bus: mhi: core: Fix some error return code
[linux/fpc-iii.git] / fs / btrfs / super.c
blob7932d8d07cffe16c0d313a236e1da97f28529817
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
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 #include "discard.h"
51 #include "qgroup.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/btrfs.h>
55 static const struct super_operations btrfs_super_ops;
58 * Types for mounting the default subvolume and a subvolume explicitly
59 * requested by subvol=/path. That way the callchain is straightforward and we
60 * don't have to play tricks with the mount options and recursive calls to
61 * btrfs_mount.
63 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
65 static struct file_system_type btrfs_fs_type;
66 static struct file_system_type btrfs_root_fs_type;
68 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
70 const char * __attribute_const__ btrfs_decode_error(int errno)
72 char *errstr = "unknown";
74 switch (errno) {
75 case -EIO:
76 errstr = "IO failure";
77 break;
78 case -ENOMEM:
79 errstr = "Out of memory";
80 break;
81 case -EROFS:
82 errstr = "Readonly filesystem";
83 break;
84 case -EEXIST:
85 errstr = "Object already exists";
86 break;
87 case -ENOSPC:
88 errstr = "No space left";
89 break;
90 case -ENOENT:
91 errstr = "No such entry";
92 break;
95 return errstr;
99 * __btrfs_handle_fs_error decodes expected errors from the caller and
100 * invokes the appropriate error response.
102 __cold
103 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
104 unsigned int line, int errno, const char *fmt, ...)
106 struct super_block *sb = fs_info->sb;
107 #ifdef CONFIG_PRINTK
108 const char *errstr;
109 #endif
112 * Special case: if the error is EROFS, and we're already
113 * under SB_RDONLY, then it is safe here.
115 if (errno == -EROFS && sb_rdonly(sb))
116 return;
118 #ifdef CONFIG_PRINTK
119 errstr = btrfs_decode_error(errno);
120 if (fmt) {
121 struct va_format vaf;
122 va_list args;
124 va_start(args, fmt);
125 vaf.fmt = fmt;
126 vaf.va = &args;
128 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
129 sb->s_id, function, line, errno, errstr, &vaf);
130 va_end(args);
131 } else {
132 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
133 sb->s_id, function, line, errno, errstr);
135 #endif
138 * Today we only save the error info to memory. Long term we'll
139 * also send it down to the disk
141 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
143 /* Don't go through full error handling during mount */
144 if (!(sb->s_flags & SB_BORN))
145 return;
147 if (sb_rdonly(sb))
148 return;
150 btrfs_discard_stop(fs_info);
152 /* btrfs handle error by forcing the filesystem readonly */
153 sb->s_flags |= SB_RDONLY;
154 btrfs_info(fs_info, "forced readonly");
156 * Note that a running device replace operation is not canceled here
157 * although there is no way to update the progress. It would add the
158 * risk of a deadlock, therefore the canceling is omitted. The only
159 * penalty is that some I/O remains active until the procedure
160 * completes. The next time when the filesystem is mounted writable
161 * again, the device replace operation continues.
165 #ifdef CONFIG_PRINTK
166 static const char * const logtypes[] = {
167 "emergency",
168 "alert",
169 "critical",
170 "error",
171 "warning",
172 "notice",
173 "info",
174 "debug",
179 * Use one ratelimit state per log level so that a flood of less important
180 * messages doesn't cause more important ones to be dropped.
182 static struct ratelimit_state printk_limits[] = {
183 RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
184 RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
185 RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
186 RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
187 RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
188 RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
189 RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
190 RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
193 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
195 char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
196 struct va_format vaf;
197 va_list args;
198 int kern_level;
199 const char *type = logtypes[4];
200 struct ratelimit_state *ratelimit = &printk_limits[4];
202 va_start(args, fmt);
204 while ((kern_level = printk_get_level(fmt)) != 0) {
205 size_t size = printk_skip_level(fmt) - fmt;
207 if (kern_level >= '0' && kern_level <= '7') {
208 memcpy(lvl, fmt, size);
209 lvl[size] = '\0';
210 type = logtypes[kern_level - '0'];
211 ratelimit = &printk_limits[kern_level - '0'];
213 fmt += size;
216 vaf.fmt = fmt;
217 vaf.va = &args;
219 if (__ratelimit(ratelimit))
220 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
221 fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
223 va_end(args);
225 #endif
228 * We only mark the transaction aborted and then set the file system read-only.
229 * This will prevent new transactions from starting or trying to join this
230 * one.
232 * This means that error recovery at the call site is limited to freeing
233 * any local memory allocations and passing the error code up without
234 * further cleanup. The transaction should complete as it normally would
235 * in the call path but will return -EIO.
237 * We'll complete the cleanup in btrfs_end_transaction and
238 * btrfs_commit_transaction.
240 __cold
241 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
242 const char *function,
243 unsigned int line, int errno)
245 struct btrfs_fs_info *fs_info = trans->fs_info;
247 WRITE_ONCE(trans->aborted, errno);
248 /* Nothing used. The other threads that have joined this
249 * transaction may be able to continue. */
250 if (!trans->dirty && list_empty(&trans->new_bgs)) {
251 const char *errstr;
253 errstr = btrfs_decode_error(errno);
254 btrfs_warn(fs_info,
255 "%s:%d: Aborting unused transaction(%s).",
256 function, line, errstr);
257 return;
259 WRITE_ONCE(trans->transaction->aborted, errno);
260 /* Wake up anybody who may be waiting on this transaction */
261 wake_up(&fs_info->transaction_wait);
262 wake_up(&fs_info->transaction_blocked_wait);
263 __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
266 * __btrfs_panic decodes unexpected, fatal errors from the caller,
267 * issues an alert, and either panics or BUGs, depending on mount options.
269 __cold
270 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
271 unsigned int line, int errno, const char *fmt, ...)
273 char *s_id = "<unknown>";
274 const char *errstr;
275 struct va_format vaf = { .fmt = fmt };
276 va_list args;
278 if (fs_info)
279 s_id = fs_info->sb->s_id;
281 va_start(args, fmt);
282 vaf.va = &args;
284 errstr = btrfs_decode_error(errno);
285 if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
286 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
287 s_id, function, line, &vaf, errno, errstr);
289 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
290 function, line, &vaf, errno, errstr);
291 va_end(args);
292 /* Caller calls BUG() */
295 static void btrfs_put_super(struct super_block *sb)
297 close_ctree(btrfs_sb(sb));
300 enum {
301 Opt_acl, Opt_noacl,
302 Opt_clear_cache,
303 Opt_commit_interval,
304 Opt_compress,
305 Opt_compress_force,
306 Opt_compress_force_type,
307 Opt_compress_type,
308 Opt_degraded,
309 Opt_device,
310 Opt_fatal_errors,
311 Opt_flushoncommit, Opt_noflushoncommit,
312 Opt_inode_cache, Opt_noinode_cache,
313 Opt_max_inline,
314 Opt_barrier, Opt_nobarrier,
315 Opt_datacow, Opt_nodatacow,
316 Opt_datasum, Opt_nodatasum,
317 Opt_defrag, Opt_nodefrag,
318 Opt_discard, Opt_nodiscard,
319 Opt_discard_mode,
320 Opt_nologreplay,
321 Opt_norecovery,
322 Opt_ratio,
323 Opt_rescan_uuid_tree,
324 Opt_skip_balance,
325 Opt_space_cache, Opt_no_space_cache,
326 Opt_space_cache_version,
327 Opt_ssd, Opt_nossd,
328 Opt_ssd_spread, Opt_nossd_spread,
329 Opt_subvol,
330 Opt_subvol_empty,
331 Opt_subvolid,
332 Opt_thread_pool,
333 Opt_treelog, Opt_notreelog,
334 Opt_usebackuproot,
335 Opt_user_subvol_rm_allowed,
337 /* Deprecated options */
338 Opt_alloc_start,
339 Opt_recovery,
340 Opt_subvolrootid,
342 /* Debugging options */
343 Opt_check_integrity,
344 Opt_check_integrity_including_extent_data,
345 Opt_check_integrity_print_mask,
346 Opt_enospc_debug, Opt_noenospc_debug,
347 #ifdef CONFIG_BTRFS_DEBUG
348 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
349 #endif
350 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
351 Opt_ref_verify,
352 #endif
353 Opt_err,
356 static const match_table_t tokens = {
357 {Opt_acl, "acl"},
358 {Opt_noacl, "noacl"},
359 {Opt_clear_cache, "clear_cache"},
360 {Opt_commit_interval, "commit=%u"},
361 {Opt_compress, "compress"},
362 {Opt_compress_type, "compress=%s"},
363 {Opt_compress_force, "compress-force"},
364 {Opt_compress_force_type, "compress-force=%s"},
365 {Opt_degraded, "degraded"},
366 {Opt_device, "device=%s"},
367 {Opt_fatal_errors, "fatal_errors=%s"},
368 {Opt_flushoncommit, "flushoncommit"},
369 {Opt_noflushoncommit, "noflushoncommit"},
370 {Opt_inode_cache, "inode_cache"},
371 {Opt_noinode_cache, "noinode_cache"},
372 {Opt_max_inline, "max_inline=%s"},
373 {Opt_barrier, "barrier"},
374 {Opt_nobarrier, "nobarrier"},
375 {Opt_datacow, "datacow"},
376 {Opt_nodatacow, "nodatacow"},
377 {Opt_datasum, "datasum"},
378 {Opt_nodatasum, "nodatasum"},
379 {Opt_defrag, "autodefrag"},
380 {Opt_nodefrag, "noautodefrag"},
381 {Opt_discard, "discard"},
382 {Opt_discard_mode, "discard=%s"},
383 {Opt_nodiscard, "nodiscard"},
384 {Opt_nologreplay, "nologreplay"},
385 {Opt_norecovery, "norecovery"},
386 {Opt_ratio, "metadata_ratio=%u"},
387 {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
388 {Opt_skip_balance, "skip_balance"},
389 {Opt_space_cache, "space_cache"},
390 {Opt_no_space_cache, "nospace_cache"},
391 {Opt_space_cache_version, "space_cache=%s"},
392 {Opt_ssd, "ssd"},
393 {Opt_nossd, "nossd"},
394 {Opt_ssd_spread, "ssd_spread"},
395 {Opt_nossd_spread, "nossd_spread"},
396 {Opt_subvol, "subvol=%s"},
397 {Opt_subvol_empty, "subvol="},
398 {Opt_subvolid, "subvolid=%s"},
399 {Opt_thread_pool, "thread_pool=%u"},
400 {Opt_treelog, "treelog"},
401 {Opt_notreelog, "notreelog"},
402 {Opt_usebackuproot, "usebackuproot"},
403 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
405 /* Deprecated options */
406 {Opt_alloc_start, "alloc_start=%s"},
407 {Opt_recovery, "recovery"},
408 {Opt_subvolrootid, "subvolrootid=%d"},
410 /* Debugging options */
411 {Opt_check_integrity, "check_int"},
412 {Opt_check_integrity_including_extent_data, "check_int_data"},
413 {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
414 {Opt_enospc_debug, "enospc_debug"},
415 {Opt_noenospc_debug, "noenospc_debug"},
416 #ifdef CONFIG_BTRFS_DEBUG
417 {Opt_fragment_data, "fragment=data"},
418 {Opt_fragment_metadata, "fragment=metadata"},
419 {Opt_fragment_all, "fragment=all"},
420 #endif
421 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
422 {Opt_ref_verify, "ref_verify"},
423 #endif
424 {Opt_err, NULL},
428 * Regular mount options parser. Everything that is needed only when
429 * reading in a new superblock is parsed here.
