
158static const struct constant_table btrfs_parameter_discard[] = {
159 { "sync", Opt_discard_sync },
160 { "async", Opt_discard_async },
161 {}
162};
163
164enum {
165 Opt_space_cache_v1,
166 Opt_space_cache_v2,
167};
168
169static const struct constant_table btrfs_parameter_space_cache[] = {
170 { "v1", Opt_space_cache_v1 },
171 { "v2", Opt_space_cache_v2 },
172 {}
173};
174
175enum {
176 Opt_rescue_usebackuproot,
177 Opt_rescue_nologreplay,
178 Opt_rescue_ignorebadroots,
179 Opt_rescue_ignoredatacsums,
180 Opt_rescue_parameter_all,
181};
182
183static const struct constant_table btrfs_parameter_rescue[] = {
184 { "usebackuproot", Opt_rescue_usebackuproot },
185 { "nologreplay", Opt_rescue_nologreplay },
186 { "ignorebadroots", Opt_rescue_ignorebadroots },
187 { "ibadroots", Opt_rescue_ignorebadroots },
188 { "ignoredatacsums", Opt_rescue_ignoredatacsums },
189 { "idatacsums", Opt_rescue_ignoredatacsums },
190 { "all", Opt_rescue_parameter_all },
191 {}
192};
193
194#ifdef CONFIG_BTRFS_DEBUG
195enum {
196 Opt_fragment_parameter_data,
197 Opt_fragment_parameter_metadata,
198 Opt_fragment_parameter_all,
199};
200
201static const struct constant_table btrfs_parameter_fragment[] = {
202 { "data", Opt_fragment_parameter_data },
203 { "metadata", Opt_fragment_parameter_metadata },
204 { "all", Opt_fragment_parameter_all },
205 {}
206};
207#endif
208
209static const struct fs_parameter_spec btrfs_fs_parameters[] = {
210 fsparam_flag_no("acl", Opt_acl),
211 fsparam_flag_no("autodefrag", Opt_defrag),
212 fsparam_flag_no("barrier", Opt_barrier),
213 fsparam_flag("clear_cache", Opt_clear_cache),
214 fsparam_u32("commit", Opt_commit_interval),
215 fsparam_flag("compress", Opt_compress),
216 fsparam_string("compress", Opt_compress_type),
217 fsparam_flag("compress-force", Opt_compress_force),
218 fsparam_string("compress-force", Opt_compress_force_type),
219 fsparam_flag_no("datacow", Opt_datacow),
220 fsparam_flag_no("datasum", Opt_datasum),
221 fsparam_flag("degraded", Opt_degraded),
222 fsparam_string("device", Opt_device),
223 fsparam_flag_no("discard", Opt_discard),
224 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
225 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
226 fsparam_flag_no("flushoncommit", Opt_flushoncommit),
227 fsparam_string("max_inline", Opt_max_inline),
228 fsparam_u32("metadata_ratio", Opt_ratio),
229 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
230 fsparam_flag("skip_balance", Opt_skip_balance),
231 fsparam_flag_no("space_cache", Opt_space_cache),
232 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
233 fsparam_flag_no("ssd", Opt_ssd),
234 fsparam_flag_no("ssd_spread", Opt_ssd_spread),
235 fsparam_string("subvol", Opt_subvol),
236 fsparam_flag("subvol=", Opt_subvol_empty),
237 fsparam_u64("subvolid", Opt_subvolid),
238 fsparam_u32("thread_pool", Opt_thread_pool),
239 fsparam_flag_no("treelog", Opt_treelog),
240 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
241
242 /* Rescue options. */
243 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
244 /* Deprecated, with alias rescue=nologreplay */
245 __fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
246 /* Deprecated, with alias rescue=usebackuproot */
247 __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
248
249 /* Debugging options. */
250 fsparam_flag_no("enospc_debug", Opt_enospc_debug),
251#ifdef CONFIG_BTRFS_DEBUG
252 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
253#endif
254#ifdef CONFIG_BTRFS_FS_REF_VERIFY
255 fsparam_flag("ref_verify", Opt_ref_verify),
256#endif
257 {}
258};
259
260/* No support for restricting writes to btrfs devices yet... */
261static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
262{
263 return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
264}
265
266static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
267{
268 struct btrfs_fs_context *ctx = fc->fs_private;
269 struct fs_parse_result result;
270 int opt;
271
272 opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
273 if (opt < 0)
274 return opt;
275
276 switch (opt) {
277 case Opt_degraded:
278 btrfs_set_opt(ctx->mount_opt, DEGRADED);
279 break;
280 case Opt_subvol_empty:
281 /*
282 * This exists because we used to allow it on accident, so we're
283 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow
284 * empty subvol= again").
285 */
286 break;
287 case Opt_subvol:
288 kfree(ctx->subvol_name);
289 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
290 if (!ctx->subvol_name)
291 return -ENOMEM;
292 break;
293 case Opt_subvolid:
294 ctx->subvol_objectid = result.uint_64;
295
296 /* subvolid=0 means give me the original fs_tree. */
297 if (!ctx->subvol_objectid)
298 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
299 break;
300 case Opt_device: {
301 struct btrfs_device *device;
302 blk_mode_t mode = btrfs_open_mode(fc);
303
304 mutex_lock(&uuid_mutex);
305 device = btrfs_scan_one_device(param->string, mode, false);
306 mutex_unlock(&uuid_mutex);
307 if (IS_ERR(device))
308 return PTR_ERR(device);
309 break;
310 }
311 case Opt_datasum:
312 if (result.negated) {
313 btrfs_set_opt(ctx->mount_opt, NODATASUM);
314 } else {
315 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
316 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
317 }
318 break;
319 case Opt_datacow:
320 if (result.negated) {
321 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
322 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
323 btrfs_set_opt(ctx->mount_opt, NODATACOW);
324 btrfs_set_opt(ctx->mount_opt, NODATASUM);
325 } else {
326 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
327 }
328 break;
329 case Opt_compress_force:
330 case Opt_compress_force_type:
331 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
332 fallthrough;
333 case Opt_compress:
334 case Opt_compress_type:
335 if (opt == Opt_compress || opt == Opt_compress_force) {
336 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
337 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
338 btrfs_set_opt(ctx->mount_opt, COMPRESS);
339 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
340 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
341 } else if (strncmp(param->string, "zlib", 4) == 0) {
342 ctx->compress_type = BTRFS_COMPRESS_ZLIB;
343 ctx->compress_level =
344 btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
345 param->string + 4);
346 btrfs_set_opt(ctx->mount_opt, COMPRESS);
347 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
348 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
349 } else if (strncmp(param->string, "lzo", 3) == 0) {
350 ctx->compress_type = BTRFS_COMPRESS_LZO;
351 ctx->compress_level = 0;
352 btrfs_set_opt(ctx->mount_opt, COMPRESS);
353 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
354 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
355 } else if (strncmp(param->string, "zstd", 4) == 0) {
356 ctx->compress_type = BTRFS_COMPRESS_ZSTD;
357 ctx->compress_level =
358 btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
359 param->string + 4);
360 btrfs_set_opt(ctx->mount_opt, COMPRESS);
361 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
362 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
363 } else if (strncmp(param->string, "no", 2) == 0) {
364 ctx->compress_level = 0;
365 ctx->compress_type = 0;
366 btrfs_clear_opt(ctx->mount_opt, COMPRESS);
367 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
368 } else {
369 btrfs_err(NULL, "unrecognized compression value %s",
370 param->string);
371 return -EINVAL;
372 }
373 break;
374 case Opt_ssd:
375 if (result.negated) {
376 btrfs_set_opt(ctx->mount_opt, NOSSD);
377 btrfs_clear_opt(ctx->mount_opt, SSD);
378 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
379 } else {
380 btrfs_set_opt(ctx->mount_opt, SSD);
381 btrfs_clear_opt(ctx->mount_opt, NOSSD);
382 }
383 break;
384 case Opt_ssd_spread:
385 if (result.negated) {
386 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
387 } else {
388 btrfs_set_opt(ctx->mount_opt, SSD);
389 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
390 btrfs_clear_opt(ctx->mount_opt, NOSSD);
391 }
392 break;
393 case Opt_barrier:
394 if (result.negated)
395 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
396 else
397 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
398 break;
399 case Opt_thread_pool:
400 if (result.uint_32 == 0) {
401 btrfs_err(NULL, "invalid value 0 for thread_pool");
402 return -EINVAL;
403 }
404 ctx->thread_pool_size = result.uint_32;
405 break;
406 case Opt_max_inline:
407 ctx->max_inline = memparse(param->string, NULL);
408 break;
409 case Opt_acl:
410 if (result.negated) {
411 fc->sb_flags &= ~SB_POSIXACL;
412 } else {
413#ifdef CONFIG_BTRFS_FS_POSIX_ACL
414 fc->sb_flags |= SB_POSIXACL;
415#else
416 btrfs_err(NULL, "support for ACL not compiled in");
417 return -EINVAL;
418#endif
419 }
420 /*
421 * VFS limits the ability to toggle ACL on and off via remount,
422 * despite every file system allowing this. This seems to be
423 * an oversight since we all do, but it'll fail if we're
424 * remounting. So don't set the mask here, we'll check it in
425 * btrfs_reconfigure and do the toggling ourselves.
