From bb36a12921e5fc76f3b26a80ab0217d3dc62a473 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 13:45:39 -0400
Subject: [PATCH 01/16] bcachefs: bch2_run_explicit_recovery_pass_printbuf()

We prefer helpers that emit log messages to printbufs rather than
printing them directly; that way, we can ensure that different log
messages from the same event are grouped together and formatted
appropriately in the dmesg log.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/buckets.c         | 33 +++++++++++++++++++++---------
 fs/bcachefs/error.c           |  2 +-
 fs/bcachefs/recovery_passes.c | 38 +++++++++++++++++++++++++++--------
 fs/bcachefs/recovery_passes.h |  3 +++
 4 files changed, 57 insertions(+), 19 deletions(-)

diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index 31fbc2716d8b..ffe957602cca 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -400,7 +400,8 @@ static int bucket_ref_update_err(struct btree_trans *trans, struct printbuf *buf
 
 	__bch2_count_fsck_err(c, id, buf->buf, &repeat, &print, &suppress);
 
-	int ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_allocations);
+	int ret = bch2_run_explicit_recovery_pass_printbuf(c, buf,
+					BCH_RECOVERY_PASS_check_allocations);
 
 	if (insert) {
 		print = true;
@@ -966,14 +967,27 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 		return PTR_ERR(a);
 
 	if (a->v.data_type && type && a->v.data_type != type) {
-		bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_allocations);
-		log_fsck_err(trans, bucket_metadata_type_mismatch,
-			"bucket %llu:%llu gen %u different types of data in same bucket: %s, %s\n"
-			"while marking %s",
-			iter.pos.inode, iter.pos.offset, a->v.gen,
-			bch2_data_type_str(a->v.data_type),
-			bch2_data_type_str(type),
-			bch2_data_type_str(type));
+		struct printbuf buf = PRINTBUF;
+		bch2_log_msg_start(c, &buf);
+		prt_printf(&buf, "bucket %llu:%llu gen %u different types of data in same bucket: %s, %s\n"
+			   "while marking %s\n",
+			   iter.pos.inode, iter.pos.offset, a->v.gen,
+			   bch2_data_type_str(a->v.data_type),
+			   bch2_data_type_str(type),
+			   bch2_data_type_str(type));
+
+		bool repeat = false, print = true, suppress = false;
+		bch2_count_fsck_err(c, bucket_metadata_type_mismatch, buf.buf,
+				    &repeat, &print, &suppress);
+
+		bch2_run_explicit_recovery_pass_printbuf(c, &buf,
+					BCH_RECOVERY_PASS_check_allocations);
+
+		if (suppress)
+			prt_printf(&buf, "Ratelimiting new instances of previous error\n");
+		if (print)
+			bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		printbuf_exit(&buf);
 		ret = -BCH_ERR_metadata_bucket_inconsistency;
 		goto err;
 	}
@@ -985,7 +999,6 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 		ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
 	}
 err:
-fsck_err:
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c
index 6b8695b1349c..faeadffa1103 100644
--- a/fs/bcachefs/error.c
+++ b/fs/bcachefs/error.c
@@ -104,7 +104,7 @@ int __bch2_topology_error(struct bch_fs *c, struct printbuf *out)
 		__bch2_inconsistent_error(c, out);
 		return -BCH_ERR_btree_need_topology_repair;
 	} else {
-		return bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology) ?:
+		return bch2_run_explicit_recovery_pass_printbuf(c, out, BCH_RECOVERY_PASS_check_topology) ?:
 			-BCH_ERR_btree_node_read_validate_error;
 	}
 }
diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index 22f72bb5b853..946428daeecc 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -101,7 +101,8 @@ u64 bch2_recovery_passes_from_stable(u64 v)
 /*
  * For when we need to rewind recovery passes and run a pass we skipped:
  */
-static int __bch2_run_explicit_recovery_pass(struct bch_fs *c,
+static int __bch2_run_explicit_recovery_pass(struct printbuf *out,
+					     struct bch_fs *c,
 					     enum bch_recovery_pass pass)
 {
 	if (c->curr_recovery_pass == ARRAY_SIZE(recovery_pass_fns))
@@ -115,15 +116,15 @@ static int __bch2_run_explicit_recovery_pass(struct bch_fs *c,
 	if (pass < BCH_RECOVERY_PASS_set_may_go_rw &&
 	    c->curr_recovery_pass >= BCH_RECOVERY_PASS_set_may_go_rw) {
 		if (print)
-			bch_info(c, "need recovery pass %s (%u), but already rw",
-				 bch2_recovery_passes[pass], pass);
+			prt_printf(out, "need recovery pass %s (%u), but already rw",
+				   bch2_recovery_passes[pass], pass);
 		return -BCH_ERR_cannot_rewind_recovery;
 	}
 
 	if (print)
-		bch_info(c, "running explicit recovery pass %s (%u), currently at %s (%u)",
-			 bch2_recovery_passes[pass], pass,
-			 bch2_recovery_passes[c->curr_recovery_pass], c->curr_recovery_pass);
+		prt_printf(out, "running explicit recovery pass %s (%u), currently at %s (%u)",
+			   bch2_recovery_passes[pass], pass,
+			   bch2_recovery_passes[c->curr_recovery_pass], c->curr_recovery_pass);
 
 	c->opts.recovery_passes |= BIT_ULL(pass);
 
@@ -136,13 +137,34 @@ static int __bch2_run_explicit_recovery_pass(struct bch_fs *c,
 	}
 }
 
-int bch2_run_explicit_recovery_pass(struct bch_fs *c,
+int bch2_run_explicit_recovery_pass_printbuf(struct bch_fs *c,
+				    struct printbuf *out,
 				    enum bch_recovery_pass pass)
 {
+	bch2_printbuf_make_room(out, 1024);
+	out->atomic++;
+
 	unsigned long flags;
 	spin_lock_irqsave(&c->recovery_pass_lock, flags);
-	int ret = __bch2_run_explicit_recovery_pass(c, pass);
+	int ret = __bch2_run_explicit_recovery_pass(out, c, pass);
 	spin_unlock_irqrestore(&c->recovery_pass_lock, flags);
+
+	--out->atomic;
+	return ret;
+}
+
+int bch2_run_explicit_recovery_pass(struct bch_fs *c,
+				    enum bch_recovery_pass pass)
+{
+	struct printbuf buf = PRINTBUF;
+	bch2_log_msg_start(c, &buf);
+	unsigned len = buf.pos;
+
+	int ret = bch2_run_explicit_recovery_pass_printbuf(c, &buf, pass);
+
+	if (len != buf.pos)
+		bch2_print_string_as_lines(KERN_NOTICE, buf.buf);
+	printbuf_exit(&buf);
 	return ret;
 }
 
diff --git a/fs/bcachefs/recovery_passes.h b/fs/bcachefs/recovery_passes.h
index 7d7339c8fa29..e19a8aaba2f8 100644
--- a/fs/bcachefs/recovery_passes.h
+++ b/fs/bcachefs/recovery_passes.h
@@ -8,6 +8,9 @@ u64 bch2_recovery_passes_from_stable(u64 v);
 
 u64 bch2_fsck_recovery_passes(void);
 
+int bch2_run_explicit_recovery_pass_printbuf(struct bch_fs *,
+				    struct printbuf *,
+				    enum bch_recovery_pass);
 int bch2_run_explicit_recovery_pass(struct bch_fs *, enum bch_recovery_pass);
 int bch2_run_explicit_recovery_pass_persistent_locked(struct bch_fs *, enum bch_recovery_pass);
 int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *, enum bch_recovery_pass);
-- 
2.51.0


From 2085325171f2f2d33a94101e58266f325c286e95 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 13:55:16 -0400
Subject: [PATCH 02/16] bcachefs: Simplify bch2_count_fsck_err()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/buckets.c  | 16 +++-------------
 fs/bcachefs/error.c    | 14 ++++++++++----
 fs/bcachefs/error.h    |  4 +---
 fs/bcachefs/fs-io.c    |  7 ++-----
 fs/bcachefs/io_write.c |  4 +---
 5 files changed, 17 insertions(+), 28 deletions(-)

diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index ffe957602cca..8e64077c15c1 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -392,29 +392,23 @@ static int bucket_ref_update_err(struct btree_trans *trans, struct printbuf *buf
 				 struct bkey_s_c k, bool insert, enum bch_sb_error_id id)
 {
 	struct bch_fs *c = trans->c;
-	bool repeat = false, print = true, suppress = false;
 
 	prt_printf(buf, "\nwhile marking ");
 	bch2_bkey_val_to_text(buf, c, k);
 	prt_newline(buf);
 
-	__bch2_count_fsck_err(c, id, buf->buf, &repeat, &print, &suppress);
+	bool print = __bch2_count_fsck_err(c, id, buf);
 
 	int ret = bch2_run_explicit_recovery_pass_printbuf(c, buf,
 					BCH_RECOVERY_PASS_check_allocations);
 
 	if (insert) {
-		print = true;
-		suppress = false;
-
 		bch2_trans_updates_to_text(buf, trans);
 		__bch2_inconsistent_error(c, buf);
 		ret = -BCH_ERR_bucket_ref_update;
 	}
 
-	if (suppress)
-		prt_printf(buf, "Ratelimiting new instances of previous error\n");
-	if (print)
+	if (print || insert)
 		bch2_print_string_as_lines(KERN_ERR, buf->buf);
 	return ret;
 }
@@ -976,15 +970,11 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 			   bch2_data_type_str(type),
 			   bch2_data_type_str(type));
 
-		bool repeat = false, print = true, suppress = false;
-		bch2_count_fsck_err(c, bucket_metadata_type_mismatch, buf.buf,
-				    &repeat, &print, &suppress);
+		bool print = bch2_count_fsck_err(c, bucket_metadata_type_mismatch, &buf);
 
 		bch2_run_explicit_recovery_pass_printbuf(c, &buf,
 					BCH_RECOVERY_PASS_check_allocations);
 
-		if (suppress)
-			prt_printf(&buf, "Ratelimiting new instances of previous error\n");
 		if (print)
 			bch2_print_string_as_lines(KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c
index faeadffa1103..4627aabd1f1a 100644
--- a/fs/bcachefs/error.c
+++ b/fs/bcachefs/error.c
@@ -376,15 +376,21 @@ static struct fsck_err_state *count_fsck_err_locked(struct bch_fs *c,
 	return s;
 }
 
-void __bch2_count_fsck_err(struct bch_fs *c,
-			   enum bch_sb_error_id id, const char *msg,
-			   bool *repeat, bool *print, bool *suppress)
+bool __bch2_count_fsck_err(struct bch_fs *c,
+			   enum bch_sb_error_id id, struct printbuf *msg)
 {
 	bch2_sb_error_count(c, id);
 
 	mutex_lock(&c->fsck_error_msgs_lock);
-	count_fsck_err_locked(c, id, msg, repeat, print, suppress);
+	bool print = true, repeat = false, suppress = false;
+
+	count_fsck_err_locked(c, id, msg->buf, &repeat, &print, &suppress);
 	mutex_unlock(&c->fsck_error_msgs_lock);
+
+	if (suppress)
+		prt_printf(msg, "Ratelimiting new instances of previous error\n");
+
+	return print && !repeat;
 }
 
 int __bch2_fsck_err(struct bch_fs *c,
diff --git a/fs/bcachefs/error.h b/fs/bcachefs/error.h
index 4a364fd44abe..0b3ede1c2015 100644
--- a/fs/bcachefs/error.h
+++ b/fs/bcachefs/error.h
@@ -76,9 +76,7 @@ struct fsck_err_state {
 
 #define fsck_err_count(_c, _err)	bch2_sb_err_count(_c, BCH_FSCK_ERR_##_err)
 