430 * XXX JDM: This needs to be cleaned up for remount.
432 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
433 unsigned long new_flags)
435 substring_t args[MAX_OPT_ARGS];
436 char *p, *num;
437 u64 cache_gen;
438 int intarg;
439 int ret = 0;
440 char *compress_type;
441 bool compress_force = false;
442 enum btrfs_compression_type saved_compress_type;
443 bool saved_compress_force;
444 int no_compress = 0;
446 cache_gen = btrfs_super_cache_generation(info->super_copy);
447 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
448 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
449 else if (cache_gen)
450 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
453 * Even the options are empty, we still need to do extra check
454 * against new flags
456 if (!options)
457 goto check;
459 while ((p = strsep(&options, ",")) != NULL) {
460 int token;
461 if (!*p)
462 continue;
464 token = match_token(p, tokens, args);
465 switch (token) {
466 case Opt_degraded:
467 btrfs_info(info, "allowing degraded mounts");
468 btrfs_set_opt(info->mount_opt, DEGRADED);
469 break;
470 case Opt_subvol:
471 case Opt_subvol_empty:
472 case Opt_subvolid:
473 case Opt_subvolrootid:
474 case Opt_device:
476 * These are parsed by btrfs_parse_subvol_options or
477 * btrfs_parse_device_options and can be ignored here.
479 break;
480 case Opt_nodatasum:
481 btrfs_set_and_info(info, NODATASUM,
482 "setting nodatasum");
483 break;
484 case Opt_datasum:
485 if (btrfs_test_opt(info, NODATASUM)) {
486 if (btrfs_test_opt(info, NODATACOW))
487 btrfs_info(info,
488 "setting datasum, datacow enabled");
489 else
490 btrfs_info(info, "setting datasum");
492 btrfs_clear_opt(info->mount_opt, NODATACOW);
493 btrfs_clear_opt(info->mount_opt, NODATASUM);
494 break;
495 case Opt_nodatacow:
496 if (!btrfs_test_opt(info, NODATACOW)) {
497 if (!btrfs_test_opt(info, COMPRESS) ||
498 !btrfs_test_opt(info, FORCE_COMPRESS)) {
499 btrfs_info(info,
500 "setting nodatacow, compression disabled");
501 } else {
502 btrfs_info(info, "setting nodatacow");
505 btrfs_clear_opt(info->mount_opt, COMPRESS);
506 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
507 btrfs_set_opt(info->mount_opt, NODATACOW);
508 btrfs_set_opt(info->mount_opt, NODATASUM);
509 break;
510 case Opt_datacow:
511 btrfs_clear_and_info(info, NODATACOW,
512 "setting datacow");
513 break;
514 case Opt_compress_force:
515 case Opt_compress_force_type:
516 compress_force = true;
517 /* Fallthrough */
518 case Opt_compress:
519 case Opt_compress_type:
520 saved_compress_type = btrfs_test_opt(info,
521 COMPRESS) ?
522 info->compress_type : BTRFS_COMPRESS_NONE;
523 saved_compress_force =
524 btrfs_test_opt(info, FORCE_COMPRESS);
525 if (token == Opt_compress ||
526 token == Opt_compress_force ||
527 strncmp(args[0].from, "zlib", 4) == 0) {
528 compress_type = "zlib";
530 info->compress_type = BTRFS_COMPRESS_ZLIB;
531 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
533 * args[0] contains uninitialized data since
534 * for these tokens we don't expect any
535 * parameter.
537 if (token != Opt_compress &&
538 token != Opt_compress_force)
539 info->compress_level =
540 btrfs_compress_str2level(
541 BTRFS_COMPRESS_ZLIB,
542 args[0].from + 4);
543 btrfs_set_opt(info->mount_opt, COMPRESS);
544 btrfs_clear_opt(info->mount_opt, NODATACOW);
545 btrfs_clear_opt(info->mount_opt, NODATASUM);
546 no_compress = 0;
547 } else if (strncmp(args[0].from, "lzo", 3) == 0) {
548 compress_type = "lzo";
549 info->compress_type = BTRFS_COMPRESS_LZO;
550 btrfs_set_opt(info->mount_opt, COMPRESS);
551 btrfs_clear_opt(info->mount_opt, NODATACOW);
552 btrfs_clear_opt(info->mount_opt, NODATASUM);
553 btrfs_set_fs_incompat(info, COMPRESS_LZO);
554 no_compress = 0;
555 } else if (strncmp(args[0].from, "zstd", 4) == 0) {
556 compress_type = "zstd";
557 info->compress_type = BTRFS_COMPRESS_ZSTD;
558 info->compress_level =
559 btrfs_compress_str2level(
560 BTRFS_COMPRESS_ZSTD,
561 args[0].from + 4);
562 btrfs_set_opt(info->mount_opt, COMPRESS);
563 btrfs_clear_opt(info->mount_opt, NODATACOW);
564 btrfs_clear_opt(info->mount_opt, NODATASUM);
565 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
566 no_compress = 0;
567 } else if (strncmp(args[0].from, "no", 2) == 0) {
568 compress_type = "no";
569 btrfs_clear_opt(info->mount_opt, COMPRESS);
570 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
571 compress_force = false;
572 no_compress++;
573 } else {
574 ret = -EINVAL;
575 goto out;
578 if (compress_force) {
579 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
580 } else {
582 * If we remount from compress-force=xxx to
583 * compress=xxx, we need clear FORCE_COMPRESS
584 * flag, otherwise, there is no way for users
585 * to disable forcible compression separately.
