From d65d126c0256cf2349e118a3e8627d8281295eee Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 21 Sep 2024 23:27:59 -0400
Subject: [PATCH 01/16] bcachefs: Add locking for bch_fs.curr_recovery_pass

Recovery can rewind in certain situations - when we discover we need to
run a pass that doesn't normally run.

This can happen from another thread for btree node read errors, so we
need a bit of locking.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs.h        |  1 +
 fs/bcachefs/recovery_passes.c | 76 ++++++++++++++++++++++++++---------
 fs/bcachefs/super.c           |  1 +
 3 files changed, 59 insertions(+), 19 deletions(-)

diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index 7db81e182c3c..fbd89f91625d 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -1060,6 +1060,7 @@ struct bch_fs {
 	u64			recovery_passes_complete;
 	/* never rewinds version of curr_recovery_pass */
 	enum bch_recovery_pass	recovery_pass_done;
+	spinlock_t		recovery_pass_lock;
 	struct semaphore	online_fsck_mutex;
 
 	/* DEBUG JUNK */
diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index 1cc010bf1695..5e7722cc0879 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -100,8 +100,8 @@ u64 bch2_recovery_passes_from_stable(u64 v)
 /*
  * For when we need to rewind recovery passes and run a pass we skipped:
  */
-int bch2_run_explicit_recovery_pass(struct bch_fs *c,
-				    enum bch_recovery_pass pass)
+static int __bch2_run_explicit_recovery_pass(struct bch_fs *c,
+					     enum bch_recovery_pass pass)
 {
 	if (c->opts.recovery_passes & BIT_ULL(pass))
 		return 0;
@@ -109,6 +109,13 @@ int bch2_run_explicit_recovery_pass(struct bch_fs *c,
 	if (c->curr_recovery_pass == ARRAY_SIZE(recovery_pass_fns))
 		return -BCH_ERR_not_in_recovery;
 
+	if (pass < BCH_RECOVERY_PASS_set_may_go_rw &&
+	    c->curr_recovery_pass >= BCH_RECOVERY_PASS_set_may_go_rw) {
+		bch_info(c, "need recovery pass %s (%u), but already rw",
+			 bch2_recovery_passes[pass], pass);
+		return -BCH_ERR_cannot_rewind_recovery;
+	}
+
 	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);
@@ -124,6 +131,16 @@ int bch2_run_explicit_recovery_pass(struct bch_fs *c,
 	}
 }
 
+int bch2_run_explicit_recovery_pass(struct bch_fs *c,
+				    enum bch_recovery_pass pass)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&c->recovery_pass_lock, flags);
+	int ret = __bch2_run_explicit_recovery_pass(c, pass);
+	spin_unlock_irqrestore(&c->recovery_pass_lock, flags);
+	return ret;
+}
+
 int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *c,
 					       enum bch_recovery_pass pass)
 {
@@ -237,30 +254,51 @@ int bch2_run_recovery_passes(struct bch_fs *c)
 	c->opts.recovery_passes_exclude &= ~BCH_RECOVERY_PASS_set_may_go_rw;
 
 	while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns)) {
+		spin_lock_irq(&c->recovery_pass_lock);
+		unsigned pass = c->curr_recovery_pass;
+
 		if (c->opts.recovery_pass_last &&
-		    c->curr_recovery_pass > c->opts.recovery_pass_last)
+		    c->curr_recovery_pass > c->opts.recovery_pass_last) {
+			spin_unlock_irq(&c->recovery_pass_lock);
 			break;
+		}
 
-		if (should_run_recovery_pass(c, c->curr_recovery_pass)) {
-			unsigned pass = c->curr_recovery_pass;
-
-			ret =   bch2_run_recovery_pass(c, c->curr_recovery_pass) ?:
-				bch2_journal_flush(&c->journal);
-			if (bch2_err_matches(ret, BCH_ERR_restart_recovery) ||
-			    (ret && c->curr_recovery_pass < pass))
-				continue;
-			if (ret)
-				break;
-
-			c->recovery_passes_complete |= BIT_ULL(c->curr_recovery_pass);
+		if (!should_run_recovery_pass(c, pass)) {
+			c->curr_recovery_pass++;
+			c->recovery_pass_done = max(c->recovery_pass_done, pass);
+			spin_unlock_irq(&c->recovery_pass_lock);
+			continue;
+		}
+		spin_unlock_irq(&c->recovery_pass_lock);
+
+		ret =   bch2_run_recovery_pass(c, pass) ?:
+			bch2_journal_flush(&c->journal);
+
+		spin_lock_irq(&c->recovery_pass_lock);
+		if (c->curr_recovery_pass < pass) {
+			/*
+			 * bch2_run_explicit_recovery_pass() was called: we
+			 * can't always catch -BCH_ERR_restart_recovery because
+			 * it may have been called from another thread (btree
+			 * node read completion)
+			 */
+			spin_unlock_irq(&c->recovery_pass_lock);
+			continue;
+		} else if (c->curr_recovery_pass == pass) {
+			c->curr_recovery_pass++;
+		} else {
+			BUG();
 		}
+		spin_unlock_irq(&c->recovery_pass_lock);
 
-		c->recovery_pass_done = max(c->recovery_pass_done, c->curr_recovery_pass);
+		if (ret)
+			break;
 
-		if (!test_bit(BCH_FS_error, &c->flags))
-			bch2_clear_recovery_pass_required(c, c->curr_recovery_pass);
+		c->recovery_passes_complete |= BIT_ULL(pass);
+		c->recovery_pass_done = max(c->recovery_pass_done, pass);
 
-		c->curr_recovery_pass++;
+		if (!test_bit(BCH_FS_error, &c->flags))
+			bch2_clear_recovery_pass_required(c, pass);
 	}
 
 	return ret;
diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c
index 17442df7326d..d6411324cd3f 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -766,6 +766,7 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 
 	refcount_set(&c->ro_ref, 1);
 	init_waitqueue_head(&c->ro_ref_wait);
+	spin_lock_init(&c->recovery_pass_lock);
 	sema_init(&c->online_fsck_mutex, 1);
 
 	init_rwsem(&c->gc_lock);
-- 
2.51.0


From e3c43dbe8e5ff64e77b6f927b32f489bccc7d75e Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 21 Sep 2024 23:40:01 -0400
Subject: [PATCH 02/16] bcachefs: bch2_btree_lost_data() now uses
 run_explicit_rceovery_pass_persistent()

Also get a bit more fine grained about which passes to run for which
btrees.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/recovery.c        | 63 +++++++++++++++++++++++------------
 fs/bcachefs/recovery.h        |  2 +-
 fs/bcachefs/recovery_passes.c | 11 ++++++
 fs/bcachefs/recovery_passes.h |  1 +
 4 files changed, 54 insertions(+), 23 deletions(-)

diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 3c7f941dde39..b1c83e72c0d8 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -34,21 +34,52 @@
 
 #define QSTR(n) { { { .len = strlen(n) } }, .name = n }
 
-void bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
+int bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
 {
-	if (btree >= BTREE_ID_NR_MAX)
-		return;
-
 	u64 b = BIT_ULL(btree);
+	int ret = 0;
+
+	mutex_lock(&c->sb_lock);
 
 	if (!(c->sb.btrees_lost_data & b)) {
 		bch_err(c, "flagging btree %s lost data", bch2_btree_id_str(btree));
-
-		mutex_lock(&c->sb_lock);
 		bch2_sb_field_get(c->disk_sb.sb, ext)->btrees_lost_data |= cpu_to_le64(b);
-		bch2_write_super(c);
-		mutex_unlock(&c->sb_lock);
 	}
+
+	switch (btree) {
+	case BTREE_ID_alloc:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_allocations) ?: ret;
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_alloc_info) ?: ret;
+		goto out;
+	case BTREE_ID_backpointers:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_btree_backpointers) ?: ret;
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_extents_to_backpointers) ?: ret;
+		goto out;
+	case BTREE_ID_need_discard:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_alloc_info) ?: ret;
+		goto out;
+	case BTREE_ID_freespace:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_alloc_info) ?: ret;
+		goto out;
+	case BTREE_ID_bucket_gens:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_alloc_info) ?: ret;
+		goto out;
+	case BTREE_ID_lru:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_alloc_info) ?: ret;
+		goto out;
+	case BTREE_ID_accounting:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_allocations) ?: ret;
+		goto out;
+	default:
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_scan_for_btree_nodes) ?: ret;
+		ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_topology) ?: ret;
+		goto out;
+	}
+out:
+	bch2_write_super(c);
+	mutex_unlock(&c->sb_lock);
+
+	return ret;
 }
 
 /* for -o reconstruct_alloc: */
@@ -524,22 +555,10 @@ static int read_btree_roots(struct bch_fs *c)
 					c, btree_root_read_error,
 					"error reading btree root %s l=%u: %s",
 					bch2_btree_id_str(i), r->level, bch2_err_str(ret))) {
-			if (btree_id_is_alloc(i)) {
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_allocations);
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_alloc_info);
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_lrus);
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_extents_to_backpointers);
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_alloc_to_lru_refs);
-				c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info);
+			if (btree_id_is_alloc(i))
 				r->error = 0;
-			} else if (!(c->opts.recovery_passes & BIT_ULL(BCH_RECOVERY_PASS_scan_for_btree_nodes))) {
-				bch_info(c, "will run btree node scan");
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_scan_for_btree_nodes);
-				c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_topology);
-			}
 
-			ret = 0;
-			bch2_btree_lost_data(c, i);
+			ret = bch2_btree_lost_data(c, i);
 		}
 	}
 
diff --git a/fs/bcachefs/recovery.h b/fs/bcachefs/recovery.h
index 4bf818de1f2f..b0d55754b21b 100644
--- a/fs/bcachefs/recovery.h
+++ b/fs/bcachefs/recovery.h
@@ -2,7 +2,7 @@
 #ifndef _BCACHEFS_RECOVERY_H
 #define _BCACHEFS_RECOVERY_H
 
-void bch2_btree_lost_data(struct bch_fs *, enum btree_id);
+int bch2_btree_lost_data(struct bch_fs *, enum btree_id);
 
 int bch2_journal_replay(struct bch_fs *);
 
diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c
index 5e7722cc0879..1240c5c19fea 100644
--- a/fs/bcachefs/recovery_passes.c
+++ b/fs/bcachefs/recovery_passes.c
@@ -141,6 +141,17 @@ int bch2_run_explicit_recovery_pass(struct bch_fs *c,
 	return ret;
 }
 