-void __bch2_count_fsck_err(struct bch_fs *,
-			   enum bch_sb_error_id, const char *,
-			   bool *, bool *, bool *);
+bool __bch2_count_fsck_err(struct bch_fs *, enum bch_sb_error_id, struct printbuf *);
 #define bch2_count_fsck_err(_c, _err, ...)				\
 	__bch2_count_fsck_err(_c, BCH_FSCK_ERR_##_err, __VA_ARGS__)
 
diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index 9657144666b8..b81117b51c69 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -151,8 +151,7 @@ void __bch2_i_sectors_acct(struct bch_fs *c, struct bch_inode_info *inode,
 			   inode->v.i_ino, (u64) inode->v.i_blocks, sectors,
 			   inode->ei_inode.bi_sectors);
 
-		bool repeat = false, print = false, suppress = false;
-		bch2_count_fsck_err(c, vfs_inode_i_blocks_underflow, buf.buf, &repeat, &print, &suppress);
+		bool print = bch2_count_fsck_err(c, vfs_inode_i_blocks_underflow, &buf);
 		if (print)
 			bch2_print_str(c, buf.buf);
 		printbuf_exit(&buf);
@@ -526,9 +525,7 @@ int bchfs_truncate(struct mnt_idmap *idmap,
 			   inode->v.i_ino, (u64) inode->v.i_blocks,
 			   inode->ei_inode.bi_sectors);
 
-		bool repeat = false, print = false, suppress = false;
-		bch2_count_fsck_err(c, vfs_inode_i_blocks_not_zero_at_truncate, buf.buf,
-				    &repeat, &print, &suppress);
+		bool print = bch2_count_fsck_err(c, vfs_inode_i_blocks_not_zero_at_truncate, &buf);
 		if (print)
 			bch2_print_str(c, buf.buf);
 		printbuf_exit(&buf);
diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c
index 401347e135b7..c738ae6fd9a5 100644
--- a/fs/bcachefs/io_write.c
+++ b/fs/bcachefs/io_write.c
@@ -263,9 +263,7 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
 			prt_printf(&buf, "inode %llu i_sectors underflow: %lli + %lli < 0",
 				   extent_iter->pos.inode, bi_sectors, i_sectors_delta);
 
-			bool repeat = false, print = false, suppress = false;
-			bch2_count_fsck_err(c, inode_i_sectors_underflow, buf.buf,
-					    &repeat, &print, &suppress);
+			bool print = bch2_count_fsck_err(c, inode_i_sectors_underflow, &buf);
 			if (print)
 				bch2_print_str(c, buf.buf);
 			printbuf_exit(&buf);
-- 
2.51.0


From 040c762152f5f4fb1b13e8a46c17ecb4e670d96d Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 14:08:42 -0400
Subject: [PATCH 03/16] bcachefs: bch2_dev_missing_bkey()

Part of the ongoing project to kill off bch2_(fs|trans)_inconsistent
calls - they generally need to be replaced with either

- a fsck_err() call that can repair the error, or

- logging an error of the appropriate type in the superblock, and
  flagging the appropriate recovery pass to repair the error

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/extents.c    | 10 +++++++++-
 fs/bcachefs/sb-members.c | 22 +++++++++++++++++++++-
 fs/bcachefs/sb-members.h |  8 +++++---
 3 files changed, 35 insertions(+), 5 deletions(-)

diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c
index a369f978ffe6..ef116c55f0a7 100644
--- a/fs/bcachefs/extents.c
+++ b/fs/bcachefs/extents.c
@@ -158,7 +158,15 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
 		if (dev >= 0 && p.ptr.dev != dev)
 			continue;
 
-		struct bch_dev *ca = bch2_dev_rcu(c, p.ptr.dev);
+		struct bch_dev *ca = bch2_dev_rcu_noerror(c, p.ptr.dev);
+
+		if (unlikely(!ca && p.ptr.dev != BCH_SB_MEMBER_INVALID)) {
+			rcu_read_unlock();
+			int ret = bch2_dev_missing_bkey(c, k, p.ptr.dev);
+			if (ret)
+				return ret;
+			rcu_read_lock();
+		}
 
 		if (p.ptr.cached && (!ca || dev_ptr_stale_rcu(ca, &p.ptr)))
 			continue;
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index 72779912939b..f776b00c3cc0 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -5,11 +5,31 @@
 #include "disk_groups.h"
 #include "error.h"
 #include "opts.h"
+#include "recovery_passes.h"
 #include "replicas.h"
 #include "sb-members.h"
 #include "super-io.h"
 
-void bch2_dev_missing(struct bch_fs *c, unsigned dev)
+int bch2_dev_missing_bkey(struct bch_fs *c, struct bkey_s_c k, unsigned dev)
+{
+	struct printbuf buf = PRINTBUF;
+	bch2_log_msg_start(c, &buf);
+
+	prt_printf(&buf, "pointer to nonexistent device %u in key\n", dev);
+	bch2_bkey_val_to_text(&buf, c, k);
+
+	bool print = bch2_count_fsck_err(c, ptr_to_invalid_device, &buf);
+
+	int ret = bch2_run_explicit_recovery_pass_printbuf(c, &buf,
+						 BCH_RECOVERY_PASS_check_allocations);
+
+	if (print)
+		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	printbuf_exit(&buf);
+	return ret;
+}
+
+void bch2_dev_missing_atomic(struct bch_fs *c, unsigned dev)
 {
 	if (dev != BCH_SB_MEMBER_INVALID)
 		bch2_fs_inconsistent(c, "pointer to nonexistent device %u", dev);
diff --git a/fs/bcachefs/sb-members.h b/fs/bcachefs/sb-members.h
index 42786657522c..0f1741fffcb6 100644
--- a/fs/bcachefs/sb-members.h
+++ b/fs/bcachefs/sb-members.h
@@ -218,13 +218,15 @@ static inline struct bch_dev *bch2_dev_rcu_noerror(struct bch_fs *c, unsigned de
 		: NULL;
 }
 
-void bch2_dev_missing(struct bch_fs *, unsigned);
+int bch2_dev_missing_bkey(struct bch_fs *, struct bkey_s_c, unsigned);
+
+void bch2_dev_missing_atomic(struct bch_fs *, unsigned);
 
 static inline struct bch_dev *bch2_dev_rcu(struct bch_fs *c, unsigned dev)
 {
 	struct bch_dev *ca = bch2_dev_rcu_noerror(c, dev);
 	if (unlikely(!ca))
-		bch2_dev_missing(c, dev);
+		bch2_dev_missing_atomic(c, dev);
 	return ca;
 }
 
@@ -242,7 +244,7 @@ static inline struct bch_dev *bch2_dev_tryget(struct bch_fs *c, unsigned dev)
 {
 	struct bch_dev *ca = bch2_dev_tryget_noerror(c, dev);
 	if (unlikely(!ca))
-		bch2_dev_missing(c, dev);
+		bch2_dev_missing_atomic(c, dev);
 	return ca;
 }
 
-- 
2.51.0


From ebf561b2083d797da4673207044855ccd764195b Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 17:31:47 -0400
Subject: [PATCH 04/16] bcachefs: print_str_as_lines() -> print_str()

bch2_print_string_as_lines() is a low level helper that allows messages
longer than 1k to be printed without truncation.

But we should always be printing with the helpers that take a filesystem
object, if we're in fsck they direct output to the userspace process
controlling fsck instead of the dmesg log.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/alloc_foreground.c      |  2 +-
 fs/bcachefs/bcachefs.h              |  3 ++-
 fs/bcachefs/btree_io.c              |  2 +-
 fs/bcachefs/btree_iter.c            |  4 ++--
 fs/bcachefs/btree_locking.c         |  2 +-
 fs/bcachefs/btree_node_scan.c       |  6 +++---
 fs/bcachefs/btree_update_interior.c |  2 +-
 fs/bcachefs/buckets.c               |  6 +++---
 fs/bcachefs/data_update.c           |  2 +-
 fs/bcachefs/error.c                 | 10 +++++-----
 fs/bcachefs/fs-io.c                 |  4 ++--
 fs/bcachefs/io_write.c              |  2 +-
 fs/bcachefs/journal.c               |  4 ++--
 fs/bcachefs/journal_io.c            |  2 +-
 fs/bcachefs/recovery_passes.c       |  2 +-
 fs/bcachefs/sb-members.c            |  2 +-
 fs/bcachefs/super.c                 | 15 +++++++++++++--
 fs/bcachefs/util.c                  | 14 ++------------
 fs/bcachefs/util.h                  |  3 +--
 19 files changed, 44 insertions(+), 43 deletions(-)

diff --git a/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c
index 31d2207a071b..d56cee7e8cb5 100644
--- a/fs/bcachefs/alloc_foreground.c
+++ b/fs/bcachefs/alloc_foreground.c
@@ -1664,7 +1664,7 @@ static noinline void bch2_print_allocator_stuck(struct bch_fs *c)
 	bch2_journal_debug_to_text(&buf, &c->journal);
 	printbuf_indent_sub(&buf, 2);
 
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	bch2_print_str(c, KERN_ERR, buf.buf);
 	printbuf_exit(&buf);
 }
 
diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index 7782e311b6e2..4fd096349790 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -269,7 +269,8 @@ do {									\
 
 #define bch2_fmt(_c, fmt)		bch2_log_msg(_c, fmt "\n")
 
-void bch2_print_str(struct bch_fs *, const char *);
+void bch2_print_str(struct bch_fs *, const char *, const char *);
+void bch2_print_str_nonblocking(struct bch_fs *, const char *, const char *);
 
 __printf(2, 3)
 void bch2_print_opts(struct bch_opts *, const char *, ...);
diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index 4832ac31392a..b6f5e0dfc9f1 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -605,7 +605,7 @@ static int __btree_err(int ret,
 	}
 
 	if (!silent)
-		bch2_print_string_as_lines(KERN_ERR, out.buf);
+		bch2_print_str(c, KERN_ERR, out.buf);
 out:
 fsck_err:
 	printbuf_exit(&out);
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index cfd6363dfc39..bd3a0bc07511 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -1591,7 +1591,7 @@ void __bch2_dump_trans_paths_updates(struct btree_trans *trans, bool nosort)
 	__bch2_trans_paths_to_text(&buf, trans, nosort);
 	bch2_trans_updates_to_text(&buf, trans);
 
-	bch2_print_str(trans->c, buf.buf);
+	bch2_print_str(trans->c, KERN_ERR, buf.buf);
 	printbuf_exit(&buf);
 }
 
@@ -3121,7 +3121,7 @@ void *__bch2_trans_kmalloc(struct btree_trans *trans, size_t size, unsigned long
 #ifdef CONFIG_BCACHEFS_TRANS_KMALLOC_TRACE
 		struct printbuf buf = PRINTBUF;
 		bch2_trans_kmalloc_trace_to_text(&buf, &trans->trans_kmalloc_trace);
-		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		bch2_print_str(c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 #endif
 	}
diff --git a/fs/bcachefs/btree_locking.c b/fs/bcachefs/btree_locking.c
index f4f563944340..baa505a9a706 100644
--- a/fs/bcachefs/btree_locking.c
+++ b/fs/bcachefs/btree_locking.c
@@ -236,7 +236,7 @@ static noinline int break_cycle(struct lock_graph *g, struct printbuf *cycle,
 			prt_newline(&buf);
 		}
 
