bcachefs: sb-clean.c

Pull code for bch_sb_field_clean out into its own file.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

8 files changed, 424 insertions, 396 deletions

diff --git a/fs/bcachefs/Makefile b/fs/bcachefs/Makefile
index 7baf27595343..0ee755864e68 100644
--- a/fs/bcachefs/Makefile
+++ b/fs/bcachefs/Makefile
@@ -64,6 +64,7 @@ bcachefs-y		:=	\
 	recovery.o		\
 	reflink.o		\
 	replicas.o		\
+	sb-clean.o		\
 	sb-members.o		\
 	siphash.o		\
 	six.o			\
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index 42c9700e6d26..378b3f9170d4 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -14,6 +14,7 @@
 #include "journal_reclaim.h"
 #include "journal_seq_blacklist.h"
 #include "replicas.h"
+#include "sb-clean.h"
 #include "trace.h"
 
 static struct nonce journal_nonce(const struct jset *jset)
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 5dbe1b273b71..048a62f90b06 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -23,6 +23,7 @@
 #include "quota.h"
 #include "recovery.h"
 #include "replicas.h"
+#include "sb-clean.h"
 #include "subvolume.h"
 #include "super-io.h"
 
@@ -846,134 +847,6 @@ static int journal_replay_early(struct bch_fs *c,
 
 /* sb clean section: */
 
-static struct bkey_i *btree_root_find(struct bch_fs *c,
-				      struct bch_sb_field_clean *clean,
-				      struct jset *j,
-				      enum btree_id id, unsigned *level)
-{
-	struct bkey_i *k;
-	struct jset_entry *entry, *start, *end;
-
-	if (clean) {
-		start = clean->start;
-		end = vstruct_end(&clean->field);
-	} else {
-		start = j->start;
-		end = vstruct_last(j);
-	}
-
-	for (entry = start; entry < end; entry = vstruct_next(entry))
-		if (entry->type == BCH_JSET_ENTRY_btree_root &&
-		    entry->btree_id == id)
-			goto found;
-
-	return NULL;
-found:
-	if (!entry->u64s)
-		return ERR_PTR(-EINVAL);
-
-	k = entry->start;
-	*level = entry->level;
-	return k;
-}
-
-static int verify_superblock_clean(struct bch_fs *c,
-				   struct bch_sb_field_clean **cleanp,
-				   struct jset *j)
-{
-	unsigned i;
-	struct bch_sb_field_clean *clean = *cleanp;
-	struct printbuf buf1 = PRINTBUF;
-	struct printbuf buf2 = PRINTBUF;
-	int ret = 0;
-
-	if (mustfix_fsck_err_on(j->seq != clean->journal_seq, c,
-			"superblock journal seq (%llu) doesn't match journal (%llu) after clean shutdown",
-			le64_to_cpu(clean->journal_seq),
-			le64_to_cpu(j->seq))) {
-		kfree(clean);
-		*cleanp = NULL;
-		return 0;
-	}
-
-	for (i = 0; i < BTREE_ID_NR; i++) {
-		struct bkey_i *k1, *k2;
-		unsigned l1 = 0, l2 = 0;
-
-		k1 = btree_root_find(c, clean, NULL, i, &l1);
-		k2 = btree_root_find(c, NULL, j, i, &l2);
-
-		if (!k1 && !k2)
-			continue;
-
-		printbuf_reset(&buf1);
-		printbuf_reset(&buf2);
-
-		if (k1)
-			bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(k1));
-		else
-			prt_printf(&buf1, "(none)");
-
-		if (k2)
-			bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(k2));
-		else
-			prt_printf(&buf2, "(none)");
-
-		mustfix_fsck_err_on(!k1 || !k2 ||
-				    IS_ERR(k1) ||
-				    IS_ERR(k2) ||
-				    k1->k.u64s != k2->k.u64s ||
-				    memcmp(k1, k2, bkey_bytes(&k1->k)) ||
-				    l1 != l2, c,
-			"superblock btree root %u doesn't match journal after clean shutdown\n"
-			"sb:      l=%u %s\n"
-			"journal: l=%u %s\n", i,
-			l1, buf1.buf,
-			l2, buf2.buf);
-	}
-fsck_err:
-	printbuf_exit(&buf2);
-	printbuf_exit(&buf1);
-	return ret;
-}
-
-static struct bch_sb_field_clean *read_superblock_clean(struct bch_fs *c)
-{
-	struct bch_sb_field_clean *clean, *sb_clean;
-	int ret;
-
-	mutex_lock(&c->sb_lock);
-	sb_clean = bch2_sb_get_clean(c->disk_sb.sb);
-
-	if (fsck_err_on(!sb_clean, c,
-			"superblock marked clean but clean section not present")) {
-		SET_BCH_SB_CLEAN(c->disk_sb.sb, false);
-		c->sb.clean = false;
-		mutex_unlock(&c->sb_lock);
-		return NULL;
-	}
-
-	clean = kmemdup(sb_clean, vstruct_bytes(&sb_clean->field),
-			GFP_KERNEL);
-	if (!clean) {
-		mutex_unlock(&c->sb_lock);
-		return ERR_PTR(-BCH_ERR_ENOMEM_read_superblock_clean);
-	}
-
-	ret = bch2_sb_clean_validate_late(c, clean, READ);
-	if (ret) {
-		mutex_unlock(&c->sb_lock);
-		return ERR_PTR(ret);
-	}
-
-	mutex_unlock(&c->sb_lock);
-
-	return clean;
-fsck_err:
-	mutex_unlock(&c->sb_lock);
-	return ERR_PTR(ret);
-}
-
 static bool btree_id_is_alloc(enum btree_id id)
 {
 	switch (id) {
@@ -1297,17 +1170,17 @@ int bch2_fs_recovery(struct bch_fs *c)
 	bool write_sb = false;
 	int ret = 0;
 
