1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/blkdev.h>
7 #include <linux/module.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>
32 #include "delayed-inode.h"
35 #include "transaction.h"
36 #include "btrfs_inode.h"
41 #include "compression.h"
42 #include "dev-replace.h"
43 #include "free-space-cache.h"
45 #include "space-info.h"
48 #include "tests/btrfs-tests.h"
49 #include "block-group.h"
54 #include "accessors.h"
61 #include "extent-tree.h"
62 #define CREATE_TRACE_POINTS
63 #include <trace/events/btrfs.h>
65 static const struct super_operations btrfs_super_ops;
66 static struct file_system_type btrfs_fs_type;
68 static void btrfs_put_super(struct super_block *sb)
70 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
72 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
76 /* Store the mount options related information. */
77 struct btrfs_fs_context {
84 unsigned long mount_opt;
85 unsigned long compress_type:4;
86 unsigned int compress_level;
96 Opt_compress_force_type,
110 Opt_rescan_uuid_tree,
113 Opt_space_cache_version,
121 Opt_user_subvol_rm_allowed,
131 /* Debugging options */
133 #ifdef CONFIG_BTRFS_DEBUG
134 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
136 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
143 Opt_fatal_errors_panic,
144 Opt_fatal_errors_bug,
147 static const struct constant_table btrfs_parameter_fatal_errors[] = {
148 { "panic", Opt_fatal_errors_panic },
149 { "bug", Opt_fatal_errors_bug },
158 static const struct constant_table btrfs_parameter_discard[] = {
159 { "sync", Opt_discard_sync },
160 { "async", Opt_discard_async },
169 static const struct constant_table btrfs_parameter_space_cache[] = {
170 { "v1", Opt_space_cache_v1 },
171 { "v2", Opt_space_cache_v2 },
176 Opt_rescue_usebackuproot,
177 Opt_rescue_nologreplay,
178 Opt_rescue_ignorebadroots,
179 Opt_rescue_ignoredatacsums,
180 Opt_rescue_parameter_all,
183 static 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 },
194 #ifdef CONFIG_BTRFS_DEBUG
196 Opt_fragment_parameter_data,
197 Opt_fragment_parameter_metadata,
198 Opt_fragment_parameter_all,
201 static 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 },
209 static 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),
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),
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),
254 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
255 fsparam_flag("ref_verify", Opt_ref_verify),
260 /* No support for restricting writes to btrfs devices yet... */
261 static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
263 return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
266 static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
268 struct btrfs_fs_context *ctx = fc->fs_private;
269 struct fs_parse_result result;
272 opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
278 btrfs_set_opt(ctx->mount_opt, DEGRADED);
280 case Opt_subvol_empty:
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").
288 kfree(ctx->subvol_name);
289 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
290 if (!ctx->subvol_name)
294 ctx->subvol_objectid = result.uint_64;
296 /* subvolid=0 means give me the original fs_tree. */
297 if (!ctx->subvol_objectid)
298 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
301 struct btrfs_device *device;
302 blk_mode_t mode = btrfs_open_mode(fc);
304 mutex_lock(&uuid_mutex);
305 device = btrfs_scan_one_device(param->string, mode, false);
306 mutex_unlock(&uuid_mutex);
308 return PTR_ERR(device);
312 if (result.negated) {
313 btrfs_set_opt(ctx->mount_opt, NODATASUM);
315 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
316 btrfs_clear_opt(ctx->mount_opt, NODATASUM);
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);
326 btrfs_clear_opt(ctx->mount_opt, NODATACOW);
329 case Opt_compress_force:
330 case Opt_compress_force_type:
331 btrfs_set_opt(ctx->mount_opt, FORCE_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,
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,
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);
369 btrfs_err(NULL, "unrecognized compression value %s",
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);
380 btrfs_set_opt(ctx->mount_opt, SSD);
381 btrfs_clear_opt(ctx->mount_opt, NOSSD);
385 if (result.negated) {
386 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
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);
395 btrfs_set_opt(ctx->mount_opt, NOBARRIER);
397 btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
399 case Opt_thread_pool:
400 if (result.uint_32 == 0) {
401 btrfs_err(NULL, "invalid value 0 for thread_pool");
404 ctx->thread_pool_size = result.uint_32;
407 ctx->max_inline = memparse(param->string, NULL);
410 if (result.negated) {
411 fc->sb_flags &= ~SB_POSIXACL;
413 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
414 fc->sb_flags |= SB_POSIXACL;
416 btrfs_err(NULL, "support for ACL not compiled in");
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.