-		bch2_print_string_as_lines_nonblocking(KERN_ERR, buf.buf);
+		bch2_print_str_nonblocking(g->g->trans->c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 		BUG();
 	}
diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c
index 86acf037590c..81ee7ae88a77 100644
--- a/fs/bcachefs/btree_node_scan.c
+++ b/fs/bcachefs/btree_node_scan.c
@@ -395,7 +395,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c)
 		printbuf_reset(&buf);
 		prt_printf(&buf, "%s: nodes found:\n", __func__);
 		found_btree_nodes_to_text(&buf, c, f->nodes);
-		bch2_print_string_as_lines(KERN_INFO, buf.buf);
+		bch2_print_str(c, KERN_INFO, buf.buf);
 	}
 
 	sort_nonatomic(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_cookie, NULL);
@@ -424,7 +424,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c)
 		printbuf_reset(&buf);
 		prt_printf(&buf, "%s: nodes after merging replicas:\n", __func__);
 		found_btree_nodes_to_text(&buf, c, f->nodes);
-		bch2_print_string_as_lines(KERN_INFO, buf.buf);
+		bch2_print_str(c, KERN_INFO, buf.buf);
 	}
 
 	swap(nodes_heap, f->nodes);
@@ -470,7 +470,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c)
 		printbuf_reset(&buf);
 		prt_printf(&buf, "%s: nodes found after overwrites:\n", __func__);
 		found_btree_nodes_to_text(&buf, c, f->nodes);
-		bch2_print_string_as_lines(KERN_INFO, buf.buf);
+		bch2_print_str(c, KERN_INFO, buf.buf);
 	} else {
 		bch_info(c, "btree node scan found %zu nodes after overwrites", f->nodes.nr);
 	}
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 00307356d7c8..2be7c10fc59c 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -1807,7 +1807,7 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t
 		bch2_btree_update_to_text(&buf, as);
 		bch2_btree_path_to_text(&buf, trans, path_idx);
 
-		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		bch2_print_str(c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 		bch2_fs_emergency_read_only(c);
 		return -EIO;
diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index 8e64077c15c1..36c1e391d4df 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -409,7 +409,7 @@ static int bucket_ref_update_err(struct btree_trans *trans, struct printbuf *buf
 	}
 
 	if (print || insert)
-		bch2_print_string_as_lines(KERN_ERR, buf->buf);
+		bch2_print_str(c, KERN_ERR, buf->buf);
 	return ret;
 }
 
@@ -706,7 +706,7 @@ err:
 				   (u64) p.ec.idx);
 			bch2_bkey_val_to_text(&buf, c, k);
 			__bch2_inconsistent_error(c, &buf);
-			bch2_print_string_as_lines(KERN_ERR, buf.buf);
+			bch2_print_str(c, KERN_ERR, buf.buf);
 			printbuf_exit(&buf);
 			return -BCH_ERR_trigger_stripe_pointer;
 		}
@@ -976,7 +976,7 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 					BCH_RECOVERY_PASS_check_allocations);
 
 		if (print)
-			bch2_print_string_as_lines(KERN_ERR, buf.buf);
+			bch2_print_str(c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 		ret = -BCH_ERR_metadata_bucket_inconsistency;
 		goto err;
diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c
index b211c97238ab..c3034338f9e4 100644
--- a/fs/bcachefs/data_update.c
+++ b/fs/bcachefs/data_update.c
@@ -358,7 +358,7 @@ restart_drop_extra_replicas:
 			prt_str(&buf, "\nnew: ");
 			bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(insert));
 
-			bch2_print_string_as_lines(KERN_ERR, buf.buf);
+			bch2_print_str(c, KERN_ERR, buf.buf);
 			printbuf_exit(&buf);
 
 			bch2_fatal_error(c);
diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c
index 4627aabd1f1a..20495062d6e1 100644
--- a/fs/bcachefs/error.c
+++ b/fs/bcachefs/error.c
@@ -34,7 +34,7 @@ bool __bch2_inconsistent_error(struct bch_fs *c, struct printbuf *out)
 				   journal_cur_seq(&c->journal));
 		return true;
 	case BCH_ON_ERROR_panic:
-		bch2_print_string_as_lines_nonblocking(KERN_ERR, out->buf);
+		bch2_print_str(c, KERN_ERR, out->buf);
 		panic(bch2_fmt(c, "panic after error"));
 		return true;
 	default:
@@ -71,7 +71,7 @@ static bool bch2_fs_trans_inconsistent(struct bch_fs *c, struct btree_trans *tra
 	if (trans)
 		bch2_trans_updates_to_text(&buf, trans);
 	bool ret = __bch2_inconsistent_error(c, &buf);
-	bch2_print_string_as_lines_nonblocking(KERN_ERR, buf.buf);
+	bch2_print_str_nonblocking(c, KERN_ERR, buf.buf);
 
 	printbuf_exit(&buf);
 	return ret;
@@ -121,7 +121,7 @@ int bch2_fs_topology_error(struct bch_fs *c, const char *fmt, ...)
 	va_end(args);
 
 	int ret = __bch2_topology_error(c, &buf);
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	bch2_print_str(c, KERN_ERR, buf.buf);
 
 	printbuf_exit(&buf);
 	return ret;
@@ -328,7 +328,7 @@ static int do_fsck_ask_yn(struct bch_fs *c,
 	if (bch2_fs_stdio_redirect(c))
 		bch2_print(c, "%s", question->buf);
 	else
-		bch2_print_string_as_lines(KERN_ERR, question->buf);
+		bch2_print_str(c, KERN_ERR, question->buf);
 
 	int ask = bch2_fsck_ask_yn(c, trans);
 
@@ -565,7 +565,7 @@ print:
 		if (bch2_fs_stdio_redirect(c))
 			bch2_print(c, "%s", out->buf);
 		else
-			bch2_print_string_as_lines(KERN_ERR, out->buf);
+			bch2_print_str(c, KERN_ERR, out->buf);
 	}
 
 	if (s)
diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index b81117b51c69..7200ec00128d 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -153,7 +153,7 @@ void __bch2_i_sectors_acct(struct bch_fs *c, struct bch_inode_info *inode,
 
 		bool print = bch2_count_fsck_err(c, vfs_inode_i_blocks_underflow, &buf);
 		if (print)
-			bch2_print_str(c, buf.buf);
+			bch2_print_str(c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 
 		if (sectors < 0)
@@ -527,7 +527,7 @@ int bchfs_truncate(struct mnt_idmap *idmap,
 
 		bool print = bch2_count_fsck_err(c, vfs_inode_i_blocks_not_zero_at_truncate, &buf);
 		if (print)
-			bch2_print_str(c, buf.buf);
+			bch2_print_str(c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 	}
 
diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c
index c738ae6fd9a5..38086c1a8e28 100644
--- a/fs/bcachefs/io_write.c
+++ b/fs/bcachefs/io_write.c
@@ -265,7 +265,7 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
 
 			bool print = bch2_count_fsck_err(c, inode_i_sectors_underflow, &buf);
 			if (print)
-				bch2_print_str(c, buf.buf);
+				bch2_print_str(c, KERN_ERR, buf.buf);
 			printbuf_exit(&buf);
 
 			if (i_sectors_delta < 0)
diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index a51ad32931b8..5442d526a448 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -173,7 +173,7 @@ journal_error_check_stuck(struct journal *j, int error, unsigned flags)
 	spin_unlock(&j->lock);
 	prt_printf(&buf, bch2_fmt(c, "Journal stuck! Hava a pre-reservation but journal full (error %s)"),
 				  bch2_err_str(error));
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	bch2_print_str(c, KERN_ERR, buf.buf);
 
 	printbuf_reset(&buf);
 	bch2_journal_pins_to_text(&buf, j);
@@ -743,7 +743,7 @@ int bch2_journal_res_get_slowpath(struct journal *j, struct journal_res *res,
 
 	struct printbuf buf = PRINTBUF;
 	bch2_journal_debug_to_text(&buf, j);
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
+	bch2_print_str(c, KERN_ERR, buf.buf);
 	prt_printf(&buf, bch2_fmt(c, "Journal stuck? Waited for 10 seconds, err %s"), bch2_err_str(ret));
 	printbuf_exit(&buf);
 
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index ded18a94ed02..28fa381cd589 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -2107,7 +2107,7 @@ CLOSURE_CALLBACK(bch2_journal_write)
 					  le64_to_cpu(w->data->seq),
 					  vstruct_sectors(w->data, c->block_bits),
 					  bch2_err_str(ret));
-		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		bch2_print_str(c, KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 	}
 	if (ret)
diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index 946428daeecc..de1a14c4bc3c 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -163,7 +163,7 @@ int bch2_run_explicit_recovery_pass(struct bch_fs *c,
 	int ret = bch2_run_explicit_recovery_pass_printbuf(c, &buf, pass);
 
 	if (len != buf.pos)
-		bch2_print_string_as_lines(KERN_NOTICE, buf.buf);
+		bch2_print_str(c, KERN_NOTICE, buf.buf);
 	printbuf_exit(&buf);
 	return ret;
 }
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index f776b00c3cc0..77809ee23c45 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -24,7 +24,7 @@ int bch2_dev_missing_bkey(struct bch_fs *c, struct bkey_s_c k, unsigned dev)
 						 BCH_RECOVERY_PASS_check_allocations);
 
 	if (print)
-		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		bch2_print_str(c, KERN_ERR, buf.buf);
 	printbuf_exit(&buf);
 	return ret;
 }
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 65aab7ea182e..b34c91dd51b1 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -84,7 +84,8 @@ const char * const bch2_fs_flag_strs[] = {
 	NULL
 };
 
-void bch2_print_str(struct bch_fs *c, const char *str)
+static void __bch2_print_str(struct bch_fs *c, const char *prefix,
+			     const char *str, bool nonblocking)
 {
 #ifdef __KERNEL__
 	struct stdio_redirect *stdio = bch2_fs_stdio_redirect(c);
@@ -94,7 +95,17 @@ void bch2_print_str(struct bch_fs *c, const char *str)
 		return;
 	}
 #endif
-	bch2_print_string_as_lines(KERN_ERR, str);
+	bch2_print_string_as_lines(KERN_ERR, str, nonblocking);
+}
+
+void bch2_print_str(struct bch_fs *c, const char *prefix, const char *str)
+{
+	__bch2_print_str(c, prefix, str, false);
+}
+
+void bch2_print_str_nonblocking(struct bch_fs *c, const char *prefix, const char *str)
+{
+	__bch2_print_str(c, prefix, str, true);
 }
 
 __printf(2, 0)
diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c
index 87af551692f4..6e5d7fc265bd 100644
--- a/fs/bcachefs/util.c
+++ b/fs/bcachefs/util.c
@@ -252,8 +252,8 @@ void bch2_prt_u64_base2(struct printbuf *out, u64 v)
 	bch2_prt_u64_base2_nbits(out, v, fls64(v) ?: 1);
 }
 
-static void __bch2_print_string_as_lines(const char *prefix, const char *lines,
-					 bool nonblocking)
+void bch2_print_string_as_lines(const char *prefix, const char *lines,
+				bool nonblocking)
 {
 	bool locked = false;
 	const char *p;
@@ -281,16 +281,6 @@ static void __bch2_print_string_as_lines(const char *prefix, const char *lines,
 		console_unlock();
 }
 
-void bch2_print_string_as_lines(const char *prefix, const char *lines)
-{
-	return __bch2_print_string_as_lines(prefix, lines, false);
-}
-
-void bch2_print_string_as_lines_nonblocking(const char *prefix, const char *lines)
-{
-	return __bch2_print_string_as_lines(prefix, lines, true);
-}
-
 int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *task, unsigned skipnr,
 			gfp_t gfp)
 {
diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index ccc1cf699c4b..50f7197c67fc 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -212,8 +212,7 @@ u64 bch2_read_flag_list(const char *, const char * const[]);
 void bch2_prt_u64_base2_nbits(struct printbuf *, u64, unsigned);
 void bch2_prt_u64_base2(struct printbuf *, u64);
 
-void bch2_print_string_as_lines(const char *prefix, const char *lines);
-void bch2_print_string_as_lines_nonblocking(const char *prefix, const char *lines);
+void bch2_print_string_as_lines(const char *, const char *, bool);
 
 typedef DARRAY(unsigned long) bch_stacktrace;
 int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *, unsigned, gfp_t);
-- 
2.51.0


From bdad8962c94d3e557317b3d0691a507177f27a22 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 19:15:43 -0400
Subject: [PATCH 05/16] bcachefs: Flag for repair on missing subvolume

Instead of going emegency read only with a bch2_fs_inconsistent() call,
log the error and recovery pass appropriately.