+int bch2_run_explicit_recovery_pass_persistent_locked(struct bch_fs *c,
+					       enum bch_recovery_pass pass)
+{
+	lockdep_assert_held(&c->sb_lock);
+
+	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
+	__set_bit_le64(bch2_recovery_pass_to_stable(pass), ext->recovery_passes_required);
+
+	return bch2_run_explicit_recovery_pass(c, pass);
+}
+
 int bch2_run_explicit_recovery_pass_persistent(struct bch_fs *c,
 					       enum bch_recovery_pass pass)
 {
diff --git a/fs/bcachefs/recovery_passes.h b/fs/bcachefs/recovery_passes.h
index 99b464e127b8..7d7339c8fa29 100644
--- a/fs/bcachefs/recovery_passes.h
+++ b/fs/bcachefs/recovery_passes.h
@@ -9,6 +9,7 @@ u64 bch2_recovery_passes_from_stable(u64 v);
 u64 bch2_fsck_recovery_passes(void);
 
 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);
 
 int bch2_run_online_recovery_passes(struct bch_fs *);
-- 
2.51.0


From 0269e27ce3f7be2bd1e565cc17a88e4074facad1 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 9 Oct 2024 21:26:05 -0400
Subject: [PATCH 03/16] bcachefs: improved bkey_val_copy()

Factor out some common code, add typechecking.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_iter.h | 28 +++++++++++++---------------
 1 file changed, 13 insertions(+), 15 deletions(-)

diff --git a/fs/bcachefs/btree_iter.h b/fs/bcachefs/btree_iter.h
index 550db3654f2c..dda07a320488 100644
--- a/fs/bcachefs/btree_iter.h
+++ b/fs/bcachefs/btree_iter.h
@@ -594,13 +594,18 @@ static inline struct bkey_s_c bch2_bkey_get_iter(struct btree_trans *trans,
 	bkey_s_c_to_##_type(__bch2_bkey_get_iter(_trans, _iter,			\
 				       _btree_id, _pos, _flags, KEY_TYPE_##_type))
 
+static inline void __bkey_val_copy(void *dst_v, unsigned dst_size, struct bkey_s_c src_k)
+{
+	unsigned b = min_t(unsigned, dst_size, bkey_val_bytes(src_k.k));
+	memcpy(dst_v, src_k.v, b);
+	if (unlikely(b < dst_size))
+		memset(dst_v + b, 0, dst_size - b);
+}
+
 #define bkey_val_copy(_dst_v, _src_k)					\
 do {									\
-	unsigned b = min_t(unsigned, sizeof(*_dst_v),			\
-			   bkey_val_bytes(_src_k.k));			\
-	memcpy(_dst_v, _src_k.v, b);					\
-	if (b < sizeof(*_dst_v))					\
-		memset((void *) (_dst_v) + b, 0, sizeof(*_dst_v) - b);	\
+	BUILD_BUG_ON(!__typecheck(*_dst_v, *_src_k.v));			\
+	__bkey_val_copy(_dst_v, sizeof(*_dst_v), _src_k.s_c);		\
 } while (0)
 
 static inline int __bch2_bkey_get_val_typed(struct btree_trans *trans,
@@ -609,17 +614,10 @@ static inline int __bch2_bkey_get_val_typed(struct btree_trans *trans,
 				unsigned val_size, void *val)
 {
 	struct btree_iter iter;
-	struct bkey_s_c k;
-	int ret;
-
-	k = __bch2_bkey_get_iter(trans, &iter, btree_id, pos, flags, type);
-	ret = bkey_err(k);
+	struct bkey_s_c k = __bch2_bkey_get_iter(trans, &iter, btree_id, pos, flags, type);
+	int ret = bkey_err(k);
 	if (!ret) {
-		unsigned b = min_t(unsigned, bkey_val_bytes(k.k), val_size);
-
-		memcpy(val, k.v, b);
-		if (unlikely(b < sizeof(*val)))
-			memset((void *) val + b, 0, sizeof(*val) - b);
+		__bkey_val_copy(val, val_size, k);
 		bch2_trans_iter_exit(trans, &iter);
 	}
 
-- 
2.51.0


From 106480e9a869e8d2dd2db34819d04e15ccfd896c Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 9 Oct 2024 21:51:05 -0400
Subject: [PATCH 04/16] bcachefs: Factor out jset_entry_log_msg_bytes()

Needed for improved userspace cmd_list_journal

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs_format.h | 9 +++++++++
 fs/bcachefs/journal_io.c      | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h
index 79a80a78c2d8..c5e3824d5771 100644
--- a/fs/bcachefs/bcachefs_format.h
+++ b/fs/bcachefs/bcachefs_format.h
@@ -1219,6 +1219,15 @@ struct jset_entry_log {
 	u8			d[];
 } __packed __aligned(8);
 
+static inline unsigned jset_entry_log_msg_bytes(struct jset_entry_log *l)
+{
+	unsigned b = vstruct_bytes(&l->entry) - offsetof(struct jset_entry_log, d);
+
+	while (b && !l->d[b - 1])
+		--b;
+	return b;
+}
+
 struct jset_entry_datetime {
 	struct jset_entry	entry;
 	__le64			seconds;
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index fb35dd336331..7c7595e5369b 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -738,9 +738,8 @@ static void journal_entry_log_to_text(struct printbuf *out, struct bch_fs *c,
 				      struct jset_entry *entry)
 {
 	struct jset_entry_log *l = container_of(entry, struct jset_entry_log, entry);
-	unsigned bytes = vstruct_bytes(entry) - offsetof(struct jset_entry_log, d);
 
-	prt_printf(out, "%.*s", bytes, l->d);
+	prt_printf(out, "%.*s", jset_entry_log_msg_bytes(l), l->d);
 }
 
 static int journal_entry_overwrite_validate(struct bch_fs *c,
-- 
2.51.0


From 9e2f5f79882b855156cd522acb7354e5a7901418 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 9 Oct 2024 21:27:11 -0400
Subject: [PATCH 05/16] bcachefs: better error message in check_snapshot_tree()

If we find a snapshot node and it didn't match the snapshot tree, we
should print it.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/snapshot.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c
index feaf2aa0d900..34e01bd8127f 100644
--- a/fs/bcachefs/snapshot.c
+++ b/fs/bcachefs/snapshot.c
@@ -506,7 +506,6 @@ static int bch2_snapshot_tree_master_subvol(struct btree_trans *trans,
 			break;
 		}
 	}
-
 	bch2_trans_iter_exit(trans, &iter);
 
 	if (!ret && !found) {
@@ -536,6 +535,7 @@ static int check_snapshot_tree(struct btree_trans *trans,
 	struct bch_snapshot s;
 	struct bch_subvolume subvol;
 	struct printbuf buf = PRINTBUF;
+	struct btree_iter snapshot_iter = {};
 	u32 root_id;
 	int ret;
 
@@ -545,16 +545,27 @@ static int check_snapshot_tree(struct btree_trans *trans,
 	st = bkey_s_c_to_snapshot_tree(k);
 	root_id = le32_to_cpu(st.v->root_snapshot);
 
-	ret = bch2_snapshot_lookup(trans, root_id, &s);
+	struct bkey_s_c_snapshot snapshot_k =
+		bch2_bkey_get_iter_typed(trans, &snapshot_iter, BTREE_ID_snapshots,
+					 POS(0, root_id), 0, snapshot);
+	ret = bkey_err(snapshot_k);
 	if (ret && !bch2_err_matches(ret, ENOENT))
 		goto err;
 
+	if (!ret)
+		bkey_val_copy(&s, snapshot_k);
+
 	if (fsck_err_on(ret ||
 			root_id != bch2_snapshot_root(c, root_id) ||
 			st.k->p.offset != le32_to_cpu(s.tree),
 			trans, snapshot_tree_to_missing_snapshot,
 			"snapshot tree points to missing/incorrect snapshot:\n  %s",
-			(bch2_bkey_val_to_text(&buf, c, st.s_c), buf.buf))) {
+			(bch2_bkey_val_to_text(&buf, c, st.s_c),
+			 prt_newline(&buf),
+			 ret
+			 ? prt_printf(&buf, "(%s)", bch2_err_str(ret))
+			 : bch2_bkey_val_to_text(&buf, c, snapshot_k.s_c),
+			 buf.buf))) {
 		ret = bch2_btree_delete_at(trans, iter, 0);
 		goto err;
 	}
@@ -605,6 +616,7 @@ static int check_snapshot_tree(struct btree_trans *trans,
 	}
 err:
 fsck_err:
+	bch2_trans_iter_exit(trans, &snapshot_iter);
 	printbuf_exit(&buf);
 	return ret;
 }
-- 
2.51.0


From db514cf6775fa58b45780969e407f678e0a5132c Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 9 Oct 2024 23:02:04 -0400
Subject: [PATCH 06/16] bcachefs: Avoid bch2_btree_id_str()

Prefer bch2_btree_id_to_text() - it prints out the integer ID when
unknown.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/backpointers.c          | 24 ++++++++-------
 fs/bcachefs/bbpos.h                 |  2 +-
 fs/bcachefs/btree_cache.c           | 37 +++++++++++++++---------
 fs/bcachefs/btree_cache.h           |  3 +-
 fs/bcachefs/btree_gc.c              | 45 +++++++++++++++++------------
 fs/bcachefs/btree_io.c              | 13 +++++----
 fs/bcachefs/btree_iter.c            | 32 ++++++++++----------
 fs/bcachefs/btree_journal_iter.c    |  5 +++-
 fs/bcachefs/btree_node_scan.c       | 10 ++++---
 fs/bcachefs/btree_update_interior.c | 23 ++++++++-------
 fs/bcachefs/debug.c                 |  4 ++-
 fs/bcachefs/disk_accounting.c       |  3 +-
 fs/bcachefs/journal_io.c            |  3 +-
 fs/bcachefs/recovery.c              | 25 +++++++++++-----
 fs/bcachefs/sysfs.c                 |  3 +-
 15 files changed, 140 insertions(+), 92 deletions(-)

diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c
index 654a58132a4d..f323ce4b0b33 100644
--- a/fs/bcachefs/backpointers.c
+++ b/fs/bcachefs/backpointers.c
@@ -81,12 +81,11 @@ fsck_err:
 
 void bch2_backpointer_to_text(struct printbuf *out, const struct bch_backpointer *bp)
 {
-	prt_printf(out, "btree=%s l=%u offset=%llu:%u len=%u pos=",
-	       bch2_btree_id_str(bp->btree_id),
-	       bp->level,
-	       (u64) (bp->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT),
-	       (u32) bp->bucket_offset & ~(~0U << MAX_EXTENT_COMPRESS_RATIO_SHIFT),
-	       bp->bucket_len);
+	bch2_btree_id_level_to_text(out, bp->btree_id, bp->level);
+	prt_printf(out, " offset=%llu:%u len=%u pos=",
+		   (u64) (bp->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT),
+		   (u32) bp->bucket_offset & ~(~0U << MAX_EXTENT_COMPRESS_RATIO_SHIFT),
+		   bp->bucket_len);
 	bch2_bpos_to_text(out, bp->pos);
 }
 
