bcachefs: Fix usage of six lock's percpu mode

Six locks have a percpu mode, which we use for interior btree nodes, as
well as btree key cache keys for the subvolumes btree. We've been
switching locks back and forth between percpu and non percpu mode as
needed, but it turns out this is racy - when we're reusing an existing
node, other threads could be attempting to lock it while we're switching
it between modes.

This patch fixes this by never switching 'struct btree' between the two
modes, and instead segragating them between two different freed lists.

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

1 files changed, 5 insertions, 5 deletions

diff --git a/fs/bcachefs/btree_key_cache.c b/fs/bcachefs/btree_key_cache.c
index 70f31b5379e7..7e41552a57df 100644
--- a/fs/bcachefs/btree_key_cache.c
+++ b/fs/bcachefs/btree_key_cache.c
@@ -166,13 +166,13 @@ btree_key_cache_create(struct bch_fs *c,
 		}
 
 		was_new = false;
+	} else {
+		if (btree_id == BTREE_ID_subvolumes)
+			six_lock_pcpu_alloc(&ck->c.lock);
+		else
+			six_lock_pcpu_free(&ck->c.lock);
 	}
 
-	if (btree_id == BTREE_ID_subvolumes)
-		six_lock_pcpu_alloc(&ck->c.lock);
-	else
-		six_lock_pcpu_free(&ck->c.lock);
-
 	ck->c.level		= 0;
 	ck->c.btree_id		= btree_id;
 	ck->key.btree_id	= btree_id;