587 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
589 if ((btrfs_test_opt(info, COMPRESS) &&
590 (info->compress_type != saved_compress_type ||
591 compress_force != saved_compress_force)) ||
592 (!btrfs_test_opt(info, COMPRESS) &&
593 no_compress == 1)) {
594 btrfs_info(info, "%s %s compression, level %d",
595 (compress_force) ? "force" : "use",
596 compress_type, info->compress_level);
598 compress_force = false;
599 break;
600 case Opt_ssd:
601 btrfs_set_and_info(info, SSD,
602 "enabling ssd optimizations");
603 btrfs_clear_opt(info->mount_opt, NOSSD);
604 break;
605 case Opt_ssd_spread:
606 btrfs_set_and_info(info, SSD,
607 "enabling ssd optimizations");
608 btrfs_set_and_info(info, SSD_SPREAD,
609 "using spread ssd allocation scheme");
610 btrfs_clear_opt(info->mount_opt, NOSSD);
611 break;
612 case Opt_nossd:
613 btrfs_set_opt(info->mount_opt, NOSSD);
614 btrfs_clear_and_info(info, SSD,
615 "not using ssd optimizations");
616 /* Fallthrough */
617 case Opt_nossd_spread:
618 btrfs_clear_and_info(info, SSD_SPREAD,
619 "not using spread ssd allocation scheme");
620 break;
621 case Opt_barrier:
622 btrfs_clear_and_info(info, NOBARRIER,
623 "turning on barriers");
624 break;
625 case Opt_nobarrier:
626 btrfs_set_and_info(info, NOBARRIER,
627 "turning off barriers");
628 break;
629 case Opt_thread_pool:
630 ret = match_int(&args[0], &intarg);
631 if (ret) {
632 goto out;
633 } else if (intarg == 0) {
634 ret = -EINVAL;
635 goto out;
637 info->thread_pool_size = intarg;
638 break;
639 case Opt_max_inline:
640 num = match_strdup(&args[0]);
641 if (num) {
642 info->max_inline = memparse(num, NULL);
643 kfree(num);
645 if (info->max_inline) {
646 info->max_inline = min_t(u64,
647 info->max_inline,
648 info->sectorsize);
650 btrfs_info(info, "max_inline at %llu",
651 info->max_inline);
652 } else {
653 ret = -ENOMEM;
654 goto out;
656 break;
657 case Opt_alloc_start:
658 btrfs_info(info,
659 "option alloc_start is obsolete, ignored");
660 break;
661 case Opt_acl:
662 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
663 info->sb->s_flags |= SB_POSIXACL;
664 break;
665 #else
666 btrfs_err(info, "support for ACL not compiled in!");
667 ret = -EINVAL;
668 goto out;
669 #endif
670 case Opt_noacl:
671 info->sb->s_flags &= ~SB_POSIXACL;
672 break;
673 case Opt_notreelog:
674 btrfs_set_and_info(info, NOTREELOG,
675 "disabling tree log");
676 break;
677 case Opt_treelog:
678 btrfs_clear_and_info(info, NOTREELOG,
679 "enabling tree log");
680 break;
681 case Opt_norecovery:
682 case Opt_nologreplay:
683 btrfs_set_and_info(info, NOLOGREPLAY,
684 "disabling log replay at mount time");
685 break;
686 case Opt_flushoncommit:
687 btrfs_set_and_info(info, FLUSHONCOMMIT,
688 "turning on flush-on-commit");
689 break;
690 case Opt_noflushoncommit:
691 btrfs_clear_and_info(info, FLUSHONCOMMIT,
692 "turning off flush-on-commit");
693 break;
694 case Opt_ratio:
695 ret = match_int(&args[0], &intarg);
696 if (ret)
697 goto out;
698 info->metadata_ratio = intarg;
699 btrfs_info(info, "metadata ratio %u",
700 info->metadata_ratio);
701 break;
702 case Opt_discard:
703 case Opt_discard_mode:
704 if (token == Opt_discard ||
705 strcmp(args[0].from, "sync") == 0) {
706 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
707 btrfs_set_and_info(info, DISCARD_SYNC,
708 "turning on sync discard");
709 } else if (strcmp(args[0].from, "async") == 0) {
710 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
711 btrfs_set_and_info(info, DISCARD_ASYNC,
712 "turning on async discard");
713 } else {
714 ret = -EINVAL;
715 goto out;
717 break;
718 case Opt_nodiscard:
719 btrfs_clear_and_info(info, DISCARD_SYNC,
720 "turning off discard");
721 btrfs_clear_and_info(info, DISCARD_ASYNC,
722 "turning off async discard");
723 break;
724 case Opt_space_cache:
725 case Opt_space_cache_version:
726 if (token == Opt_space_cache ||
727 strcmp(args[0].from, "v1") == 0) {
728 btrfs_clear_opt(info->mount_opt,
729 FREE_SPACE_TREE);
730 btrfs_set_and_info(info, SPACE_CACHE,
731 "enabling disk space caching");
732 } else if (strcmp(args[0].from, "v2") == 0) {
733 btrfs_clear_opt(info->mount_opt,
734 SPACE_CACHE);
735 btrfs_set_and_info(info, FREE_SPACE_TREE,
736 "enabling free space tree");
737 } else {
738 ret = -EINVAL;
739 goto out;
741 break;
742 case Opt_rescan_uuid_tree:
743 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
744 break;
745 case Opt_no_space_cache:
746 if (btrfs_test_opt(info, SPACE_CACHE)) {
747 btrfs_clear_and_info(info, SPACE_CACHE,
748 "disabling disk space caching");
750 if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
751 btrfs_clear_and_info(info, FREE_SPACE_TREE,
752 "disabling free space tree");
754 break;
755 case Opt_inode_cache:
756 btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
757 "enabling inode map caching");
758 break;
759 case Opt_noinode_cache:
760 btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
761 "disabling inode map caching");
762 break;
763 case Opt_clear_cache:
764 btrfs_set_and_info(info, CLEAR_CACHE,
765 "force clearing of disk cache");
766 break;
767 case Opt_user_subvol_rm_allowed:
768 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
769 break;
770 case Opt_enospc_debug:
771 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
772 break;
773 case Opt_noenospc_debug:
774 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
775 break;
776 case Opt_defrag:
777 btrfs_set_and_info(info, AUTO_DEFRAG,
778 "enabling auto defrag");
779 break;
780 case Opt_nodefrag:
781 btrfs_clear_and_info(info, AUTO_DEFRAG,
782 "disabling auto defrag");
783 break;
784 case Opt_recovery:
785 btrfs_warn(info,
786 "'recovery' is deprecated, use 'usebackuproot' instead");
787 /* fall through */
788 case Opt_usebackuproot:
789 btrfs_info(info,
790 "trying to use backup root at mount time");
791 btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
792 break;
793 case Opt_skip_balance:
794 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
795 break;
796 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
797 case Opt_check_integrity_including_extent_data:
798 btrfs_info(info,
799 "enabling check integrity including extent data");
800 btrfs_set_opt(info->mount_opt,
801 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
802 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
803 break;
804 case Opt_check_integrity:
805 btrfs_info(info, "enabling check integrity");
806 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
807 break;
808 case Opt_check_integrity_print_mask:
809 ret = match_int(&args[0], &intarg);
810 if (ret)
811 goto out;
812 info->check_integrity_print_mask = intarg;
813 btrfs_info(info, "check_integrity_print_mask 0x%x",
814 info->check_integrity_print_mask);
815 break;
816 #else
817 case Opt_check_integrity_including_extent_data:
818 case Opt_check_integrity:
819 case Opt_check_integrity_print_mask:
820 btrfs_err(info,
821 "support for check_integrity* not compiled in!");
822 ret = -EINVAL;
823 goto out;
824 #endif
825 case Opt_fatal_errors:
826 if (strcmp(args[0].from, "panic") == 0)
827 btrfs_set_opt(info->mount_opt,
828 PANIC_ON_FATAL_ERROR);
829 else if (strcmp(args[0].from, "bug") == 0)
830 btrfs_clear_opt(info->mount_opt,
831 PANIC_ON_FATAL_ERROR);
832 else {
833 ret = -EINVAL;
834 goto out;
836 break;
837 case Opt_commit_interval:
838 intarg = 0;
839 ret = match_int(&args[0], &intarg);
840 if (ret)
841 goto out;
842 if (intarg == 0) {
843 btrfs_info(info,
844 "using default commit interval %us",
845 BTRFS_DEFAULT_COMMIT_INTERVAL);
846 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
847 } else if (intarg > 300) {
848 btrfs_warn(info, "excessive commit interval %d",
849 intarg);
851 info->commit_interval = intarg;
852 break;
853 #ifdef CONFIG_BTRFS_DEBUG
854 case Opt_fragment_all:
855 btrfs_info(info, "fragmenting all space");
856 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
857 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
858 break;
859 case Opt_fragment_metadata:
860 btrfs_info(info, "fragmenting metadata");
861 btrfs_set_opt(info->mount_opt,
862 FRAGMENT_METADATA);
863 break;
864 case Opt_fragment_data:
865 btrfs_info(info, "fragmenting data");
866 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
867 break;
868 #endif
869 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
870 case Opt_ref_verify:
871 btrfs_info(info, "doing ref verification");
872 btrfs_set_opt(info->mount_opt, REF_VERIFY);
873 break;
874 #endif
875 case Opt_err:
876 btrfs_err(info, "unrecognized mount option '%s'", p);
877 ret = -EINVAL;
878 goto out;
879 default:
880 break;
883 check:
885 * Extra check for current option against current flag
887 if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
888 btrfs_err(info,
889 "nologreplay must be used with ro mount option");
890 ret = -EINVAL;
892 out:
893 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
894 !btrfs_test_opt(info, FREE_SPACE_TREE) &&
895 !btrfs_test_opt(info, CLEAR_CACHE)) {
896 btrfs_err(info, "cannot disable free space tree");
897 ret = -EINVAL;
900 if (!ret && btrfs_test_opt(info, SPACE_CACHE))
901 btrfs_info(info, "disk space caching is enabled");
902 if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
903 btrfs_info(info, "using free space tree");
904 return ret;
908 * Parse mount options that are required early in the mount process.
910 * All other options will be parsed on much later in the mount process and
911 * only when we need to allocate a new super block.