@@ -501,9 +500,13 @@ found:
 		goto err;
 
 	prt_str(&buf, "extents pointing to same space, but first extent checksum bad:");
-	prt_printf(&buf, "\n  %s ", bch2_btree_id_str(btree));
+	prt_printf(&buf, "\n  ");
+	bch2_btree_id_to_text(&buf, btree);
+	prt_str(&buf, " ");
 	bch2_bkey_val_to_text(&buf, c, extent);
-	prt_printf(&buf, "\n  %s ", bch2_btree_id_str(o_btree));
+	prt_printf(&buf, "\n  ");
+	bch2_btree_id_to_text(&buf, o_btree);
+	prt_str(&buf, " ");
 	bch2_bkey_val_to_text(&buf, c, extent2);
 
 	struct nonce nonce = extent_nonce(extent.k->bversion, p.crc);
@@ -638,8 +641,9 @@ check_existing_bp:
 	goto err;
 missing:
 	printbuf_reset(&buf);
-	prt_printf(&buf, "missing backpointer for btree=%s l=%u ",
-	       bch2_btree_id_str(bp.btree_id), bp.level);
+	prt_str(&buf, "missing backpointer for btree=");
+	bch2_btree_id_to_text(&buf, bp.btree_id);
+	prt_printf(&buf, " l=%u ", bp.level);
 	bch2_bkey_val_to_text(&buf, c, orig_k);
 	prt_printf(&buf, "\n  got:   ");
 	bch2_bkey_val_to_text(&buf, c, bp_k);
diff --git a/fs/bcachefs/bbpos.h b/fs/bcachefs/bbpos.h
index be2edced5213..63abe17f35ea 100644
--- a/fs/bcachefs/bbpos.h
+++ b/fs/bcachefs/bbpos.h
@@ -29,7 +29,7 @@ static inline struct bbpos bbpos_successor(struct bbpos pos)
 
 static inline void bch2_bbpos_to_text(struct printbuf *out, struct bbpos pos)
 {
-	prt_str(out, bch2_btree_id_str(pos.btree));
+	bch2_btree_id_to_text(out, pos.btree);
 	prt_char(out, ':');
 	bch2_bpos_to_text(out, pos.pos);
 }
diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index 7123019ab3bc..a0a406b0c7bc 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -1004,16 +1004,14 @@ static noinline void btree_bad_header(struct bch_fs *c, struct btree *b)
 		return;
 
 	prt_printf(&buf,
-	       "btree node header doesn't match ptr\n"
-	       "btree %s level %u\n"
-	       "ptr: ",
-	       bch2_btree_id_str(b->c.btree_id), b->c.level);
+		   "btree node header doesn't match ptr: ");
+	bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+	prt_str(&buf, "\nptr: ");
 	bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 
-	prt_printf(&buf, "\nheader: btree %s level %llu\n"
-	       "min ",
-	       bch2_btree_id_str(BTREE_NODE_ID(b->data)),
-	       BTREE_NODE_LEVEL(b->data));
+	prt_str(&buf, "\nheader: ");
+	bch2_btree_id_level_to_text(&buf, BTREE_NODE_ID(b->data), BTREE_NODE_LEVEL(b->data));
+	prt_str(&buf, "\nmin ");
 	bch2_bpos_to_text(&buf, b->data->min_key);
 
 	prt_printf(&buf, "\nmax ");
@@ -1398,12 +1396,19 @@ void bch2_btree_id_to_text(struct printbuf *out, enum btree_id btree)
 		prt_printf(out, "(unknown btree %u)", btree);
 }
 
+void bch2_btree_id_level_to_text(struct printbuf *out, enum btree_id btree, unsigned level)
+{
+	prt_str(out, "btree=");
+	bch2_btree_id_to_text(out, btree);
+	prt_printf(out, " level=%u", level);
+}
+
 void bch2_btree_pos_to_text(struct printbuf *out, struct bch_fs *c, const struct btree *b)
 {
-	prt_printf(out, "%s level %u/%u\n  ",
-	       bch2_btree_id_str(b->c.btree_id),
-	       b->c.level,
-	       bch2_btree_id_root(c, b->c.btree_id)->level);
+	bch2_btree_id_to_text(out, b->c.btree_id);
+	prt_printf(out, " level %u/%u\n  ",
+		   b->c.level,
+		   bch2_btree_id_root(c, b->c.btree_id)->level);
 	bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(&b->key));
 }
 
@@ -1478,8 +1483,12 @@ void bch2_btree_cache_to_text(struct printbuf *out, const struct btree_cache *bc
 	prt_printf(out, "cannibalize lock:\t%p\n",	bc->alloc_lock);
 	prt_newline(out);
 
-	for (unsigned i = 0; i < ARRAY_SIZE(bc->nr_by_btree); i++)
-		prt_btree_cache_line(out, c, bch2_btree_id_str(i), bc->nr_by_btree[i]);
+	for (unsigned i = 0; i < ARRAY_SIZE(bc->nr_by_btree); i++) {
+		bch2_btree_id_to_text(out, i);
+		prt_printf(out, "\t");
+		prt_human_readable_u64(out, bc->nr_by_btree[i] * c->opts.btree_node_size);
+		prt_printf(out, " (%zu)\n", bc->nr_by_btree[i]);
+	}
 
 	prt_newline(out);
 	prt_printf(out, "freed:\t%zu\n", bc->nr_freed);
diff --git a/fs/bcachefs/btree_cache.h b/fs/bcachefs/btree_cache.h
index 66e86d1a178d..6cfacacb6769 100644
--- a/fs/bcachefs/btree_cache.h
+++ b/fs/bcachefs/btree_cache.h
@@ -138,8 +138,9 @@ static inline struct btree *btree_node_root(struct bch_fs *c, struct btree *b)
 	return bch2_btree_id_root(c, b->c.btree_id)->b;
 }
 
-const char *bch2_btree_id_str(enum btree_id);
+const char *bch2_btree_id_str(enum btree_id);	/* avoid */
 void bch2_btree_id_to_text(struct printbuf *, enum btree_id);
+void bch2_btree_id_level_to_text(struct printbuf *, enum btree_id, unsigned);
 
 void bch2_btree_pos_to_text(struct printbuf *, struct bch_fs *, const struct btree *);
 void bch2_btree_node_to_text(struct printbuf *, struct bch_fs *, const struct btree *);
diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 81dcf9e512c0..3c4e66da1ca4 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -56,8 +56,8 @@ void bch2_gc_pos_to_text(struct printbuf *out, struct gc_pos *p)
 {
 	prt_str(out, bch2_gc_phase_strs[p->phase]);
 	prt_char(out, ' ');
-	bch2_btree_id_to_text(out, p->btree);
-	prt_printf(out, " l=%u ", p->level);
+	bch2_btree_id_level_to_text(out, p->btree, p->level);
+	prt_char(out, ' ');
 	bch2_bpos_to_text(out, p->pos);
 }
 
@@ -209,8 +209,9 @@ static int btree_check_node_boundaries(struct btree_trans *trans, struct btree *
 	if (bpos_eq(expected_start, cur->data->min_key))
 		return 0;
 
-	prt_printf(&buf, "  at btree %s level %u:\n  parent: ",
-		   bch2_btree_id_str(b->c.btree_id), b->c.level);
+	prt_printf(&buf, "  at ");
+	bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+	prt_printf(&buf, ":\n  parent: ");
 	bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 
 	if (prev) {
@@ -277,8 +278,9 @@ static int btree_repair_node_end(struct btree_trans *trans, struct btree *b,
 	if (bpos_eq(child->key.k.p, b->key.k.p))
 		return 0;
 
-	prt_printf(&buf, "at btree %s level %u:\n  parent: ",
-		   bch2_btree_id_str(b->c.btree_id), b->c.level);
+	prt_printf(&buf, "  at ");
+	bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+	prt_printf(&buf, ":\n  parent: ");
 	bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 
 	prt_str(&buf, "\n  child: ");
@@ -341,14 +343,14 @@ again:
 		ret = PTR_ERR_OR_ZERO(cur);
 
 		printbuf_reset(&buf);
+		bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level - 1);
+		prt_char(&buf, ' ');
 		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur_k.k));
 
 		if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO),
 				trans, btree_node_unreadable,
-				"Topology repair: unreadable btree node at btree %s level %u:\n"
+				"Topology repair: unreadable btree node at\n"
 				"  %s",
-				bch2_btree_id_str(b->c.btree_id),
-				b->c.level - 1,
 				buf.buf)) {
 			bch2_btree_node_evict(trans, cur_k.k);
 			cur = NULL;
@@ -370,7 +372,7 @@ again:
 			break;
 
 		if (bch2_btree_node_is_stale(c, cur)) {
-			bch_info(c, "btree node %s older than nodes found by scanning", buf.buf);
+			bch_info(c, "btree node older than nodes found by scanning\n  %s", buf.buf);
 			six_unlock_read(&cur->c.lock);
 			bch2_btree_node_evict(trans, cur_k.k);
 			ret = bch2_journal_key_delete(c, b->c.btree_id,
@@ -478,14 +480,13 @@ again:
 	}
 
 	printbuf_reset(&buf);
+	bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+	prt_newline(&buf);
 	bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 
 	if (mustfix_fsck_err_on(!have_child,
 			trans, btree_node_topology_interior_node_empty,
-			"empty interior btree node at btree %s level %u\n"
-			"  %s",
-			bch2_btree_id_str(b->c.btree_id),
-			b->c.level, buf.buf))
+			"empty interior btree node at %s", buf.buf))
 		ret = DROP_THIS_NODE;
 err:
 fsck_err:
@@ -511,6 +512,7 @@ int bch2_check_topology(struct bch_fs *c)
 {
 	struct btree_trans *trans = bch2_trans_get(c);
 	struct bpos pulled_from_scan = POS_MIN;
+	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
 	bch2_trans_srcu_unlock(trans);
@@ -519,19 +521,21 @@ int bch2_check_topology(struct bch_fs *c)
 		struct btree_root *r = bch2_btree_id_root(c, i);
 		bool reconstructed_root = false;
 