-	if (c->sb.clean)
-		clean = read_superblock_clean(c);
-	ret = PTR_ERR_OR_ZERO(clean);
-	if (ret)
-		goto err;
+	if (c->sb.clean) {
+		clean = bch2_read_superblock_clean(c);
+		ret = PTR_ERR_OR_ZERO(clean);
+		if (ret)
+			goto err;
 
-	if (c->sb.clean)
 		bch_info(c, "recovering from clean shutdown, journal seq %llu",
 			 le64_to_cpu(clean->journal_seq));
-	else
+	} else {
 		bch_info(c, "recovering from unclean shutdown");
+	}
 
 	if (!(c->sb.features & (1ULL << BCH_FEATURE_new_extent_overwrite))) {
 		bch_err(c, "feature new_extent_overwrite not set, filesystem no longer supported");
@@ -1386,7 +1259,7 @@ int bch2_fs_recovery(struct bch_fs *c)
 			goto err;
 
 		if (c->sb.clean && last_journal_entry) {
-			ret = verify_superblock_clean(c, &clean,
+			ret = bch2_verify_superblock_clean(c, &clean,
 						      last_journal_entry);
 			if (ret)
 				goto err;
diff --git a/fs/bcachefs/sb-clean.c b/fs/bcachefs/sb-clean.c
new file mode 100644
index 000000000000..a3695e56a155
--- /dev/null
+++ b/fs/bcachefs/sb-clean.c
@@ -0,0 +1,395 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include "bcachefs.h"
+#include "btree_update_interior.h"
+#include "buckets.h"
+#include "error.h"
+#include "journal_io.h"
+#include "replicas.h"
+#include "sb-clean.h"
+#include "super-io.h"
+
+/*
+ * BCH_SB_FIELD_clean:
+ *
+ * Btree roots, and a few other things, are recovered from the journal after an
+ * unclean shutdown - but after a clean shutdown, to avoid having to read the
+ * journal, we can store them in the superblock.
+ *
+ * bch_sb_field_clean simply contains a list of journal entries, stored exactly
+ * as they would be in the journal:
+ */
+
+int bch2_sb_clean_validate_late(struct bch_fs *c, struct bch_sb_field_clean *clean,
+				int write)
+{
+	struct jset_entry *entry;
+	int ret;
+
+	for (entry = clean->start;
+	     entry < (struct jset_entry *) vstruct_end(&clean->field);
+	     entry = vstruct_next(entry)) {
+		ret = bch2_journal_entry_validate(c, NULL, entry,
+						  le16_to_cpu(c->disk_sb.sb->version),
+						  BCH_SB_BIG_ENDIAN(c->disk_sb.sb),
+						  write);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static struct bkey_i *btree_root_find(struct bch_fs *c,
+				      struct bch_sb_field_clean *clean,
+				      struct jset *j,
+				      enum btree_id id, unsigned *level)
+{
+	struct bkey_i *k;
+	struct jset_entry *entry, *start, *end;
+
+	if (clean) {
+		start = clean->start;
+		end = vstruct_end(&clean->field);
+	} else {
+		start = j->start;
+		end = vstruct_last(j);
+	}
+
+	for (entry = start; entry < end; entry = vstruct_next(entry))
+		if (entry->type == BCH_JSET_ENTRY_btree_root &&
+		    entry->btree_id == id)
+			goto found;
+
+	return NULL;
+found:
+	if (!entry->u64s)
+		return ERR_PTR(-EINVAL);
+
+	k = entry->start;
+	*level = entry->level;
+	return k;
+}
+
+int bch2_verify_superblock_clean(struct bch_fs *c,
+				 struct bch_sb_field_clean **cleanp,
+				 struct jset *j)
+{
+	unsigned i;
+	struct bch_sb_field_clean *clean = *cleanp;
+	struct printbuf buf1 = PRINTBUF;
+	struct printbuf buf2 = PRINTBUF;
+	int ret = 0;
+
+	if (mustfix_fsck_err_on(j->seq != clean->journal_seq, c,
+			"superblock journal seq (%llu) doesn't match journal (%llu) after clean shutdown",
+			le64_to_cpu(clean->journal_seq),