913 static int btrfs_parse_device_options(const char *options, fmode_t flags,
914 void *holder)
916 substring_t args[MAX_OPT_ARGS];
917 char *device_name, *opts, *orig, *p;
918 struct btrfs_device *device = NULL;
919 int error = 0;
921 lockdep_assert_held(&uuid_mutex);
923 if (!options)
924 return 0;
927 * strsep changes the string, duplicate it because btrfs_parse_options
928 * gets called later
930 opts = kstrdup(options, GFP_KERNEL);
931 if (!opts)
932 return -ENOMEM;
933 orig = opts;
935 while ((p = strsep(&opts, ",")) != NULL) {
936 int token;
938 if (!*p)
939 continue;
941 token = match_token(p, tokens, args);
942 if (token == Opt_device) {
943 device_name = match_strdup(&args[0]);
944 if (!device_name) {
945 error = -ENOMEM;
946 goto out;
948 device = btrfs_scan_one_device(device_name, flags,
949 holder);
950 kfree(device_name);
951 if (IS_ERR(device)) {
952 error = PTR_ERR(device);
953 goto out;
958 out:
959 kfree(orig);
960 return error;
964 * Parse mount options that are related to subvolume id
966 * The value is later passed to mount_subvol()
968 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
969 u64 *subvol_objectid)
971 substring_t args[MAX_OPT_ARGS];
972 char *opts, *orig, *p;
973 int error = 0;
974 u64 subvolid;
976 if (!options)
977 return 0;
980 * strsep changes the string, duplicate it because
981 * btrfs_parse_device_options gets called later
983 opts = kstrdup(options, GFP_KERNEL);
984 if (!opts)
985 return -ENOMEM;
986 orig = opts;
988 while ((p = strsep(&opts, ",")) != NULL) {
989 int token;
990 if (!*p)
991 continue;
993 token = match_token(p, tokens, args);
994 switch (token) {
995 case Opt_subvol:
996 kfree(*subvol_name);
997 *subvol_name = match_strdup(&args[0]);
998 if (!*subvol_name) {
999 error = -ENOMEM;
1000 goto out;
1002 break;
1003 case Opt_subvolid:
1004 error = match_u64(&args[0], &subvolid);
1005 if (error)
1006 goto out;
1008 /* we want the original fs_tree */
1009 if (subvolid == 0)
1010 subvolid = BTRFS_FS_TREE_OBJECTID;
1012 *subvol_objectid = subvolid;
1013 break;
1014 case Opt_subvolrootid:
1015 pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
1016 break;
1017 default:
1018 break;
1022 out:
1023 kfree(orig);
1024 return error;
1027 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1028 u64 subvol_objectid)
1030 struct btrfs_root *root = fs_info->tree_root;
1031 struct btrfs_root *fs_root = NULL;
1032 struct btrfs_root_ref *root_ref;
1033 struct btrfs_inode_ref *inode_ref;
1034 struct btrfs_key key;
1035 struct btrfs_path *path = NULL;
1036 char *name = NULL, *ptr;
1037 u64 dirid;
1038 int len;
1039 int ret;
1041 path = btrfs_alloc_path();
1042 if (!path) {
1043 ret = -ENOMEM;
1044 goto err;
1046 path->leave_spinning = 1;
1048 name = kmalloc(PATH_MAX, GFP_KERNEL);
1049 if (!name) {
1050 ret = -ENOMEM;
1051 goto err;
1053 ptr = name + PATH_MAX - 1;
1054 ptr[0] = '\0';
1057 * Walk up the subvolume trees in the tree of tree roots by root
1058 * backrefs until we hit the top-level subvolume.
1060 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1061 key.objectid = subvol_objectid;
1062 key.type = BTRFS_ROOT_BACKREF_KEY;
1063 key.offset = (u64)-1;
1065 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1066 if (ret < 0) {
1067 goto err;
1068 } else if (ret > 0) {
1069 ret = btrfs_previous_item(root, path, subvol_objectid,
1070 BTRFS_ROOT_BACKREF_KEY);
1071 if (ret < 0) {
1072 goto err;
1073 } else if (ret > 0) {
1074 ret = -ENOENT;
1075 goto err;
1079 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1080 subvol_objectid = key.offset;
1082 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1083 struct btrfs_root_ref);
1084 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1085 ptr -= len + 1;
1086 if (ptr < name) {
1087 ret = -ENAMETOOLONG;
1088 goto err;
1090 read_extent_buffer(path->nodes[0], ptr + 1,
1091 (unsigned long)(root_ref + 1), len);
1092 ptr[0] = '/';
1093 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1094 btrfs_release_path(path);
1096 key.objectid = subvol_objectid;
1097 key.type = BTRFS_ROOT_ITEM_KEY;
1098 key.offset = (u64)-1;
1099 fs_root = btrfs_get_fs_root(fs_info, &key, true);
1100 if (IS_ERR(fs_root)) {
1101 ret = PTR_ERR(fs_root);
1102 fs_root = NULL;
1103 goto err;
1107 * Walk up the filesystem tree by inode refs until we hit the
1108 * root directory.
1110 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1111 key.objectid = dirid;
1112 key.type = BTRFS_INODE_REF_KEY;
1113 key.offset = (u64)-1;
1115 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1116 if (ret < 0) {
1117 goto err;
1118 } else if (ret > 0) {
1119 ret = btrfs_previous_item(fs_root, path, dirid,
1120 BTRFS_INODE_REF_KEY);
1121 if (ret < 0) {
1122 goto err;
1123 } else if (ret > 0) {
1124 ret = -ENOENT;
1125 goto err;
1129 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1130 dirid = key.offset;
1132 inode_ref = btrfs_item_ptr(path->nodes[0],
1133 path->slots[0],
1134 struct btrfs_inode_ref);
1135 len = btrfs_inode_ref_name_len(path->nodes[0],
1136 inode_ref);
1137 ptr -= len + 1;
1138 if (ptr < name) {
1139 ret = -ENAMETOOLONG;
1140 goto err;
1142 read_extent_buffer(path->nodes[0], ptr + 1,
1143 (unsigned long)(inode_ref + 1), len);
1144 ptr[0] = '/';
1145 btrfs_release_path(path);
1147 btrfs_put_root(fs_root);
1148 fs_root = NULL;
1151 btrfs_free_path(path);
1152 if (ptr == name + PATH_MAX - 1) {
1153 name[0] = '/';
1154 name[1] = '\0';
1155 } else {
1156 memmove(name, ptr, name + PATH_MAX - ptr);
1158 return name;
1160 err:
1161 btrfs_put_root(fs_root);
1162 btrfs_free_path(path);
1163 kfree(name);
1164 return ERR_PTR(ret);
1167 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1169 struct btrfs_root *root = fs_info->tree_root;
1170 struct btrfs_dir_item *di;
1171 struct btrfs_path *path;
1172 struct btrfs_key location;
1173 u64 dir_id;
1175 path = btrfs_alloc_path();
1176 if (!path)
1177 return -ENOMEM;
1178 path->leave_spinning = 1;
1181 * Find the "default" dir item which points to the root item that we
1182 * will mount by default if we haven't been given a specific subvolume
1183 * to mount.
1185 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1186 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1187 if (IS_ERR(di)) {
1188 btrfs_free_path(path);
1189 return PTR_ERR(di);
1191 if (!di) {
1193 * Ok the default dir item isn't there. This is weird since
1194 * it's always been there, but don't freak out, just try and
1195 * mount the top-level subvolume.
1197 btrfs_free_path(path);
1198 *objectid = BTRFS_FS_TREE_OBJECTID;
1199 return 0;
1202 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1203 btrfs_free_path(path);
1204 *objectid = location.objectid;
1205 return 0;
1208 static int btrfs_fill_super(struct super_block *sb,
1209 struct btrfs_fs_devices *fs_devices,
1210 void *data)
1212 struct inode *inode;
1213 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1214 struct btrfs_key key;
1215 int err;
1217 sb->s_maxbytes = MAX_LFS_FILESIZE;
1218 sb->s_magic = BTRFS_SUPER_MAGIC;
1219 sb->s_op = &btrfs_super_ops;
1220 sb->s_d_op = &btrfs_dentry_operations;
1221 sb->s_export_op = &btrfs_export_ops;
1222 sb->s_xattr = btrfs_xattr_handlers;
1223 sb->s_time_gran = 1;
1224 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1225 sb->s_flags |= SB_POSIXACL;
1226 #endif
1227 sb->s_flags |= SB_I_VERSION;
1228 sb->s_iflags |= SB_I_CGROUPWB;
1230 err = super_setup_bdi(sb);
1231 if (err) {
1232 btrfs_err(fs_info, "super_setup_bdi failed");
1233 return err;
1236 err = open_ctree(sb, fs_devices, (char *)data);
1237 if (err) {
1238 btrfs_err(fs_info, "open_ctree failed");
1239 return err;
1242 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1243 key.type = BTRFS_INODE_ITEM_KEY;
1244 key.offset = 0;
1245 inode = btrfs_iget(sb, &key, fs_info->fs_root);
1246 if (IS_ERR(inode)) {
1247 err = PTR_ERR(inode);
1248 goto fail_close;
1251 sb->s_root = d_make_root(inode);
1252 if (!sb->s_root) {
1253 err = -ENOMEM;
1254 goto fail_close;
1257 cleancache_init_fs(sb);
1258 sb->s_flags |= SB_ACTIVE;
1259 return 0;
1261 fail_close:
1262 close_ctree(fs_info);
1263 return err;
1266 int btrfs_sync_fs(struct super_block *sb, int wait)
1268 struct btrfs_trans_handle *trans;
1269 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1270 struct btrfs_root *root = fs_info->tree_root;
1272 trace_btrfs_sync_fs(fs_info, wait);
1274 if (!wait) {
1275 filemap_flush(fs_info->btree_inode->i_mapping);
1276 return 0;
1279 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1281 trans = btrfs_attach_transaction_barrier(root);
1282 if (IS_ERR(trans)) {
1283 /* no transaction, don't bother */
1284 if (PTR_ERR(trans) == -ENOENT) {
1286 * Exit unless we have some pending changes
1287 * that need to go through commit
1289 if (fs_info->pending_changes == 0)
1290 return 0;
1292 * A non-blocking test if the fs is frozen. We must not
1293 * start a new transaction here otherwise a deadlock
1294 * happens. The pending operations are delayed to the
1295 * next commit after thawing.