+		bch2_btree_id_to_text(&buf, i);
+
 		if (r->error) {
 			ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes);
 			if (ret)
 				break;
 reconstruct_root:
-			bch_info(c, "btree root %s unreadable, must recover from scan", bch2_btree_id_str(i));
+			bch_info(c, "btree root %s unreadable, must recover from scan", buf.buf);
 
 			r->alive = false;
 			r->error = 0;
 
 			if (!bch2_btree_has_scanned_nodes(c, i)) {
 				mustfix_fsck_err(trans, btree_root_unreadable_and_scan_found_nothing,
-						 "no nodes found for btree %s, continue?", bch2_btree_id_str(i));
+						 "no nodes found for btree %s, continue?", buf.buf);
 				bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			} else {
 				bch2_btree_root_alloc_fake_trans(trans, i, 1);
@@ -560,13 +564,14 @@ reconstruct_root:
 			if (!reconstructed_root)
 				goto reconstruct_root;
 
-			bch_err(c, "empty btree root %s", bch2_btree_id_str(i));
+			bch_err(c, "empty btree root %s", buf.buf);
 			bch2_btree_root_alloc_fake_trans(trans, i, 0);
 			r->alive = false;
 			ret = 0;
 		}
 	}
 fsck_err:
+	printbuf_exit(&buf);
 	bch2_trans_put(trans);
 	return ret;
 }
@@ -713,6 +718,7 @@ static int bch2_gc_btrees(struct bch_fs *c)
 {
 	struct btree_trans *trans = bch2_trans_get(c);
 	enum btree_id ids[BTREE_ID_NR];
+	struct printbuf buf = PRINTBUF;
 	unsigned i;
 	int ret = 0;
 
@@ -731,10 +737,13 @@ static int bch2_gc_btrees(struct bch_fs *c)
 		if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO),
 					trans, btree_node_read_error,
 			       "btree node read error for %s",
-			       bch2_btree_id_str(btree)))
+			       (printbuf_reset(&buf),
+				bch2_btree_id_to_text(&buf, btree),
+				buf.buf)))
 			ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology);
 	}
 fsck_err:
+	printbuf_exit(&buf);
 	bch2_trans_put(trans);
 	bch_err_fn(c, ret);
 	return ret;
diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c
index 839d68802e42..89a42ee81e5c 100644
--- a/fs/bcachefs/btree_io.c
+++ b/fs/bcachefs/btree_io.c
@@ -25,9 +25,8 @@
 
 static void bch2_btree_node_header_to_text(struct printbuf *out, struct btree_node *bn)
 {
-	prt_printf(out, "btree=%s l=%u seq %llux\n",
-		   bch2_btree_id_str(BTREE_NODE_ID(bn)),
-		   (unsigned) BTREE_NODE_LEVEL(bn), bn->keys.seq);
+	bch2_btree_id_level_to_text(out, BTREE_NODE_ID(bn), BTREE_NODE_LEVEL(bn));
+	prt_printf(out, " seq %llux\n", bn->keys.seq);
 	prt_str(out, "min: ");
 	bch2_bpos_to_text(out, bn->min_key);
 	prt_newline(out);
@@ -1343,9 +1342,11 @@ start:
 	    !btree_node_read_error(b) &&
 	    c->curr_recovery_pass != BCH_RECOVERY_PASS_scan_for_btree_nodes) {
 		printbuf_reset(&buf);
-		bch2_bpos_to_text(&buf, b->key.k.p);
-		bch_err_ratelimited(c, "%s: rewriting btree node at btree=%s level=%u %s due to error",
-			 __func__, bch2_btree_id_str(b->c.btree_id), b->c.level, buf.buf);
+		bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+		prt_str(&buf, " ");
+		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
+		bch_err_ratelimited(c, "%s: rewriting btree node at due to error\n  %s",
+				    __func__, buf.buf);
 
 		bch2_btree_node_rewrite_async(c, b);
 	}
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index 01152fd5ac57..07bce85dafaf 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -1448,10 +1448,11 @@ void bch2_trans_updates_to_text(struct printbuf *buf, struct btree_trans *trans)
 	trans_for_each_update(trans, i) {
 		struct bkey_s_c old = { &i->old_k, i->old_v };
 
-		prt_printf(buf, "update: btree=%s cached=%u %pS\n",
-		       bch2_btree_id_str(i->btree_id),
-		       i->cached,
-		       (void *) i->ip_allocated);
+		prt_str(buf, "update: btree=");
+		bch2_btree_id_to_text(buf, i->btree_id);
+		prt_printf(buf, " cached=%u %pS\n",
+			   i->cached,
+			   (void *) i->ip_allocated);
 
 		prt_printf(buf, "  old ");
 		bch2_bkey_val_to_text(buf, trans->c, old);
@@ -1484,13 +1485,13 @@ static void bch2_btree_path_to_text_short(struct printbuf *out, struct btree_tra
 {
 	struct btree_path *path = trans->paths + path_idx;
 
-	prt_printf(out, "path: idx %3u ref %u:%u %c %c %c btree=%s l=%u pos ",
+	prt_printf(out, "path: idx %3u ref %u:%u %c %c %c ",
 		   path_idx, path->ref, path->intent_ref,
 		   path->preserve ? 'P' : ' ',
 		   path->should_be_locked ? 'S' : ' ',
-		   path->cached ? 'C' : 'B',
-		   bch2_btree_id_str(path->btree_id),
-		   path->level);
+		   path->cached ? 'C' : 'B');
+	bch2_btree_id_level_to_text(out, path->btree_id, path->level);
+	prt_str(out, " pos ");
 	bch2_bpos_to_text(out, path->pos);
 
 	if (!path->cached && btree_node_locked(path, path->level)) {
@@ -3336,8 +3337,9 @@ bch2_btree_bkey_cached_common_to_text(struct printbuf *out,
 	pid = owner ? owner->pid : 0;
 	rcu_read_unlock();
 
-	prt_printf(out, "\t%px %c l=%u %s:", b, b->cached ? 'c' : 'b',
-		   b->level, bch2_btree_id_str(b->btree_id));
+	prt_printf(out, "\t%px %c ", b, b->cached ? 'c' : 'b');
+	bch2_btree_id_to_text(out, b->btree_id);
+	prt_printf(out, " l=%u:", b->level);
 	bch2_bpos_to_text(out, btree_node_pos(b));
 
 	prt_printf(out, "\t locks %u:%u:%u held by pid %u",
@@ -3376,11 +3378,11 @@ void bch2_btree_trans_to_text(struct printbuf *out, struct btree_trans *trans)
 		if (!path->nodes_locked)
 			continue;
 
-		prt_printf(out, "  path %u %c l=%u %s:",
-		       idx,
-		       path->cached ? 'c' : 'b',
-		       path->level,
-		       bch2_btree_id_str(path->btree_id));
+		prt_printf(out, "  path %u %c ",
+			   idx,
+			   path->cached ? 'c' : 'b');
+		bch2_btree_id_to_text(out, path->btree_id);
+		prt_printf(out, " l=%u:", path->level);
 		bch2_bpos_to_text(out, path->pos);
 		prt_newline(out);
 
diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c
index c1657182c275..924b5e3a4390 100644
--- a/fs/bcachefs/btree_journal_iter.c
+++ b/fs/bcachefs/btree_journal_iter.c
@@ -628,8 +628,11 @@ void bch2_journal_keys_dump(struct bch_fs *c)
 
 	darray_for_each(*keys, i) {
 		printbuf_reset(&buf);
+		prt_printf(&buf, "btree=");
+		bch2_btree_id_to_text(&buf, i->btree_id);
+		prt_printf(&buf, " l=%u ", i->level);
 		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(i->k));
-		pr_err("%s l=%u %s", bch2_btree_id_str(i->btree_id), i->level, buf.buf);
+		pr_err("%s", buf.buf);
 	}
 	printbuf_exit(&buf);
 }
diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c
index 30131c3bdd97..4b4df31d4b95 100644
--- a/fs/bcachefs/btree_node_scan.c
+++ b/fs/bcachefs/btree_node_scan.c
@@ -22,9 +22,9 @@ struct find_btree_nodes_worker {
 
 static void found_btree_node_to_text(struct printbuf *out, struct bch_fs *c, const struct found_btree_node *n)
 {
-	prt_printf(out, "%s l=%u seq=%u journal_seq=%llu cookie=%llx ",
-		   bch2_btree_id_str(n->btree_id), n->level, n->seq,
-		   n->journal_seq, n->cookie);
+	bch2_btree_id_level_to_text(out, n->btree_id, n->level);
+	prt_printf(out, " seq=%u journal_seq=%llu cookie=%llx ",
+		   n->seq, n->journal_seq, n->cookie);
 	bch2_bpos_to_text(out, n->min_key);
 	prt_str(out, "-");
 	bch2_bpos_to_text(out, n->max_key);
@@ -499,7 +499,9 @@ int bch2_get_scanned_nodes(struct bch_fs *c, enum btree_id btree,
 	if (c->opts.verbose) {
 		struct printbuf buf = PRINTBUF;
 
-		prt_printf(&buf, "recovering %s l=%u ", bch2_btree_id_str(btree), level);
+		prt_str(&buf, "recovery ");
+		bch2_btree_id_level_to_text(&buf, btree, level);
+		prt_str(&buf, " ");
 		bch2_bpos_to_text(&buf, node_min);
 		prt_str(&buf, " - ");
 		bch2_bpos_to_text(&buf, node_max);
diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index d596ef93239f..d62de3f79b29 100644
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -97,9 +97,9 @@ int bch2_btree_node_check_topology(struct btree_trans *trans, struct btree *b)
 			bch2_topology_error(c);
 
 			printbuf_reset(&buf);
-			prt_str(&buf, "end of prev node doesn't match start of next node\n"),
-			prt_printf(&buf, "  in btree %s level %u node ",
-				   bch2_btree_id_str(b->c.btree_id), b->c.level);
+			prt_str(&buf, "end of prev node doesn't match start of next node\n  in ");
+			bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+			prt_str(&buf, " node ");
 			bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 			prt_str(&buf, "\n  prev ");
 			bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(prev.k));
@@ -118,9 +118,9 @@ int bch2_btree_node_check_topology(struct btree_trans *trans, struct btree *b)
 		bch2_topology_error(c);
 
 		printbuf_reset(&buf);
-		prt_str(&buf, "empty interior node\n");
-		prt_printf(&buf, "  in btree %s level %u node ",
-			   bch2_btree_id_str(b->c.btree_id), b->c.level);
+		prt_str(&buf, "empty interior node\n  in ");
+		bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+		prt_str(&buf, " node ");
 		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 
 		need_fsck_err(trans, btree_node_topology_empty_interior_node, "%s", buf.buf);
@@ -129,9 +129,9 @@ int bch2_btree_node_check_topology(struct btree_trans *trans, struct btree *b)
 		bch2_topology_error(c);
 
 		printbuf_reset(&buf);
-		prt_str(&buf, "last child node doesn't end at end of parent node\n");
-		prt_printf(&buf, "  in btree %s level %u node ",
-			   bch2_btree_id_str(b->c.btree_id), b->c.level);
+		prt_str(&buf, "last child node doesn't end at end of parent node\n  in ");
+		bch2_btree_id_level_to_text(&buf, b->c.btree_id, b->c.level);
+		prt_str(&buf, " node ");
 		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
 		prt_str(&buf, "\n  last key ");
 		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(prev.k));
@@ -2575,8 +2575,9 @@ static void bch2_btree_update_to_text(struct printbuf *out, struct btree_update
 	prt_printf(out, "%ps: ", (void *) as->ip_started);
 	bch2_trans_commit_flags_to_text(out, as->flags);
 