+			le64_to_cpu(j->seq))) {
+		kfree(clean);
+		*cleanp = NULL;
+		return 0;
+	}
+
+	for (i = 0; i < BTREE_ID_NR; i++) {
+		struct bkey_i *k1, *k2;
+		unsigned l1 = 0, l2 = 0;
+
+		k1 = btree_root_find(c, clean, NULL, i, &l1);
+		k2 = btree_root_find(c, NULL, j, i, &l2);
+
+		if (!k1 && !k2)
+			continue;
+
+		printbuf_reset(&buf1);
+		printbuf_reset(&buf2);
+
+		if (k1)
+			bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(k1));
+		else
+			prt_printf(&buf1, "(none)");
+
+		if (k2)
+			bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(k2));
+		else
+			prt_printf(&buf2, "(none)");
+
+		mustfix_fsck_err_on(!k1 || !k2 ||
+				    IS_ERR(k1) ||
+				    IS_ERR(k2) ||
+				    k1->k.u64s != k2->k.u64s ||
+				    memcmp(k1, k2, bkey_bytes(&k1->k)) ||
+				    l1 != l2, c,
+			"superblock btree root %u doesn't match journal after clean shutdown\n"
+			"sb:      l=%u %s\n"
+			"journal: l=%u %s\n", i,
+			l1, buf1.buf,
+			l2, buf2.buf);
+	}
+fsck_err:
+	printbuf_exit(&buf2);
+	printbuf_exit(&buf1);
+	return ret;
+}
+
+struct bch_sb_field_clean *bch2_read_superblock_clean(struct bch_fs *c)
+{
+	struct bch_sb_field_clean *clean, *sb_clean;
+	int ret;
+
+	mutex_lock(&c->sb_lock);
+	sb_clean = bch2_sb_get_clean(c->disk_sb.sb);
+
+	if (fsck_err_on(!sb_clean, c,
+			"superblock marked clean but clean section not present")) {
+		SET_BCH_SB_CLEAN(c->disk_sb.sb, false);
+		c->sb.clean = false;
+		mutex_unlock(&c->sb_lock);
+		return NULL;
+	}
+
+	clean = kmemdup(sb_clean, vstruct_bytes(&sb_clean->field),
+			GFP_KERNEL);
+	if (!clean) {
+		mutex_unlock(&c->sb_lock);
+		return ERR_PTR(-BCH_ERR_ENOMEM_read_superblock_clean);
+	}
+
+	ret = bch2_sb_clean_validate_late(c, clean, READ);
+	if (ret) {
+		mutex_unlock(&c->sb_lock);
+		return ERR_PTR(ret);
+	}
+
+	mutex_unlock(&c->sb_lock);
+
+	return clean;
+fsck_err:
+	mutex_unlock(&c->sb_lock);
+	return ERR_PTR(ret);
+}
+
+static struct jset_entry *jset_entry_init(struct jset_entry **end, size_t size)
+{
+	struct jset_entry *entry = *end;
+	unsigned u64s = DIV_ROUND_UP(size, sizeof(u64));
+
+	memset(entry, 0, u64s * sizeof(u64));
+	/*
+	 * The u64s field counts from the start of data, ignoring the shared
+	 * fields.
+	 */
+	entry->u64s = cpu_to_le16(u64s - 1);
+
+	*end = vstruct_next(*end);
+	return entry;
+}
+
+void bch2_journal_super_entries_add_common(struct bch_fs *c,
+					   struct jset_entry **end,
+					   u64 journal_seq)
+{
+	struct bch_dev *ca;
+	unsigned i, dev;
+
+	percpu_down_read(&c->mark_lock);
+
+	if (!journal_seq) {
+		for (i = 0; i < ARRAY_SIZE(c->usage); i++)
+			bch2_fs_usage_acc_to_base(c, i);
+	} else {
+		bch2_fs_usage_acc_to_base(c, journal_seq & JOURNAL_BUF_MASK);
+	}
+
+	{
+		struct jset_entry_usage *u =
+			container_of(jset_entry_init(end, sizeof(*u)),
+				     struct jset_entry_usage, entry);
+
+		u->entry.type	= BCH_JSET_ENTRY_usage;
+		u->entry.btree_id = BCH_FS_USAGE_inodes;
+		u->v		= cpu_to_le64(c->usage_base->nr_inodes);
+	}
+
+	{
+		struct jset_entry_usage *u =
+			container_of(jset_entry_init(end, sizeof(*u)),
+				     struct jset_entry_usage, entry);
+
+		u->entry.type	= BCH_JSET_ENTRY_usage;
+		u->entry.btree_id = BCH_FS_USAGE_key_version;
+		u->v		= cpu_to_le64(atomic64_read(&c->key_version));
+	}
+
+	for (i = 0; i < BCH_REPLICAS_MAX; i++) {
+		struct jset_entry_usage *u =