1297 if (sb_start_write_trylock(sb))
1298 sb_end_write(sb);
1299 else
1300 return 0;
1301 trans = btrfs_start_transaction(root, 0);
1303 if (IS_ERR(trans))
1304 return PTR_ERR(trans);
1306 return btrfs_commit_transaction(trans);
1309 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1311 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1312 const char *compress_type;
1314 if (btrfs_test_opt(info, DEGRADED))
1315 seq_puts(seq, ",degraded");
1316 if (btrfs_test_opt(info, NODATASUM))
1317 seq_puts(seq, ",nodatasum");
1318 if (btrfs_test_opt(info, NODATACOW))
1319 seq_puts(seq, ",nodatacow");
1320 if (btrfs_test_opt(info, NOBARRIER))
1321 seq_puts(seq, ",nobarrier");
1322 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1323 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1324 if (info->thread_pool_size != min_t(unsigned long,
1325 num_online_cpus() + 2, 8))
1326 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1327 if (btrfs_test_opt(info, COMPRESS)) {
1328 compress_type = btrfs_compress_type2str(info->compress_type);
1329 if (btrfs_test_opt(info, FORCE_COMPRESS))
1330 seq_printf(seq, ",compress-force=%s", compress_type);
1331 else
1332 seq_printf(seq, ",compress=%s", compress_type);
1333 if (info->compress_level)
1334 seq_printf(seq, ":%d", info->compress_level);
1336 if (btrfs_test_opt(info, NOSSD))
1337 seq_puts(seq, ",nossd");
1338 if (btrfs_test_opt(info, SSD_SPREAD))
1339 seq_puts(seq, ",ssd_spread");
1340 else if (btrfs_test_opt(info, SSD))
1341 seq_puts(seq, ",ssd");
1342 if (btrfs_test_opt(info, NOTREELOG))
1343 seq_puts(seq, ",notreelog");
1344 if (btrfs_test_opt(info, NOLOGREPLAY))
1345 seq_puts(seq, ",nologreplay");
1346 if (btrfs_test_opt(info, FLUSHONCOMMIT))
1347 seq_puts(seq, ",flushoncommit");
1348 if (btrfs_test_opt(info, DISCARD_SYNC))
1349 seq_puts(seq, ",discard");
1350 if (btrfs_test_opt(info, DISCARD_ASYNC))
1351 seq_puts(seq, ",discard=async");
1352 if (!(info->sb->s_flags & SB_POSIXACL))
1353 seq_puts(seq, ",noacl");
1354 if (btrfs_test_opt(info, SPACE_CACHE))
1355 seq_puts(seq, ",space_cache");
1356 else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1357 seq_puts(seq, ",space_cache=v2");
1358 else
1359 seq_puts(seq, ",nospace_cache");
1360 if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1361 seq_puts(seq, ",rescan_uuid_tree");
1362 if (btrfs_test_opt(info, CLEAR_CACHE))
1363 seq_puts(seq, ",clear_cache");
1364 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1365 seq_puts(seq, ",user_subvol_rm_allowed");
1366 if (btrfs_test_opt(info, ENOSPC_DEBUG))
1367 seq_puts(seq, ",enospc_debug");
1368 if (btrfs_test_opt(info, AUTO_DEFRAG))
1369 seq_puts(seq, ",autodefrag");
1370 if (btrfs_test_opt(info, INODE_MAP_CACHE))
1371 seq_puts(seq, ",inode_cache");
1372 if (btrfs_test_opt(info, SKIP_BALANCE))
1373 seq_puts(seq, ",skip_balance");
1374 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1375 if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1376 seq_puts(seq, ",check_int_data");
1377 else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1378 seq_puts(seq, ",check_int");
1379 if (info->check_integrity_print_mask)
1380 seq_printf(seq, ",check_int_print_mask=%d",
1381 info->check_integrity_print_mask);
1382 #endif
1383 if (info->metadata_ratio)
1384 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1385 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1386 seq_puts(seq, ",fatal_errors=panic");
1387 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1388 seq_printf(seq, ",commit=%u", info->commit_interval);
1389 #ifdef CONFIG_BTRFS_DEBUG
1390 if (btrfs_test_opt(info, FRAGMENT_DATA))
1391 seq_puts(seq, ",fragment=data");
1392 if (btrfs_test_opt(info, FRAGMENT_METADATA))
1393 seq_puts(seq, ",fragment=metadata");
1394 #endif
1395 if (btrfs_test_opt(info, REF_VERIFY))
1396 seq_puts(seq, ",ref_verify");
1397 seq_printf(seq, ",subvolid=%llu",
1398 BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1399 seq_puts(seq, ",subvol=");
1400 seq_dentry(seq, dentry, " \t\n\\");
1401 return 0;
1404 static int btrfs_test_super(struct super_block *s, void *data)
1406 struct btrfs_fs_info *p = data;
1407 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1409 return fs_info->fs_devices == p->fs_devices;
1412 static int btrfs_set_super(struct super_block *s, void *data)
1414 int err = set_anon_super(s, data);
1415 if (!err)
1416 s->s_fs_info = data;
1417 return err;
1421 * subvolumes are identified by ino 256
1423 static inline int is_subvolume_inode(struct inode *inode)
1425 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1426 return 1;
1427 return 0;
1430 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1431 struct vfsmount *mnt)
1433 struct dentry *root;
1434 int ret;
1436 if (!subvol_name) {
1437 if (!subvol_objectid) {
1438 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1439 &subvol_objectid);
1440 if (ret) {
1441 root = ERR_PTR(ret);
1442 goto out;
1445 subvol_name = btrfs_get_subvol_name_from_objectid(
1446 btrfs_sb(mnt->mnt_sb), subvol_objectid);
1447 if (IS_ERR(subvol_name)) {
1448 root = ERR_CAST(subvol_name);
1449 subvol_name = NULL;
1450 goto out;
1455 root = mount_subtree(mnt, subvol_name);
1456 /* mount_subtree() drops our reference on the vfsmount. */
1457 mnt = NULL;
1459 if (!IS_ERR(root)) {
1460 struct super_block *s = root->d_sb;
1461 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1462 struct inode *root_inode = d_inode(root);
1463 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1465 ret = 0;
1466 if (!is_subvolume_inode(root_inode)) {
1467 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1468 subvol_name);
1469 ret = -EINVAL;
1471 if (subvol_objectid && root_objectid != subvol_objectid) {
1473 * This will also catch a race condition where a
1474 * subvolume which was passed by ID is renamed and
1475 * another subvolume is renamed over the old location.
1477 btrfs_err(fs_info,
1478 "subvol '%s' does not match subvolid %llu",
1479 subvol_name, subvol_objectid);
1480 ret = -EINVAL;
1482 if (ret) {
1483 dput(root);
1484 root = ERR_PTR(ret);
1485 deactivate_locked_super(s);
1489 out:
1490 mntput(mnt);
1491 kfree(subvol_name);
1492 return root;
1496 * Find a superblock for the given device / mount point.
1498 * Note: This is based on mount_bdev from fs/super.c with a few additions
1499 * for multiple device setup. Make sure to keep it in sync.
1501 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1502 int flags, const char *device_name, void *data)
1504 struct block_device *bdev = NULL;
1505 struct super_block *s;
1506 struct btrfs_device *device = NULL;
1507 struct btrfs_fs_devices *fs_devices = NULL;
1508 struct btrfs_fs_info *fs_info = NULL;
1509 void *new_sec_opts = NULL;
1510 fmode_t mode = FMODE_READ;
1511 int error = 0;
1513 if (!(flags & SB_RDONLY))
1514 mode |= FMODE_WRITE;
1516 if (data) {
1517 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1518 if (error)
1519 return ERR_PTR(error);
1523 * Setup a dummy root and fs_info for test/set super. This is because
1524 * we don't actually fill this stuff out until open_ctree, but we need
1525 * then open_ctree will properly initialize the file system specific
1526 * settings later. btrfs_init_fs_info initializes the static elements
1527 * of the fs_info (locks and such) to make cleanup easier if we find a
1528 * superblock with our given fs_devices later on at sget() time.