427 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
428 fc->sb_flags_mask |= SB_POSIXACL;
432 btrfs_set_opt(ctx->mount_opt, NOTREELOG);
434 btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
436 case Opt_nologreplay:
438 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
439 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
441 case Opt_flushoncommit:
443 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
445 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
448 ctx->metadata_ratio = result.uint_32;
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);
456 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
457 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
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);
466 case Opt_discard_async:
467 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
468 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
471 btrfs_err(NULL, "unrecognized discard mode value %s",
475 btrfs_clear_opt(ctx->mount_opt, NODISCARD);
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);
483 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
484 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
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);
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);
498 btrfs_err(NULL, "unrecognized space_cache value %s",
503 case Opt_rescan_uuid_tree:
504 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
506 case Opt_clear_cache:
507 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
509 case Opt_user_subvol_rm_allowed:
510 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
512 case Opt_enospc_debug:
514 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
516 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
520 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
522 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
524 case Opt_usebackuproot:
526 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
527 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
529 /* If we're loading the backup roots we can't trust the space cache. */
530 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
532 case Opt_skip_balance:
533 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
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);
540 case Opt_fatal_errors_bug:
541 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
544 btrfs_err(NULL, "unrecognized fatal_errors value %s",
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;
555 switch (result.uint_32) {
556 case Opt_rescue_usebackuproot:
557 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
559 case Opt_rescue_nologreplay:
560 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
562 case Opt_rescue_ignorebadroots:
563 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
565 case Opt_rescue_ignoredatacsums:
566 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
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);
574 btrfs_info(NULL, "unrecognized rescue option '%s'",
579 #ifdef CONFIG_BTRFS_DEBUG
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);
586 case Opt_fragment_parameter_metadata:
587 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
589 case Opt_fragment_parameter_data:
590 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
593 btrfs_info(NULL, "unrecognized fragment option '%s'",
599 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
601 btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
605 btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
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
617 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
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);
624 static bool check_ro_option(struct btrfs_fs_info *fs_info,
625 unsigned long mount_opt, unsigned long opt,
626 const char *opt_name)
628 if (mount_opt & opt) {
629 btrfs_err(fs_info, "%s must be used with ro mount option",
636 bool btrfs_check_options(struct btrfs_fs_info *info, unsigned long *mount_opt,
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")))
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");
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");
659 if (btrfs_check_mountopts_zoned(info, mount_opt))
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");
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.
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.
684 void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
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)) {
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);
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.
700 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
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);
710 if (btrfs_test_opt(fs_info, SPACE_CACHE))
713 if (btrfs_test_opt(fs_info, NOSPACECACHE))
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.
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);
726 static void set_device_specific_options(struct btrfs_fs_info *fs_info)
728 if (!btrfs_test_opt(fs_info, NOSSD) &&
729 !fs_info->fs_devices->rotating)
730 btrfs_set_opt(fs_info->mount_opt, SSD);
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.
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.
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);
749 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
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;
763 path = btrfs_alloc_path();
769 name = kmalloc(PATH_MAX, GFP_KERNEL);
774 ptr = name + PATH_MAX - 1;
778 * Walk up the subvolume trees in the tree of tree roots by root
779 * backrefs until we hit the top-level subvolume.
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;
786 ret = btrfs_search_backwards(root, &key, path);
789 } else if (ret > 0) {
794 subvol_objectid = key.offset;
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);
804 read_extent_buffer(path->nodes[0], ptr + 1,
805 (unsigned long)(root_ref + 1), len);
807 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
808 btrfs_release_path(path);
810 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
811 if (IS_ERR(fs_root)) {
812 ret = PTR_ERR(fs_root);
818 * Walk up the filesystem tree by inode refs until we hit the
821 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
822 key.objectid = dirid;
823 key.type = BTRFS_INODE_REF_KEY;
824 key.offset = (u64)-1;
826 ret = btrfs_search_backwards(fs_root, &key, path);
829 } else if (ret > 0) {
836 inode_ref = btrfs_item_ptr(path->nodes[0],
838 struct btrfs_inode_ref);
839 len = btrfs_inode_ref_name_len(path->nodes[0],
846 read_extent_buffer(path->nodes[0], ptr + 1,
847 (unsigned long)(inode_ref + 1), len);
849 btrfs_release_path(path);
851 btrfs_put_root(fs_root);
855 btrfs_free_path(path);
856 if (ptr == name + PATH_MAX - 1) {
860 memmove(name, ptr, name + PATH_MAX - ptr);
865 btrfs_put_root(fs_root);
866 btrfs_free_path(path);
871 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
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);
880 path = btrfs_alloc_path();
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
889 dir_id = btrfs_super_root_dir(fs_info->super_copy);
890 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
892 btrfs_free_path(path);
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.