-	prt_printf(out, " btree=%s l=%u-%u mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
-		   bch2_btree_id_str(as->btree_id),
+	prt_str(out, " ");
+	bch2_btree_id_to_text(out, as->btree_id);
+	prt_printf(out, " l=%u-%u mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
 		   as->update_level_start,
 		   as->update_level_end,
 		   bch2_btree_update_modes[as->mode],
diff --git a/fs/bcachefs/debug.c b/fs/bcachefs/debug.c
index 45aec1afdb0e..b5de52a50d10 100644
--- a/fs/bcachefs/debug.c
+++ b/fs/bcachefs/debug.c
@@ -472,7 +472,9 @@ static void bch2_cached_btree_node_to_text(struct printbuf *out, struct bch_fs *
 	if (!out->nr_tabstops)
 		printbuf_tabstop_push(out, 32);
 
-	prt_printf(out, "%px btree=%s l=%u\n", b, bch2_btree_id_str(b->c.btree_id), b->c.level);
+	prt_printf(out, "%px ", b);
+	bch2_btree_id_level_to_text(out, b->c.btree_id, b->c.level);
+	prt_printf(out, "\n");
 
 	printbuf_indent_add(out, 2);
 
diff --git a/fs/bcachefs/disk_accounting.c b/fs/bcachefs/disk_accounting.c
index 07eb8fa1b026..38b563113cfb 100644
--- a/fs/bcachefs/disk_accounting.c
+++ b/fs/bcachefs/disk_accounting.c
@@ -217,7 +217,8 @@ void bch2_accounting_key_to_text(struct printbuf *out, struct disk_accounting_po
 		prt_printf(out, "id=%u", k->snapshot.id);
 		break;
 	case BCH_DISK_ACCOUNTING_btree:
-		prt_printf(out, "btree=%s", bch2_btree_id_str(k->btree.id));
+		prt_str(out, "btree=");
+		bch2_btree_id_to_text(out, k->btree.id);
 		break;
 	}
 }
diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index 7c7595e5369b..9bc0caa9d5e4 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -421,7 +421,8 @@ static void journal_entry_btree_keys_to_text(struct printbuf *out, struct bch_fs
 			bch2_prt_jset_entry_type(out, entry->type);
 			prt_str(out, ": ");
 		}
-		prt_printf(out, "btree=%s l=%u ", bch2_btree_id_str(entry->btree_id), entry->level);
+		bch2_btree_id_level_to_text(out, entry->btree_id, entry->level);
+		prt_char(out, ' ');
 		bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(k));
 		first = false;
 	}
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index b1c83e72c0d8..0e5a53541ce4 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -42,7 +42,10 @@ int bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
 	mutex_lock(&c->sb_lock);
 
 	if (!(c->sb.btrees_lost_data & b)) {
-		bch_err(c, "flagging btree %s lost data", bch2_btree_id_str(btree));
+		struct printbuf buf = PRINTBUF;
+		bch2_btree_id_to_text(&buf, btree);
+		bch_err(c, "flagging btree %s lost data", buf.buf);
+		printbuf_exit(&buf);
 		bch2_sb_field_get(c->disk_sb.sb, ext)->btrees_lost_data |= cpu_to_le64(b);
 	}
 
@@ -385,10 +388,13 @@ int bch2_journal_replay(struct bch_fs *c)
 				 ? BCH_TRANS_COMMIT_no_journal_res|BCH_WATERMARK_reclaim
 				 : 0),
 			     bch2_journal_replay_key(trans, k));
-		bch_err_msg(c, ret, "while replaying key at btree %s level %u:",
-			    bch2_btree_id_str(k->btree_id), k->level);
-		if (ret)
+		if (ret) {
+			struct printbuf buf = PRINTBUF;
+			bch2_btree_id_level_to_text(&buf, k->btree_id, k->level);
+			bch_err_msg(c, ret, "while replaying key at %s:", buf.buf);
+			printbuf_exit(&buf);
 			goto err;
+		}
 
 		BUG_ON(k->btree_id != BTREE_ID_accounting && !k->overwritten);
 	}
@@ -536,6 +542,7 @@ static int journal_replay_early(struct bch_fs *c,
 
 static int read_btree_roots(struct bch_fs *c)
 {
+	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
 	for (unsigned i = 0; i < btree_id_nr_alive(c); i++) {
@@ -547,14 +554,17 @@ static int read_btree_roots(struct bch_fs *c)
 		if (btree_id_is_alloc(i) && c->opts.reconstruct_alloc)
 			continue;
 
+		printbuf_reset(&buf);
+		bch2_btree_id_level_to_text(&buf, i, r->level);
+
 		if (mustfix_fsck_err_on((ret = r->error),
 					c, btree_root_bkey_invalid,
 					"invalid btree root %s",
-					bch2_btree_id_str(i)) ||
+					buf.buf) ||
 		    mustfix_fsck_err_on((ret = r->error = bch2_btree_root_read(c, i, &r->key, r->level)),
 					c, btree_root_read_error,
-					"error reading btree root %s l=%u: %s",
-					bch2_btree_id_str(i), r->level, bch2_err_str(ret))) {
+					"error reading btree root %s: %s",
+					buf.buf, bch2_err_str(ret))) {
 			if (btree_id_is_alloc(i))
 				r->error = 0;
 
@@ -572,6 +582,7 @@ static int read_btree_roots(struct bch_fs *c)
 		}
 	}
 fsck_err:
+	printbuf_exit(&buf);
 	return ret;
 }
 
diff --git a/fs/bcachefs/sysfs.c b/fs/bcachefs/sysfs.c
index 03e59f86f360..3270bfab9466 100644
--- a/fs/bcachefs/sysfs.c
+++ b/fs/bcachefs/sysfs.c
@@ -302,7 +302,8 @@ static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c
 
 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 {
-	prt_printf(out, "%s: ", bch2_btree_id_str(c->gc_gens_btree));
+	bch2_btree_id_to_text(out, c->gc_gens_btree);
+	prt_printf(out, ": ");
 	bch2_bpos_to_text(out, c->gc_gens_pos);
 	prt_printf(out, "\n");
 }
-- 
2.51.0


From d55d4a0ca27adea2e6bb404eb9b65a19036dd047 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 1 Sep 2024 14:57:26 -0400
Subject: [PATCH 07/16] bcachefs: Refactor new stripe path to reduce
 dependencies on ec_stripe_head

We need to add a path for reshaping existing stripes (for e.g. device
removal), and this new path won't necessarily use ec_stripe_head.

Refactor the code to avoid unnecessary references to it for clarity.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/ec.c | 196 +++++++++++++++++++++++++----------------------
 1 file changed, 104 insertions(+), 92 deletions(-)

diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index 075bfd1cbb15..8b727d63af3e 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -1707,7 +1707,7 @@ static void ec_stripe_key_init(struct bch_fs *c,
 	set_bkey_val_u64s(&s->k, u64s);
 }
 
-static int ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
+static struct ec_stripe_new *ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
 {
 	struct ec_stripe_new *s;
 
@@ -1715,7 +1715,7 @@ static int ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
 
 	s = kzalloc(sizeof(*s), GFP_KERNEL);
 	if (!s)
-		return -BCH_ERR_ENOMEM_ec_new_stripe_alloc;
+		return NULL;
 
 	mutex_init(&s->lock);
 	closure_init(&s->iodone, NULL);
@@ -1730,10 +1730,7 @@ static int ec_new_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
 	ec_stripe_key_init(c, &s->new_stripe.key,
 			   s->nr_data, s->nr_parity,
 			   h->blocksize, h->disk_label);
-
-	h->s = s;
-	h->nr_created++;
-	return 0;
+	return s;
 }
 
 static void ec_stripe_head_devs_update(struct bch_fs *c, struct ec_stripe_head *h)
@@ -1878,25 +1875,26 @@ err:
 	return h;
 }
 
-static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_head *h,
+static int new_stripe_alloc_buckets(struct btree_trans *trans,
+				    struct ec_stripe_head *h, struct ec_stripe_new *s,
 				    enum bch_watermark watermark, struct closure *cl)
 {
 	struct bch_fs *c = trans->c;
 	struct bch_devs_mask devs = h->devs;
 	struct open_bucket *ob;
 	struct open_buckets buckets;
-	struct bch_stripe *v = &bkey_i_to_stripe(&h->s->new_stripe.key)->v;
+	struct bch_stripe *v = &bkey_i_to_stripe(&s->new_stripe.key)->v;
 	unsigned i, j, nr_have_parity = 0, nr_have_data = 0;
 	bool have_cache = true;
 	int ret = 0;
 
-	BUG_ON(v->nr_blocks	!= h->s->nr_data + h->s->nr_parity);
-	BUG_ON(v->nr_redundant	!= h->s->nr_parity);
+	BUG_ON(v->nr_blocks	!= s->nr_data + s->nr_parity);
+	BUG_ON(v->nr_redundant	!= s->nr_parity);
 
 	/* * We bypass the sector allocator which normally does this: */
 	bitmap_and(devs.d, devs.d, c->rw_devs[BCH_DATA_user].d, BCH_SB_MEMBERS_MAX);
 
-	for_each_set_bit(i, h->s->blocks_gotten, v->nr_blocks) {
+	for_each_set_bit(i, s->blocks_gotten, v->nr_blocks) {
 		/*
 		 * Note: we don't yet repair invalid blocks (failed/removed
 		 * devices) when reusing stripes - we still need a codepath to
@@ -1906,21 +1904,21 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
 		if (v->ptrs[i].dev != BCH_SB_MEMBER_INVALID)
 			__clear_bit(v->ptrs[i].dev, devs.d);
 
-		if (i < h->s->nr_data)
+		if (i < s->nr_data)
 			nr_have_data++;
 		else
 			nr_have_parity++;
 	}
 