1530 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1531 if (!fs_info) {
1532 error = -ENOMEM;
1533 goto error_sec_opts;
1535 btrfs_init_fs_info(fs_info);
1537 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1538 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1539 if (!fs_info->super_copy || !fs_info->super_for_commit) {
1540 error = -ENOMEM;
1541 goto error_fs_info;
1544 mutex_lock(&uuid_mutex);
1545 error = btrfs_parse_device_options(data, mode, fs_type);
1546 if (error) {
1547 mutex_unlock(&uuid_mutex);
1548 goto error_fs_info;
1551 device = btrfs_scan_one_device(device_name, mode, fs_type);
1552 if (IS_ERR(device)) {
1553 mutex_unlock(&uuid_mutex);
1554 error = PTR_ERR(device);
1555 goto error_fs_info;
1558 fs_devices = device->fs_devices;
1559 fs_info->fs_devices = fs_devices;
1561 error = btrfs_open_devices(fs_devices, mode, fs_type);
1562 mutex_unlock(&uuid_mutex);
1563 if (error)
1564 goto error_fs_info;
1566 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1567 error = -EACCES;
1568 goto error_close_devices;
1571 bdev = fs_devices->latest_bdev;
1572 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1573 fs_info);
1574 if (IS_ERR(s)) {
1575 error = PTR_ERR(s);
1576 goto error_close_devices;
1579 if (s->s_root) {
1580 btrfs_close_devices(fs_devices);
1581 btrfs_free_fs_info(fs_info);
1582 if ((flags ^ s->s_flags) & SB_RDONLY)
1583 error = -EBUSY;
1584 } else {
1585 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1586 btrfs_sb(s)->bdev_holder = fs_type;
1587 if (!strstr(crc32c_impl(), "generic"))
1588 set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1589 error = btrfs_fill_super(s, fs_devices, data);
1591 if (!error)
1592 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1593 security_free_mnt_opts(&new_sec_opts);
1594 if (error) {
1595 deactivate_locked_super(s);
1596 return ERR_PTR(error);
1599 return dget(s->s_root);
1601 error_close_devices:
1602 btrfs_close_devices(fs_devices);
1603 error_fs_info:
1604 btrfs_free_fs_info(fs_info);
1605 error_sec_opts:
1606 security_free_mnt_opts(&new_sec_opts);
1607 return ERR_PTR(error);
1611 * Mount function which is called by VFS layer.
1613 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1614 * which needs vfsmount* of device's root (/). This means device's root has to
1615 * be mounted internally in any case.
1617 * Operation flow:
1618 * 1. Parse subvol id related options for later use in mount_subvol().
1620 * 2. Mount device's root (/) by calling vfs_kern_mount().
1622 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1623 * first place. In order to avoid calling btrfs_mount() again, we use
1624 * different file_system_type which is not registered to VFS by
1625 * register_filesystem() (btrfs_root_fs_type). As a result,
1626 * btrfs_mount_root() is called. The return value will be used by
1627 * mount_subtree() in mount_subvol().
1629 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is
1630 * "btrfs subvolume set-default", mount_subvol() is called always.
1632 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1633 const char *device_name, void *data)
1635 struct vfsmount *mnt_root;
1636 struct dentry *root;
1637 char *subvol_name = NULL;
1638 u64 subvol_objectid = 0;
1639 int error = 0;
1641 error = btrfs_parse_subvol_options(data, &subvol_name,
1642 &subvol_objectid);
1643 if (error) {
1644 kfree(subvol_name);
1645 return ERR_PTR(error);
1648 /* mount device's root (/) */
1649 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1650 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1651 if (flags & SB_RDONLY) {
1652 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1653 flags & ~SB_RDONLY, device_name, data);
1654 } else {
1655 mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1656 flags | SB_RDONLY, device_name, data);
1657 if (IS_ERR(mnt_root)) {
1658 root = ERR_CAST(mnt_root);
1659 kfree(subvol_name);
1660 goto out;
1663 down_write(&mnt_root->mnt_sb->s_umount);
1664 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1665 up_write(&mnt_root->mnt_sb->s_umount);
1666 if (error < 0) {
1667 root = ERR_PTR(error);
1668 mntput(mnt_root);
1669 kfree(subvol_name);
1670 goto out;
1674 if (IS_ERR(mnt_root)) {
1675 root = ERR_CAST(mnt_root);
1676 kfree(subvol_name);
1677 goto out;
1680 /* mount_subvol() will free subvol_name and mnt_root */
1681 root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1683 out:
1684 return root;
1687 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1688 u32 new_pool_size, u32 old_pool_size)
1690 if (new_pool_size == old_pool_size)
1691 return;
1693 fs_info->thread_pool_size = new_pool_size;
1695 btrfs_info(fs_info, "resize thread pool %d -> %d",
1696 old_pool_size, new_pool_size);
1698 btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1699 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1700 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1701 btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1702 btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1703 btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1704 new_pool_size);
1705 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1706 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1707 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1708 btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1709 btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1710 new_pool_size);
1713 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1715 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1718 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1719 unsigned long old_opts, int flags)
1721 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1722 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1723 (flags & SB_RDONLY))) {
1724 /* wait for any defraggers to finish */
1725 wait_event(fs_info->transaction_wait,
1726 (atomic_read(&fs_info->defrag_running) == 0));
1727 if (flags & SB_RDONLY)
1728 sync_filesystem(fs_info->sb);
1732 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1733 unsigned long old_opts)
1736 * We need to cleanup all defragable inodes if the autodefragment is
1737 * close or the filesystem is read only.
1739 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1740 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1741 btrfs_cleanup_defrag_inodes(fs_info);
1744 /* If we toggled discard async */
1745 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1746 btrfs_test_opt(fs_info, DISCARD_ASYNC))
1747 btrfs_discard_resume(fs_info);
1748 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1749 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1750 btrfs_discard_cleanup(fs_info);
1752 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1755 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1757 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1758 struct btrfs_root *root = fs_info->tree_root;
1759 unsigned old_flags = sb->s_flags;
1760 unsigned long old_opts = fs_info->mount_opt;
1761 unsigned long old_compress_type = fs_info->compress_type;
1762 u64 old_max_inline = fs_info->max_inline;
1763 u32 old_thread_pool_size = fs_info->thread_pool_size;
1764 u32 old_metadata_ratio = fs_info->metadata_ratio;
1765 int ret;
1767 sync_filesystem(sb);
1768 btrfs_remount_prepare(fs_info);
1770 if (data) {
1771 void *new_sec_opts = NULL;
1773 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1774 if (!ret)
1775 ret = security_sb_remount(sb, new_sec_opts);
1776 security_free_mnt_opts(&new_sec_opts);
1777 if (ret)
1778 goto restore;
1781 ret = btrfs_parse_options(fs_info, data, *flags);
1782 if (ret)
1783 goto restore;
1785 btrfs_remount_begin(fs_info, old_opts, *flags);
1786 btrfs_resize_thread_pool(fs_info,
1787 fs_info->thread_pool_size, old_thread_pool_size);
1789 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1790 goto out;
1792 if (*flags & SB_RDONLY) {
1794 * this also happens on 'umount -rf' or on shutdown, when
1795 * the filesystem is busy.
1797 cancel_work_sync(&fs_info->async_reclaim_work);
1799 btrfs_discard_cleanup(fs_info);
1801 /* wait for the uuid_scan task to finish */
1802 down(&fs_info->uuid_tree_rescan_sem);
1803 /* avoid complains from lockdep et al. */
1804 up(&fs_info->uuid_tree_rescan_sem);
1806 sb->s_flags |= SB_RDONLY;
1809 * Setting SB_RDONLY will put the cleaner thread to
1810 * sleep at the next loop if it's already active.
1811 * If it's already asleep, we'll leave unused block
1812 * groups on disk until we're mounted read-write again
1813 * unless we clean them up here.
1815 btrfs_delete_unused_bgs(fs_info);
1817 btrfs_dev_replace_suspend_for_unmount(fs_info);
1818 btrfs_scrub_cancel(fs_info);
1819 btrfs_pause_balance(fs_info);
1821 ret = btrfs_commit_super(fs_info);
1822 if (ret)
1823 goto restore;
1824 } else {
1825 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1826 btrfs_err(fs_info,
1827 "Remounting read-write after error is not allowed");
1828 ret = -EINVAL;
1829 goto restore;
1831 if (fs_info->fs_devices->rw_devices == 0) {
1832 ret = -EACCES;
1833 goto restore;
1836 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1837 btrfs_warn(fs_info,
1838 "too many missing devices, writable remount is not allowed");
1839 ret = -EACCES;
1840 goto restore;
1843 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1844 btrfs_warn(fs_info,
1845 "mount required to replay tree-log, cannot remount read-write");
1846 ret = -EINVAL;
1847 goto restore;
1850 ret = btrfs_cleanup_fs_roots(fs_info);
1851 if (ret)
1852 goto restore;
1854 /* recover relocation */
1855 mutex_lock(&fs_info->cleaner_mutex);
1856 ret = btrfs_recover_relocation(root);
1857 mutex_unlock(&fs_info->cleaner_mutex);
1858 if (ret)
1859 goto restore;
1861 ret = btrfs_resume_balance_async(fs_info);
1862 if (ret)
1863 goto restore;
1865 ret = btrfs_resume_dev_replace_async(fs_info);
1866 if (ret) {
1867 btrfs_warn(fs_info, "failed to resume dev_replace");
1868 goto restore;
1871 btrfs_qgroup_rescan_resume(fs_info);
1873 if (!fs_info->uuid_root) {
1874 btrfs_info(fs_info, "creating UUID tree");
1875 ret = btrfs_create_uuid_tree(fs_info);
1876 if (ret) {
1877 btrfs_warn(fs_info,
1878 "failed to create the UUID tree %d",
1879 ret);
1880 goto restore;
1883 sb->s_flags &= ~SB_RDONLY;
1885 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1887 out:
1888 wake_up_process(fs_info->transaction_kthread);
1889 btrfs_remount_cleanup(fs_info, old_opts);
1890 return 0;
1892 restore:
1893 /* We've hit an error - don't reset SB_RDONLY */
1894 if (sb_rdonly(sb))
1895 old_flags |= SB_RDONLY;
1896 sb->s_flags = old_flags;
1897 fs_info->mount_opt = old_opts;
1898 fs_info->compress_type = old_compress_type;
1899 fs_info->max_inline = old_max_inline;
1900 btrfs_resize_thread_pool(fs_info,
1901 old_thread_pool_size, fs_info->thread_pool_size);
1902 fs_info->metadata_ratio = old_metadata_ratio;
1903 btrfs_remount_cleanup(fs_info, old_opts);
1904 return ret;
1907 /* Used to sort the devices by max_avail(descending sort) */
1908 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
1909 const void *dev_info2)
1911 if (((struct btrfs_device_info *)dev_info1)->max_avail >
1912 ((struct btrfs_device_info *)dev_info2)->max_avail)
1913 return -1;
1914 else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1915 ((struct btrfs_device_info *)dev_info2)->max_avail)
1916 return 1;
1917 else
1918 return 0;
1922 * sort the devices by max_avail, in which max free extent size of each device
1923 * is stored.(Descending Sort)
1925 static inline void btrfs_descending_sort_devices(
1926 struct btrfs_device_info *devices,
1927 size_t nr_devices)
1929 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1930 btrfs_cmp_device_free_bytes, NULL);
1934 * The helper to calc the free space on the devices that can be used to store
1935 * file data.