+			container_of(jset_entry_init(end, sizeof(*u)),
+				     struct jset_entry_usage, entry);
+
+		u->entry.type	= BCH_JSET_ENTRY_usage;
+		u->entry.btree_id = BCH_FS_USAGE_reserved;
+		u->entry.level	= i;
+		u->v		= cpu_to_le64(c->usage_base->persistent_reserved[i]);
+	}
+
+	for (i = 0; i < c->replicas.nr; i++) {
+		struct bch_replicas_entry *e =
+			cpu_replicas_entry(&c->replicas, i);
+		struct jset_entry_data_usage *u =
+			container_of(jset_entry_init(end, sizeof(*u) + e->nr_devs),
+				     struct jset_entry_data_usage, entry);
+
+		u->entry.type	= BCH_JSET_ENTRY_data_usage;
+		u->v		= cpu_to_le64(c->usage_base->replicas[i]);
+		unsafe_memcpy(&u->r, e, replicas_entry_bytes(e),
+			      "embedded variable length struct");
+	}
+
+	for_each_member_device(ca, c, dev) {
+		unsigned b = sizeof(struct jset_entry_dev_usage) +
+			sizeof(struct jset_entry_dev_usage_type) * BCH_DATA_NR;
+		struct jset_entry_dev_usage *u =
+			container_of(jset_entry_init(end, b),
+				     struct jset_entry_dev_usage, entry);
+
+		u->entry.type = BCH_JSET_ENTRY_dev_usage;
+		u->dev = cpu_to_le32(dev);
+		u->buckets_ec		= cpu_to_le64(ca->usage_base->buckets_ec);
+
+		for (i = 0; i < BCH_DATA_NR; i++) {
+			u->d[i].buckets = cpu_to_le64(ca->usage_base->d[i].buckets);
+			u->d[i].sectors	= cpu_to_le64(ca->usage_base->d[i].sectors);
+			u->d[i].fragmented = cpu_to_le64(ca->usage_base->d[i].fragmented);
+		}
+	}
+
+	percpu_up_read(&c->mark_lock);
+
+	for (i = 0; i < 2; i++) {
+		struct jset_entry_clock *clock =
+			container_of(jset_entry_init(end, sizeof(*clock)),
+				     struct jset_entry_clock, entry);
+
+		clock->entry.type = BCH_JSET_ENTRY_clock;
+		clock->rw	= i;
+		clock->time	= cpu_to_le64(atomic64_read(&c->io_clock[i].now));
+	}
+}
+
+static int bch2_sb_clean_validate(struct bch_sb *sb,
+				  struct bch_sb_field *f,
+				  struct printbuf *err)
+{
+	struct bch_sb_field_clean *clean = field_to_type(f, clean);
+
+	if (vstruct_bytes(&clean->field) < sizeof(*clean)) {
+		prt_printf(err, "wrong size (got %zu should be %zu)",
+		       vstruct_bytes(&clean->field), sizeof(*clean));
+		return -BCH_ERR_invalid_sb_clean;
+	}
+
+	return 0;
+}
+
+static void bch2_sb_clean_to_text(struct printbuf *out, struct bch_sb *sb,
+				  struct bch_sb_field *f)
+{
+	struct bch_sb_field_clean *clean = field_to_type(f, clean);
+	struct jset_entry *entry;
+
+	prt_printf(out, "flags:          %x",	le32_to_cpu(clean->flags));
+	prt_newline(out);
+	prt_printf(out, "journal_seq:    %llu",	le64_to_cpu(clean->journal_seq));
+	prt_newline(out);
+
+	for (entry = clean->start;
+	     entry != vstruct_end(&clean->field);
+	     entry = vstruct_next(entry)) {
+		if (entry->type == BCH_JSET_ENTRY_btree_keys &&
+		    !entry->u64s)
+			continue;
+
+		bch2_journal_entry_to_text(out, NULL, entry);
+		prt_newline(out);
+	}
+}
+
+const struct bch_sb_field_ops bch_sb_field_ops_clean = {
+	.validate	= bch2_sb_clean_validate,
+	.to_text	= bch2_sb_clean_to_text,
+};
+
+int bch2_fs_mark_dirty(struct bch_fs *c)
+{
+	int ret;
+
+	/*
+	 * Unconditionally write superblock, to verify it hasn't changed before
+	 * we go rw:
+	 */
+
+	mutex_lock(&c->sb_lock);
+	SET_BCH_SB_CLEAN(c->disk_sb.sb, false);
+
+	bch2_sb_maybe_downgrade(c);
+	c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALWAYS);