426 */
427 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
428 fc->sb_flags_mask |= SB_POSIXACL;
429 break;
430 case Opt_treelog:
431 if (result.negated)
432 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
433 else
434 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
435 break;
436 case Opt_nologreplay:
437 btrfs_warn(NULL,
438 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
439 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
440 break;
441 case Opt_flushoncommit:
442 if (result.negated)
443 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
444 else
445 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
446 break;
447 case Opt_ratio:
448 ctx->metadata_ratio = result.uint_32;
449 break;
450 case Opt_discard:
451 if (result.negated) {
452 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
453 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
454 btrfs_set_opt(ctx->mount_opt, NODISCARD);
455 } else {
456 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
457 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
458 }
459 break;
460 case Opt_discard_mode:
461 switch (result.uint_32) {
462 case Opt_discard_sync:
463 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
464 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
465 break;
466 case Opt_discard_async:
467 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
468 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
469 break;
470 default:
471 btrfs_err(NULL, "unrecognized discard mode value %s",
472 param->key);
473 return -EINVAL;
474 }
475 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
476 break;
477 case Opt_space_cache:
478 if (result.negated) {
479 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
480 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
481 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
482 } else {
483 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
484 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
485 }
486 break;
487 case Opt_space_cache_version:
488 switch (result.uint_32) {
489 case Opt_space_cache_v1:
490 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
491 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
492 break;
493 case Opt_space_cache_v2:
494 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
495 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
496 break;
497 default:
498 btrfs_err(NULL, "unrecognized space_cache value %s",
499 param->key);
500 return -EINVAL;
501 }
502 break;
503 case Opt_rescan_uuid_tree:
504 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
505 break;
506 case Opt_clear_cache:
507 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
508 break;
509 case Opt_user_subvol_rm_allowed:
510 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
511 break;
512 case Opt_enospc_debug:
513 if (result.negated)
514 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
515 else
516 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
517 break;
518 case Opt_defrag:
519 if (result.negated)
520 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
521 else
522 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
523 break;
524 case Opt_usebackuproot:
525 btrfs_warn(NULL,
526 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
527 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
528
529 /* If we're loading the backup roots we can't trust the space cache. */
530 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
531 break;
532 case Opt_skip_balance:
533 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
534 break;
535 case Opt_fatal_errors:
536 switch (result.uint_32) {
537 case Opt_fatal_errors_panic:
538 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
539 break;
540 case Opt_fatal_errors_bug:
541 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
542 break;
543 default:
544 btrfs_err(NULL, "unrecognized fatal_errors value %s",
545 param->key);
546 return -EINVAL;
547 }
548 break;
549 case Opt_commit_interval:
550 ctx->commit_interval = result.uint_32;
551 if (ctx->commit_interval == 0)
552 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
553 break;
554 case Opt_rescue:
555 switch (result.uint_32) {
556 case Opt_rescue_usebackuproot:
557 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
558 break;
559 case Opt_rescue_nologreplay:
560 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
561 break;
562 case Opt_rescue_ignorebadroots:
563 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
564 break;
565 case Opt_rescue_ignoredatacsums:
566 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
567 break;
568 case Opt_rescue_parameter_all:
569 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
570 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
571 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
572 break;
573 default:
574 btrfs_info(NULL, "unrecognized rescue option '%s'",
575 param->key);
576 return -EINVAL;
577 }
578 break;
579#ifdef CONFIG_BTRFS_DEBUG
580 case Opt_fragment:
581 switch (result.uint_32) {
582 case Opt_fragment_parameter_all:
583 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
584 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
585 break;
586 case Opt_fragment_parameter_metadata:
587 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
588 break;
589 case Opt_fragment_parameter_data:
590 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
591 break;
592 default:
593 btrfs_info(NULL, "unrecognized fragment option '%s'",
594 param->key);
595 return -EINVAL;
596 }
597 break;
598#endif
599#ifdef CONFIG_BTRFS_FS_REF_VERIFY
600 case Opt_ref_verify:
601 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
602 break;
603#endif
604 default:
605 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
606 return -EINVAL;
607 }
608
609 return 0;
610}
611
612/*
613 * Some options only have meaning at mount time and shouldn't persist across
614 * remounts, or be displayed. Clear these at the end of mount and remount code
615 * paths.
616 */
617static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
618{
619 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
620 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
621 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
622}
623
624static bool check_ro_option(struct btrfs_fs_info *fs_info,
625 unsigned long mount_opt, unsigned long opt,
626 const char *opt_name)
627{
628 if (mount_opt & opt) {
629 btrfs_err(fs_info, "%s must be used with ro mount option",
630 opt_name);
631 return true;
632 }
633 return false;
634}
635
636bool btrfs_check_options(struct btrfs_fs_info *info, unsigned long *mount_opt,
637 unsigned long flags)
638{
639 bool ret = true;
640
641 if (!(flags & SB_RDONLY) &&
642 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
643 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
644 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums")))
645 ret = false;
646
647 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
648 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
649 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
650 btrfs_err(info, "cannot disable free-space-tree");
651 ret = false;
652 }
653 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
654 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
655 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
656 ret = false;
657 }
658
659 if (btrfs_check_mountopts_zoned(info, mount_opt))
660 ret = false;
661
662 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
663 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE))
664 btrfs_info(info, "disk space caching is enabled");
665 if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
666 btrfs_info(info, "using free-space-tree");
667 }
668
669 return ret;
670}
671
672/*
673 * This is subtle, we only call this during open_ctree(). We need to pre-load
674 * the mount options with the on-disk settings. Before the new mount API took
675 * effect we would do this on mount and remount. With the new mount API we'll
676 * only do this on the initial mount.
677 *
678 * This isn't a change in behavior, because we're using the current state of the
679 * file system to set the current mount options. If you mounted with special
680 * options to disable these features and then remounted we wouldn't revert the
681 * settings, because mounting without these features cleared the on-disk
682 * settings, so this being called on re-mount is not needed.
683 */
684void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
685{
686 if (fs_info->sectorsize < PAGE_SIZE) {
687 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
688 if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
689 btrfs_info(fs_info,
690 "forcing free space tree for sector size %u with page size %lu",
691 fs_info->sectorsize, PAGE_SIZE);
692 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
693 }
694 }
695
696 /*
697 * At this point our mount options are populated, so we only mess with
698 * these settings if we don't have any settings already.
699 */
700 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
701 return;
702
703 if (btrfs_is_zoned(fs_info) &&
704 btrfs_free_space_cache_v1_active(fs_info)) {
705 btrfs_info(fs_info, "zoned: clearing existing space cache");
706 btrfs_set_super_cache_generation(fs_info->super_copy, 0);
707 return;
708 }
709
710 if (btrfs_test_opt(fs_info, SPACE_CACHE))
711 return;
712
713 if (btrfs_test_opt(fs_info, NOSPACECACHE))
714 return;
715
716 /*
717 * At this point we don't have explicit options set by the user, set
718 * them ourselves based on the state of the file system.
719 */
720 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
721 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
722 else if (btrfs_free_space_cache_v1_active(fs_info))
723 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
724}
725
726static void set_device_specific_options(struct btrfs_fs_info *fs_info)
727{
728 if (!btrfs_test_opt(fs_info, NOSSD) &&
729 !fs_info->fs_devices->rotating)
730 btrfs_set_opt(fs_info->mount_opt, SSD);
731
732 /*
733 * For devices supporting discard turn on discard=async automatically,
734 * unless it's already set or disabled. This could be turned off by
735 * nodiscard for the same mount.
736 *
737 * The zoned mode piggy backs on the discard functionality for
738 * resetting a zone. There is no reason to delay the zone reset as it is
739 * fast enough. So, do not enable async discard for zoned mode.
740 */
741 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
742 btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
743 btrfs_test_opt(fs_info, NODISCARD)) &&
744 fs_info->fs_devices->discardable &&
745 !btrfs_is_zoned(fs_info))
746 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
747}
748
749char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
750 u64 subvol_objectid)
751{
752 struct btrfs_root *root = fs_info->tree_root;
753 struct btrfs_root *fs_root = NULL;
754 struct btrfs_root_ref *root_ref;
755 struct btrfs_inode_ref *inode_ref;
756 struct btrfs_key key;
757 struct btrfs_path *path = NULL;
758 char *name = NULL, *ptr;
759 u64 dirid;
760 int len;
761 int ret;
762
763 path = btrfs_alloc_path();
764 if (!path) {
765 ret = -ENOMEM;
766 goto err;
767 }
768
769 name = kmalloc(PATH_MAX, GFP_KERNEL);
770 if (!name) {
771 ret = -ENOMEM;
772 goto err;
773 }
774 ptr = name + PATH_MAX - 1;
775 ptr[0] = '\0';
776
777 /*
778 * Walk up the subvolume trees in the tree of tree roots by root
779 * backrefs until we hit the top-level subvolume.