901 btrfs_free_path(path);
902 *objectid = BTRFS_FS_TREE_OBJECTID;
906 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
907 btrfs_free_path(path);
908 *objectid = location.objectid;
912 static int btrfs_fill_super(struct super_block *sb,
913 struct btrfs_fs_devices *fs_devices,
917 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
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;
928 sb->s_xattr = btrfs_xattr_handlers;
930 sb->s_iflags |= SB_I_CGROUPWB;
932 err = super_setup_bdi(sb);
934 btrfs_err(fs_info, "super_setup_bdi failed");
938 err = open_ctree(sb, fs_devices, (char *)data);
940 btrfs_err(fs_info, "open_ctree failed");
944 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
946 err = PTR_ERR(inode);
947 btrfs_handle_fs_error(fs_info, err, NULL);
951 sb->s_root = d_make_root(inode);
957 sb->s_flags |= SB_ACTIVE;
961 close_ctree(fs_info);
965 int btrfs_sync_fs(struct super_block *sb, int wait)
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;
971 trace_btrfs_sync_fs(fs_info, wait);
974 filemap_flush(fs_info->btree_inode->i_mapping);
978 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
980 trans = btrfs_attach_transaction_barrier(root);
982 /* no transaction, don't bother */
983 if (PTR_ERR(trans) == -ENOENT) {
985 * Exit unless we have some pending changes
986 * that need to go through commit
988 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
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.
997 if (sb_start_write_trylock(sb))
1001 trans = btrfs_start_transaction(root, 0);
1004 return PTR_ERR(trans);
1006 return btrfs_commit_transaction(trans);
1009 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1011 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1015 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
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;
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);
1040 seq_printf(seq, ",compress=%s", compress_type);
1041 if (info->compress_level)
1042 seq_printf(seq, ":%d", info->compress_level);
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");
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");
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\\");
1113 * subvolumes are identified by ino 256
1115 static inline int is_subvolume_inode(struct inode *inode)
1117 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1122 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1123 struct vfsmount *mnt)
1125 struct dentry *root;
1129 if (!subvol_objectid) {
1130 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1133 root = ERR_PTR(ret);
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);
1147 root = mount_subtree(mnt, subvol_name);
1148 /* mount_subtree() drops our reference on the vfsmount. */
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;
1158 if (!is_subvolume_inode(root_inode)) {
1159 btrfs_err(fs_info, "'%s' is not a valid subvolume",
1163 if (subvol_objectid && root_objectid != subvol_objectid) {
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.
1170 "subvol '%s' does not match subvolid %llu",
1171 subvol_name, subvol_objectid);
1176 root = ERR_PTR(ret);
1177 deactivate_locked_super(s);
1187 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1188 u32 new_pool_size, u32 old_pool_size)
1190 if (new_pool_size == old_pool_size)
1193 fs_info->thread_pool_size = new_pool_size;
1195 btrfs_info(fs_info, "resize thread pool %d -> %d",
1196 old_pool_size, new_pool_size);
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);
1208 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1209 unsigned long old_opts, int flags)
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);
1222 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1223 unsigned long old_opts)
1225 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1228 * We need to cleanup all defragable inodes if the autodefragment is
1229 * close or the filesystem is read only.
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);
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);
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);
1249 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1253 if (BTRFS_FS_ERROR(fs_info)) {
1255 "remounting read-write after error is not allowed");
1259 if (fs_info->fs_devices->rw_devices == 0)
1262 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1264 "too many missing devices, writable remount is not allowed");
1268 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1270 "mount required to replay tree-log, cannot remount read-write");
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.