If we're still in recovery it'll be repaired immediately, otherwise
it'll be repaired on the next mount.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/subvolume.c | 43 +++++++++++++++++++++++++++--------------
 1 file changed, 28 insertions(+), 15 deletions(-)

diff --git a/fs/bcachefs/subvolume.c b/fs/bcachefs/subvolume.c
index 0421ffc1128f..c9d7209f0cb1 100644
--- a/fs/bcachefs/subvolume.c
+++ b/fs/bcachefs/subvolume.c
@@ -14,6 +14,22 @@
 
 static int bch2_subvolume_delete(struct btree_trans *, u32);
 
+static int bch2_subvolume_missing(struct bch_fs *c, u32 subvolid)
+{
+	struct printbuf buf = PRINTBUF;
+	bch2_log_msg_start(c, &buf);
+
+	prt_printf(&buf, "missing subvolume %u", subvolid);
+	bool print = bch2_count_fsck_err(c, subvol_missing, &buf);
+
+	int ret = bch2_run_explicit_recovery_pass_printbuf(c, &buf,
+					BCH_RECOVERY_PASS_check_inodes);
+	if (print)
+		bch2_print_str(c, KERN_ERR, buf.buf);
+	printbuf_exit(&buf);
+	return ret;
+}
+
 static struct bpos subvolume_children_pos(struct bkey_s_c k)
 {
 	if (k.k->type != KEY_TYPE_subvolume)
@@ -292,9 +308,8 @@ bch2_subvolume_get_inlined(struct btree_trans *trans, unsigned subvol,
 	int ret = bch2_bkey_get_val_typed(trans, BTREE_ID_subvolumes, POS(0, subvol),
 					  BTREE_ITER_cached|
 					  BTREE_ITER_with_updates, subvolume, s);
-	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT) &&
-				inconsistent_if_not_found,
-				trans->c, "missing subvolume %u", subvol);
+	if (bch2_err_matches(ret, ENOENT) && inconsistent_if_not_found)
+		ret = bch2_subvolume_missing(trans->c, subvol) ?: ret;
 	return ret;
 }
 
@@ -344,8 +359,8 @@ int __bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
 					  subvolume);
 	ret = bkey_err(subvol);
 
-	bch2_fs_inconsistent_on(warn && bch2_err_matches(ret, ENOENT), trans->c,
-				"missing subvolume %u", subvolid);
+	if (bch2_err_matches(ret, ENOENT))
+		ret = bch2_subvolume_missing(trans->c, subvolid) ?: ret;
 
 	if (likely(!ret))
 		*snapid = le32_to_cpu(subvol.v->snapshot);
@@ -418,8 +433,8 @@ static int __bch2_subvolume_delete(struct btree_trans *trans, u32 subvolid)
 				BTREE_ITER_cached|BTREE_ITER_intent,
 				subvolume);
 	int ret = bkey_err(subvol);
-	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), trans->c,
-				"missing subvolume %u", subvolid);
+	if (bch2_err_matches(ret, ENOENT))
+		ret = bch2_subvolume_missing(trans->c, subvolid) ?: ret;
 	if (ret)
 		goto err;
 
@@ -553,11 +568,10 @@ int bch2_subvolume_unlink(struct btree_trans *trans, u32 subvolid)
 			BTREE_ID_subvolumes, POS(0, subvolid),
 			BTREE_ITER_cached, subvolume);
 	ret = PTR_ERR_OR_ZERO(n);
-	if (unlikely(ret)) {
-		bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), trans->c,
-					"missing subvolume %u", subvolid);
+	if (bch2_err_matches(ret, ENOENT))
+		ret = bch2_subvolume_missing(trans->c, subvolid) ?: ret;
+	if (unlikely(ret))
 		return ret;
-	}
 
 	SET_BCH_SUBVOLUME_UNLINKED(&n->v, true);
 	n->v.fs_path_parent = 0;
@@ -596,11 +610,10 @@ int bch2_subvolume_create(struct btree_trans *trans, u64 inode,
 				BTREE_ID_subvolumes, POS(0, src_subvolid),
 				BTREE_ITER_cached, subvolume);
 		ret = PTR_ERR_OR_ZERO(src_subvol);
-		if (unlikely(ret)) {
-			bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), c,
-						"subvolume %u not found", src_subvolid);
+		if (bch2_err_matches(ret, ENOENT))
+			ret = bch2_subvolume_missing(trans->c, src_subvolid) ?: ret;
+		if (unlikely(ret))
 			goto err;
-		}
 
 		parent = le32_to_cpu(src_subvol->v.snapshot);
 	}
-- 
2.51.0


From d12bd4101825c14222ecbe4c6fba9c03ce42f624 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 19:21:52 -0400
Subject: [PATCH 06/16] bcachefs: Add a recovery pass for making sure root
 inode is readable

If the root inode/subvolume is unreadable we can repair automatically -
but only if we're still in recovery, so that we can rewind to the
appropriate recovery pass.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/recovery_passes.c       | 15 +++++++++++++++
 fs/bcachefs/recovery_passes_types.h |  3 ++-
 2 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index de1a14c4bc3c..b4de21f80811 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -52,6 +52,21 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
 	return 0;
 }
 
+/*
+ * Make sure root inode is readable while we're still in recovery and can rewind
+ * for repair:
+ */
+static int bch2_lookup_root_inode(struct bch_fs *c)
+{
+	subvol_inum inum = BCACHEFS_ROOT_SUBVOL_INUM;
+	struct bch_inode_unpacked inode_u;
+	struct bch_subvolume subvol;
+
+	return bch2_trans_do(c,
+		bch2_subvolume_get(trans, inum.subvol, true, &subvol) ?:
+		bch2_inode_find_by_inum_trans(trans, inum, &inode_u));
+}
+
 struct recovery_pass_fn {
 	int		(*fn)(struct bch_fs *);
 	unsigned	when;
diff --git a/fs/bcachefs/recovery_passes_types.h b/fs/bcachefs/recovery_passes_types.h
index e89b9c783285..4671ccf2d560 100644
--- a/fs/bcachefs/recovery_passes_types.h
+++ b/fs/bcachefs/recovery_passes_types.h
@@ -60,7 +60,8 @@
 	x(resume_logged_ops,			23, PASS_ALWAYS)			\
 	x(delete_dead_inodes,			32, PASS_ALWAYS)			\
 	x(fix_reflink_p,			33, 0)					\
-	x(set_fs_needs_rebalance,		34, 0)
+	x(set_fs_needs_rebalance,		34, 0)					\
+	x(lookup_root_inode,			42, PASS_ALWAYS|PASS_SILENT)
 
 /* We normally enumerate recovery passes in the order we run them: */
 enum bch_recovery_pass {
-- 
2.51.0


From 1c8dfd7ba50dbbb72113caf4fa7868512cdad2f4 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 23:35:48 -0400
Subject: [PATCH 07/16] bcachefs: sb_validate() no longer requires members_v1

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/super-io.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c
index 8ebc5e3f3ea3..872707e5fa95 100644
--- a/fs/bcachefs/super-io.c
+++ b/fs/bcachefs/super-io.c
@@ -384,7 +384,6 @@ static int bch2_sb_compatible(struct bch_sb *sb, struct printbuf *out)
 int bch2_sb_validate(struct bch_sb *sb, u64 read_offset,
 		     enum bch_validate_flags flags, struct printbuf *out)
 {
-	struct bch_sb_field_members_v1 *mi;
 	enum bch_opt_id opt_id;
 	int ret;
 
@@ -539,14 +538,17 @@ int bch2_sb_validate(struct bch_sb *sb, u64 read_offset,
 		}
 	}
 
+	struct bch_sb_field *mi =
+		bch2_sb_field_get_id(sb, BCH_SB_FIELD_members_v2) ?:
+		bch2_sb_field_get_id(sb, BCH_SB_FIELD_members_v1);
+
 	/* members must be validated first: */
-	mi = bch2_sb_field_get(sb, members_v1);
 	if (!mi) {
 		prt_printf(out, "Invalid superblock: member info area missing");
 		return -BCH_ERR_invalid_sb_members_missing;
 	}
 
-	ret = bch2_sb_field_validate(sb, &mi->field, flags, out);
+	ret = bch2_sb_field_validate(sb, mi, flags, out);
 	if (ret)
 		return ret;
 
-- 
2.51.0


From 0dc73809e93aeb905acf9fa88502c73534cfa83d Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 16 Apr 2025 06:48:31 -0400
Subject: [PATCH 08/16] bcachefs: Shrink superblock downgrade table

Don't generate entries for versions that won't be able to mount.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/sb-downgrade.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/bcachefs/sb-downgrade.c b/fs/bcachefs/sb-downgrade.c
index badd0e17ada5..296c6c925386 100644
--- a/fs/bcachefs/sb-downgrade.c
+++ b/fs/bcachefs/sb-downgrade.c
@@ -374,6 +374,9 @@ int bch2_sb_downgrade_update(struct bch_fs *c)
 		if (BCH_VERSION_MAJOR(src->version) != BCH_VERSION_MAJOR(le16_to_cpu(c->disk_sb.sb->version)))
 			continue;
 
+		if (src->version < c->sb.version_incompat)
+			continue;
+
 		struct bch_sb_field_downgrade_entry *dst;
 		unsigned bytes = sizeof(*dst) + sizeof(dst->errors[0]) * src->nr_errors;
 
-- 
2.51.0


From 576493133f26a172b8db4313448206d30750c9b2 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 16 Apr 2025 09:23:15 -0400
Subject: [PATCH 09/16] bcachefs: Print features on startup with -o verbose

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/super.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index b34c91dd51b1..60e632e22b98 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -1054,6 +1054,11 @@ static void print_mount_opts(struct bch_fs *c)
 		bch2_version_to_text(&p, c->sb.version_incompat_allowed);
 	}
 
+	if (c->opts.verbose) {
+		prt_printf(&p, "\n  features: ");
+		prt_bitflags(&p, bch2_sb_features, c->sb.features);
+	}
+
 	bch_info(c, "%s", p.buf);
 	printbuf_exit(&p);
 }
-- 
2.51.0


From 203852d9db68e14b50b119cbd123def7e7c9efd0 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 3 Apr 2025 14:19:23 -0400
Subject: [PATCH 10/16] bcachefs: BCH_FEATURE_no_alloc_info

If a filesystem is going to only be used read-only, and will be a
deployable image, we can strip out alloc info for a substantial
reduction in metadata size - around half, due to backpointers.

Alloc info will be regenerated on first read-write mount.