780 */
781 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
782 key.objectid = subvol_objectid;
783 key.type = BTRFS_ROOT_BACKREF_KEY;
784 key.offset = (u64)-1;
785
786 ret = btrfs_search_backwards(root, &key, path);
787 if (ret < 0) {
788 goto err;
789 } else if (ret > 0) {
790 ret = -ENOENT;
791 goto err;
792 }
793
794 subvol_objectid = key.offset;
795
796 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
797 struct btrfs_root_ref);
798 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
799 ptr -= len + 1;
800 if (ptr < name) {
801 ret = -ENAMETOOLONG;
802 goto err;
803 }
804 read_extent_buffer(path->nodes[0], ptr + 1,
805 (unsigned long)(root_ref + 1), len);
806 ptr[0] = '/';
807 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
808 btrfs_release_path(path);
809
810 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
811 if (IS_ERR(fs_root)) {
812 ret = PTR_ERR(fs_root);
813 fs_root = NULL;
814 goto err;
815 }
816
817 /*
818 * Walk up the filesystem tree by inode refs until we hit the
819 * root directory.
820 */
821 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
822 key.objectid = dirid;
823 key.type = BTRFS_INODE_REF_KEY;
824 key.offset = (u64)-1;
825
826 ret = btrfs_search_backwards(fs_root, &key, path);
827 if (ret < 0) {
828 goto err;
829 } else if (ret > 0) {
830 ret = -ENOENT;
831 goto err;
832 }
833
834 dirid = key.offset;
835
836 inode_ref = btrfs_item_ptr(path->nodes[0],
837 path->slots[0],
838 struct btrfs_inode_ref);
839 len = btrfs_inode_ref_name_len(path->nodes[0],
840 inode_ref);
841 ptr -= len + 1;
842 if (ptr < name) {
843 ret = -ENAMETOOLONG;
844 goto err;
845 }
846 read_extent_buffer(path->nodes[0], ptr + 1,
847 (unsigned long)(inode_ref + 1), len);
848 ptr[0] = '/';
849 btrfs_release_path(path);
850 }
851 btrfs_put_root(fs_root);
852 fs_root = NULL;
853 }
854
855 btrfs_free_path(path);
856 if (ptr == name + PATH_MAX - 1) {
857 name[0] = '/';
858 name[1] = '\0';
859 } else {
860 memmove(name, ptr, name + PATH_MAX - ptr);
861 }
862 return name;
863
864err:
865 btrfs_put_root(fs_root);
866 btrfs_free_path(path);
867 kfree(name);
868 return ERR_PTR(ret);
869}
870
871static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
872{
873 struct btrfs_root *root = fs_info->tree_root;
874 struct btrfs_dir_item *di;
875 struct btrfs_path *path;
876 struct btrfs_key location;
877 struct fscrypt_str name = FSTR_INIT("default", 7);
878 u64 dir_id;
879
880 path = btrfs_alloc_path();
881 if (!path)
882 return -ENOMEM;
883
884 /*
885 * Find the "default" dir item which points to the root item that we
886 * will mount by default if we haven't been given a specific subvolume
887 * to mount.
888 */
889 dir_id = btrfs_super_root_dir(fs_info->super_copy);
890 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
891 if (IS_ERR(di)) {
892 btrfs_free_path(path);
893 return PTR_ERR(di);
894 }
895 if (!di) {
896 /*
897 * Ok the default dir item isn't there. This is weird since
898 * it's always been there, but don't freak out, just try and
899 * mount the top-level subvolume.
900 */
901 btrfs_free_path(path);
902 *objectid = BTRFS_FS_TREE_OBJECTID;
903 return 0;
904 }
905
906 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
907 btrfs_free_path(path);
908 *objectid = location.objectid;
909 return 0;
910}
911
912static int btrfs_fill_super(struct super_block *sb,
913 struct btrfs_fs_devices *fs_devices,
914 void *data)
915{
916 struct inode *inode;
917 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
918 int err;
919
920 sb->s_maxbytes = MAX_LFS_FILESIZE;
921 sb->s_magic = BTRFS_SUPER_MAGIC;
922 sb->s_op = &btrfs_super_ops;
923 sb->s_d_op = &btrfs_dentry_operations;
924 sb->s_export_op = &btrfs_export_ops;
925#ifdef CONFIG_FS_VERITY
926 sb->s_vop = &btrfs_verityops;
927#endif
928 sb->s_xattr = btrfs_xattr_handlers;
929 sb->s_time_gran = 1;
930 sb->s_iflags |= SB_I_CGROUPWB;
931
932 err = super_setup_bdi(sb);
933 if (err) {
934 btrfs_err(fs_info, "super_setup_bdi failed");
935 return err;
936 }
937
938 err = open_ctree(sb, fs_devices, (char *)data);
939 if (err) {
940 btrfs_err(fs_info, "open_ctree failed");
941 return err;
942 }
943
944 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
945 if (IS_ERR(inode)) {
946 err = PTR_ERR(inode);
947 btrfs_handle_fs_error(fs_info, err, NULL);
948 goto fail_close;
949 }
950
951 sb->s_root = d_make_root(inode);
952 if (!sb->s_root) {
953 err = -ENOMEM;
954 goto fail_close;
955 }
956
957 sb->s_flags |= SB_ACTIVE;
958 return 0;
959
960fail_close:
961 close_ctree(fs_info);
962 return err;
963}
964
965int btrfs_sync_fs(struct super_block *sb, int wait)
966{
967 struct btrfs_trans_handle *trans;
968 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
969 struct btrfs_root *root = fs_info->tree_root;
970
971 trace_btrfs_sync_fs(fs_info, wait);
972
973 if (!wait) {
974 filemap_flush(fs_info->btree_inode->i_mapping);
975 return 0;
976 }
977
978 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
979
980 trans = btrfs_attach_transaction_barrier(root);
981 if (IS_ERR(trans)) {
982 /* no transaction, don't bother */
983 if (PTR_ERR(trans) == -ENOENT) {
984 /*
985 * Exit unless we have some pending changes
986 * that need to go through commit
987 */
988 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
989 &fs_info->flags))
990 return 0;
991 /*
992 * A non-blocking test if the fs is frozen. We must not
993 * start a new transaction here otherwise a deadlock
994 * happens. The pending operations are delayed to the
995 * next commit after thawing.
996 */
997 if (sb_start_write_trylock(sb))
998 sb_end_write(sb);
999 else
1000 return 0;
1001 trans = btrfs_start_transaction(root, 0);
1002 }
1003 if (IS_ERR(trans))
1004 return PTR_ERR(trans);
1005 }
1006 return btrfs_commit_transaction(trans);
1007}
1008
1009static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1010{
1011 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1012 *printed = true;
1013}
1014
1015static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1016{
1017 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1018 const char *compress_type;
1019 const char *subvol_name;
1020 bool printed = false;
1021
1022 if (btrfs_test_opt(info, DEGRADED))
1023 seq_puts(seq, ",degraded");
1024 if (btrfs_test_opt(info, NODATASUM))
1025 seq_puts(seq, ",nodatasum");
1026 if (btrfs_test_opt(info, NODATACOW))
1027 seq_puts(seq, ",nodatacow");
1028 if (btrfs_test_opt(info, NOBARRIER))
1029 seq_puts(seq, ",nobarrier");
1030 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1031 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1032 if (info->thread_pool_size != min_t(unsigned long,
1033 num_online_cpus() + 2, 8))
1034 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1035 if (btrfs_test_opt(info, COMPRESS)) {
1036 compress_type = btrfs_compress_type2str(info->compress_type);
1037 if (btrfs_test_opt(info, FORCE_COMPRESS))
1038 seq_printf(seq, ",compress-force=%s", compress_type);
1039 else
1040 seq_printf(seq, ",compress=%s", compress_type);
1041 if (info->compress_level)
1042 seq_printf(seq, ":%d", info->compress_level);
1043 }
1044 if (btrfs_test_opt(info, NOSSD))
1045 seq_puts(seq, ",nossd");
1046 if (btrfs_test_opt(info, SSD_SPREAD))
1047 seq_puts(seq, ",ssd_spread");
1048 else if (btrfs_test_opt(info, SSD))
1049 seq_puts(seq, ",ssd");
1050 if (btrfs_test_opt(info, NOTREELOG))
1051 seq_puts(seq, ",notreelog");
1052 if (btrfs_test_opt(info, NOLOGREPLAY))
1053 print_rescue_option(seq, "nologreplay", &printed);
1054 if (btrfs_test_opt(info, USEBACKUPROOT))
1055 print_rescue_option(seq, "usebackuproot", &printed);
1056 if (btrfs_test_opt(info, IGNOREBADROOTS))
1057 print_rescue_option(seq, "ignorebadroots", &printed);
1058 if (btrfs_test_opt(info, IGNOREDATACSUMS))
1059 print_rescue_option(seq, "ignoredatacsums", &printed);
1060 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1061 seq_puts(seq, ",flushoncommit");
1062 if (btrfs_test_opt(info, DISCARD_SYNC))
1063 seq_puts(seq, ",discard");
1064 if (btrfs_test_opt(info, DISCARD_ASYNC))
1065 seq_puts(seq, ",discard=async");
1066 if (!(info->sb->s_flags & SB_POSIXACL))
1067 seq_puts(seq, ",noacl");
1068 if (btrfs_free_space_cache_v1_active(info))
1069 seq_puts(seq, ",space_cache");
1070 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1071 seq_puts(seq, ",space_cache=v2");
1072 else
1073 seq_puts(seq, ",nospace_cache");
1074 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1075 seq_puts(seq, ",rescan_uuid_tree");
1076 if (btrfs_test_opt(info, CLEAR_CACHE))
1077 seq_puts(seq, ",clear_cache");
1078 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1079 seq_puts(seq, ",user_subvol_rm_allowed");
1080 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1081 seq_puts(seq, ",enospc_debug");
1082 if (btrfs_test_opt(info, AUTO_DEFRAG))
1083 seq_puts(seq, ",autodefrag");