-	BUG_ON(nr_have_data	> h->s->nr_data);
-	BUG_ON(nr_have_parity	> h->s->nr_parity);
+	BUG_ON(nr_have_data	> s->nr_data);
+	BUG_ON(nr_have_parity	> s->nr_parity);
 
 	buckets.nr = 0;
-	if (nr_have_parity < h->s->nr_parity) {
+	if (nr_have_parity < s->nr_parity) {
 		ret = bch2_bucket_alloc_set_trans(trans, &buckets,
 					    &h->parity_stripe,
 					    &devs,
-					    h->s->nr_parity,
+					    s->nr_parity,
 					    &nr_have_parity,
 					    &have_cache, 0,
 					    BCH_DATA_parity,
@@ -1928,14 +1926,14 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
 					    cl);
 
 		open_bucket_for_each(c, &buckets, ob, i) {
-			j = find_next_zero_bit(h->s->blocks_gotten,
-					       h->s->nr_data + h->s->nr_parity,
-					       h->s->nr_data);
-			BUG_ON(j >= h->s->nr_data + h->s->nr_parity);
+			j = find_next_zero_bit(s->blocks_gotten,
+					       s->nr_data + s->nr_parity,
+					       s->nr_data);
+			BUG_ON(j >= s->nr_data + s->nr_parity);
 
-			h->s->blocks[j] = buckets.v[i];
+			s->blocks[j] = buckets.v[i];
 			v->ptrs[j] = bch2_ob_ptr(c, ob);
-			__set_bit(j, h->s->blocks_gotten);
+			__set_bit(j, s->blocks_gotten);
 		}
 
 		if (ret)
@@ -1943,11 +1941,11 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
 	}
 
 	buckets.nr = 0;
-	if (nr_have_data < h->s->nr_data) {
+	if (nr_have_data < s->nr_data) {
 		ret = bch2_bucket_alloc_set_trans(trans, &buckets,
 					    &h->block_stripe,
 					    &devs,
-					    h->s->nr_data,
+					    s->nr_data,
 					    &nr_have_data,
 					    &have_cache, 0,
 					    BCH_DATA_user,
@@ -1955,13 +1953,13 @@ static int new_stripe_alloc_buckets(struct btree_trans *trans, struct ec_stripe_
 					    cl);
 
 		open_bucket_for_each(c, &buckets, ob, i) {
-			j = find_next_zero_bit(h->s->blocks_gotten,
-					       h->s->nr_data, 0);
-			BUG_ON(j >= h->s->nr_data);
+			j = find_next_zero_bit(s->blocks_gotten,
+					       s->nr_data, 0);
+			BUG_ON(j >= s->nr_data);
 
-			h->s->blocks[j] = buckets.v[i];
+			s->blocks[j] = buckets.v[i];
 			v->ptrs[j] = bch2_ob_ptr(c, ob);
-			__set_bit(j, h->s->blocks_gotten);
+			__set_bit(j, s->blocks_gotten);
 		}
 
 		if (ret)
@@ -2007,73 +2005,78 @@ static s64 get_existing_stripe(struct bch_fs *c,
 	return ret;
 }
 
-static int __bch2_ec_stripe_head_reuse(struct btree_trans *trans, struct ec_stripe_head *h)
+static int init_new_stripe_from_existing(struct bch_fs *c, struct ec_stripe_new *s)
 {
-	struct bch_fs *c = trans->c;
-	struct bch_stripe *new_v = &bkey_i_to_stripe(&h->s->new_stripe.key)->v;
-	struct bch_stripe *existing_v;
+	struct bch_stripe *new_v = &bkey_i_to_stripe(&s->new_stripe.key)->v;
+	struct bch_stripe *existing_v = &bkey_i_to_stripe(&s->existing_stripe.key)->v;
 	unsigned i;
-	s64 idx;
-	int ret;
 
-	/*
-	 * If we can't allocate a new stripe, and there's no stripes with empty
-	 * blocks for us to reuse, that means we have to wait on copygc:
-	 */
-	idx = get_existing_stripe(c, h);
-	if (idx < 0)
-		return -BCH_ERR_stripe_alloc_blocked;
-
-	ret = get_stripe_key_trans(trans, idx, &h->s->existing_stripe);
-	bch2_fs_fatal_err_on(ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart), c,
-			     "reading stripe key: %s", bch2_err_str(ret));
-	if (ret) {
-		bch2_stripe_close(c, h->s);
-		return ret;
-	}
-
-	existing_v = &bkey_i_to_stripe(&h->s->existing_stripe.key)->v;
-
-	BUG_ON(existing_v->nr_redundant != h->s->nr_parity);
-	h->s->nr_data = existing_v->nr_blocks -
+	BUG_ON(existing_v->nr_redundant != s->nr_parity);
+	s->nr_data = existing_v->nr_blocks -
 		existing_v->nr_redundant;
 
-	ret = ec_stripe_buf_init(&h->s->existing_stripe, 0, h->blocksize);
+	int ret = ec_stripe_buf_init(&s->existing_stripe, 0, le16_to_cpu(existing_v->sectors));
 	if (ret) {
-		bch2_stripe_close(c, h->s);
+		bch2_stripe_close(c, s);
 		return ret;
 	}
 
-	BUG_ON(h->s->existing_stripe.size != h->blocksize);
-	BUG_ON(h->s->existing_stripe.size != le16_to_cpu(existing_v->sectors));
+	BUG_ON(s->existing_stripe.size != le16_to_cpu(existing_v->sectors));
 
 	/*
 	 * Free buckets we initially allocated - they might conflict with
 	 * blocks from the stripe we're reusing:
 	 */
-	for_each_set_bit(i, h->s->blocks_gotten, new_v->nr_blocks) {
-		bch2_open_bucket_put(c, c->open_buckets + h->s->blocks[i]);
-		h->s->blocks[i] = 0;
+	for_each_set_bit(i, s->blocks_gotten, new_v->nr_blocks) {
+		bch2_open_bucket_put(c, c->open_buckets + s->blocks[i]);
+		s->blocks[i] = 0;
 	}
-	memset(h->s->blocks_gotten, 0, sizeof(h->s->blocks_gotten));
-	memset(h->s->blocks_allocated, 0, sizeof(h->s->blocks_allocated));
+	memset(s->blocks_gotten, 0, sizeof(s->blocks_gotten));
+	memset(s->blocks_allocated, 0, sizeof(s->blocks_allocated));
 
-	for (i = 0; i < existing_v->nr_blocks; i++) {
+	for (unsigned i = 0; i < existing_v->nr_blocks; i++) {
 		if (stripe_blockcount_get(existing_v, i)) {
-			__set_bit(i, h->s->blocks_gotten);
-			__set_bit(i, h->s->blocks_allocated);
+			__set_bit(i, s->blocks_gotten);
+			__set_bit(i, s->blocks_allocated);
 		}
 
-		ec_block_io(c, &h->s->existing_stripe, READ, i, &h->s->iodone);
+		ec_block_io(c, &s->existing_stripe, READ, i, &s->iodone);
 	}
 
-	bkey_copy(&h->s->new_stripe.key, &h->s->existing_stripe.key);
-	h->s->have_existing_stripe = true;
+	bkey_copy(&s->new_stripe.key, &s->existing_stripe.key);
+	s->have_existing_stripe = true;
 
 	return 0;
 }
 
-static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_stripe_head *h)
+static int __bch2_ec_stripe_head_reuse(struct btree_trans *trans, struct ec_stripe_head *h,
+				       struct ec_stripe_new *s)
+{
+	struct bch_fs *c = trans->c;
+	s64 idx;
+	int ret;
+
+	/*
+	 * If we can't allocate a new stripe, and there's no stripes with empty
+	 * blocks for us to reuse, that means we have to wait on copygc:
+	 */
+	idx = get_existing_stripe(c, h);
+	if (idx < 0)
+		return -BCH_ERR_stripe_alloc_blocked;
+
+	ret = get_stripe_key_trans(trans, idx, &s->existing_stripe);
+	bch2_fs_fatal_err_on(ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart), c,
+			     "reading stripe key: %s", bch2_err_str(ret));
+	if (ret) {
+		bch2_stripe_close(c, s);
+		return ret;
+	}
+
+	return init_new_stripe_from_existing(c, s);
+}
+
+static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_stripe_head *h,
+					 struct ec_stripe_new *s)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
@@ -2082,15 +2085,19 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
 	struct bpos start_pos = bpos_max(min_pos, POS(0, c->ec_stripe_hint));
 	int ret;
 
-	if (!h->s->res.sectors) {
-		ret = bch2_disk_reservation_get(c, &h->s->res,
+	if (!s->res.sectors) {
+		ret = bch2_disk_reservation_get(c, &s->res,
 					h->blocksize,
-					h->s->nr_parity,
+					s->nr_parity,
 					BCH_DISK_RESERVATION_NOFAIL);
 		if (ret)
 			return ret;
 	}
 
+	/*
+	 * Allocate stripe slot
+	 * XXX: we're going to need a bitrange btree of free stripes
+	 */
 	for_each_btree_key_norestart(trans, iter, BTREE_ID_stripes, start_pos,
 			   BTREE_ITER_slots|BTREE_ITER_intent, k, ret) {
 		if (bkey_gt(k.k->p, POS(0, U32_MAX))) {
@@ -2105,7 +2112,7 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
 		}
 
 		if (bkey_deleted(k.k) &&
-		    bch2_try_open_stripe(c, h->s, k.k->p.offset))
+		    bch2_try_open_stripe(c, s, k.k->p.offset))
 			break;
 	}
 
@@ -2116,16 +2123,16 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
 
 	ret = ec_stripe_mem_alloc(trans, &iter);
 	if (ret) {
-		bch2_stripe_close(c, h->s);
+		bch2_stripe_close(c, s);
 		goto err;
 	}
 
-	h->s->new_stripe.key.k.p = iter.pos;
+	s->new_stripe.key.k.p = iter.pos;
 out:
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 err:
-	bch2_disk_reservation_put(c, &h->s->res);
+	bch2_disk_reservation_put(c, &s->res);
 	goto out;
 }
 
@@ -2156,22 +2163,27 @@ struct ec_stripe_head *bch2_ec_stripe_head_get(struct btree_trans *trans,
 		return h;
 
 	if (!h->s) {
-		ret = ec_new_stripe_alloc(c, h);
-		if (ret) {
+		h->s = ec_new_stripe_alloc(c, h);
+		if (!h->s) {
+			ret = -BCH_ERR_ENOMEM_ec_new_stripe_alloc;
 			bch_err(c, "failed to allocate new stripe");
 			goto err;
 		}
+
+		h->nr_created++;
 	}
 
-	if (h->s->allocated)
+	struct ec_stripe_new *s = h->s;
+
+	if (s->allocated)
 		goto allocated;
 
-	if (h->s->have_existing_stripe)
+	if (s->have_existing_stripe)
 		goto alloc_existing;
 