1279 ret = btrfs_start_pre_rw_mount(fs_info);
1283 btrfs_clear_sb_rdonly(fs_info->sb);
1285 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1288 * If we've gone from readonly -> read-write, we need to get our
1289 * sync/async discard lists in the right state.
1291 btrfs_discard_resume(fs_info);
1296 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1299 * This also happens on 'umount -rf' or on shutdown, when the
1300 * filesystem is busy.
1302 cancel_work_sync(&fs_info->async_reclaim_work);
1303 cancel_work_sync(&fs_info->async_data_reclaim_work);
1305 btrfs_discard_cleanup(fs_info);
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);
1312 btrfs_set_sb_rdonly(fs_info->sb);
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.
1320 btrfs_delete_unused_bgs(fs_info);
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()).
1332 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
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.
1341 btrfs_run_delayed_iputs(fs_info);
1343 btrfs_dev_replace_suspend_for_unmount(fs_info);
1344 btrfs_scrub_cancel(fs_info);
1345 btrfs_pause_balance(fs_info);
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.
1353 btrfs_qgroup_wait_for_completion(fs_info, false);
1355 return btrfs_commit_super(fs_info);
1358 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
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;
1369 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
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;
1380 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
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); \
1387 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
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); \
1394 static void btrfs_emit_options(struct btrfs_fs_info *info,
1395 struct btrfs_fs_context *old)
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");
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");
1429 /* Did the compression settings change? */
1430 if (btrfs_test_opt(info, COMPRESS) &&
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);
1438 btrfs_info(info, "%s %s compression, level %d",
1439 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1440 compress_type, info->compress_level);
1443 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1444 btrfs_info(info, "max_inline set to %llu", info->max_inline);
1447 static int btrfs_reconfigure(struct fs_context *fc)
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;
1454 bool mount_reconfigure = (fc->s_fs_info != NULL);
1456 btrfs_info_to_ctx(fs_info, &old_ctx);
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.
1463 if (mount_reconfigure)
1464 ctx->mount_opt = old_ctx.mount_opt;
1466 sync_filesystem(sb);
1467 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1469 if (!mount_reconfigure &&
1470 !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1473 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
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);
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))) {
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);
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);
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);
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
1511 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1512 fc->sb_flags_mask |= SB_POSIXACL;
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);
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);
1528 /* Used to sort the devices by max_avail(descending sort) */
1529 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1531 const struct btrfs_device_info *dev_info1 = a;
1532 const struct btrfs_device_info *dev_info2 = b;
1534 if (dev_info1->max_avail > dev_info2->max_avail)
1536 else if (dev_info1->max_avail < dev_info2->max_avail)
1542 * sort the devices by max_avail, in which max free extent size of each device
1543 * is stored.(Descending Sort)
1545 static inline void btrfs_descending_sort_devices(
1546 struct btrfs_device_info *devices,
1549 sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1550 btrfs_cmp_device_free_bytes, NULL);
1554 * The helper to calc the free space on the devices that can be used to store
1557 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1560 struct btrfs_device_info *devices_info;
1561 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1562 struct btrfs_device *device;
1565 u64 min_stripe_size;
1566 int num_stripes = 1;
1567 int i = 0, nr_devices;
1568 const struct btrfs_raid_attr *rattr;
1571 * We aren't under the device list lock, so this is racy-ish, but good
1572 * enough for our purposes.
1574 nr_devices = fs_info->fs_devices->open_devices;
1577 nr_devices = fs_info->fs_devices->open_devices;
1585 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
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)];
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)
1601 /* Adjust for more than 1 stripe per device */
1602 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
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) ||
1609 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1612 if (i >= nr_devices)
1615 avail_space = device->total_bytes - device->bytes_used;
1617 /* align with stripe_len */
1618 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1621 * Ensure we have at least min_stripe_size on top of the
1622 * reserved space on the device.
1624 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1627 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1629 devices_info[i].dev = device;
1630 devices_info[i].max_avail = avail_space;
1638 btrfs_descending_sort_devices(devices_info, nr_devices);
1642 while (nr_devices >= rattr->devs_min) {
1643 num_stripes = min(num_stripes, nr_devices);
1645 if (devices_info[i].max_avail >= min_stripe_size) {
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;
1658 kfree(devices_info);
1659 *free_bytes = avail_space;
1664 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1666 * If there's a redundant raid level at DATA block groups, use the respective
1667 * multiplier to scale the sizes.