1084 if (btrfs_test_opt(info, SKIP_BALANCE))
1085 seq_puts(seq, ",skip_balance");
1086 if (info->metadata_ratio)
1087 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1088 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1089 seq_puts(seq, ",fatal_errors=panic");
1090 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1091 seq_printf(seq, ",commit=%u", info->commit_interval);
1092#ifdef CONFIG_BTRFS_DEBUG
1093 if (btrfs_test_opt(info, FRAGMENT_DATA))
1094 seq_puts(seq, ",fragment=data");
1095 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1096 seq_puts(seq, ",fragment=metadata");
1097#endif
1098 if (btrfs_test_opt(info, REF_VERIFY))
1099 seq_puts(seq, ",ref_verify");
1100 seq_printf(seq, ",subvolid=%llu",
1101 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1102 subvol_name = btrfs_get_subvol_name_from_objectid(info,
1103 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1104 if (!IS_ERR(subvol_name)) {
1105 seq_puts(seq, ",subvol=");
1106 seq_escape(seq, subvol_name, " \t\n\\");
1107 kfree(subvol_name);
1108 }
1109 return 0;
1110}
1111
1112/*
1113 * subvolumes are identified by ino 256
1114 */
1115static inline int is_subvolume_inode(struct inode *inode)
1116{
1117 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1118 return 1;
1119 return 0;
1120}
1121
1122static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1123 struct vfsmount *mnt)
1124{
1125 struct dentry *root;
1126 int ret;
1127
1128 if (!subvol_name) {
1129 if (!subvol_objectid) {
1130 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1131 &subvol_objectid);
1132 if (ret) {
1133 root = ERR_PTR(ret);
1134 goto out;
1135 }
1136 }
1137 subvol_name = btrfs_get_subvol_name_from_objectid(
1138 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1139 if (IS_ERR(subvol_name)) {
1140 root = ERR_CAST(subvol_name);
1141 subvol_name = NULL;
1142 goto out;
1143 }
1144
1145 }
1146
1147 root = mount_subtree(mnt, subvol_name);
1148 /* mount_subtree() drops our reference on the vfsmount. */
1149 mnt = NULL;
1150
1151 if (!IS_ERR(root)) {
1152 struct super_block *s = root->d_sb;
1153 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1154 struct inode *root_inode = d_inode(root);
1155 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1156
1157 ret = 0;
1158 if (!is_subvolume_inode(root_inode)) {
1159 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1160 subvol_name);
1161 ret = -EINVAL;
1162 }
1163 if (subvol_objectid && root_objectid != subvol_objectid) {
1164 /*
1165 * This will also catch a race condition where a
1166 * subvolume which was passed by ID is renamed and
1167 * another subvolume is renamed over the old location.
1168 */
1169 btrfs_err(fs_info,
1170 "subvol '%s' does not match subvolid %llu",
1171 subvol_name, subvol_objectid);
1172 ret = -EINVAL;
1173 }
1174 if (ret) {
1175 dput(root);
1176 root = ERR_PTR(ret);
1177 deactivate_locked_super(s);
1178 }
1179 }
1180
1181out:
1182 mntput(mnt);
1183 kfree(subvol_name);
1184 return root;
1185}
1186
1187static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1188 u32 new_pool_size, u32 old_pool_size)
1189{
1190 if (new_pool_size == old_pool_size)
1191 return;
1192
1193 fs_info->thread_pool_size = new_pool_size;
1194
1195 btrfs_info(fs_info, "resize thread pool %d -> %d",
1196 old_pool_size, new_pool_size);
1197
1198 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1199 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1200 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1201 workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1202 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1203 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1204 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1205 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1206}
1207
1208static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1209 unsigned long old_opts, int flags)
1210{
1211 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1212 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1213 (flags & SB_RDONLY))) {
1214 /* wait for any defraggers to finish */
1215 wait_event(fs_info->transaction_wait,
1216 (atomic_read(&fs_info->defrag_running) == 0));
1217 if (flags & SB_RDONLY)
1218 sync_filesystem(fs_info->sb);
1219 }
1220}
1221
1222static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1223 unsigned long old_opts)
1224{
1225 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1226
1227 /*
1228 * We need to cleanup all defragable inodes if the autodefragment is
1229 * close or the filesystem is read only.
1230 */
1231 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1232 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1233 btrfs_cleanup_defrag_inodes(fs_info);
1234 }
1235
1236 /* If we toggled discard async */
1237 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1238 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1239 btrfs_discard_resume(fs_info);
1240 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1241 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1242 btrfs_discard_cleanup(fs_info);
1243
1244 /* If we toggled space cache */
1245 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1246 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1247}
1248
1249static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1250{
1251 int ret;
1252
1253 if (BTRFS_FS_ERROR(fs_info)) {
1254 btrfs_err(fs_info,
1255 "remounting read-write after error is not allowed");
1256 return -EINVAL;
1257 }
1258
1259 if (fs_info->fs_devices->rw_devices == 0)
1260 return -EACCES;
1261
1262 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1263 btrfs_warn(fs_info,
1264 "too many missing devices, writable remount is not allowed");
1265 return -EACCES;
1266 }
1267
1268 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1269 btrfs_warn(fs_info,
1270 "mount required to replay tree-log, cannot remount read-write");
1271 return -EINVAL;
1272 }
1273
1274 /*
1275 * NOTE: when remounting with a change that does writes, don't put it
1276 * anywhere above this point, as we are not sure to be safe to write
1277 * until we pass the above checks.
1278 */
1279 ret = btrfs_start_pre_rw_mount(fs_info);
1280 if (ret)
1281 return ret;
1282
1283 btrfs_clear_sb_rdonly(fs_info->sb);
1284
1285 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1286
1287 /*
1288 * If we've gone from readonly -> read-write, we need to get our
1289 * sync/async discard lists in the right state.
1290 */
1291 btrfs_discard_resume(fs_info);
1292
1293 return 0;
1294}
1295
1296static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1297{
1298 /*
1299 * This also happens on 'umount -rf' or on shutdown, when the
1300 * filesystem is busy.
1301 */
1302 cancel_work_sync(&fs_info->async_reclaim_work);
1303 cancel_work_sync(&fs_info->async_data_reclaim_work);
1304
1305 btrfs_discard_cleanup(fs_info);
1306
1307 /* Wait for the uuid_scan task to finish */
1308 down(&fs_info->uuid_tree_rescan_sem);
1309 /* Avoid complains from lockdep et al. */
1310 up(&fs_info->uuid_tree_rescan_sem);
1311
1312 btrfs_set_sb_rdonly(fs_info->sb);
1313
1314 /*
1315 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1316 * loop if it's already active. If it's already asleep, we'll leave
1317 * unused block groups on disk until we're mounted read-write again
1318 * unless we clean them up here.
1319 */
1320 btrfs_delete_unused_bgs(fs_info);
1321
1322 /*
1323 * The cleaner task could be already running before we set the flag
1324 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1325 * sure that after we finish the remount, i.e. after we call
1326 * btrfs_commit_super(), the cleaner can no longer start a transaction
1327 * - either because it was dropping a dead root, running delayed iputs
1328 * or deleting an unused block group (the cleaner picked a block
1329 * group from the list of unused block groups before we were able to
1330 * in the previous call to btrfs_delete_unused_bgs()).
1331 */
1332 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1333
1334 /*
1335 * We've set the superblock to RO mode, so we might have made the
1336 * cleaner task sleep without running all pending delayed iputs. Go
1337 * through all the delayed iputs here, so that if an unmount happens
1338 * without remounting RW we don't end up at finishing close_ctree()
1339 * with a non-empty list of delayed iputs.
1340 */
1341 btrfs_run_delayed_iputs(fs_info);
1342
1343 btrfs_dev_replace_suspend_for_unmount(fs_info);
1344 btrfs_scrub_cancel(fs_info);
1345 btrfs_pause_balance(fs_info);
1346
1347 /*
1348 * Pause the qgroup rescan worker if it is running. We don't want it to
1349 * be still running after we are in RO mode, as after that, by the time
1350 * we unmount, it might have left a transaction open, so we would leak
1351 * the transaction and/or crash.