Remounting RW is disallowed for now, since we don't yet have
check_allocations running in RW mode.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs_format.h       |  3 ++-
 fs/bcachefs/errcode.h               |  1 +
 fs/bcachefs/recovery.c              | 28 ++++++++++++++++++++++------
 fs/bcachefs/recovery.h              |  1 +
 fs/bcachefs/recovery_passes.c       | 14 +++++++++++++-
 fs/bcachefs/recovery_passes_types.h | 18 ++++++++++--------
 fs/bcachefs/sb-members.c            |  6 ++++++
 fs/bcachefs/super.c                 |  6 ++++++
 8 files changed, 61 insertions(+), 16 deletions(-)

diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h
index 22ee49408d11..c0041391e2e8 100644
--- a/fs/bcachefs/bcachefs_format.h
+++ b/fs/bcachefs/bcachefs_format.h
@@ -923,7 +923,8 @@ static inline void SET_BCH_SB_BACKGROUND_COMPRESSION_TYPE(struct bch_sb *sb, __u
 	x(alloc_v2,			17)	\
 	x(extents_across_btree_nodes,	18)	\
 	x(incompat_version_field,	19)	\
-	x(casefolding,			20)
+	x(casefolding,			20)	\
+	x(no_alloc_info,		21)
 
 #define BCH_SB_FEATURES_ALWAYS				\
 	(BIT_ULL(BCH_FEATURE_new_extent_overwrite)|	\
diff --git a/fs/bcachefs/errcode.h b/fs/bcachefs/errcode.h
index 051938657cc9..8a4435660d86 100644
--- a/fs/bcachefs/errcode.h
+++ b/fs/bcachefs/errcode.h
@@ -221,6 +221,7 @@
 	x(EROFS,			erofs_unfixed_errors)			\
 	x(EROFS,			erofs_norecovery)			\
 	x(EROFS,			erofs_nochanges)			\
+	x(EROFS,			erofs_no_alloc_info)			\
 	x(EROFS,			insufficient_devices)			\
 	x(0,				operation_blocked)			\
 	x(BCH_ERR_operation_blocked,	btree_cache_cannibalize_lock_blocked)	\
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 4c336f20d5eb..b5ab77f3c692 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -33,7 +33,6 @@
 #include <linux/sort.h>
 #include <linux/stat.h>
 
-
 int bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
 {
 	u64 b = BIT_ULL(btree);
@@ -114,11 +113,8 @@ static void kill_btree(struct bch_fs *c, enum btree_id btree)
 }
 
 /* for -o reconstruct_alloc: */
-static void bch2_reconstruct_alloc(struct bch_fs *c)
+void bch2_reconstruct_alloc(struct bch_fs *c)
 {
-	bch2_journal_log_msg(c, "dropping alloc info");
-	bch_info(c, "dropping and reconstructing all alloc info");
-
 	mutex_lock(&c->sb_lock);
 	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
 
@@ -160,6 +156,8 @@ static void bch2_reconstruct_alloc(struct bch_fs *c)
 
 	c->opts.recovery_passes |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
 
+	c->disk_sb.sb->features[0] &= ~cpu_to_le64(BIT_ULL(BCH_FEATURE_no_alloc_info));
+
 	bch2_write_super(c);
 	mutex_unlock(&c->sb_lock);
 
@@ -889,8 +887,26 @@ use_clean:
 	if (ret)
 		goto err;
 
-	if (c->opts.reconstruct_alloc)
+	if (!c->opts.read_only &&
+	    (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info))) {
+		bch_info(c, "mounting a filesystem with no alloc info read-write; will recreate");
+
 		bch2_reconstruct_alloc(c);
+	} else if (c->opts.reconstruct_alloc) {
+		bch2_journal_log_msg(c, "dropping alloc info");
+		bch_info(c, "dropping and reconstructing all alloc info");
+
+		bch2_reconstruct_alloc(c);
+	}
+
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)) {
+		/* We can't go RW to fix errors without alloc info */
+		if (c->opts.fix_errors == FSCK_FIX_yes ||
+		    c->opts.fix_errors == FSCK_FIX_ask)
+			c->opts.fix_errors = FSCK_FIX_no;
+		if (c->opts.errors == BCH_ON_ERROR_fix_safe)
+			c->opts.errors = BCH_ON_ERROR_continue;
+	}
 
 	/*
 	 * After an unclean shutdown, skip then next few journal sequence
diff --git a/fs/bcachefs/recovery.h b/fs/bcachefs/recovery.h
index b0d55754b21b..d858ba674eaa 100644
--- a/fs/bcachefs/recovery.h
+++ b/fs/bcachefs/recovery.h
@@ -3,6 +3,7 @@
 #define _BCACHEFS_RECOVERY_H
 
 int bch2_btree_lost_data(struct bch_fs *, enum btree_id);
+void bch2_reconstruct_alloc(struct bch_fs *);
 
 int bch2_journal_replay(struct bch_fs *);
 
diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index b4de21f80811..87150dd30f4b 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -47,8 +47,18 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
 
 	set_bit(BCH_FS_may_go_rw, &c->flags);
 
-	if (keys->nr || !c->opts.read_only || c->opts.fsck || !c->sb.clean || c->opts.recovery_passes)
+	if (keys->nr ||
+	    !c->opts.read_only ||
+	    !c->sb.clean ||
+	    c->opts.recovery_passes ||
+	    (c->opts.fsck && !(c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)))) {
+		if (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)) {
+			bch_info(c, "mounting a filesystem with no alloc info read-write; will recreate");
+			bch2_reconstruct_alloc(c);
+		}
+
 		return bch2_fs_read_write_early(c);
+	}
 	return 0;
 }
 
@@ -240,6 +250,8 @@ static bool should_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pa
 {
 	struct recovery_pass_fn *p = recovery_pass_fns + pass;
 
+	if ((p->when & PASS_ALLOC) && (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)))
+		return false;
 	if (c->opts.recovery_passes_exclude & BIT_ULL(pass))
 		return false;
 	if (c->opts.recovery_passes & BIT_ULL(pass))
diff --git a/fs/bcachefs/recovery_passes_types.h b/fs/bcachefs/recovery_passes_types.h
index 4671ccf2d560..f9d565bb50dd 100644
--- a/fs/bcachefs/recovery_passes_types.h
+++ b/fs/bcachefs/recovery_passes_types.h
@@ -7,6 +7,8 @@
 #define PASS_UNCLEAN		BIT(2)
 #define PASS_ALWAYS		BIT(3)
 #define PASS_ONLINE		BIT(4)
+#define PASS_ALLOC		BIT(5)
+#define PASS_FSCK_ALLOC		(PASS_FSCK|PASS_ALLOC)
 
 #ifdef CONFIG_BCACHEFS_DEBUG
 #define PASS_FSCK_DEBUG		BIT(1)
@@ -27,17 +29,17 @@
 	x(stripes_read,				 1, 0)					\
 	x(initialize_subvolumes,		 2, 0)					\
 	x(snapshots_read,			 3, PASS_ALWAYS)			\
-	x(check_allocations,			 5, PASS_FSCK)				\
-	x(trans_mark_dev_sbs,			 6, PASS_ALWAYS|PASS_SILENT)		\
-	x(fs_journal_alloc,			 7, PASS_ALWAYS|PASS_SILENT)		\
+	x(check_allocations,			 5, PASS_FSCK_ALLOC)			\
+	x(trans_mark_dev_sbs,			 6, PASS_ALWAYS|PASS_SILENT|PASS_ALLOC)	\
+	x(fs_journal_alloc,			 7, PASS_ALWAYS|PASS_SILENT|PASS_ALLOC)	\
 	x(set_may_go_rw,			 8, PASS_ALWAYS|PASS_SILENT)		\
 	x(journal_replay,			 9, PASS_ALWAYS)			\
-	x(check_alloc_info,			10, PASS_ONLINE|PASS_FSCK)		\
-	x(check_lrus,				11, PASS_ONLINE|PASS_FSCK)		\
-	x(check_btree_backpointers,		12, PASS_ONLINE|PASS_FSCK)		\
+	x(check_alloc_info,			10, PASS_ONLINE|PASS_FSCK_ALLOC)	\
+	x(check_lrus,				11, PASS_ONLINE|PASS_FSCK_ALLOC)	\
+	x(check_btree_backpointers,		12, PASS_ONLINE|PASS_FSCK_ALLOC)	\
 	x(check_backpointers_to_extents,	13, PASS_ONLINE|PASS_FSCK_DEBUG)	\
-	x(check_extents_to_backpointers,	14, PASS_ONLINE|PASS_FSCK)		\
-	x(check_alloc_to_lru_refs,		15, PASS_ONLINE|PASS_FSCK)		\
+	x(check_extents_to_backpointers,	14, PASS_ONLINE|PASS_FSCK_ALLOC)	\
+	x(check_alloc_to_lru_refs,		15, PASS_ONLINE|PASS_FSCK_ALLOC)	\
 	x(fs_freespace_init,			16, PASS_ALWAYS|PASS_SILENT)		\
 	x(bucket_gens_init,			17, 0)					\
 	x(reconstruct_snapshots,		38, 0)					\
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index 77809ee23c45..39ce94875dde 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -190,6 +190,12 @@ static int validate_member(struct printbuf *err,
 		return -BCH_ERR_invalid_sb_members;
 	}
 
+	if (BCH_MEMBER_FREESPACE_INITIALIZED(&m) &&
+	    sb->features[0] & cpu_to_le64(BIT_ULL(BCH_FEATURE_no_alloc_info))) {
+		prt_printf(err, "device %u: freespace initialized but fs has no alloc info", i);
+		return -BCH_ERR_invalid_sb_members;
+	}
+
 	return 0;
 }
 
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 60e632e22b98..6ab3e63ef139 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -443,6 +443,9 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early)
 
 	BUG_ON(!test_bit(BCH_FS_may_go_rw, &c->flags));
 
+	if (WARN_ON(c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)))
+		return -BCH_ERR_erofs_no_alloc_info;
+
 	if (test_bit(BCH_FS_initial_gc_unfixed, &c->flags)) {
 		bch_err(c, "cannot go rw, unfixed btree errors");
 		return -BCH_ERR_erofs_unfixed_errors;
@@ -535,6 +538,9 @@ int bch2_fs_read_write(struct bch_fs *c)
 	if (c->opts.nochanges)
 		return -BCH_ERR_erofs_nochanges;
 
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info))
+		return -BCH_ERR_erofs_no_alloc_info;
+
 	return __bch2_fs_read_write(c, false);
 }
 
-- 
2.51.0


From 530112d88ebd7405fb61711892b2a680048984c7 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 15:15:36 -0400
Subject: [PATCH 11/16] bcachefs: BCH_FEATURE_small_image

We can't go RW if it's an image file that hasn't been resized.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/alloc_background.c |  8 +++++---
 fs/bcachefs/bcachefs_format.h  |  3 ++-
 fs/bcachefs/errcode.h          |  1 +
 fs/bcachefs/journal.c          |  9 ++++++++-
 fs/bcachefs/journal_reclaim.c  | 26 ++++++++++++++------------
 fs/bcachefs/recovery.c         |  5 +++++
 fs/bcachefs/super.c            |  5 +++++
 7 files changed, 40 insertions(+), 17 deletions(-)

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 8b8c2344855f..6ac8bd49c629 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -2392,14 +2392,16 @@ bkey_err:
 
 int bch2_fs_freespace_init(struct bch_fs *c)
 {
-	int ret = 0;
-	bool doing_init = false;
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image))
+		return 0;
+
 
 	/*
 	 * We can crash during the device add path, so we need to check this on
 	 * every mount:
 	 */
 
+	bool doing_init = false;
 	for_each_member_device(c, ca) {
 		if (ca->mi.freespace_initialized)
 			continue;
@@ -2409,7 +2411,7 @@ int bch2_fs_freespace_init(struct bch_fs *c)
 			doing_init = true;
 		}
 
-		ret = bch2_dev_freespace_init(c, ca, 0, ca->mi.nbuckets);
+		int ret = bch2_dev_freespace_init(c, ca, 0, ca->mi.nbuckets);
 		if (ret) {
 			bch2_dev_put(ca);
 			bch_err_fn(c, ret);
diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h
index c0041391e2e8..7ce475c565b5 100644
--- a/fs/bcachefs/bcachefs_format.h
+++ b/fs/bcachefs/bcachefs_format.h
@@ -924,7 +924,8 @@ static inline void SET_BCH_SB_BACKGROUND_COMPRESSION_TYPE(struct bch_sb *sb, __u
 	x(extents_across_btree_nodes,	18)	\
 	x(incompat_version_field,	19)	\
 	x(casefolding,			20)	\
-	x(no_alloc_info,		21)
+	x(no_alloc_info,		21)	\
+	x(small_image,			22)
 