+
+	ret = bch2_write_super(c);
+	mutex_unlock(&c->sb_lock);
+
+	return ret;
+}
+
+void bch2_fs_mark_clean(struct bch_fs *c)
+{
+	struct bch_sb_field_clean *sb_clean;
+	struct jset_entry *entry;
+	unsigned u64s;
+	int ret;
+
+	mutex_lock(&c->sb_lock);
+	if (BCH_SB_CLEAN(c->disk_sb.sb))
+		goto out;
+
+	SET_BCH_SB_CLEAN(c->disk_sb.sb, true);
+
+	c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_info);
+	c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_metadata);
+	c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_extents_above_btree_updates));
+	c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_btree_updates_journalled));
+
+	u64s = sizeof(*sb_clean) / sizeof(u64) + c->journal.entry_u64s_reserved;
+
+	sb_clean = bch2_sb_resize_clean(&c->disk_sb, u64s);
+	if (!sb_clean) {
+		bch_err(c, "error resizing superblock while setting filesystem clean");
+		goto out;
+	}
+
+	sb_clean->flags		= 0;
+	sb_clean->journal_seq	= cpu_to_le64(atomic64_read(&c->journal.seq));
+
+	/* Trying to catch outstanding bug: */
+	BUG_ON(le64_to_cpu(sb_clean->journal_seq) > S64_MAX);
+
+	entry = sb_clean->start;
+	bch2_journal_super_entries_add_common(c, &entry, 0);
+	entry = bch2_btree_roots_to_journal_entries(c, entry, entry);
+	BUG_ON((void *) entry > vstruct_end(&sb_clean->field));
+
+	memset(entry, 0,
+	       vstruct_end(&sb_clean->field) - (void *) entry);
+
+	/*
+	 * this should be in the write path, and we should be validating every
+	 * superblock section:
+	 */
+	ret = bch2_sb_clean_validate_late(c, sb_clean, WRITE);
+	if (ret) {
+		bch_err(c, "error writing marking filesystem clean: validate error");
+		goto out;
+	}
+
+	bch2_write_super(c);
+out:
+	mutex_unlock(&c->sb_lock);
+}
diff --git a/fs/bcachefs/sb-clean.h b/fs/bcachefs/sb-clean.h
new file mode 100644
index 000000000000..71caef281239
--- /dev/null
+++ b/fs/bcachefs/sb-clean.h
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BCACHEFS_SB_CLEAN_H
+#define _BCACHEFS_SB_CLEAN_H
+
+int bch2_sb_clean_validate_late(struct bch_fs *, struct bch_sb_field_clean *, int);
+int bch2_verify_superblock_clean(struct bch_fs *, struct bch_sb_field_clean **,
+				 struct jset *);
+struct bch_sb_field_clean *bch2_read_superblock_clean(struct bch_fs *);
+void bch2_journal_super_entries_add_common(struct bch_fs *, struct jset_entry **, u64);
+
+extern const struct bch_sb_field_ops bch_sb_field_ops_clean;
+
+int bch2_fs_mark_dirty(struct bch_fs *);
+void bch2_fs_mark_clean(struct bch_fs *);
+
+#endif /* _BCACHEFS_SB_CLEAN_H */
diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c
index ff86088022fc..f01883e785a5 100644
--- a/fs/bcachefs/super-io.c
+++ b/fs/bcachefs/super-io.c
@@ -1,8 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
-#include "btree_update_interior.h"
-#include "buckets.h"
 #include "checksum.h"
 #include "counters.h"
 #include "disk_groups.h"
@@ -10,12 +8,12 @@
 #include "error.h"
 #include "io.h"
 #include "journal.h"
-#include "journal_io.h"
 #include "journal_sb.h"
 #include "journal_seq_blacklist.h"
 #include "recovery.h"
 #include "replicas.h"
 #include "quota.h"
+#include "sb-clean.h"
 #include "sb-members.h"
 #include "super-io.h"
 #include "super.h"
@@ -1016,27 +1014,6 @@ void __bch2_check_set_feature(struct bch_fs *c, unsigned feat)
 	mutex_unlock(&c->sb_lock);
 }
 