1352 */
1353 btrfs_qgroup_wait_for_completion(fs_info, false);
1354
1355 return btrfs_commit_super(fs_info);
1356}
1357
1358static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1359{
1360 fs_info->max_inline = ctx->max_inline;
1361 fs_info->commit_interval = ctx->commit_interval;
1362 fs_info->metadata_ratio = ctx->metadata_ratio;
1363 fs_info->thread_pool_size = ctx->thread_pool_size;
1364 fs_info->mount_opt = ctx->mount_opt;
1365 fs_info->compress_type = ctx->compress_type;
1366 fs_info->compress_level = ctx->compress_level;
1367}
1368
1369static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1370{
1371 ctx->max_inline = fs_info->max_inline;
1372 ctx->commit_interval = fs_info->commit_interval;
1373 ctx->metadata_ratio = fs_info->metadata_ratio;
1374 ctx->thread_pool_size = fs_info->thread_pool_size;
1375 ctx->mount_opt = fs_info->mount_opt;
1376 ctx->compress_type = fs_info->compress_type;
1377 ctx->compress_level = fs_info->compress_level;
1378}
1379
1380#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1381do { \
1382 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1383 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1384 btrfs_info(fs_info, fmt, ##args); \
1385} while (0)
1386
1387#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1388do { \
1389 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1390 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1391 btrfs_info(fs_info, fmt, ##args); \
1392} while (0)
1393
1394static void btrfs_emit_options(struct btrfs_fs_info *info,
1395 struct btrfs_fs_context *old)
1396{
1397 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1398 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1399 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1400 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1401 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1402 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1403 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1404 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1405 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1406 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1407 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1408 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1409 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1410 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1411 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1412 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1413 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1414 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1415 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1416 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1417 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1418
1419 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1420 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1421 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1422 btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1423 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1424 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1425 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1426 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1427 btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1428
1429 /* Did the compression settings change? */
1430 if (btrfs_test_opt(info, COMPRESS) &&
1431 (!old ||
1432 old->compress_type != info->compress_type ||
1433 old->compress_level != info->compress_level ||
1434 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1435 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1436 const char *compress_type = btrfs_compress_type2str(info->compress_type);
1437
1438 btrfs_info(info, "%s %s compression, level %d",
1439 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1440 compress_type, info->compress_level);
1441 }
1442
1443 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1444 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1445}
1446
1447static int btrfs_reconfigure(struct fs_context *fc)
1448{
1449 struct super_block *sb = fc->root->d_sb;
1450 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1451 struct btrfs_fs_context *ctx = fc->fs_private;
1452 struct btrfs_fs_context old_ctx;
1453 int ret = 0;
1454 bool mount_reconfigure = (fc->s_fs_info != NULL);
1455
1456 btrfs_info_to_ctx(fs_info, &old_ctx);
1457
1458 /*
1459 * This is our "bind mount" trick, we don't want to allow the user to do
1460 * anything other than mount a different ro/rw and a different subvol,
1461 * all of the mount options should be maintained.
1462 */
1463 if (mount_reconfigure)
1464 ctx->mount_opt = old_ctx.mount_opt;
1465
1466 sync_filesystem(sb);
1467 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1468
1469 if (!mount_reconfigure &&
1470 !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1471 return -EINVAL;
1472
1473 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1474 if (ret < 0)
1475 return ret;
1476
1477 btrfs_ctx_to_info(fs_info, ctx);
1478 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1479 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1480 old_ctx.thread_pool_size);
1481
1482 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1483 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1484 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1485 btrfs_warn(fs_info,
1486 "remount supports changing free space tree only from RO to RW");
1487 /* Make sure free space cache options match the state on disk. */
1488 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1489 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1490 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1491 }
1492 if (btrfs_free_space_cache_v1_active(fs_info)) {
1493 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1494 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1495 }
1496 }
1497
1498 ret = 0;
1499 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1500 ret = btrfs_remount_ro(fs_info);
1501 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1502 ret = btrfs_remount_rw(fs_info);
1503 if (ret)
1504 goto restore;
1505
1506 /*
1507 * If we set the mask during the parameter parsing VFS would reject the
1508 * remount. Here we can set the mask and the value will be updated
1509 * appropriately.
1510 */
1511 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1512 fc->sb_flags_mask |= SB_POSIXACL;
1513
1514 btrfs_emit_options(fs_info, &old_ctx);
1515 wake_up_process(fs_info->transaction_kthread);
1516 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1517 btrfs_clear_oneshot_options(fs_info);
1518 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1519
1520 return 0;
1521restore:
1522 btrfs_ctx_to_info(fs_info, &old_ctx);
1523 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1524 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1525 return ret;
1526}
1527
1528/* Used to sort the devices by max_avail(descending sort) */
1529static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1530{
1531 const struct btrfs_device_info *dev_info1 = a;
1532 const struct btrfs_device_info *dev_info2 = b;
1533
1534 if (dev_info1->max_avail > dev_info2->max_avail)
1535 return -1;
1536 else if (dev_info1->max_avail < dev_info2->max_avail)
1537 return 1;
1538 return 0;
1539}
1540
1541/*
1542 * sort the devices by max_avail, in which max free extent size of each device
1543 * is stored.(Descending Sort)
1544 */
1545static inline void btrfs_descending_sort_devices(
1546 struct btrfs_device_info *devices,
1547 size_t nr_devices)
1548{
1549 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1550 btrfs_cmp_device_free_bytes, NULL);
1551}
1552
1553/*
1554 * The helper to calc the free space on the devices that can be used to store
1555 * file data.
1556 */
1557static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1558 u64 *free_bytes)
1559{
1560 struct btrfs_device_info *devices_info;
1561 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1562 struct btrfs_device *device;
1563 u64 type;
1564 u64 avail_space;
1565 u64 min_stripe_size;
1566 int num_stripes = 1;
1567 int i = 0, nr_devices;
1568 const struct btrfs_raid_attr *rattr;
1569
1570 /*
1571 * We aren't under the device list lock, so this is racy-ish, but good
1572 * enough for our purposes.
1573 */
1574 nr_devices = fs_info->fs_devices->open_devices;
1575 if (!nr_devices) {
1576 smp_mb();
1577 nr_devices = fs_info->fs_devices->open_devices;
1578 ASSERT(nr_devices);
1579 if (!nr_devices) {
1580 *free_bytes = 0;
1581 return 0;
1582 }
1583 }
1584
1585 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1586 GFP_KERNEL);
1587 if (!devices_info)
1588 return -ENOMEM;
1589
1590 /* calc min stripe number for data space allocation */
1591 type = btrfs_data_alloc_profile(fs_info);
1592 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1593
1594 if (type & BTRFS_BLOCK_GROUP_RAID0)
1595 num_stripes = nr_devices;
1596 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1597 num_stripes = rattr->ncopies;
1598 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1599 num_stripes = 4;
1600
1601 /* Adjust for more than 1 stripe per device */
1602 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1603
1604 rcu_read_lock();
1605 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1606 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1607 &device->dev_state) ||
1608 !device->bdev ||
1609 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1610 continue;
1611
1612 if (i >= nr_devices)
1613 break;
1614
1615 avail_space = device->total_bytes - device->bytes_used;
1616
1617 /* align with stripe_len */
1618 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1619
1620 /*
1621 * Ensure we have at least min_stripe_size on top of the
1622 * reserved space on the device.
1623 */
1624 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1625 continue;
1626
1627 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1628
1629 devices_info[i].dev = device;
1630 devices_info[i].max_avail = avail_space;
1631
1632 i++;
1633 }
1634 rcu_read_unlock();
1635
1636 nr_devices = i;
1637
1638 btrfs_descending_sort_devices(devices_info, nr_devices);
1639
1640 i = nr_devices - 1;
1641 avail_space = 0;
1642 while (nr_devices >= rattr->devs_min) {
1643 num_stripes = min(num_stripes, nr_devices);
1644
1645 if (devices_info[i].max_avail >= min_stripe_size) {
1646 int j;
1647 u64 alloc_size;
1648
1649 avail_space += devices_info[i].max_avail * num_stripes;
1650 alloc_size = devices_info[i].max_avail;
1651 for (j = i + 1 - num_stripes; j <= i; j++)
1652 devices_info[j].max_avail -= alloc_size;
1653 }
1654 i--;
1655 nr_devices--;
1656 }
1657
1658 kfree(devices_info);
1659 *free_bytes = avail_space;
1660 return 0;
1661}
1662
1663/*
1664 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1665 *
1666 * If there's a redundant raid level at DATA block groups, use the respective
1667 * multiplier to scale the sizes.
1668 *
1669 * Unused device space usage is based on simulating the chunk allocator
1670 * algorithm that respects the device sizes and order of allocations. This is
1671 * a close approximation of the actual use but there are other factors that may
1672 * change the result (like a new metadata chunk).
1673 *
1674 * If metadata is exhausted, f_bavail will be 0.
1675 */
1676static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1677{
1678 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1679 struct btrfs_super_block *disk_super = fs_info->super_copy;
1680 struct btrfs_space_info *found;
1681 u64 total_used = 0;
1682 u64 total_free_data = 0;
1683 u64 total_free_meta = 0;
1684 u32 bits = fs_info->sectorsize_bits;
1685 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1686 unsigned factor = 1;
1687 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1688 int ret;
1689 u64 thresh = 0;
1690 int mixed = 0;
1691
1692 list_for_each_entry(found, &fs_info->space_info, list) {
1693 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1694 int i;
1695
1696 total_free_data += found->disk_total - found->disk_used;
1697 total_free_data -=
1698 btrfs_account_ro_block_groups_free_space(found);
1699
1700 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1701 if (!list_empty(&found->block_groups[i]))
1702 factor = btrfs_bg_type_to_factor(
1703 btrfs_raid_array[i].bg_flag);
1704 }
1705 }
1706
1707 /*
1708 * Metadata in mixed block group profiles are accounted in data
1709 */
1710 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1711 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1712 mixed = 1;
1713 else
1714 total_free_meta += found->disk_total -
1715 found->disk_used;
1716 }
1717
1718 total_used += found->disk_used;
1719 }
1720
1721 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1722 buf->f_blocks >>= bits;
1723 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1724
1725 /* Account global block reserve as used, it's in logical size already */
1726 spin_lock(&block_rsv->lock);
1727 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
1728 if (buf->f_bfree >= block_rsv->size >> bits)
1729 buf->f_bfree -= block_rsv->size >> bits;
1730 else
1731 buf->f_bfree = 0;
1732 spin_unlock(&block_rsv->lock);
1733
1734 buf->f_bavail = div_u64(total_free_data, factor);
1735 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1736 if (ret)
1737 return ret;
1738 buf->f_bavail += div_u64(total_free_data, factor);
1739 buf->f_bavail = buf->f_bavail >> bits;
1740
1741 /*
1742 * We calculate the remaining metadata space minus global reserve. If
1743 * this is (supposedly) smaller than zero, there's no space. But this
1744 * does not hold in practice, the exhausted state happens where's still
1745 * some positive delta. So we apply some guesswork and compare the
1746 * delta to a 4M threshold. (Practically observed delta was ~2M.)