1937 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1938 u64 *free_bytes)
1940 struct btrfs_device_info *devices_info;
1941 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1942 struct btrfs_device *device;
1943 u64 type;
1944 u64 avail_space;
1945 u64 min_stripe_size;
1946 int num_stripes = 1;
1947 int i = 0, nr_devices;
1948 const struct btrfs_raid_attr *rattr;
1951 * We aren't under the device list lock, so this is racy-ish, but good
1952 * enough for our purposes.
1954 nr_devices = fs_info->fs_devices->open_devices;
1955 if (!nr_devices) {
1956 smp_mb();
1957 nr_devices = fs_info->fs_devices->open_devices;
1958 ASSERT(nr_devices);
1959 if (!nr_devices) {
1960 *free_bytes = 0;
1961 return 0;
1965 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1966 GFP_KERNEL);
1967 if (!devices_info)
1968 return -ENOMEM;
1970 /* calc min stripe number for data space allocation */
1971 type = btrfs_data_alloc_profile(fs_info);
1972 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1974 if (type & BTRFS_BLOCK_GROUP_RAID0)
1975 num_stripes = nr_devices;
1976 else if (type & BTRFS_BLOCK_GROUP_RAID1)
1977 num_stripes = 2;
1978 else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
1979 num_stripes = 3;
1980 else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
1981 num_stripes = 4;
1982 else if (type & BTRFS_BLOCK_GROUP_RAID10)
1983 num_stripes = 4;
1985 /* Adjust for more than 1 stripe per device */
1986 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1988 rcu_read_lock();
1989 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1990 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1991 &device->dev_state) ||
1992 !device->bdev ||
1993 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1994 continue;
1996 if (i >= nr_devices)
1997 break;
1999 avail_space = device->total_bytes - device->bytes_used;
2001 /* align with stripe_len */
2002 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2005 * In order to avoid overwriting the superblock on the drive,
2006 * btrfs starts at an offset of at least 1MB when doing chunk
2007 * allocation.
2009 * This ensures we have at least min_stripe_size free space
2010 * after excluding 1MB.
2012 if (avail_space <= SZ_1M + min_stripe_size)
2013 continue;
2015 avail_space -= SZ_1M;
2017 devices_info[i].dev = device;
2018 devices_info[i].max_avail = avail_space;
2020 i++;
2022 rcu_read_unlock();
2024 nr_devices = i;
2026 btrfs_descending_sort_devices(devices_info, nr_devices);
2028 i = nr_devices - 1;
2029 avail_space = 0;
2030 while (nr_devices >= rattr->devs_min) {
2031 num_stripes = min(num_stripes, nr_devices);
2033 if (devices_info[i].max_avail >= min_stripe_size) {
2034 int j;
2035 u64 alloc_size;
2037 avail_space += devices_info[i].max_avail * num_stripes;
2038 alloc_size = devices_info[i].max_avail;
2039 for (j = i + 1 - num_stripes; j <= i; j++)
2040 devices_info[j].max_avail -= alloc_size;
2042 i--;
2043 nr_devices--;
2046 kfree(devices_info);
2047 *free_bytes = avail_space;
2048 return 0;
2052 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2054 * If there's a redundant raid level at DATA block groups, use the respective
2055 * multiplier to scale the sizes.
2057 * Unused device space usage is based on simulating the chunk allocator
2058 * algorithm that respects the device sizes and order of allocations. This is
2059 * a close approximation of the actual use but there are other factors that may
2060 * change the result (like a new metadata chunk).
2062 * If metadata is exhausted, f_bavail will be 0.
2064 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2066 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2067 struct btrfs_super_block *disk_super = fs_info->super_copy;
2068 struct btrfs_space_info *found;
2069 u64 total_used = 0;
2070 u64 total_free_data = 0;
2071 u64 total_free_meta = 0;
2072 int bits = dentry->d_sb->s_blocksize_bits;
2073 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2074 unsigned factor = 1;
2075 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2076 int ret;
2077 u64 thresh = 0;
2078 int mixed = 0;
2080 rcu_read_lock();
2081 list_for_each_entry_rcu(found, &fs_info->space_info, list) {
2082 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2083 int i;
2085 total_free_data += found->disk_total - found->disk_used;
2086 total_free_data -=
2087 btrfs_account_ro_block_groups_free_space(found);
2089 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2090 if (!list_empty(&found->block_groups[i]))
2091 factor = btrfs_bg_type_to_factor(
2092 btrfs_raid_array[i].bg_flag);
2097 * Metadata in mixed block goup profiles are accounted in data
2099 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2100 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2101 mixed = 1;
2102 else
2103 total_free_meta += found->disk_total -
2104 found->disk_used;
2107 total_used += found->disk_used;
2110 rcu_read_unlock();
2112 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2113 buf->f_blocks >>= bits;
2114 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2116 /* Account global block reserve as used, it's in logical size already */
2117 spin_lock(&block_rsv->lock);
2118 /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2119 if (buf->f_bfree >= block_rsv->size >> bits)
2120 buf->f_bfree -= block_rsv->size >> bits;
2121 else
2122 buf->f_bfree = 0;
2123 spin_unlock(&block_rsv->lock);
2125 buf->f_bavail = div_u64(total_free_data, factor);
2126 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2127 if (ret)
2128 return ret;
2129 buf->f_bavail += div_u64(total_free_data, factor);
2130 buf->f_bavail = buf->f_bavail >> bits;
2133 * We calculate the remaining metadata space minus global reserve. If
2134 * this is (supposedly) smaller than zero, there's no space. But this
2135 * does not hold in practice, the exhausted state happens where's still
2136 * some positive delta. So we apply some guesswork and compare the
2137 * delta to a 4M threshold. (Practically observed delta was ~2M.)
2139 * We probably cannot calculate the exact threshold value because this
2140 * depends on the internal reservations requested by various
2141 * operations, so some operations that consume a few metadata will
2142 * succeed even if the Avail is zero. But this is better than the other
2143 * way around.
2145 thresh = SZ_4M;
2148 * We only want to claim there's no available space if we can no longer
2149 * allocate chunks for our metadata profile and our global reserve will
2150 * not fit in the free metadata space. If we aren't ->full then we
2151 * still can allocate chunks and thus are fine using the currently
2152 * calculated f_bavail.
2154 if (!mixed && block_rsv->space_info->full &&
2155 total_free_meta - thresh < block_rsv->size)
2156 buf->f_bavail = 0;
2158 buf->f_type = BTRFS_SUPER_MAGIC;
2159 buf->f_bsize = dentry->d_sb->s_blocksize;
2160 buf->f_namelen = BTRFS_NAME_LEN;
2162 /* We treat it as constant endianness (it doesn't matter _which_)
2163 because we want the fsid to come out the same whether mounted
2164 on a big-endian or little-endian host */
2165 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2166 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2167 /* Mask in the root object ID too, to disambiguate subvols */
2168 buf->f_fsid.val[0] ^=
2169 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2170 buf->f_fsid.val[1] ^=
2171 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2173 return 0;
2176 static void btrfs_kill_super(struct super_block *sb)
2178 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2179 kill_anon_super(sb);
2180 btrfs_free_fs_info(fs_info);
2183 static struct file_system_type btrfs_fs_type = {
2184 .owner = THIS_MODULE,
2185 .name = "btrfs",
2186 .mount = btrfs_mount,
2187 .kill_sb = btrfs_kill_super,
2188 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2191 static struct file_system_type btrfs_root_fs_type = {
2192 .owner = THIS_MODULE,
2193 .name = "btrfs",
2194 .mount = btrfs_mount_root,
2195 .kill_sb = btrfs_kill_super,
2196 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2199 MODULE_ALIAS_FS("btrfs");
2201 static int btrfs_control_open(struct inode *inode, struct file *file)
2204 * The control file's private_data is used to hold the
2205 * transaction when it is started and is used to keep
2206 * track of whether a transaction is already in progress.
