bcachefs: bch2_btree_key_cache_drop() now evicts

As part of improving btree key cache coherency, the bkey_cached.valid
flag is going away.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/btree_key_cache.c

@@ -792,6 +792,7 @@ void bch2_btree_key_cache_drop(struct btree_trans *trans,
 			       struct btree_path *path)
 {
 	struct bch_fs *c = trans->c;
+	struct btree_key_cache *bc = &c->btree_key_cache;
 	struct bkey_cached *ck = (void *) path->l[0].b;
 
 	BUG_ON(!ck->valid);
@@ -806,7 +807,11 @@ void bch2_btree_key_cache_drop(struct btree_trans *trans,
 		bch2_journal_pin_drop(&c->journal, &ck->journal);
 	}
 
-	ck->valid = false;
+	bkey_cached_evict(bc, ck);
+	bkey_cached_free_fast(bc, ck);
+
+	mark_btree_node_locked(trans, path, 0, BTREE_NODE_UNLOCKED);
+	btree_path_set_dirty(path, BTREE_ITER_NEED_TRAVERSE);
 }
 
 static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,

fs/bcachefs/btree_trans_commit.c

@@ -137,7 +137,8 @@ static inline void bch2_trans_unlock_write(struct btree_trans *trans)
 {
 	if (likely(trans->write_locked)) {
 		trans_for_each_update(trans, i)
-			if (!same_leaf_as_prev(trans, i))
+			if (btree_node_locked_type(trans->paths + i->path, i->level) ==
+			    BTREE_NODE_WRITE_LOCKED)
 				bch2_btree_node_unlock_write_inlined(trans,
 						trans->paths + i->path, insert_l(trans, i)->b);
 		trans->write_locked = false;
@@ -777,14 +778,12 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
 	trans_for_each_update(trans, i) {
 		struct btree_path *path = trans->paths + i->path;
 
-		if (!i->cached) {
+		if (!i->cached)
 			bch2_btree_insert_key_leaf(trans, path, i->k, trans->journal_res.seq);
-		} else if (!i->key_cache_already_flushed)
+		else if (!i->key_cache_already_flushed)
 			bch2_btree_insert_key_cached(trans, flags, i);
-		else {
+		else
 			bch2_btree_key_cache_drop(trans, path);
-			btree_path_set_dirty(path, BTREE_ITER_NEED_TRAVERSE);
-		}
 	}
 
 	return 0;