1747 *
1748 * We probably cannot calculate the exact threshold value because this
1749 * depends on the internal reservations requested by various
1750 * operations, so some operations that consume a few metadata will
1751 * succeed even if the Avail is zero. But this is better than the other
1752 * way around.
1753 */
1754 thresh = SZ_4M;
1755
1756 /*
1757 * We only want to claim there's no available space if we can no longer
1758 * allocate chunks for our metadata profile and our global reserve will
1759 * not fit in the free metadata space. If we aren't ->full then we
1760 * still can allocate chunks and thus are fine using the currently
1761 * calculated f_bavail.
1762 */
1763 if (!mixed && block_rsv->space_info->full &&
1764 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1765 buf->f_bavail = 0;
1766
1767 buf->f_type = BTRFS_SUPER_MAGIC;
1768 buf->f_bsize = fs_info->sectorsize;
1769 buf->f_namelen = BTRFS_NAME_LEN;
1770
1771 /* We treat it as constant endianness (it doesn't matter _which_)
1772 because we want the fsid to come out the same whether mounted
1773 on a big-endian or little-endian host */
1774 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1775 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1776 /* Mask in the root object ID too, to disambiguate subvols */
1777 buf->f_fsid.val[0] ^=
1778 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
1779 buf->f_fsid.val[1] ^=
1780 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
1781
1782 return 0;
1783}
1784
1785static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1786{
1787 struct btrfs_fs_info *p = fc->s_fs_info;
1788 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1789
1790 return fs_info->fs_devices == p->fs_devices;
1791}
1792
1793static int btrfs_get_tree_super(struct fs_context *fc)
1794{
1795 struct btrfs_fs_info *fs_info = fc->s_fs_info;
1796 struct btrfs_fs_context *ctx = fc->fs_private;
1797 struct btrfs_fs_devices *fs_devices = NULL;
1798 struct block_device *bdev;
1799 struct btrfs_device *device;
1800 struct super_block *sb;
1801 blk_mode_t mode = btrfs_open_mode(fc);
1802 int ret;
1803
1804 btrfs_ctx_to_info(fs_info, ctx);
1805 mutex_lock(&uuid_mutex);
1806
1807 /*
1808 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1809 * either a valid device or an error.
1810 */
1811 device = btrfs_scan_one_device(fc->source, mode, true);
1812 ASSERT(device != NULL);
1813 if (IS_ERR(device)) {
1814 mutex_unlock(&uuid_mutex);
1815 return PTR_ERR(device);
1816 }
1817
1818 fs_devices = device->fs_devices;
1819 fs_info->fs_devices = fs_devices;
1820
1821 ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1822 mutex_unlock(&uuid_mutex);
1823 if (ret)
1824 return ret;
1825
1826 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1827 ret = -EACCES;
1828 goto error;
1829 }
1830
1831 bdev = fs_devices->latest_dev->bdev;
1832
1833 /*
1834 * From now on the error handling is not straightforward.
1835 *
1836 * If successful, this will transfer the fs_info into the super block,
1837 * and fc->s_fs_info will be NULL. However if there's an existing
1838 * super, we'll still have fc->s_fs_info populated. If we error
1839 * completely out it'll be cleaned up when we drop the fs_context,
1840 * otherwise it's tied to the lifetime of the super_block.
1841 */
1842 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1843 if (IS_ERR(sb)) {
1844 ret = PTR_ERR(sb);
1845 goto error;
1846 }
1847
1848 set_device_specific_options(fs_info);
1849
1850 if (sb->s_root) {
1851 btrfs_close_devices(fs_devices);
1852 if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1853 ret = -EBUSY;
1854 } else {
1855 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1856 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1857 btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1858 ret = btrfs_fill_super(sb, fs_devices, NULL);
1859 }
1860
1861 if (ret) {
1862 deactivate_locked_super(sb);
1863 return ret;
1864 }
1865
1866 btrfs_clear_oneshot_options(fs_info);
1867
1868 fc->root = dget(sb->s_root);
1869 return 0;
1870
1871error:
1872 btrfs_close_devices(fs_devices);
1873 return ret;
1874}
1875
1876/*
1877 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1878 * with different ro/rw options") the following works:
1879 *
1880 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1881 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1882 *
1883 * which looks nice and innocent but is actually pretty intricate and deserves
1884 * a long comment.
1885 *
1886 * On another filesystem a subvolume mount is close to something like:
1887 *
1888 * (iii) # create rw superblock + initial mount
1889 * mount -t xfs /dev/sdb /opt/
1890 *
1891 * # create ro bind mount
1892 * mount --bind -o ro /opt/foo /mnt/foo
1893 *
1894 * # unmount initial mount
1895 * umount /opt
1896 *
1897 * Of course, there's some special subvolume sauce and there's the fact that the
1898 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1899 * it's very close and will help us understand the issue.
1900 *
1901 * The old mount API didn't cleanly distinguish between a mount being made ro
1902 * and a superblock being made ro. The only way to change the ro state of
1903 * either object was by passing ms_rdonly. If a new mount was created via
1904 * mount(2) such as:
1905 *
1906 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1907 *
1908 * the MS_RDONLY flag being specified had two effects:
1909 *
1910 * (1) MNT_READONLY was raised -> the resulting mount got
1911 * @mnt->mnt_flags |= MNT_READONLY raised.
1912 *
1913 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1914 * made the superblock ro. Note, how SB_RDONLY has the same value as
1915 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1916 *
1917 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1918 * subtree mounted ro.
1919 *
1920 * But consider the effect on the old mount API on btrfs subvolume mounting
1921 * which combines the distinct step in (iii) into a single step.
1922 *
1923 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1924 * is issued the superblock is ro and thus even if the mount created for (ii) is
1925 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1926 * to rw for (ii) which it did using an internal remount call.
1927 *
1928 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1929 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1930 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1931 * passed by mount(8) to mount(2).
1932 *
1933 * Enter the new mount API. The new mount API disambiguates making a mount ro
1934 * and making a superblock ro.
1935 *
1936 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1937 * fsmount() or mount_setattr() this is a pure VFS level change for a
1938 * specific mount or mount tree that is never seen by the filesystem itself.
1939 *
1940 * (4) To turn a superblock ro the "ro" flag must be used with
1941 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1942 * in fc->sb_flags.
1943 *
1944 * This disambiguation has rather positive consequences. Mounting a subvolume
1945 * ro will not also turn the superblock ro. Only the mount for the subvolume
1946 * will become ro.
1947 *
1948 * So, if the superblock creation request comes from the new mount API the
1949 * caller must have explicitly done:
1950 *
1951 * fsconfig(FSCONFIG_SET_FLAG, "ro")
1952 * fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1953 *
1954 * IOW, at some point the caller must have explicitly turned the whole
1955 * superblock ro and we shouldn't just undo it like we did for the old mount
1956 * API. In any case, it lets us avoid the hack in the new mount API.
1957 *
1958 * Consequently, the remounting hack must only be used for requests originating
1959 * from the old mount API and should be marked for full deprecation so it can be
1960 * turned off in a couple of years.
1961 *
1962 * The new mount API has no reason to support this hack.
1963 */
1964static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1965{
1966 struct vfsmount *mnt;
1967 int ret;
1968 const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
1969
1970 /*
1971 * We got an EBUSY because our SB_RDONLY flag didn't match the existing
1972 * super block, so invert our setting here and retry the mount so we
1973 * can get our vfsmount.
1974 */
1975 if (ro2rw)
1976 fc->sb_flags |= SB_RDONLY;
1977 else
1978 fc->sb_flags &= ~SB_RDONLY;
1979
1980 mnt = fc_mount(fc);
1981 if (IS_ERR(mnt))
1982 return mnt;
1983
1984 if (!fc->oldapi || !ro2rw)
1985 return mnt;
1986
1987 /* We need to convert to rw, call reconfigure. */
1988 fc->sb_flags &= ~SB_RDONLY;
1989 down_write(&mnt->mnt_sb->s_umount);
1990 ret = btrfs_reconfigure(fc);
1991 up_write(&mnt->mnt_sb->s_umount);
1992 if (ret) {
1993 mntput(mnt);
1994 return ERR_PTR(ret);
1995 }
1996 return mnt;
1997}
1998
1999static int btrfs_get_tree_subvol(struct fs_context *fc)
2000{
2001 struct btrfs_fs_info *fs_info = NULL;
2002 struct btrfs_fs_context *ctx = fc->fs_private;
2003 struct fs_context *dup_fc;
2004 struct dentry *dentry;
2005 struct vfsmount *mnt;
2006
2007 /*
2008 * Setup a dummy root and fs_info for test/set super. This is because
2009 * we don't actually fill this stuff out until open_ctree, but we need
2010 * then open_ctree will properly initialize the file system specific
2011 * settings later. btrfs_init_fs_info initializes the static elements
2012 * of the fs_info (locks and such) to make cleanup easier if we find a
2013 * superblock with our given fs_devices later on at sget() time.