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).
1674 * If metadata is exhausted, f_bavail will be 0.
1676 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
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;
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;
1692 list_for_each_entry(found, &fs_info->space_info, list) {
1693 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1696 total_free_data += found->disk_total - found->disk_used;
1698 btrfs_account_ro_block_groups_free_space(found);
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);
1708 * Metadata in mixed block group profiles are accounted in data
1710 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1711 if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1714 total_free_meta += found->disk_total -
1718 total_used += found->disk_used;
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);
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;
1732 spin_unlock(&block_rsv->lock);
1734 buf->f_bavail = div_u64(total_free_data, factor);
1735 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1738 buf->f_bavail += div_u64(total_free_data, factor);
1739 buf->f_bavail = buf->f_bavail >> bits;
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.)
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
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.
1763 if (!mixed && block_rsv->space_info->full &&
1764 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1767 buf->f_type = BTRFS_SUPER_MAGIC;
1768 buf->f_bsize = fs_info->sectorsize;
1769 buf->f_namelen = BTRFS_NAME_LEN;
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;
1785 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1787 struct btrfs_fs_info *p = fc->s_fs_info;
1788 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1790 return fs_info->fs_devices == p->fs_devices;
1793 static int btrfs_get_tree_super(struct fs_context *fc)
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);
1804 btrfs_ctx_to_info(fs_info, ctx);
1805 mutex_lock(&uuid_mutex);
1808 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1809 * either a valid device or an error.
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);
1818 fs_devices = device->fs_devices;
1819 fs_info->fs_devices = fs_devices;
1821 ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1822 mutex_unlock(&uuid_mutex);
1826 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1831 bdev = fs_devices->latest_dev->bdev;
1834 * From now on the error handling is not straightforward.
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.
1842 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1848 set_device_specific_options(fs_info);
1851 btrfs_close_devices(fs_devices);
1852 if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
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);
1862 deactivate_locked_super(sb);
1866 btrfs_clear_oneshot_options(fs_info);
1868 fc->root = dget(sb->s_root);
1872 btrfs_close_devices(fs_devices);
1877 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1878 * with different ro/rw options") the following works:
1880 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1881 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1883 * which looks nice and innocent but is actually pretty intricate and deserves
1886 * On another filesystem a subvolume mount is close to something like:
1888 * (iii) # create rw superblock + initial mount
1889 * mount -t xfs /dev/sdb /opt/
1891 * # create ro bind mount
1892 * mount --bind -o ro /opt/foo /mnt/foo
1894 * # unmount initial mount
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.
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
1906 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1908 * the MS_RDONLY flag being specified had two effects:
1910 * (1) MNT_READONLY was raised -> the resulting mount got
1911 * @mnt->mnt_flags |= MNT_READONLY raised.
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).
1917 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1918 * subtree mounted ro.
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.
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.
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).
1933 * Enter the new mount API. The new mount API disambiguates making a mount ro
1934 * and making a superblock ro.
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.
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
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
1948 * So, if the superblock creation request comes from the new mount API the
1949 * caller must have explicitly done:
1951 * fsconfig(FSCONFIG_SET_FLAG, "ro")
1952 * fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
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.
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.
1962 * The new mount API has no reason to support this hack.
1964 static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1966 struct vfsmount *mnt;
1968 const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
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.
1976 fc->sb_flags |= SB_RDONLY;
1978 fc->sb_flags &= ~SB_RDONLY;
1984 if (!fc->oldapi || !ro2rw)
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);
1994 return ERR_PTR(ret);
1999 static int btrfs_get_tree_subvol(struct fs_context *fc)
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;
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.
2015 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
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);
2025 btrfs_init_fs_info(fs_info);
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);
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.
2037 dup_fc->s_fs_info = fs_info;
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
2044 security_free_mnt_opts(&fc->security);
2045 fc->security = NULL;
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);
2052 return PTR_ERR(mnt);
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.
2059 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2060 ctx->subvol_name = NULL;
2062 return PTR_ERR(dentry);
2068 static int btrfs_get_tree(struct fs_context *fc)
2071 * Since we use mount_subtree to mount the default/specified subvol, we
2072 * have to do mounts in two steps.
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.