2208 file->private_data = NULL;
2209 return 0;
2213 * Used by /dev/btrfs-control for devices ioctls.
2215 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2216 unsigned long arg)
2218 struct btrfs_ioctl_vol_args *vol;
2219 struct btrfs_device *device = NULL;
2220 int ret = -ENOTTY;
2222 if (!capable(CAP_SYS_ADMIN))
2223 return -EPERM;
2225 vol = memdup_user((void __user *)arg, sizeof(*vol));
2226 if (IS_ERR(vol))
2227 return PTR_ERR(vol);
2228 vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2230 switch (cmd) {
2231 case BTRFS_IOC_SCAN_DEV:
2232 mutex_lock(&uuid_mutex);
2233 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2234 &btrfs_root_fs_type);
2235 ret = PTR_ERR_OR_ZERO(device);
2236 mutex_unlock(&uuid_mutex);
2237 break;
2238 case BTRFS_IOC_FORGET_DEV:
2239 ret = btrfs_forget_devices(vol->name);
2240 break;
2241 case BTRFS_IOC_DEVICES_READY:
2242 mutex_lock(&uuid_mutex);
2243 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2244 &btrfs_root_fs_type);
2245 if (IS_ERR(device)) {
2246 mutex_unlock(&uuid_mutex);
2247 ret = PTR_ERR(device);
2248 break;
2250 ret = !(device->fs_devices->num_devices ==
2251 device->fs_devices->total_devices);
2252 mutex_unlock(&uuid_mutex);
2253 break;
2254 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2255 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2256 break;
2259 kfree(vol);
2260 return ret;
2263 static int btrfs_freeze(struct super_block *sb)
2265 struct btrfs_trans_handle *trans;
2266 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2267 struct btrfs_root *root = fs_info->tree_root;
2269 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2271 * We don't need a barrier here, we'll wait for any transaction that
2272 * could be in progress on other threads (and do delayed iputs that
2273 * we want to avoid on a frozen filesystem), or do the commit
2274 * ourselves.
2276 trans = btrfs_attach_transaction_barrier(root);
2277 if (IS_ERR(trans)) {
2278 /* no transaction, don't bother */
2279 if (PTR_ERR(trans) == -ENOENT)
2280 return 0;
2281 return PTR_ERR(trans);
2283 return btrfs_commit_transaction(trans);
2286 static int btrfs_unfreeze(struct super_block *sb)
2288 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2290 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2291 return 0;
2294 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2296 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2297 struct btrfs_fs_devices *cur_devices;
2298 struct btrfs_device *dev, *first_dev = NULL;
2299 struct list_head *head;
2302 * Lightweight locking of the devices. We should not need
2303 * device_list_mutex here as we only read the device data and the list
2304 * is protected by RCU. Even if a device is deleted during the list
2305 * traversals, we'll get valid data, the freeing callback will wait at
2306 * least until the rcu_read_unlock.
2308 rcu_read_lock();
2309 cur_devices = fs_info->fs_devices;
2310 while (cur_devices) {
2311 head = &cur_devices->devices;
2312 list_for_each_entry_rcu(dev, head, dev_list) {
2313 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2314 continue;
2315 if (!dev->name)
2316 continue;
2317 if (!first_dev || dev->devid < first_dev->devid)
2318 first_dev = dev;
2320 cur_devices = cur_devices->seed;
2323 if (first_dev)
2324 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2325 else
2326 WARN_ON(1);
2327 rcu_read_unlock();
2328 return 0;
2331 static const struct super_operations btrfs_super_ops = {
2332 .drop_inode = btrfs_drop_inode,
2333 .evict_inode = btrfs_evict_inode,
2334 .put_super = btrfs_put_super,
2335 .sync_fs = btrfs_sync_fs,
2336 .show_options = btrfs_show_options,
2337 .show_devname = btrfs_show_devname,
2338 .alloc_inode = btrfs_alloc_inode,
2339 .destroy_inode = btrfs_destroy_inode,
2340 .free_inode = btrfs_free_inode,
2341 .statfs = btrfs_statfs,
2342 .remount_fs = btrfs_remount,
2343 .freeze_fs = btrfs_freeze,
2344 .unfreeze_fs = btrfs_unfreeze,
2347 static const struct file_operations btrfs_ctl_fops = {
2348 .open = btrfs_control_open,
2349 .unlocked_ioctl = btrfs_control_ioctl,
2350 .compat_ioctl = compat_ptr_ioctl,
2351 .owner = THIS_MODULE,
2352 .llseek = noop_llseek,
2355 static struct miscdevice btrfs_misc = {
2356 .minor = BTRFS_MINOR,
2357 .name = "btrfs-control",
2358 .fops = &btrfs_ctl_fops
2361 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2362 MODULE_ALIAS("devname:btrfs-control");
2364 static int __init btrfs_interface_init(void)
2366 return misc_register(&btrfs_misc);
2369 static __cold void btrfs_interface_exit(void)
2371 misc_deregister(&btrfs_misc);
2374 static void __init btrfs_print_mod_info(void)
2376 static const char options[] = ""
2377 #ifdef CONFIG_BTRFS_DEBUG
2378 ", debug=on"
2379 #endif
2380 #ifdef CONFIG_BTRFS_ASSERT
2381 ", assert=on"
2382 #endif
2383 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2384 ", integrity-checker=on"
2385 #endif
2386 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2387 ", ref-verify=on"
2388 #endif
2390 pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2393 static int __init init_btrfs_fs(void)
2395 int err;
2397 btrfs_props_init();
2399 err = btrfs_init_sysfs();
2400 if (err)
2401 return err;
2403 btrfs_init_compress();
2405 err = btrfs_init_cachep();
2406 if (err)
2407 goto free_compress;
2409 err = extent_io_init();
2410 if (err)
2411 goto free_cachep;
2413 err = extent_state_cache_init();
2414 if (err)
2415 goto free_extent_io;
2417 err = extent_map_init();
2418 if (err)
2419 goto free_extent_state_cache;
2421 err = ordered_data_init();
2422 if (err)
2423 goto free_extent_map;
2425 err = btrfs_delayed_inode_init();
2426 if (err)
2427 goto free_ordered_data;
2429 err = btrfs_auto_defrag_init();
2430 if (err)
2431 goto free_delayed_inode;
2433 err = btrfs_delayed_ref_init();
2434 if (err)
2435 goto free_auto_defrag;
2437 err = btrfs_prelim_ref_init();
2438 if (err)
2439 goto free_delayed_ref;
2441 err = btrfs_end_io_wq_init();
2442 if (err)
2443 goto free_prelim_ref;
2445 err = btrfs_interface_init();
2446 if (err)
2447 goto free_end_io_wq;
2449 btrfs_init_lockdep();
2451 btrfs_print_mod_info();
2453 err = btrfs_run_sanity_tests();
2454 if (err)
2455 goto unregister_ioctl;
2457 err = register_filesystem(&btrfs_fs_type);
2458 if (err)
2459 goto unregister_ioctl;
2461 return 0;
2463 unregister_ioctl:
2464 btrfs_interface_exit();
2465 free_end_io_wq:
2466 btrfs_end_io_wq_exit();
2467 free_prelim_ref:
2468 btrfs_prelim_ref_exit();
2469 free_delayed_ref:
2470 btrfs_delayed_ref_exit();
2471 free_auto_defrag:
2472 btrfs_auto_defrag_exit();
2473 free_delayed_inode:
2474 btrfs_delayed_inode_exit();
2475 free_ordered_data:
2476 ordered_data_exit();
2477 free_extent_map:
2478 extent_map_exit();
2479 free_extent_state_cache:
2480 extent_state_cache_exit();
2481 free_extent_io:
2482 extent_io_exit();
2483 free_cachep:
2484 btrfs_destroy_cachep();
2485 free_compress:
2486 btrfs_exit_compress();
2487 btrfs_exit_sysfs();
2489 return err;
2492 static void __exit exit_btrfs_fs(void)
2494 btrfs_destroy_cachep();
2495 btrfs_delayed_ref_exit();
2496 btrfs_auto_defrag_exit();
2497 btrfs_delayed_inode_exit();
2498 btrfs_prelim_ref_exit();
2499 ordered_data_exit();
2500 extent_map_exit();
2501 extent_state_cache_exit();
2502 extent_io_exit();
2503 btrfs_interface_exit();
2504 btrfs_end_io_wq_exit();
2505 unregister_filesystem(&btrfs_fs_type);
2506 btrfs_exit_sysfs();
2507 btrfs_cleanup_fs_uuids();
2508 btrfs_exit_compress();
2511 late_initcall(init_btrfs_fs);
2512 module_exit(exit_btrfs_fs)
2514 MODULE_LICENSE("GPL");
2515 MODULE_SOFTDEP("pre: crc32c");
2516 MODULE_SOFTDEP("pre: xxhash64");
2517 MODULE_SOFTDEP("pre: sha256");
2518 MODULE_SOFTDEP("pre: blake2b-256");