2014 */
2015 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2016 if (!fs_info)
2017 return -ENOMEM;
2018
2019 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2020 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2021 if (!fs_info->super_copy || !fs_info->super_for_commit) {
2022 btrfs_free_fs_info(fs_info);
2023 return -ENOMEM;
2024 }
2025 btrfs_init_fs_info(fs_info);
2026
2027 dup_fc = vfs_dup_fs_context(fc);
2028 if (IS_ERR(dup_fc)) {
2029 btrfs_free_fs_info(fs_info);
2030 return PTR_ERR(dup_fc);
2031 }
2032
2033 /*
2034 * When we do the sget_fc this gets transferred to the sb, so we only
2035 * need to set it on the dup_fc as this is what creates the super block.
2036 */
2037 dup_fc->s_fs_info = fs_info;
2038
2039 /*
2040 * We'll do the security settings in our btrfs_get_tree_super() mount
2041 * loop, they were duplicated into dup_fc, we can drop the originals
2042 * here.
2043 */
2044 security_free_mnt_opts(&fc->security);
2045 fc->security = NULL;
2046
2047 mnt = fc_mount(dup_fc);
2048 if (PTR_ERR_OR_ZERO(mnt) == -EBUSY)
2049 mnt = btrfs_reconfigure_for_mount(dup_fc);
2050 put_fs_context(dup_fc);
2051 if (IS_ERR(mnt))
2052 return PTR_ERR(mnt);
2053
2054 /*
2055 * This free's ->subvol_name, because if it isn't set we have to
2056 * allocate a buffer to hold the subvol_name, so we just drop our
2057 * reference to it here.
2058 */
2059 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2060 ctx->subvol_name = NULL;
2061 if (IS_ERR(dentry))
2062 return PTR_ERR(dentry);
2063
2064 fc->root = dentry;
2065 return 0;
2066}
2067
2068static int btrfs_get_tree(struct fs_context *fc)
2069{
2070 /*
2071 * Since we use mount_subtree to mount the default/specified subvol, we
2072 * have to do mounts in two steps.
2073 *
2074 * First pass through we call btrfs_get_tree_subvol(), this is just a
2075 * wrapper around fc_mount() to call back into here again, and this time
2076 * we'll call btrfs_get_tree_super(). This will do the open_ctree() and
2077 * everything to open the devices and file system. Then we return back
2078 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2079 * from there we can do our mount_subvol() call, which will lookup
2080 * whichever subvol we're mounting and setup this fc with the
2081 * appropriate dentry for the subvol.
2082 */
2083 if (fc->s_fs_info)
2084 return btrfs_get_tree_super(fc);
2085 return btrfs_get_tree_subvol(fc);
2086}
2087
2088static void btrfs_kill_super(struct super_block *sb)
2089{
2090 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2091 kill_anon_super(sb);
2092 btrfs_free_fs_info(fs_info);
2093}
2094
2095static void btrfs_free_fs_context(struct fs_context *fc)
2096{
2097 struct btrfs_fs_context *ctx = fc->fs_private;
2098 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2099
2100 if (fs_info)
2101 btrfs_free_fs_info(fs_info);
2102
2103 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2104 kfree(ctx->subvol_name);
2105 kfree(ctx);
2106 }
2107}
2108
2109static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2110{
2111 struct btrfs_fs_context *ctx = src_fc->fs_private;
2112
2113 /*
2114 * Give a ref to our ctx to this dup, as we want to keep it around for
2115 * our original fc so we can have the subvolume name or objectid.
2116 *
2117 * We unset ->source in the original fc because the dup needs it for
2118 * mounting, and then once we free the dup it'll free ->source, so we
2119 * need to make sure we're only pointing to it in one fc.
2120 */
2121 refcount_inc(&ctx->refs);
2122 fc->fs_private = ctx;
2123 fc->source = src_fc->source;
2124 src_fc->source = NULL;
2125 return 0;
2126}
2127
2128static const struct fs_context_operations btrfs_fs_context_ops = {
2129 .parse_param = btrfs_parse_param,
2130 .reconfigure = btrfs_reconfigure,
2131 .get_tree = btrfs_get_tree,
2132 .dup = btrfs_dup_fs_context,
2133 .free = btrfs_free_fs_context,
2134};
2135
2136static int btrfs_init_fs_context(struct fs_context *fc)
2137{
2138 struct btrfs_fs_context *ctx;
2139
2140 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2141 if (!ctx)
2142 return -ENOMEM;
2143
2144 refcount_set(&ctx->refs, 1);
2145 fc->fs_private = ctx;
2146 fc->ops = &btrfs_fs_context_ops;
2147
2148 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2149 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2150 } else {
2151 ctx->thread_pool_size =
2152 min_t(unsigned long, num_online_cpus() + 2, 8);
2153 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2154 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2155 }
2156
2157#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2158 fc->sb_flags |= SB_POSIXACL;
2159#endif
2160 fc->sb_flags |= SB_I_VERSION;
2161
2162 return 0;
2163}
2164
2165static struct file_system_type btrfs_fs_type = {
2166 .owner = THIS_MODULE,
2167 .name = "btrfs",
2168 .init_fs_context = btrfs_init_fs_context,
2169 .parameters = btrfs_fs_parameters,
2170 .kill_sb = btrfs_kill_super,
2171 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2172 };
2173
2174MODULE_ALIAS_FS("btrfs");
2175
2176static int btrfs_control_open(struct inode *inode, struct file *file)
2177{
2178 /*
2179 * The control file's private_data is used to hold the
2180 * transaction when it is started and is used to keep
2181 * track of whether a transaction is already in progress.
2182 */
2183 file->private_data = NULL;
2184 return 0;
2185}
2186
2187/*
2188 * Used by /dev/btrfs-control for devices ioctls.
2189 */
2190static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2191 unsigned long arg)
2192{
2193 struct btrfs_ioctl_vol_args *vol;
2194 struct btrfs_device *device = NULL;
2195 dev_t devt = 0;
2196 int ret = -ENOTTY;
2197
2198 if (!capable(CAP_SYS_ADMIN))
2199 return -EPERM;
2200
2201 vol = memdup_user((void __user *)arg, sizeof(*vol));
2202 if (IS_ERR(vol))
2203 return PTR_ERR(vol);
2204 ret = btrfs_check_ioctl_vol_args_path(vol);
2205 if (ret < 0)
2206 goto out;
2207
2208 switch (cmd) {
2209 case BTRFS_IOC_SCAN_DEV:
2210 mutex_lock(&uuid_mutex);
2211 /*
2212 * Scanning outside of mount can return NULL which would turn
2213 * into 0 error code.
2214 */
2215 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2216 ret = PTR_ERR_OR_ZERO(device);
2217 mutex_unlock(&uuid_mutex);
2218 break;
2219 case BTRFS_IOC_FORGET_DEV:
2220 if (vol->name[0] != 0) {
2221 ret = lookup_bdev(vol->name, &devt);
2222 if (ret)
2223 break;
2224 }
2225 ret = btrfs_forget_devices(devt);
2226 break;
2227 case BTRFS_IOC_DEVICES_READY:
2228 mutex_lock(&uuid_mutex);
2229 /*
2230 * Scanning outside of mount can return NULL which would turn
2231 * into 0 error code.
2232 */
2233 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2234 if (IS_ERR_OR_NULL(device)) {
2235 mutex_unlock(&uuid_mutex);
2236 ret = PTR_ERR(device);
2237 break;
2238 }
2239 ret = !(device->fs_devices->num_devices ==
2240 device->fs_devices->total_devices);
2241 mutex_unlock(&uuid_mutex);
2242 break;
2243 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2244 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2245 break;
2246 }
2247
2248out:
2249 kfree(vol);
2250 return ret;
2251}
2252
2253static int btrfs_freeze(struct super_block *sb)
2254{
2255 struct btrfs_trans_handle *trans;
2256 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2257 struct btrfs_root *root = fs_info->tree_root;
2258
2259 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2260 /*
2261 * We don't need a barrier here, we'll wait for any transaction that
2262 * could be in progress on other threads (and do delayed iputs that
2263 * we want to avoid on a frozen filesystem), or do the commit
2264 * ourselves.