 	/* First, try to allocate a full stripe: */
-	ret =   new_stripe_alloc_buckets(trans, h, BCH_WATERMARK_stripe, NULL) ?:
-		__bch2_ec_stripe_head_reserve(trans, h);
+	ret =   new_stripe_alloc_buckets(trans, h, s, BCH_WATERMARK_stripe, NULL) ?:
+		__bch2_ec_stripe_head_reserve(trans, h, s);
 	if (!ret)
 		goto allocate_buf;
 	if (bch2_err_matches(ret, BCH_ERR_transaction_restart) ||
@@ -2183,15 +2195,15 @@ struct ec_stripe_head *bch2_ec_stripe_head_get(struct btree_trans *trans,
 	 * existing stripe:
 	 */
 	while (1) {
-		ret = __bch2_ec_stripe_head_reuse(trans, h);
+		ret = __bch2_ec_stripe_head_reuse(trans, h, s);
 		if (!ret)
 			break;
 		if (waiting || !cl || ret != -BCH_ERR_stripe_alloc_blocked)
 			goto err;
 
 		if (watermark == BCH_WATERMARK_copygc) {
-			ret =   new_stripe_alloc_buckets(trans, h, watermark, NULL) ?:
-				__bch2_ec_stripe_head_reserve(trans, h);
+			ret =   new_stripe_alloc_buckets(trans, h, s, watermark, NULL) ?:
+				__bch2_ec_stripe_head_reserve(trans, h, s);
 			if (ret)
 				goto err;
 			goto allocate_buf;
@@ -2209,19 +2221,19 @@ alloc_existing:
 	 * Retry allocating buckets, with the watermark for this
 	 * particular write:
 	 */
-	ret = new_stripe_alloc_buckets(trans, h, watermark, cl);
+	ret = new_stripe_alloc_buckets(trans, h, s, watermark, cl);
 	if (ret)
 		goto err;
 
 allocate_buf:
-	ret = ec_stripe_buf_init(&h->s->new_stripe, 0, h->blocksize);
+	ret = ec_stripe_buf_init(&s->new_stripe, 0, h->blocksize);
 	if (ret)
 		goto err;
 
-	h->s->allocated = true;
+	s->allocated = true;
 allocated:
-	BUG_ON(!h->s->idx);
-	BUG_ON(!h->s->new_stripe.data[0]);
+	BUG_ON(!s->idx);
+	BUG_ON(!s->new_stripe.data[0]);
 	BUG_ON(trans->restarted);
 	return h;
 err:
-- 
2.51.0


From 8b22abb4c84058e9533d71a4814e54316ba2621f Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 11 Oct 2024 22:53:09 -0400
Subject: [PATCH 08/16] bcachefs: -o norecovery now bails out of recovery
 earlier

-o norecovery (used by the dump tool) should be doing the absolute
minimum amount of work to get the filesystem up and readable; we
shouldn't be running check and repair code, or going read-write.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/recovery.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index 0e5a53541ce4..bc2fd174bb32 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -690,8 +690,13 @@ int bch2_fs_recovery(struct bch_fs *c)
 		goto err;
 	}
 
-	if (c->opts.norecovery)
-		c->opts.recovery_pass_last = BCH_RECOVERY_PASS_journal_replay - 1;
+	if (c->opts.norecovery) {
+		c->opts.recovery_pass_last = c->opts.recovery_pass_last
+			? min(c->opts.recovery_pass_last, BCH_RECOVERY_PASS_snapshots_read)
+			: BCH_RECOVERY_PASS_snapshots_read;
+		c->opts.nochanges = true;
+		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);
-- 
2.51.0


From fb8c835b18d48dac953a5d755a8e90b0d8fb9c29 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 11 Oct 2024 22:50:48 -0400
Subject: [PATCH 09/16] bcachefs: bch2_journal_meta() takes ref on c->writes

This part of addressing
https://github.com/koverstreet/bcachefs/issues/656

where we're getting stuck in bch2_journal_meta() in the dump tool.

We shouldn't be invoking the journal without a ref on c->writes (if
we're not RW), and there's no reason for the dump tool to be going
read-write.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/bcachefs.h |  1 +
 fs/bcachefs/journal.c  | 27 +++++++++++++++++----------
 fs/bcachefs/recovery.c |  4 +---
 3 files changed, 19 insertions(+), 13 deletions(-)

diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index fbd89f91625d..d4d95ef6791f 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -688,6 +688,7 @@ struct btree_trans_buf {
 	((subvol_inum) { BCACHEFS_ROOT_SUBVOL,	BCACHEFS_ROOT_INO })
 
 #define BCH_WRITE_REFS()						\
+	x(journal)							\
 	x(trans)							\
 	x(write)							\
 	x(promote)							\
diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index 2dc0d60c1745..2cf8f24d50cc 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -831,19 +831,14 @@ out:
 	return ret;
 }
 
-int bch2_journal_meta(struct journal *j)
+static int __bch2_journal_meta(struct journal *j)
 {
-	struct journal_buf *buf;
-	struct journal_res res;
-	int ret;
-
-	memset(&res, 0, sizeof(res));
-
-	ret = bch2_journal_res_get(j, &res, jset_u64s(0), 0);
+	struct journal_res res = {};
+	int ret = bch2_journal_res_get(j, &res, jset_u64s(0), 0);
 	if (ret)
 		return ret;
 
-	buf = j->buf + (res.seq & JOURNAL_BUF_MASK);
+	struct journal_buf *buf = j->buf + (res.seq & JOURNAL_BUF_MASK);
 	buf->must_flush = true;
 
 	if (!buf->flush_time) {
@@ -856,6 +851,18 @@ int bch2_journal_meta(struct journal *j)
 	return bch2_journal_flush_seq(j, res.seq, TASK_UNINTERRUPTIBLE);
 }
 
+int bch2_journal_meta(struct journal *j)
+{
+	struct bch_fs *c = container_of(j, struct bch_fs, journal);
+
+	if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_journal))
+		return -EROFS;
+
+	int ret = __bch2_journal_meta(j);
+	bch2_write_ref_put(c, BCH_WRITE_REF_journal);
+	return ret;
+}
+
 /* block/unlock the journal: */
 
 void bch2_journal_unblock(struct journal *j)
@@ -1193,7 +1200,7 @@ void bch2_fs_journal_stop(struct journal *j)
 	 * Always write a new journal entry, to make sure the clock hands are up
 	 * to date (and match the superblock)
 	 */
-	bch2_journal_meta(j);
+	__bch2_journal_meta(j);
 
 	journal_quiesce(j);
 	cancel_delayed_work_sync(&j->write_work);
diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index bc2fd174bb32..431698189090 100644
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -910,11 +910,9 @@ use_clean:
 	set_bit(BCH_FS_accounting_replay_done, &c->flags);
 
 	/* fsync if we fixed errors */
-	if (test_bit(BCH_FS_errors_fixed, &c->flags) &&
-	    bch2_write_ref_tryget(c, BCH_WRITE_REF_fsync)) {
+	if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
 		bch2_journal_flush_all_pins(&c->journal);
 		bch2_journal_meta(&c->journal);
-		bch2_write_ref_put(c, BCH_WRITE_REF_fsync);
 	}
 
 	/* If we fixed errors, verify that fs is actually clean now: */
-- 
2.51.0


From be5a7be1062b2e588519d7ed68ff2e8f4ed0a42a Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sat, 12 Oct 2024 14:07:44 -0400
Subject: [PATCH 10/16] bcachefs: Fix warning about passing flex array member
 by value

this showed up when building in userspace

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/disk_accounting.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/fs/bcachefs/disk_accounting.c b/fs/bcachefs/disk_accounting.c
index 38b563113cfb..55a00018dc8b 100644
--- a/fs/bcachefs/disk_accounting.c
+++ b/fs/bcachefs/disk_accounting.c
@@ -244,10 +244,10 @@ void bch2_accounting_swab(struct bkey_s k)
 }
 
 static inline void __accounting_to_replicas(struct bch_replicas_entry_v1 *r,
-					    struct disk_accounting_pos acc)
+					    struct disk_accounting_pos *acc)
 {
-	unsafe_memcpy(r, &acc.replicas,
-		      replicas_entry_bytes(&acc.replicas),
+	unsafe_memcpy(r, &acc->replicas,
+		      replicas_entry_bytes(&acc->replicas),
 		      "variable length struct");
 }
 
@@ -258,7 +258,7 @@ static inline bool accounting_to_replicas(struct bch_replicas_entry_v1 *r, struc
 
 	switch (acc_k.type) {
 	case BCH_DISK_ACCOUNTING_replicas:
-		__accounting_to_replicas(r, acc_k);
+		__accounting_to_replicas(r, &acc_k);
 		return true;
 	default:
 		return false;
@@ -626,7 +626,7 @@ static int bch2_disk_accounting_validate_late(struct btree_trans *trans,
 	switch (acc.type) {
 	case BCH_DISK_ACCOUNTING_replicas: {
 		struct bch_replicas_padded r;
-		__accounting_to_replicas(&r.e, acc);
+		__accounting_to_replicas(&r.e, &acc);
 
 		for (unsigned i = 0; i < r.e.nr_devs; i++)
 			if (r.e.devs[i] != BCH_SB_MEMBER_INVALID &&
-- 
2.51.0


From 27de0ee39f810dab2e948d2c465f8fcf8cbf9f8c Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 14 Oct 2024 21:35:44 -0400
Subject: [PATCH 11/16] bcachefs: Add block plugging to read paths

This will help with some of the btree_trans srcu lock hold time warnings
that are still turning up; submit_bio() can block for awhile if the
device is sufficiently congested.