 #define BCH_SB_FEATURES_ALWAYS				\
 	(BIT_ULL(BCH_FEATURE_new_extent_overwrite)|	\
diff --git a/fs/bcachefs/errcode.h b/fs/bcachefs/errcode.h
index 8a4435660d86..6a4b3fe9ea99 100644
--- a/fs/bcachefs/errcode.h
+++ b/fs/bcachefs/errcode.h
@@ -222,6 +222,7 @@
 	x(EROFS,			erofs_norecovery)			\
 	x(EROFS,			erofs_nochanges)			\
 	x(EROFS,			erofs_no_alloc_info)			\
+	x(EROFS,			erofs_filesystem_full)			\
 	x(EROFS,			insufficient_devices)			\
 	x(0,				operation_blocked)			\
 	x(BCH_ERR_operation_blocked,	btree_cache_cannibalize_lock_blocked)	\
diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index 5442d526a448..3694b83af8cc 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -1295,9 +1295,16 @@ int bch2_set_nr_journal_buckets(struct bch_fs *c, struct bch_dev *ca,
 
 int bch2_dev_journal_alloc(struct bch_dev *ca, bool new_fs)
 {
+	struct bch_fs *c = ca->fs;
+
 	if (!(ca->mi.data_allowed & BIT(BCH_DATA_journal)))
 		return 0;
 
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image)) {
+		bch_err(c, "cannot allocate journal, filesystem is an unresized image file");
+		return -BCH_ERR_erofs_filesystem_full;
+	}
+
 	unsigned nr;
 	int ret;
 
@@ -1318,7 +1325,7 @@ int bch2_dev_journal_alloc(struct bch_dev *ca, bool new_fs)
 		     min(1 << 13,
 			 (1 << 24) / ca->mi.bucket_size));
 
-	ret = bch2_set_nr_journal_buckets_loop(ca->fs, ca, nr, new_fs);
+	ret = bch2_set_nr_journal_buckets_loop(c, ca, nr, new_fs);
 err:
 	bch_err_fn(ca, ret);
 	return ret;
diff --git a/fs/bcachefs/journal_reclaim.c b/fs/bcachefs/journal_reclaim.c
index cc00b0fc40d8..a02f483a016a 100644
--- a/fs/bcachefs/journal_reclaim.c
+++ b/fs/bcachefs/journal_reclaim.c
@@ -215,18 +215,20 @@ void bch2_journal_space_available(struct journal *j)
 	j->can_discard = can_discard;
 
 	if (nr_online < metadata_replicas_required(c)) {
-		struct printbuf buf = PRINTBUF;
-		buf.atomic++;
-		prt_printf(&buf, "insufficient writeable journal devices available: have %u, need %u\n"
-			   "rw journal devs:", nr_online, metadata_replicas_required(c));
-
-		rcu_read_lock();
-		for_each_member_device_rcu(c, ca, &c->rw_devs[BCH_DATA_journal])
-			prt_printf(&buf, " %s", ca->name);
-		rcu_read_unlock();
-
-		bch_err(c, "%s", buf.buf);
-		printbuf_exit(&buf);
+		if (!(c->sb.features & BIT_ULL(BCH_FEATURE_small_image))) {
+			struct printbuf buf = PRINTBUF;
+			buf.atomic++;
+			prt_printf(&buf, "insufficient writeable journal devices available: have %u, need %u\n"
+				   "rw journal devs:", nr_online, metadata_replicas_required(c));
+
+			rcu_read_lock();
+			for_each_member_device_rcu(c, ca, &c->rw_devs[BCH_DATA_journal])
+				prt_printf(&buf, " %s", ca->name);
+			rcu_read_unlock();
+
+			bch_err(c, "%s", buf.buf);
+			printbuf_exit(&buf);
+		}
 		ret = -BCH_ERR_insufficient_journal_devices;
 		goto out;
 	}
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index b5ab77f3c692..2436f334dde4 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -734,6 +734,11 @@ int bch2_fs_recovery(struct bch_fs *c)
 		c->opts.read_only = true;
 	}
 
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image)) {
+		bch_info(c, "filesystem is an unresized image file, mounting ro");
+		c->opts.read_only = true;
+	}
+
 	mutex_lock(&c->sb_lock);
 	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
 	bool write_sb = false;
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 6ab3e63ef139..7cd075303f95 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -451,6 +451,11 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early)
 		return -BCH_ERR_erofs_unfixed_errors;
 	}
 
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image)) {
+		bch_err(c, "cannot go rw, filesystem is an unresized image file");
+		return -BCH_ERR_erofs_filesystem_full;
+	}
+
 	if (test_bit(BCH_FS_rw, &c->flags))
 		return 0;
 
-- 
2.51.0


From 0ca375b1779f22f703f956b22ea2bdbc69c247eb Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Tue, 15 Apr 2025 14:09:34 -0400
Subject: [PATCH 12/16] bcachefs: BCH_MEMBER_RESIZE_ON_MOUNT

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/recovery.c          | 18 +++++++---
 fs/bcachefs/sb-members.c        |  1 +
 fs/bcachefs/sb-members.h        |  1 +
 fs/bcachefs/sb-members_format.h |  2 ++
 fs/bcachefs/sb-members_types.h  |  1 +
 fs/bcachefs/super.c             | 64 +++++++++++++++++++++++++++++----
 fs/bcachefs/super.h             |  2 ++
 7 files changed, 77 insertions(+), 12 deletions(-)

diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 2436f334dde4..2a8bcb9b1dd2 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -734,11 +734,6 @@ int bch2_fs_recovery(struct bch_fs *c)
 		c->opts.read_only = true;
 	}
 
-	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image)) {
-		bch_info(c, "filesystem is an unresized image file, mounting ro");
-		c->opts.read_only = true;
-	}
-
 	mutex_lock(&c->sb_lock);
 	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
 	bool write_sb = false;
@@ -892,6 +887,17 @@ use_clean:
 	if (ret)
 		goto err;
 
+	ret = bch2_fs_resize_on_mount(c);
+	if (ret) {
+		up_write(&c->state_lock);
+		goto err;
+	}
+
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_small_image)) {
+		bch_info(c, "filesystem is an unresized image file, mounting ro");
+		c->opts.read_only = true;
+	}
+
 	if (!c->opts.read_only &&
 	    (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info))) {
 		bch_info(c, "mounting a filesystem with no alloc info read-write; will recreate");
@@ -954,6 +960,8 @@ use_clean:
 	set_bit(BCH_FS_btree_running, &c->flags);
 
 	ret = bch2_sb_set_upgrade_extra(c);
+	if (ret)
+		goto err;
 
 	ret = bch2_run_recovery_passes(c);
 	if (ret)
diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c
index 39ce94875dde..462a2c21a9de 100644
--- a/fs/bcachefs/sb-members.c
+++ b/fs/bcachefs/sb-members.c
@@ -294,6 +294,7 @@ static void member_to_text(struct printbuf *out,
 
 	prt_printf(out, "Discard:\t%llu\n", BCH_MEMBER_DISCARD(&m));
 	prt_printf(out, "Freespace initialized:\t%llu\n", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
+	prt_printf(out, "Resize on mount:\t%llu\n", BCH_MEMBER_RESIZE_ON_MOUNT(&m));
 
 	printbuf_indent_sub(out, 2);
 }
diff --git a/fs/bcachefs/sb-members.h b/fs/bcachefs/sb-members.h
index 0f1741fffcb6..424143f5e330 100644
--- a/fs/bcachefs/sb-members.h
+++ b/fs/bcachefs/sb-members.h
@@ -353,6 +353,7 @@ static inline struct bch_member_cpu bch2_mi_to_cpu(struct bch_member *mi)
 			? BCH_MEMBER_DURABILITY(mi) - 1
 			: 1,
 		.freespace_initialized = BCH_MEMBER_FREESPACE_INITIALIZED(mi),
+		.resize_on_mount	= BCH_MEMBER_RESIZE_ON_MOUNT(mi),
 		.valid		= bch2_member_alive(mi),
 		.btree_bitmap_shift	= mi->btree_bitmap_shift,
 		.btree_allocated_bitmap = le64_to_cpu(mi->btree_allocated_bitmap),
diff --git a/fs/bcachefs/sb-members_format.h b/fs/bcachefs/sb-members_format.h
index 3affec823b3f..472218a59102 100644
--- a/fs/bcachefs/sb-members_format.h
+++ b/fs/bcachefs/sb-members_format.h
@@ -88,6 +88,8 @@ LE64_BITMASK(BCH_MEMBER_GROUP,		struct bch_member, flags, 20, 28)
 LE64_BITMASK(BCH_MEMBER_DURABILITY,	struct bch_member, flags, 28, 30)
 LE64_BITMASK(BCH_MEMBER_FREESPACE_INITIALIZED,
 					struct bch_member, flags, 30, 31)
+LE64_BITMASK(BCH_MEMBER_RESIZE_ON_MOUNT,
+					struct bch_member, flags, 31, 32)
 
 #if 0
 LE64_BITMASK(BCH_MEMBER_NR_READ_ERRORS,	struct bch_member, flags[1], 0,  20);
diff --git a/fs/bcachefs/sb-members_types.h b/fs/bcachefs/sb-members_types.h
index c0eda888fe39..d6443e186872 100644
--- a/fs/bcachefs/sb-members_types.h
+++ b/fs/bcachefs/sb-members_types.h
@@ -13,6 +13,7 @@ struct bch_member_cpu {
 	u8			data_allowed;
 	u8			durability;
 	u8			freespace_initialized;
+	u8			resize_on_mount;
 	u8			valid;
 	u8			btree_bitmap_shift;
 	u64			btree_allocated_bitmap;
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 7cd075303f95..839b1582c1f1 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -1141,6 +1141,9 @@ int bch2_fs_start(struct bch_fs *c)
 	for_each_online_member(c, ca)
 		bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx)->last_mount = cpu_to_le64(now);
 
+	/*
+	 * Dno't write superblock yet: recovery might have to downgrade
+	 */
 	mutex_unlock(&c->sb_lock);
 
 	for_each_rw_member(c, ca)
@@ -2039,6 +2042,18 @@ int bch2_dev_offline(struct bch_fs *c, struct bch_dev *ca, int flags)
 	return 0;
 }
 
+static int __bch2_dev_resize_alloc(struct bch_dev *ca, u64 old_nbuckets, u64 new_nbuckets)
+{
+	struct bch_fs *c = ca->fs;
+	u64 v[3] = { new_nbuckets - old_nbuckets, 0, 0 };
+
+	return bch2_trans_commit_do(ca->fs, NULL, NULL, 0,
+			bch2_disk_accounting_mod2(trans, false, v, dev_data_type,
+						  .dev = ca->dev_idx,
+						  .data_type = BCH_DATA_free)) ?:
+		bch2_dev_freespace_init(c, ca, old_nbuckets, new_nbuckets);
+}
+
 int bch2_dev_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 {
 	struct bch_member *m;
@@ -2086,13 +2101,7 @@ int bch2_dev_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 	mutex_unlock(&c->sb_lock);
 
 	if (ca->mi.freespace_initialized) {
-		u64 v[3] = { nbuckets - old_nbuckets, 0, 0 };
-
-		ret   = bch2_trans_commit_do(ca->fs, NULL, NULL, 0,
-				bch2_disk_accounting_mod2(trans, false, v, dev_data_type,
-							  .dev = ca->dev_idx,
-							  .data_type = BCH_DATA_free)) ?:
-			bch2_dev_freespace_init(c, ca, old_nbuckets, nbuckets);
+		ret = __bch2_dev_resize_alloc(ca, old_nbuckets, nbuckets);
 		if (ret)
 			goto err;
 	}
@@ -2103,6 +2112,47 @@ err:
 	return ret;
 }
 