-/* BCH_SB_FIELD_clean: */
-
-int bch2_sb_clean_validate_late(struct bch_fs *c, struct bch_sb_field_clean *clean, int write)
-{
-	struct jset_entry *entry;
-	int ret;
-
-	for (entry = clean->start;
-	     entry < (struct jset_entry *) vstruct_end(&clean->field);
-	     entry = vstruct_next(entry)) {
-		ret = bch2_journal_entry_validate(c, NULL, entry,
-						  le16_to_cpu(c->disk_sb.sb->version),
-						  BCH_SB_BIG_ENDIAN(c->disk_sb.sb),
-						  write);
-		if (ret)
-			return ret;
-	}
-
-	return 0;
-}
-
 /* Downgrade if superblock is at a higher version than currently supported: */
 void bch2_sb_maybe_downgrade(struct bch_fs *c)
 {
@@ -1063,232 +1040,6 @@ void bch2_sb_upgrade(struct bch_fs *c, unsigned new_version)
 	c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL);
 }
 
-int bch2_fs_mark_dirty(struct bch_fs *c)
-{
-	int ret;
-
-	/*
-	 * Unconditionally write superblock, to verify it hasn't changed before
-	 * we go rw:
-	 */
-
-	mutex_lock(&c->sb_lock);
-	SET_BCH_SB_CLEAN(c->disk_sb.sb, false);
-
-	bch2_sb_maybe_downgrade(c);
-	c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALWAYS);
-
-	ret = bch2_write_super(c);
-	mutex_unlock(&c->sb_lock);
-
-	return ret;
-}
-
-static struct jset_entry *jset_entry_init(struct jset_entry **end, size_t size)
-{
-	struct jset_entry *entry = *end;
-	unsigned u64s = DIV_ROUND_UP(size, sizeof(u64));
-
-	memset(entry, 0, u64s * sizeof(u64));
-	/*
-	 * The u64s field counts from the start of data, ignoring the shared
-	 * fields.
-	 */
-	entry->u64s = cpu_to_le16(u64s - 1);
-
-	*end = vstruct_next(*end);
-	return entry;
-}
-
-void bch2_journal_super_entries_add_common(struct bch_fs *c,
-					   struct jset_entry **end,
-					   u64 journal_seq)
-{
-	struct bch_dev *ca;
-	unsigned i, dev;
-
-	percpu_down_read(&c->mark_lock);
-
-	if (!journal_seq) {
-		for (i = 0; i < ARRAY_SIZE(c->usage); i++)
-			bch2_fs_usage_acc_to_base(c, i);
-	} else {
-		bch2_fs_usage_acc_to_base(c, journal_seq & JOURNAL_BUF_MASK);
-	}
-
-	{
-		struct jset_entry_usage *u =
-			container_of(jset_entry_init(end, sizeof(*u)),
-				     struct jset_entry_usage, entry);
-
-		u->entry.type	= BCH_JSET_ENTRY_usage;
-		u->entry.btree_id = BCH_FS_USAGE_inodes;
-		u->v		= cpu_to_le64(c->usage_base->nr_inodes);
-	}
-
-	{
-		struct jset_entry_usage *u =
-			container_of(jset_entry_init(end, sizeof(*u)),
-				     struct jset_entry_usage, entry);
-
-		u->entry.type	= BCH_JSET_ENTRY_usage;
-		u->entry.btree_id = BCH_FS_USAGE_key_version;
-		u->v		= cpu_to_le64(atomic64_read(&c->key_version));
-	}
-
-	for (i = 0; i < BCH_REPLICAS_MAX; i++) {
-		struct jset_entry_usage *u =
-			container_of(jset_entry_init(end, sizeof(*u)),
-				     struct jset_entry_usage, entry);
-
-		u->entry.type	= BCH_JSET_ENTRY_usage;
-		u->entry.btree_id = BCH_FS_USAGE_reserved;
-		u->entry.level	= i;