It's not a perfect solution since blk_plug bios are submitted when
scheduling; we might want a way to disable the "submit on context
switch" behaviour, or switch to our own plugging in the future.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/fs-io-buffered.c | 19 ++++++++++++++++++-
 fs/bcachefs/fs-io-direct.c   |  5 +++++
 2 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/fs-io-buffered.c b/fs/bcachefs/fs-io-buffered.c
index 95972809e76d..0923f38a2fcd 100644
--- a/fs/bcachefs/fs-io-buffered.c
+++ b/fs/bcachefs/fs-io-buffered.c
@@ -248,6 +248,7 @@ void bch2_readahead(struct readahead_control *ractl)
 	struct bch_io_opts opts;
 	struct folio *folio;
 	struct readpages_iter readpages_iter;
+	struct blk_plug plug;
 
 	bch2_inode_opts_get(&opts, c, &inode->ei_inode);
 
@@ -255,6 +256,16 @@ void bch2_readahead(struct readahead_control *ractl)
 	if (ret)
 		return;
 
+	/*
+	 * Besides being a general performance optimization, plugging helps with
+	 * avoiding btree transaction srcu warnings - submitting a bio can
+	 * block, and we don't want todo that with the transaction locked.
+	 *
+	 * However, plugged bios are submitted when we schedule; we ideally
+	 * would have our own scheduler hook to call unlock_long() before
+	 * scheduling.
+	 */
+	blk_start_plug(&plug);
 	bch2_pagecache_add_get(inode);
 
 	struct btree_trans *trans = bch2_trans_get(c);
@@ -281,7 +292,7 @@ void bch2_readahead(struct readahead_control *ractl)
 	bch2_trans_put(trans);
 
 	bch2_pagecache_add_put(inode);
-
+	blk_finish_plug(&plug);
 	darray_exit(&readpages_iter.folios);
 }
 
@@ -296,9 +307,13 @@ int bch2_read_single_folio(struct folio *folio, struct address_space *mapping)
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	struct bch_read_bio *rbio;
 	struct bch_io_opts opts;
+	struct blk_plug plug;
 	int ret;
 	DECLARE_COMPLETION_ONSTACK(done);
 
+	BUG_ON(folio_test_uptodate(folio));
+	BUG_ON(folio_test_dirty(folio));
+
 	if (!bch2_folio_create(folio, GFP_KERNEL))
 		return -ENOMEM;
 
@@ -313,7 +328,9 @@ int bch2_read_single_folio(struct folio *folio, struct address_space *mapping)
 	rbio->bio.bi_iter.bi_sector = folio_sector(folio);
 	BUG_ON(!bio_add_folio(&rbio->bio, folio, folio_size(folio), 0));
 
+	blk_start_plug(&plug);
 	bch2_trans_run(c, (bchfs_read(trans, rbio, inode_inum(inode), NULL), 0));
+	blk_finish_plug(&plug);
 	wait_for_completion(&done);
 
 	ret = blk_status_to_errno(rbio->bio.bi_status);
diff --git a/fs/bcachefs/fs-io-direct.c b/fs/bcachefs/fs-io-direct.c
index 6d3a05ae5da8..2089c36b5866 100644
--- a/fs/bcachefs/fs-io-direct.c
+++ b/fs/bcachefs/fs-io-direct.c
@@ -70,6 +70,7 @@ static int bch2_direct_IO_read(struct kiocb *req, struct iov_iter *iter)
 	struct bch_io_opts opts;
 	struct dio_read *dio;
 	struct bio *bio;
+	struct blk_plug plug;
 	loff_t offset = req->ki_pos;
 	bool sync = is_sync_kiocb(req);
 	size_t shorten;
@@ -128,6 +129,8 @@ static int bch2_direct_IO_read(struct kiocb *req, struct iov_iter *iter)
 	 */
 	dio->should_dirty = iter_is_iovec(iter);
 
+	blk_start_plug(&plug);
+
 	goto start;
 	while (iter->count) {
 		bio = bio_alloc_bioset(NULL,
@@ -160,6 +163,8 @@ start:
 		bch2_read(c, rbio_init(bio, opts), inode_inum(inode));
 	}
 
+	blk_finish_plug(&plug);
+
 	iter->count += shorten;
 
 	if (sync) {
-- 
2.51.0


From e0c8369bc8444daf0d68d23bcae472d11680d49f Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Sun, 20 Oct 2024 19:02:44 -0400
Subject: [PATCH 12/16] bcachefs: Add version check for
 bch_btree_ptr_v2.sectors_written validate

A user popped up with a very old (0.11) filesystem that needed repair
and wasn't recently backed up.

Reported-by: Manoa <manoa@mail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/extents.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c
index 37e3d69bec06..85b98c782e1b 100644
--- a/fs/bcachefs/extents.c
+++ b/fs/bcachefs/extents.c
@@ -203,7 +203,8 @@ int bch2_btree_ptr_v2_validate(struct bch_fs *c, struct bkey_s_c k,
 			 c, btree_ptr_v2_min_key_bad,
 			 "min_key > key");
 
-	if (flags & BCH_VALIDATE_write)
+	if ((flags & BCH_VALIDATE_write) &&
+	    c->sb.version_min >= bcachefs_metadata_version_btree_ptr_sectors_written)
 		bkey_fsck_err_on(!bp.v->sectors_written,
 				 c, btree_ptr_v2_written_0,
 				 "sectors_written == 0");
-- 
2.51.0


From de902e3b4a9881ae02d414d427ad56cc384c9bf1 Mon Sep 17 00:00:00 2001
From: Thorsten Blum <thorsten.blum@linux.dev>
Date: Sat, 19 Oct 2024 14:25:27 +0200
Subject: [PATCH 13/16] bcachefs: Use str_write_read() helper function

Remove hard-coded strings by using the helper function str_write_read().

Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/journal_io.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index 9bc0caa9d5e4..768a3b950997 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/journal_io.c
@@ -17,6 +17,8 @@
 #include "sb-clean.h"
 #include "trace.h"
 
+#include <linux/string_choices.h>
+
 void bch2_journal_pos_from_member_info_set(struct bch_fs *c)
 {
 	lockdep_assert_held(&c->sb_lock);
@@ -666,7 +668,7 @@ static void journal_entry_clock_to_text(struct printbuf *out, struct bch_fs *c,
 	struct jset_entry_clock *clock =
 		container_of(entry, struct jset_entry_clock, entry);
 
-	prt_printf(out, "%s=%llu", clock->rw ? "write" : "read", le64_to_cpu(clock->time));
+	prt_printf(out, "%s=%llu", str_write_read(clock->rw), le64_to_cpu(clock->time));
 }
 
 static int journal_entry_dev_usage_validate(struct bch_fs *c,
-- 
2.51.0


From 751d869710ca91b6b2c6f4235137c71eb054ce02 Mon Sep 17 00:00:00 2001
From: Thorsten Blum <thorsten.blum@linux.dev>
Date: Sun, 20 Oct 2024 13:20:46 +0200
Subject: [PATCH 14/16] bcachefs: Use str_write_read() helper in
 ec_block_endio()

Remove hard-coded strings by using the helper function str_write_read().

Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/ec.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c
index 8b727d63af3e..b46bf00c4a67 100644
--- a/fs/bcachefs/ec.c
+++ b/fs/bcachefs/ec.c
@@ -26,6 +26,7 @@
 #include "util.h"
 
 #include <linux/sort.h>
+#include <linux/string_choices.h>
 
 #ifdef __KERNEL__
 
@@ -732,7 +733,7 @@ static void ec_block_endio(struct bio *bio)
 			       ? BCH_MEMBER_ERROR_write
 			       : BCH_MEMBER_ERROR_read,
 			       "erasure coding %s error: %s",
-			       bio_data_dir(bio) ? "write" : "read",
+			       str_write_read(bio_data_dir(bio)),
 			       bch2_blk_status_to_str(bio->bi_status)))
 		clear_bit(ec_bio->idx, ec_bio->buf->valid);
 
-- 
2.51.0


From ac9826f14739023bccf1345e6e4ddb0461fa9a2e Mon Sep 17 00:00:00 2001
From: Thorsten Blum <thorsten.blum@linux.dev>
Date: Sat, 26 Oct 2024 12:47:23 +0200
Subject: [PATCH 15/16] bcachefs: Use str_write_read() helper in
 write_super_endio()

Remove hard-coded strings by using the str_write_read() helper function.

Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/super-io.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c
index 7c71594f6a8b..c83bd3dedb1b 100644
--- a/fs/bcachefs/super-io.c
+++ b/fs/bcachefs/super-io.c
@@ -23,6 +23,7 @@
 
 #include <linux/backing-dev.h>
 #include <linux/sort.h>
+#include <linux/string_choices.h>
 
 static const struct blk_holder_ops bch2_sb_handle_bdev_ops = {
 };
@@ -878,7 +879,7 @@ static void write_super_endio(struct bio *bio)
 			       ? BCH_MEMBER_ERROR_write
 			       : BCH_MEMBER_ERROR_read,
 			       "superblock %s error: %s",
-			       bio_data_dir(bio) ? "write" : "read",
+			       str_write_read(bio_data_dir(bio)),
 			       bch2_blk_status_to_str(bio->bi_status)))
 		ca->sb_write_error = 1;
 
-- 
2.51.0


From 901ff6555ba02dd917aa65b1105c9715e25dc994 Mon Sep 17 00:00:00 2001
From: Thorsten Blum <thorsten.blum@linux.dev>
Date: Sat, 26 Oct 2024 17:47:04 +0200
Subject: [PATCH 16/16] bcachefs: Annotate struct bucket_gens with
 __counted_by()

Add the __counted_by compiler attribute to the flexible array member b
to improve access bounds-checking via CONFIG_UBSAN_BOUNDS and
CONFIG_FORTIFY_SOURCE.

Use struct_size() to calculate the number of bytes to be allocated.

Update bucket_gens->nbuckets and bucket_gens->nbuckets_minus_first when
resizing.

Compile-tested only.

Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/buckets.c       | 13 ++++++++-----
 fs/bcachefs/buckets_types.h |  2 +-
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index ec7d9a59bea9..8bd17667e243 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -1266,8 +1266,9 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 
 	BUG_ON(resize && ca->buckets_nouse);
 
-	if (!(bucket_gens	= kvmalloc(sizeof(struct bucket_gens) + nbuckets,
-					   GFP_KERNEL|__GFP_ZERO))) {
+	bucket_gens = kvmalloc(struct_size(bucket_gens, b, nbuckets),
+			       GFP_KERNEL|__GFP_ZERO);
+	if (!bucket_gens) {
 		ret = -BCH_ERR_ENOMEM_bucket_gens;
 		goto err;
 	}
@@ -1285,11 +1286,13 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
 	old_bucket_gens = rcu_dereference_protected(ca->bucket_gens, 1);
 
 	if (resize) {
-		size_t n = min(bucket_gens->nbuckets, old_bucket_gens->nbuckets);
-
+		bucket_gens->nbuckets = min(bucket_gens->nbuckets,
+					    old_bucket_gens->nbuckets);
+		bucket_gens->nbuckets_minus_first =
+			bucket_gens->nbuckets - bucket_gens->first_bucket;
 		memcpy(bucket_gens->b,
 		       old_bucket_gens->b,
-		       n);
+		       bucket_gens->nbuckets);
 	}
 
 	rcu_assign_pointer(ca->bucket_gens, bucket_gens);
diff --git a/fs/bcachefs/buckets_types.h b/fs/bcachefs/buckets_types.h
index 28bd09a253c8..7174047b8e92 100644
--- a/fs/bcachefs/buckets_types.h
+++ b/fs/bcachefs/buckets_types.h
@@ -24,7 +24,7 @@ struct bucket_gens {
 	u16			first_bucket;
 	size_t			nbuckets;
 	size_t			nbuckets_minus_first;
-	u8			b[];
+	u8			b[] __counted_by(nbuckets);
 };
 
 struct bch_dev_usage {
-- 
2.51.0