2084 return btrfs_get_tree_super(fc);
2085 return btrfs_get_tree_subvol(fc);
2088 static void btrfs_kill_super(struct super_block *sb)
2090 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2091 kill_anon_super(sb);
2092 btrfs_free_fs_info(fs_info);
2095 static void btrfs_free_fs_context(struct fs_context *fc)
2097 struct btrfs_fs_context *ctx = fc->fs_private;
2098 struct btrfs_fs_info *fs_info = fc->s_fs_info;
2101 btrfs_free_fs_info(fs_info);
2103 if (ctx && refcount_dec_and_test(&ctx->refs)) {
2104 kfree(ctx->subvol_name);
2109 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2111 struct btrfs_fs_context *ctx = src_fc->fs_private;
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.
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.
2121 refcount_inc(&ctx->refs);
2122 fc->fs_private = ctx;
2123 fc->source = src_fc->source;
2124 src_fc->source = NULL;
2128 static 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,
2136 static int btrfs_init_fs_context(struct fs_context *fc)
2138 struct btrfs_fs_context *ctx;
2140 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2144 refcount_set(&ctx->refs, 1);
2145 fc->fs_private = ctx;
2146 fc->ops = &btrfs_fs_context_ops;
2148 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2149 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
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;
2157 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
2158 fc->sb_flags |= SB_POSIXACL;
2160 fc->sb_flags |= SB_I_VERSION;
2165 static struct file_system_type btrfs_fs_type = {
2166 .owner = THIS_MODULE,
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,
2174 MODULE_ALIAS_FS("btrfs");
2176 static int btrfs_control_open(struct inode *inode, struct file *file)
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.
2183 file->private_data = NULL;
2188 * Used by /dev/btrfs-control for devices ioctls.
2190 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2193 struct btrfs_ioctl_vol_args *vol;
2194 struct btrfs_device *device = NULL;
2198 if (!capable(CAP_SYS_ADMIN))
2201 vol = memdup_user((void __user *)arg, sizeof(*vol));
2203 return PTR_ERR(vol);
2204 ret = btrfs_check_ioctl_vol_args_path(vol);
2209 case BTRFS_IOC_SCAN_DEV:
2210 mutex_lock(&uuid_mutex);
2212 * Scanning outside of mount can return NULL which would turn
2213 * into 0 error code.
2215 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2216 ret = PTR_ERR_OR_ZERO(device);
2217 mutex_unlock(&uuid_mutex);
2219 case BTRFS_IOC_FORGET_DEV:
2220 if (vol->name[0] != 0) {
2221 ret = lookup_bdev(vol->name, &devt);
2225 ret = btrfs_forget_devices(devt);
2227 case BTRFS_IOC_DEVICES_READY:
2228 mutex_lock(&uuid_mutex);
2230 * Scanning outside of mount can return NULL which would turn
2231 * into 0 error code.
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);
2239 ret = !(device->fs_devices->num_devices ==
2240 device->fs_devices->total_devices);
2241 mutex_unlock(&uuid_mutex);
2243 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2244 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2253 static int btrfs_freeze(struct super_block *sb)
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;
2259 set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
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
2266 trans = btrfs_attach_transaction_barrier(root);
2267 if (IS_ERR(trans)) {
2268 /* no transaction, don't bother */
2269 if (PTR_ERR(trans) == -ENOENT)
2271 return PTR_ERR(trans);
2273 return btrfs_commit_transaction(trans);
2276 static int check_dev_super(struct btrfs_device *dev)
2278 struct btrfs_fs_info *fs_info = dev->fs_info;
2279 struct btrfs_super_block *sb;
2284 /* This should be called with fs still frozen. */
2285 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2287 /* Missing dev, no need to check. */
2291 /* Only need to check the primary super block. */
2292 sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
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));
2305 if (btrfs_check_super_csum(fs_info, sb)) {
2306 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2311 /* Btrfs_validate_super() includes fsid check against super->fsid. */
2312 ret = btrfs_validate_super(fs_info, sb, 0);
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);
2324 btrfs_release_disk_super(sb);
2328 static int btrfs_unfreeze(struct super_block *sb)
2330 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2331 struct btrfs_device *device;
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.
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.
2342 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2343 ret = check_dev_super(device);
2345 btrfs_handle_fs_error(fs_info, ret,
2346 "super block on devid %llu got modified unexpectedly",
2351 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
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.
2361 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2363 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
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.