-		u->v		= cpu_to_le64(c->usage_base->persistent_reserved[i]);
-	}
-
-	for (i = 0; i < c->replicas.nr; i++) {
-		struct bch_replicas_entry *e =
-			cpu_replicas_entry(&c->replicas, i);
-		struct jset_entry_data_usage *u =
-			container_of(jset_entry_init(end, sizeof(*u) + e->nr_devs),
-				     struct jset_entry_data_usage, entry);
-
-		u->entry.type	= BCH_JSET_ENTRY_data_usage;
-		u->v		= cpu_to_le64(c->usage_base->replicas[i]);
-		unsafe_memcpy(&u->r, e, replicas_entry_bytes(e),
-			      "embedded variable length struct");
-	}
-
-	for_each_member_device(ca, c, dev) {
-		unsigned b = sizeof(struct jset_entry_dev_usage) +
-			sizeof(struct jset_entry_dev_usage_type) * BCH_DATA_NR;
-		struct jset_entry_dev_usage *u =
-			container_of(jset_entry_init(end, b),
-				     struct jset_entry_dev_usage, entry);
-
-		u->entry.type = BCH_JSET_ENTRY_dev_usage;
-		u->dev = cpu_to_le32(dev);
-		u->buckets_ec		= cpu_to_le64(ca->usage_base->buckets_ec);
-
-		for (i = 0; i < BCH_DATA_NR; i++) {
-			u->d[i].buckets = cpu_to_le64(ca->usage_base->d[i].buckets);
-			u->d[i].sectors	= cpu_to_le64(ca->usage_base->d[i].sectors);
-			u->d[i].fragmented = cpu_to_le64(ca->usage_base->d[i].fragmented);
-		}
-	}
-
-	percpu_up_read(&c->mark_lock);
-
-	for (i = 0; i < 2; i++) {
-		struct jset_entry_clock *clock =
-			container_of(jset_entry_init(end, sizeof(*clock)),
-				     struct jset_entry_clock, entry);
-
-		clock->entry.type = BCH_JSET_ENTRY_clock;
-		clock->rw	= i;
-		clock->time	= cpu_to_le64(atomic64_read(&c->io_clock[i].now));
-	}
-}
-
-void bch2_fs_mark_clean(struct bch_fs *c)
-{
-	struct bch_sb_field_clean *sb_clean;
-	struct jset_entry *entry;
-	unsigned u64s;
-	int ret;
-
-	mutex_lock(&c->sb_lock);
-	if (BCH_SB_CLEAN(c->disk_sb.sb))
-		goto out;
-
-	SET_BCH_SB_CLEAN(c->disk_sb.sb, true);
-
-	c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_info);
-	c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_metadata);
-	c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_extents_above_btree_updates));
-	c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_btree_updates_journalled));
-
-	u64s = sizeof(*sb_clean) / sizeof(u64) + c->journal.entry_u64s_reserved;
-
-	sb_clean = bch2_sb_resize_clean(&c->disk_sb, u64s);
-	if (!sb_clean) {
-		bch_err(c, "error resizing superblock while setting filesystem clean");
-		goto out;
-	}
-
-	sb_clean->flags		= 0;
-	sb_clean->journal_seq	= cpu_to_le64(atomic64_read(&c->journal.seq));
-
-	/* Trying to catch outstanding bug: */
-	BUG_ON(le64_to_cpu(sb_clean->journal_seq) > S64_MAX);
-
-	entry = sb_clean->start;
-	bch2_journal_super_entries_add_common(c, &entry, 0);
-	entry = bch2_btree_roots_to_journal_entries(c, entry, entry);
-	BUG_ON((void *) entry > vstruct_end(&sb_clean->field));
-
-	memset(entry, 0,
-	       vstruct_end(&sb_clean->field) - (void *) entry);
-