2265 */
2266 trans = btrfs_attach_transaction_barrier(root);
2267 if (IS_ERR(trans)) {
2268 /* no transaction, don't bother */
2269 if (PTR_ERR(trans) == -ENOENT)
2270 return 0;
2271 return PTR_ERR(trans);
2272 }
2273 return btrfs_commit_transaction(trans);
2274}
2275
2276static int check_dev_super(struct btrfs_device *dev)
2277{
2278 struct btrfs_fs_info *fs_info = dev->fs_info;
2279 struct btrfs_super_block *sb;
2280 u64 last_trans;
2281 u16 csum_type;
2282 int ret = 0;
2283
2284 /* This should be called with fs still frozen. */
2285 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2286
2287 /* Missing dev, no need to check. */
2288 if (!dev->bdev)
2289 return 0;
2290
2291 /* Only need to check the primary super block. */
2292 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2293 if (IS_ERR(sb))
2294 return PTR_ERR(sb);
2295
2296 /* Verify the checksum. */
2297 csum_type = btrfs_super_csum_type(sb);
2298 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2299 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2300 csum_type, btrfs_super_csum_type(fs_info->super_copy));
2301 ret = -EUCLEAN;
2302 goto out;
2303 }
2304
2305 if (btrfs_check_super_csum(fs_info, sb)) {
2306 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2307 ret = -EUCLEAN;
2308 goto out;
2309 }
2310
2311 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2312 ret = btrfs_validate_super(fs_info, sb, 0);
2313 if (ret < 0)
2314 goto out;
2315
2316 last_trans = btrfs_get_last_trans_committed(fs_info);
2317 if (btrfs_super_generation(sb) != last_trans) {
2318 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2319 btrfs_super_generation(sb), last_trans);
2320 ret = -EUCLEAN;
2321 goto out;
2322 }
2323out:
2324 btrfs_release_disk_super(sb);
2325 return ret;
2326}
2327
2328static int btrfs_unfreeze(struct super_block *sb)
2329{
2330 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2331 struct btrfs_device *device;
2332 int ret = 0;
2333
2334 /*
2335 * Make sure the fs is not changed by accident (like hibernation then
2336 * modified by other OS).
2337 * If we found anything wrong, we mark the fs error immediately.
2338 *
2339 * And since the fs is frozen, no one can modify the fs yet, thus
2340 * we don't need to hold device_list_mutex.
2341 */
2342 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2343 ret = check_dev_super(device);
2344 if (ret < 0) {
2345 btrfs_handle_fs_error(fs_info, ret,
2346 "super block on devid %llu got modified unexpectedly",
2347 device->devid);
2348 break;
2349 }
2350 }
2351 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2352
2353 /*
2354 * We still return 0, to allow VFS layer to unfreeze the fs even the
2355 * above checks failed. Since the fs is either fine or read-only, we're
2356 * safe to continue, without causing further damage.
2357 */
2358 return 0;
2359}
2360
2361static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2362{
2363 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2364
2365 /*
2366 * There should be always a valid pointer in latest_dev, it may be stale
2367 * for a short moment in case it's being deleted but still valid until
2368 * the end of RCU grace period.
2369 */
2370 rcu_read_lock();
2371 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2372 rcu_read_unlock();
2373
2374 return 0;
2375}
2376
2377static const struct super_operations btrfs_super_ops = {
2378 .drop_inode = btrfs_drop_inode,
2379 .evict_inode = btrfs_evict_inode,
2380 .put_super = btrfs_put_super,
2381 .sync_fs = btrfs_sync_fs,
2382 .show_options = btrfs_show_options,
2383 .show_devname = btrfs_show_devname,
2384 .alloc_inode = btrfs_alloc_inode,
2385 .destroy_inode = btrfs_destroy_inode,
2386 .free_inode = btrfs_free_inode,
2387 .statfs = btrfs_statfs,
2388 .freeze_fs = btrfs_freeze,
2389 .unfreeze_fs = btrfs_unfreeze,
2390};
2391
2392static const struct file_operations btrfs_ctl_fops = {
2393 .open = btrfs_control_open,
2394 .unlocked_ioctl = btrfs_control_ioctl,
2395 .compat_ioctl = compat_ptr_ioctl,
2396 .owner = THIS_MODULE,
2397 .llseek = noop_llseek,
2398};
2399
2400static struct miscdevice btrfs_misc = {
2401 .minor = BTRFS_MINOR,
2402 .name = "btrfs-control",
2403 .fops = &btrfs_ctl_fops
2404};
2405
2406MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2407MODULE_ALIAS("devname:btrfs-control");
2408
2409static int __init btrfs_interface_init(void)
2410{
2411 return misc_register(&btrfs_misc);
2412}
2413
2414static __cold void btrfs_interface_exit(void)
2415{
2416 misc_deregister(&btrfs_misc);
2417}
2418
2419static int __init btrfs_print_mod_info(void)
2420{
2421 static const char options[] = ""
2422#ifdef CONFIG_BTRFS_DEBUG
2423 ", debug=on"
2424#endif
2425#ifdef CONFIG_BTRFS_ASSERT
2426 ", assert=on"
2427#endif
2428#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2429 ", ref-verify=on"
2430#endif
2431#ifdef CONFIG_BLK_DEV_ZONED
2432 ", zoned=yes"
2433#else
2434 ", zoned=no"
2435#endif
2436#ifdef CONFIG_FS_VERITY
2437 ", fsverity=yes"
2438#else
2439 ", fsverity=no"
2440#endif
2441 ;
2442 pr_info("Btrfs loaded%s\n", options);
2443 return 0;
2444}
2445
2446static int register_btrfs(void)
2447{
2448 return register_filesystem(&btrfs_fs_type);
2449}
2450
2451static void unregister_btrfs(void)
2452{
2453 unregister_filesystem(&btrfs_fs_type);
2454}
2455
2456/* Helper structure for long init/exit functions. */
2457struct init_sequence {
2458 int (*init_func)(void);
2459 /* Can be NULL if the init_func doesn't need cleanup. */
2460 void (*exit_func)(void);
2461};
2462
2463static const struct init_sequence mod_init_seq[] = {
2464 {
2465 .init_func = btrfs_props_init,
2466 .exit_func = NULL,
2467 }, {
2468 .init_func = btrfs_init_sysfs,
2469 .exit_func = btrfs_exit_sysfs,
2470 }, {
2471 .init_func = btrfs_init_compress,
2472 .exit_func = btrfs_exit_compress,
2473 }, {
2474 .init_func = btrfs_init_cachep,
2475 .exit_func = btrfs_destroy_cachep,
2476 }, {
2477 .init_func = btrfs_transaction_init,
2478 .exit_func = btrfs_transaction_exit,
2479 }, {
2480 .init_func = btrfs_ctree_init,
2481 .exit_func = btrfs_ctree_exit,
2482 }, {
2483 .init_func = btrfs_free_space_init,
2484 .exit_func = btrfs_free_space_exit,
2485 }, {
2486 .init_func = extent_state_init_cachep,
2487 .exit_func = extent_state_free_cachep,
2488 }, {
2489 .init_func = extent_buffer_init_cachep,
2490 .exit_func = extent_buffer_free_cachep,
2491 }, {
2492 .init_func = btrfs_bioset_init,
2493 .exit_func = btrfs_bioset_exit,
2494 }, {
2495 .init_func = extent_map_init,
2496 .exit_func = extent_map_exit,
2497 }, {
2498 .init_func = ordered_data_init,
2499 .exit_func = ordered_data_exit,
2500 }, {
2501 .init_func = btrfs_delayed_inode_init,
2502 .exit_func = btrfs_delayed_inode_exit,
2503 }, {
2504 .init_func = btrfs_auto_defrag_init,
2505 .exit_func = btrfs_auto_defrag_exit,
2506 }, {
2507 .init_func = btrfs_delayed_ref_init,
2508 .exit_func = btrfs_delayed_ref_exit,
2509 }, {
2510 .init_func = btrfs_prelim_ref_init,
2511 .exit_func = btrfs_prelim_ref_exit,
2512 }, {
2513 .init_func = btrfs_interface_init,
2514 .exit_func = btrfs_interface_exit,
2515 }, {
2516 .init_func = btrfs_print_mod_info,
2517 .exit_func = NULL,
2518 }, {
2519 .init_func = btrfs_run_sanity_tests,
2520 .exit_func = NULL,
2521 }, {
2522 .init_func = register_btrfs,
2523 .exit_func = unregister_btrfs,
2524 }
2525};
2526
2527static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2528
2529static __always_inline void btrfs_exit_btrfs_fs(void)
2530{
2531 int i;
2532
2533 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2534 if (!mod_init_result[i])
2535 continue;
2536 if (mod_init_seq[i].exit_func)
2537 mod_init_seq[i].exit_func();
2538 mod_init_result[i] = false;
2539 }
2540}
2541
2542static void __exit exit_btrfs_fs(void)
2543{
2544 btrfs_exit_btrfs_fs();
2545 btrfs_cleanup_fs_uuids();
2546}
2547
2548static int __init init_btrfs_fs(void)
2549{
2550 int ret;
2551 int i;
2552
2553 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2554 ASSERT(!mod_init_result[i]);
2555 ret = mod_init_seq[i].init_func();
2556 if (ret < 0) {
2557 btrfs_exit_btrfs_fs();
2558 return ret;
2559 }
2560 mod_init_result[i] = true;
2561 }
2562 return 0;
2563}
2564
2565late_initcall(init_btrfs_fs);
2566module_exit(exit_btrfs_fs)
2567
2568MODULE_LICENSE("GPL");
2569MODULE_SOFTDEP("pre: crc32c");
2570MODULE_SOFTDEP("pre: xxhash64");
2571MODULE_SOFTDEP("pre: sha256");
2572MODULE_SOFTDEP("pre: blake2b-256");