2371 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2377 static 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,
2392 static 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,
2400 static struct miscdevice btrfs_misc = {
2401 .minor = BTRFS_MINOR,
2402 .name = "btrfs-control",
2403 .fops = &btrfs_ctl_fops
2406 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2407 MODULE_ALIAS("devname:btrfs-control");
2409 static int __init btrfs_interface_init(void)
2411 return misc_register(&btrfs_misc);
2414 static __cold void btrfs_interface_exit(void)
2416 misc_deregister(&btrfs_misc);
2419 static int __init btrfs_print_mod_info(void)
2421 static const char options[] = ""
2422 #ifdef CONFIG_BTRFS_DEBUG
2425 #ifdef CONFIG_BTRFS_ASSERT
2428 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2431 #ifdef CONFIG_BLK_DEV_ZONED
2436 #ifdef CONFIG_FS_VERITY
2442 pr_info("Btrfs loaded%s\n", options);
2446 static int register_btrfs(void)
2448 return register_filesystem(&btrfs_fs_type);
2451 static void unregister_btrfs(void)
2453 unregister_filesystem(&btrfs_fs_type);
2456 /* Helper structure for long init/exit functions. */
2457 struct init_sequence {
2458 int (*init_func)(void);
2459 /* Can be NULL if the init_func doesn't need cleanup. */
2460 void (*exit_func)(void);
2463 static const struct init_sequence mod_init_seq[] = {
2465 .init_func = btrfs_props_init,
2468 .init_func = btrfs_init_sysfs,
2469 .exit_func = btrfs_exit_sysfs,
2471 .init_func = btrfs_init_compress,
2472 .exit_func = btrfs_exit_compress,
2474 .init_func = btrfs_init_cachep,
2475 .exit_func = btrfs_destroy_cachep,
2477 .init_func = btrfs_transaction_init,
2478 .exit_func = btrfs_transaction_exit,
2480 .init_func = btrfs_ctree_init,
2481 .exit_func = btrfs_ctree_exit,
2483 .init_func = btrfs_free_space_init,
2484 .exit_func = btrfs_free_space_exit,
2486 .init_func = extent_state_init_cachep,
2487 .exit_func = extent_state_free_cachep,
2489 .init_func = extent_buffer_init_cachep,
2490 .exit_func = extent_buffer_free_cachep,
2492 .init_func = btrfs_bioset_init,
2493 .exit_func = btrfs_bioset_exit,
2495 .init_func = extent_map_init,
2496 .exit_func = extent_map_exit,
2498 .init_func = ordered_data_init,
2499 .exit_func = ordered_data_exit,
2501 .init_func = btrfs_delayed_inode_init,
2502 .exit_func = btrfs_delayed_inode_exit,
2504 .init_func = btrfs_auto_defrag_init,
2505 .exit_func = btrfs_auto_defrag_exit,
2507 .init_func = btrfs_delayed_ref_init,
2508 .exit_func = btrfs_delayed_ref_exit,
2510 .init_func = btrfs_prelim_ref_init,
2511 .exit_func = btrfs_prelim_ref_exit,
2513 .init_func = btrfs_interface_init,
2514 .exit_func = btrfs_interface_exit,
2516 .init_func = btrfs_print_mod_info,
2519 .init_func = btrfs_run_sanity_tests,
2522 .init_func = register_btrfs,
2523 .exit_func = unregister_btrfs,
2527 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2529 static __always_inline void btrfs_exit_btrfs_fs(void)
2533 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2534 if (!mod_init_result[i])
2536 if (mod_init_seq[i].exit_func)
2537 mod_init_seq[i].exit_func();
2538 mod_init_result[i] = false;
2542 static void __exit exit_btrfs_fs(void)
2544 btrfs_exit_btrfs_fs();
2545 btrfs_cleanup_fs_uuids();
2548 static int __init init_btrfs_fs(void)
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();
2557 btrfs_exit_btrfs_fs();
2560 mod_init_result[i] = true;
2565 late_initcall(init_btrfs_fs);
2566 module_exit(exit_btrfs_fs)
2568 MODULE_LICENSE("GPL");
2569 MODULE_SOFTDEP("pre: crc32c");
2570 MODULE_SOFTDEP("pre: xxhash64");
2571 MODULE_SOFTDEP("pre: sha256");
2572 MODULE_SOFTDEP("pre: blake2b-256");