-	/*
-	 * this should be in the write path, and we should be validating every
-	 * superblock section:
-	 */
-	ret = bch2_sb_clean_validate_late(c, sb_clean, WRITE);
-	if (ret) {
-		bch_err(c, "error writing marking filesystem clean: validate error");
-		goto out;
-	}
-
-	bch2_write_super(c);
-out:
-	mutex_unlock(&c->sb_lock);
-}
-
-static int bch2_sb_clean_validate(struct bch_sb *sb,
-				  struct bch_sb_field *f,
-				  struct printbuf *err)
-{
-	struct bch_sb_field_clean *clean = field_to_type(f, clean);
-
-	if (vstruct_bytes(&clean->field) < sizeof(*clean)) {
-		prt_printf(err, "wrong size (got %zu should be %zu)",
-		       vstruct_bytes(&clean->field), sizeof(*clean));
-		return -BCH_ERR_invalid_sb_clean;
-	}
-
-	return 0;
-}
-
-static void bch2_sb_clean_to_text(struct printbuf *out, struct bch_sb *sb,
-				  struct bch_sb_field *f)
-{
-	struct bch_sb_field_clean *clean = field_to_type(f, clean);
-	struct jset_entry *entry;
-
-	prt_printf(out, "flags:          %x",	le32_to_cpu(clean->flags));
-	prt_newline(out);
-	prt_printf(out, "journal_seq:    %llu",	le64_to_cpu(clean->journal_seq));
-	prt_newline(out);
-
-	for (entry = clean->start;
-	     entry != vstruct_end(&clean->field);
-	     entry = vstruct_next(entry)) {
-		if (entry->type == BCH_JSET_ENTRY_btree_keys &&
-		    !entry->u64s)
-			continue;
-
-		bch2_journal_entry_to_text(out, NULL, entry);
-		prt_newline(out);
-	}
-}
-
-static const struct bch_sb_field_ops bch_sb_field_ops_clean = {
-	.validate	= bch2_sb_clean_validate,
-	.to_text	= bch2_sb_clean_to_text,
-};
-
 static const struct bch_sb_field_ops *bch2_sb_field_ops[] = {
 #define x(f, nr)					\
 	[BCH_SB_FIELD_##f] = &bch_sb_field_ops_##f,
diff --git a/fs/bcachefs/super-io.h b/fs/bcachefs/super-io.h
index 6e59b0148f8d..d51c0a19586f 100644
--- a/fs/bcachefs/super-io.h
+++ b/fs/bcachefs/super-io.h
@@ -122,19 +122,9 @@ static inline struct bch_member_cpu bch2_mi_to_cpu(struct bch_member *mi)
 	};
 }
 
-/* BCH_SB_FIELD_clean: */
-
-void bch2_journal_super_entries_add_common(struct bch_fs *,
-					   struct jset_entry **, u64);
-
-int bch2_sb_clean_validate_late(struct bch_fs *, struct bch_sb_field_clean *, int);
-
 void bch2_sb_maybe_downgrade(struct bch_fs *);
 void bch2_sb_upgrade(struct bch_fs *, unsigned);
 
-int bch2_fs_mark_dirty(struct bch_fs *);
-void bch2_fs_mark_clean(struct bch_fs *);
-
 void bch2_sb_field_to_text(struct printbuf *, struct bch_sb *,
 			   struct bch_sb_field *);
 void bch2_sb_layout_to_text(struct printbuf *, struct bch_sb_layout *);
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index de7bc0192c3d..46f0be3d0cef 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -46,6 +46,7 @@
 #include "rebalance.h"
 #include "recovery.h"
 #include "replicas.h"
+#include "sb-clean.h"
 #include "subvolume.h"
 #include "super.h"
 #include "super-io.h"