+int bch2_fs_resize_on_mount(struct bch_fs *c)
+{
+	for_each_online_member(c, ca) {
+		u64 old_nbuckets = ca->mi.nbuckets;
+		u64 new_nbuckets = div64_u64(get_capacity(ca->disk_sb.bdev->bd_disk),
+					 ca->mi.bucket_size);
+
+		if (ca->mi.resize_on_mount &&
+		    new_nbuckets > ca->mi.nbuckets) {
+			bch_info(ca, "resizing to size %llu", new_nbuckets * ca->mi.bucket_size);
+			int ret = bch2_dev_buckets_resize(c, ca, new_nbuckets);
+			bch_err_fn(ca, ret);
+			if (ret) {
+				percpu_ref_put(&ca->io_ref[READ]);
+				up_write(&c->state_lock);
+				return ret;
+			}
+
+			mutex_lock(&c->sb_lock);
+			struct bch_member *m =
+				bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
+			m->nbuckets = cpu_to_le64(new_nbuckets);
+			SET_BCH_MEMBER_RESIZE_ON_MOUNT(m, false);
+
+			c->disk_sb.sb->features[0] &= ~cpu_to_le64(BIT_ULL(BCH_FEATURE_small_image));
+			bch2_write_super(c);
+			mutex_unlock(&c->sb_lock);
+
+			if (ca->mi.freespace_initialized) {
+				ret = __bch2_dev_resize_alloc(ca, old_nbuckets, new_nbuckets);
+				if (ret) {
+					percpu_ref_put(&ca->io_ref[READ]);
+					up_write(&c->state_lock);
+					return ret;
+				}
+			}
+		}
+	}
+	return 0;
+}
+
 /* return with ref on ca->ref: */
 struct bch_dev *bch2_dev_lookup(struct bch_fs *c, const char *name)
 {
diff --git a/fs/bcachefs/super.h b/fs/bcachefs/super.h
index 23533bce5709..50588ab20be2 100644
--- a/fs/bcachefs/super.h
+++ b/fs/bcachefs/super.h
@@ -35,6 +35,8 @@ void bch2_fs_read_only(struct bch_fs *);
 int bch2_fs_read_write(struct bch_fs *);
 int bch2_fs_read_write_early(struct bch_fs *);
 
+int bch2_fs_resize_on_mount(struct bch_fs *);
+
 void __bch2_fs_stop(struct bch_fs *);
 void bch2_fs_free(struct bch_fs *);
 void bch2_fs_stop(struct bch_fs *);
-- 
2.51.0


From ecedc87cfaf016e7e857a209e1b2685a28d59566 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 3 Apr 2025 19:33:54 -0400
Subject: [PATCH 13/16] bcachefs: export bch2_move_data_phys()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/move.c | 20 ++++++++++----------
 fs/bcachefs/move.h |  5 +++++
 2 files changed, 15 insertions(+), 10 deletions(-)

diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index a4678a205da6..29981ebcb972 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -900,16 +900,16 @@ err:
 	return ret;
 }
 
-static int bch2_move_data_phys(struct bch_fs *c,
-			       unsigned dev,
-			       u64 start,
-			       u64 end,
-			       unsigned data_types,
-			       struct bch_ratelimit *rate,
-			       struct bch_move_stats *stats,
-			       struct write_point_specifier wp,
-			       bool wait_on_copygc,
-			       move_pred_fn pred, void *arg)
+int bch2_move_data_phys(struct bch_fs *c,
+			unsigned dev,
+			u64 start,
+			u64 end,
+			unsigned data_types,
+			struct bch_ratelimit *rate,
+			struct bch_move_stats *stats,
+			struct write_point_specifier wp,
+			bool wait_on_copygc,
+			move_pred_fn pred, void *arg)
 {
 	struct moving_context ctxt;
 
diff --git a/fs/bcachefs/move.h b/fs/bcachefs/move.h
index 51e0505a8156..1ab6dd4621d6 100644
--- a/fs/bcachefs/move.h
+++ b/fs/bcachefs/move.h
@@ -135,6 +135,11 @@ int bch2_move_data(struct bch_fs *,
 		   bool,
 		   move_pred_fn, void *);
 
+int bch2_move_data_phys(struct bch_fs *, unsigned, u64, u64, unsigned,
+			struct bch_ratelimit *, struct bch_move_stats *,
+			struct write_point_specifier, bool,
+			move_pred_fn, void *);
+
 int bch2_evacuate_bucket(struct moving_context *,
 			   struct move_bucket_in_flight *,
 			   struct bpos, int,
-- 
2.51.0


From f3c8eaf7a133ef122dfd97e6f6f972265cc84fb0 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 3 Apr 2025 19:42:02 -0400
Subject: [PATCH 14/16] bcachefs: Plumb target parameter through
 btree_node_rewrite_pos()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_io.c              |  2 +-
 fs/bcachefs/btree_update_interior.c | 37 +++++++++++++++++++----------
 fs/bcachefs/btree_update_interior.h |  4 ++--
 fs/bcachefs/move.c                  |  5 ++--
 4 files changed, 30 insertions(+), 18 deletions(-)

diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index b6f5e0dfc9f1..c1c671e340c7 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -1918,7 +1918,7 @@ static void btree_node_scrub_work(struct work_struct *work)
 			bch_err(c, "error validating btree node during scrub on %s at btree %s",
 				scrub->ca->name, err.buf);
 
-			ret = bch2_btree_node_rewrite(trans, &iter, b, 0);
+			ret = bch2_btree_node_rewrite(trans, &iter, b, 0, 0);
 		}
 err:
 		bch2_trans_iter_exit(trans, &iter);
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index 2be7c10fc59c..3155b4360fbc 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -284,6 +284,7 @@ static struct btree *__bch2_btree_node_alloc(struct btree_trans *trans,
 					     struct disk_reservation *res,
 					     struct closure *cl,
 					     bool interior_node,
+					     unsigned target,
 					     unsigned flags)
 {
 	struct bch_fs *c = trans->c;
@@ -317,6 +318,7 @@ static struct btree *__bch2_btree_node_alloc(struct btree_trans *trans,
 	mutex_unlock(&c->btree_reserve_cache_lock);
 retry:
 	ret = bch2_alloc_sectors_start_trans(trans,
+				      target ?:
 				      c->opts.metadata_target ?:
 				      c->opts.foreground_target,
 				      0,
@@ -325,7 +327,9 @@ retry:
 				      res->nr_replicas,
 				      min(res->nr_replicas,
 					  c->opts.metadata_replicas_required),
-				      watermark, 0, cl, &wp);
+				      watermark,
+				      target ? BCH_WRITE_only_specified_devs : 0,
+				      cl, &wp);
 	if (unlikely(ret))
 		goto err;
 
@@ -505,6 +509,7 @@ static void bch2_btree_reserve_put(struct btree_update *as, struct btree_trans *
 static int bch2_btree_reserve_get(struct btree_trans *trans,
 				  struct btree_update *as,
 				  unsigned nr_nodes[2],
+				  unsigned target,
 				  unsigned flags,
 				  struct closure *cl)
 {
@@ -527,7 +532,7 @@ static int bch2_btree_reserve_get(struct btree_trans *trans,
 
 		while (p->nr < nr_nodes[interior]) {
 			b = __bch2_btree_node_alloc(trans, &as->disk_res, cl,
-						    interior, flags);
+						    interior, target, flags);
 			if (IS_ERR(b)) {
 				ret = PTR_ERR(b);
 				goto err;
@@ -1116,7 +1121,8 @@ static void bch2_btree_update_done(struct btree_update *as, struct btree_trans *
 
 static struct btree_update *
 bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
-			unsigned level_start, bool split, unsigned flags)
+			unsigned level_start, bool split,
+			unsigned target, unsigned flags)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_update *as;
@@ -1226,7 +1232,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
 	if (ret)
 		goto err;
 
-	ret = bch2_btree_reserve_get(trans, as, nr_nodes, flags, NULL);
+	ret = bch2_btree_reserve_get(trans, as, nr_nodes, target, flags, NULL);
 	if (bch2_err_matches(ret, ENOSPC) ||
 	    bch2_err_matches(ret, ENOMEM)) {
 		struct closure cl;
@@ -1245,7 +1251,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
 		closure_init_stack(&cl);
 
 		do {
-			ret = bch2_btree_reserve_get(trans, as, nr_nodes, flags, &cl);
+			ret = bch2_btree_reserve_get(trans, as, nr_nodes, target, flags, &cl);
 
 			bch2_trans_unlock(trans);
 			bch2_wait_on_allocator(c, &cl);
@@ -1878,7 +1884,7 @@ int bch2_btree_split_leaf(struct btree_trans *trans,
 
 	as = bch2_btree_update_start(trans, trans->paths + path,
 				     trans->paths[path].level,
-				     true, flags);
+				     true, 0, flags);
 	if (IS_ERR(as))
 		return PTR_ERR(as);
 
@@ -1948,7 +1954,8 @@ int bch2_btree_increase_depth(struct btree_trans *trans, btree_path_idx_t path,
 		return bch2_btree_split_leaf(trans, path, flags);
 
 	struct btree_update *as =
-		bch2_btree_update_start(trans, trans->paths + path, b->c.level, true, flags);
+		bch2_btree_update_start(trans, trans->paths + path, b->c.level,
+					true, 0, flags);
 	if (IS_ERR(as))
 		return PTR_ERR(as);
 
@@ -2077,7 +2084,7 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
 
 	parent = btree_node_parent(trans->paths + path, b);
 	as = bch2_btree_update_start(trans, trans->paths + path, level, false,
-				     BCH_TRANS_COMMIT_no_enospc|flags);
+				     0, BCH_TRANS_COMMIT_no_enospc|flags);
 	ret = PTR_ERR_OR_ZERO(as);
 	if (ret)
 		goto err;
@@ -2184,6 +2191,7 @@ err:
 int bch2_btree_node_rewrite(struct btree_trans *trans,
 			    struct btree_iter *iter,
 			    struct btree *b,
+			    unsigned target,
 			    unsigned flags)
 {
 	struct bch_fs *c = trans->c;
@@ -2196,7 +2204,8 @@ int bch2_btree_node_rewrite(struct btree_trans *trans,
 
 	struct btree_path *path = btree_iter_path(trans, iter);
 	parent = btree_node_parent(path, b);
-	as = bch2_btree_update_start(trans, path, b->c.level, false, flags);
+	as = bch2_btree_update_start(trans, path, b->c.level,
+				     false, target, flags);
 	ret = PTR_ERR_OR_ZERO(as);
 	if (ret)
 		goto out;
@@ -2261,7 +2270,7 @@ static int bch2_btree_node_rewrite_key(struct btree_trans *trans,
 
 	bool found = b && btree_ptr_hash_val(&b->key) == btree_ptr_hash_val(k);
 	ret = found
-		? bch2_btree_node_rewrite(trans, &iter, b, flags)
+		? bch2_btree_node_rewrite(trans, &iter, b, 0, flags)
 		: -ENOENT;
 out:
 	bch2_trans_iter_exit(trans, &iter);
@@ -2270,7 +2279,9 @@ out:
 
 int bch2_btree_node_rewrite_pos(struct btree_trans *trans,
 				enum btree_id btree, unsigned level,
-				struct bpos pos, unsigned flags)
+				struct bpos pos,
+				unsigned target,
+				unsigned flags)
 {
 	BUG_ON(!level);
 
@@ -2282,7 +2293,7 @@ int bch2_btree_node_rewrite_pos(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	ret = bch2_btree_node_rewrite(trans, &iter, b, flags);
+	ret = bch2_btree_node_rewrite(trans, &iter, b, target, flags);
 err:
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
@@ -2296,7 +2307,7 @@ int bch2_btree_node_rewrite_key_get_iter(struct btree_trans *trans,
 	if (ret)
 		return ret == -BCH_ERR_btree_node_dying ? 0 : ret;
 
-	ret = bch2_btree_node_rewrite(trans, &iter, b, flags);
+	ret = bch2_btree_node_rewrite(trans, &iter, b, 0, flags);
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
diff --git a/fs/bcachefs/btree_update_interior.h b/fs/bcachefs/btree_update_interior.h
index be71cd73b864..ff9b95aac554 100644
--- a/fs/bcachefs/btree_update_interior.h
+++ b/fs/bcachefs/btree_update_interior.h
@@ -168,10 +168,10 @@ static inline int bch2_foreground_maybe_merge(struct btree_trans *trans,
 }
 
 int bch2_btree_node_rewrite(struct btree_trans *, struct btree_iter *,
-			    struct btree *, unsigned);
+			    struct btree *, unsigned, unsigned);
 int bch2_btree_node_rewrite_pos(struct btree_trans *,
 				enum btree_id, unsigned,
-				struct bpos, unsigned);
+				struct bpos, unsigned, unsigned);
 int bch2_btree_node_rewrite_key_get_iter(struct btree_trans *,
 					 struct btree *, unsigned);
 
diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index 29981ebcb972..d40e2d14ec52 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -872,7 +872,8 @@ static int __bch2_move_data_phys(struct moving_context *ctxt,
 		if (!bp.v->level)
 			ret = bch2_move_extent(ctxt, bucket_in_flight, &iter, k, io_opts, data_opts);
 		else if (!data_opts.scrub)
-			ret = bch2_btree_node_rewrite_pos(trans, bp.v->btree_id, bp.v->level, k.k->p, 0);
+			ret = bch2_btree_node_rewrite_pos(trans, bp.v->btree_id, bp.v->level,
+							  k.k->p, data_opts.target, 0);
 		else
 			ret = bch2_btree_node_scrub(trans, bp.v->btree_id, bp.v->level, k, data_opts.read_dev);
 
@@ -1022,7 +1023,7 @@ retry:
 			if (!pred(c, arg, b, &io_opts, &data_opts))
 				goto next;
 
-			ret = bch2_btree_node_rewrite(trans, &iter, b, 0) ?: ret;
+			ret = bch2_btree_node_rewrite(trans, &iter, b, 0, 0) ?: ret;
 			if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 				continue;
 			if (ret)
-- 
2.51.0


From 7a274285d3706608d788efcbd9982f08531dd9ec Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Thu, 3 Apr 2025 19:51:05 -0400
Subject: [PATCH 15/16] bcachefs: plumb btree_id through move_pred_fd

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/move.c      | 20 +++++++++++---------
 fs/bcachefs/move.h      |  2 +-
 fs/bcachefs/rebalance.c |  2 +-
 3 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index d40e2d14ec52..07cea68b04f0 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -667,7 +667,7 @@ static int bch2_move_data_btree(struct moving_context *ctxt,
 			continue;
 
 		memset(&data_opts, 0, sizeof(data_opts));
-		if (!pred(c, arg, k, io_opts, &data_opts))
+		if (!pred(c, arg, extent_iter->btree_id, k, io_opts, &data_opts))
 			goto next;
 
 		/*
@@ -851,7 +851,7 @@ static int __bch2_move_data_phys(struct moving_context *ctxt,
 		}
 
 		struct data_update_opts data_opts = {};
-		if (!pred(c, arg, k, &io_opts, &data_opts)) {
+		if (!pred(c, arg, bp.v->btree_id, k, &io_opts, &data_opts)) {
 			bch2_trans_iter_exit(trans, &iter);
 			goto next;
 		}
@@ -934,7 +934,8 @@ struct evacuate_bucket_arg {
 	struct data_update_opts	data_opts;
 };
 
-static bool evacuate_bucket_pred(struct bch_fs *c, void *_arg, struct bkey_s_c k,
+static bool evacuate_bucket_pred(struct bch_fs *c, void *_arg,
+				 enum btree_id btree, struct bkey_s_c k,
 				 struct bch_io_opts *io_opts,
 				 struct data_update_opts *data_opts)
 {
@@ -1048,7 +1049,7 @@ next:
 }
 
 static bool rereplicate_pred(struct bch_fs *c, void *arg,
-			     struct bkey_s_c k,
+			     enum btree_id btree, struct bkey_s_c k,
 			     struct bch_io_opts *io_opts,
 			     struct data_update_opts *data_opts)
 {
@@ -1080,7 +1081,7 @@ static bool rereplicate_pred(struct bch_fs *c, void *arg,
 }
 
 static bool migrate_pred(struct bch_fs *c, void *arg,
-			 struct bkey_s_c k,
+			 enum btree_id btree, struct bkey_s_c k,
 			 struct bch_io_opts *io_opts,
 			 struct data_update_opts *data_opts)
 {
@@ -1107,7 +1108,7 @@ static bool rereplicate_btree_pred(struct bch_fs *c, void *arg,
 				   struct bch_io_opts *io_opts,
 				   struct data_update_opts *data_opts)
 {
-	return rereplicate_pred(c, arg, bkey_i_to_s_c(&b->key), io_opts, data_opts);
+	return rereplicate_pred(c, arg, b->c.btree_id, bkey_i_to_s_c(&b->key), io_opts, data_opts);
 }
 
 /*
@@ -1163,7 +1164,7 @@ int bch2_scan_old_btree_nodes(struct bch_fs *c, struct bch_move_stats *stats)
 }
 
 static bool drop_extra_replicas_pred(struct bch_fs *c, void *arg,
-			     struct bkey_s_c k,
+			     enum btree_id btree, struct bkey_s_c k,
 			     struct bch_io_opts *io_opts,
 			     struct data_update_opts *data_opts)
 {
@@ -1196,11 +1197,12 @@ static bool drop_extra_replicas_btree_pred(struct bch_fs *c, void *arg,
 				   struct bch_io_opts *io_opts,
 				   struct data_update_opts *data_opts)
 {
-	return drop_extra_replicas_pred(c, arg, bkey_i_to_s_c(&b->key), io_opts, data_opts);
+	return drop_extra_replicas_pred(c, arg, b->c.btree_id, bkey_i_to_s_c(&b->key),
+					io_opts, data_opts);
 }
 
 static bool scrub_pred(struct bch_fs *c, void *_arg,
-		       struct bkey_s_c k,
+		       enum btree_id btree, struct bkey_s_c k,
 		       struct bch_io_opts *io_opts,
 		       struct data_update_opts *data_opts)
 {
diff --git a/fs/bcachefs/move.h b/fs/bcachefs/move.h
index 1ab6dd4621d6..9c6c229e583e 100644
--- a/fs/bcachefs/move.h
+++ b/fs/bcachefs/move.h
@@ -72,7 +72,7 @@ do {									\
 		break;							\
 } while (1)
 
-typedef bool (*move_pred_fn)(struct bch_fs *, void *, struct bkey_s_c,
+typedef bool (*move_pred_fn)(struct bch_fs *, void *, enum btree_id, struct bkey_s_c,
 			     struct bch_io_opts *, struct data_update_opts *);
 
 extern const char * const bch2_data_ops_strs[];
diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c
index 3c45500c1a28..d2a7001cf872 100644
--- a/fs/bcachefs/rebalance.c
+++ b/fs/bcachefs/rebalance.c
@@ -454,7 +454,7 @@ out:
 }
 
 static bool rebalance_pred(struct bch_fs *c, void *arg,
-			   struct bkey_s_c k,
+			   enum btree_id btree, struct bkey_s_c k,
 			   struct bch_io_opts *io_opts,
 			   struct data_update_opts *data_opts)
 {
-- 
2.51.0


From 3484840ece849ee700c7cf8e0d44d5536b29fa08 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 13 Apr 2025 16:31:34 -0400
Subject: [PATCH 16/16] bcachefs: bch2_move_data_btree() can move btree nodes

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/move.c | 37 ++++++++++++++++++++++++-------------
 fs/bcachefs/move.h |  2 ++
 2 files changed, 26 insertions(+), 13 deletions(-)

diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index 07cea68b04f0..a8ad8d4538e0 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -423,6 +423,9 @@ static struct bch_io_opts *bch2_move_get_io_opts(struct btree_trans *trans,
 	struct bch_io_opts *opts_ret = &io_opts->fs_io_opts;
 	int ret = 0;
 
+	if (extent_iter->min_depth)
+		return opts_ret;
+
 	if (extent_k.k->type == KEY_TYPE_reflink_v)
 		goto out;
 
@@ -573,11 +576,11 @@ static struct bkey_s_c bch2_lookup_indirect_extent_for_move(struct btree_trans *
 	return k;
 }
 
-static int bch2_move_data_btree(struct moving_context *ctxt,
-				struct bpos start,
-				struct bpos end,
-				move_pred_fn pred, void *arg,
-				enum btree_id btree_id)
+int bch2_move_data_btree(struct moving_context *ctxt,
+			 struct bpos start,
+			 struct bpos end,
+			 move_pred_fn pred, void *arg,
+			 enum btree_id btree_id, unsigned level)
 {
 	struct btree_trans *trans = ctxt->trans;
 	struct bch_fs *c = trans->c;
@@ -604,10 +607,10 @@ static int bch2_move_data_btree(struct moving_context *ctxt,
 	}
 
 	bch2_trans_begin(trans);
-	bch2_trans_iter_init(trans, &iter, btree_id, start,
-			     BTREE_ITER_prefetch|
-			     BTREE_ITER_not_extents|
-			     BTREE_ITER_all_snapshots);
+	bch2_trans_node_iter_init(trans, &iter, btree_id, start, 0, level,
+				  BTREE_ITER_prefetch|
+				  BTREE_ITER_not_extents|
+				  BTREE_ITER_all_snapshots);
 
 	if (ctxt->rate)
 		bch2_ratelimit_reset(ctxt->rate);
@@ -627,7 +630,7 @@ static int bch2_move_data_btree(struct moving_context *ctxt,
 		if (ret)
 			break;
 
-		if (bkey_ge(bkey_start_pos(k.k), end))
+		if (bkey_gt(bkey_start_pos(k.k), end))
 			break;
 
 		if (ctxt->stats)
@@ -677,7 +680,14 @@ static int bch2_move_data_btree(struct moving_context *ctxt,
 		bch2_bkey_buf_reassemble(&sk, c, k);
 		k = bkey_i_to_s_c(sk.k);
 
-		ret2 = bch2_move_extent(ctxt, NULL, extent_iter, k, *io_opts, data_opts);
+		if (!level)
+			ret2 = bch2_move_extent(ctxt, NULL, extent_iter, k, *io_opts, data_opts);
+		else if (!data_opts.scrub)
+			ret2 = bch2_btree_node_rewrite_pos(trans, btree_id, level,
+							  k.k->p, data_opts.target, 0);
+		else
+			ret2 = bch2_btree_node_scrub(trans, btree_id, level, k, data_opts.read_dev);
+
 		if (ret2) {
 			if (bch2_err_matches(ret2, BCH_ERR_transaction_restart))
 				continue;
@@ -695,7 +705,8 @@ next:
 		if (ctxt->stats)
 			atomic64_add(k.k->size, &ctxt->stats->sectors_seen);
 next_nondata:
-		bch2_btree_iter_advance(trans, &iter);
+		if (!bch2_btree_iter_advance(trans, &iter))
+			break;
 	}
 
 	bch2_trans_iter_exit(trans, &reflink_iter);
@@ -727,7 +738,7 @@ int __bch2_move_data(struct moving_context *ctxt,
 		ret = bch2_move_data_btree(ctxt,
 				       id == start.btree ? start.pos : POS_MIN,
 				       id == end.btree   ? end.pos   : POS_MAX,
-				       pred, arg, id);
+				       pred, arg, id, 0);
 		if (ret)
 			break;
 	}
diff --git a/fs/bcachefs/move.h b/fs/bcachefs/move.h
index 9c6c229e583e..0c620a5f728d 100644
--- a/fs/bcachefs/move.h
+++ b/fs/bcachefs/move.h
@@ -122,6 +122,8 @@ int bch2_move_extent(struct moving_context *,
 		     struct bch_io_opts,
 		     struct data_update_opts);
 
+int bch2_move_data_btree(struct moving_context *, struct bpos, struct bpos,
+			 move_pred_fn, void *, enum btree_id, unsigned);
 int __bch2_move_data(struct moving_context *,
 		     struct bbpos,
 		     struct bbpos,
-- 
2.51.0