af_unix: Skip GC if no cycle exists.

We do not need to run GC if there is no possible cyclic reference.
We use unix_graph_maybe_cyclic to decide if we should run GC.

If a fd of an AF_UNIX socket is passed to an already inflight AF_UNIX
socket, they could form a cyclic reference.  Then, we set true to
unix_graph_maybe_cyclic and later run Tarjan's algorithm to group
them into SCC.

Once we run Tarjan's algorithm, we are 100% sure whether cyclic
references exist or not.  If there is no cycle, we set false to
unix_graph_maybe_cyclic and can skip the entire garbage collection
next time.

When finalising SCC, we set true to unix_graph_maybe_cyclic if SCC
consists of multiple vertices.

Even if SCC is a single vertex, a cycle might exist as self-fd passing.
Given the corner case is rare, we detect it by checking all edges of
the vertex and set true to unix_graph_maybe_cyclic.

With this change, __unix_gc() is just a spin_lock() dance in the normal
usage.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20240325202425.60930-11-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

1 files changed, 47 insertions, 1 deletions

diff --git a/net/unix/garbage.c b/net/unix/garbage.c
index cec6189126ba..5a1fae78d6dc 100644
--- a/net/unix/garbage.c
+++ b/net/unix/garbage.c
@@ -112,6 +112,19 @@ static struct unix_vertex *unix_edge_successor(struct unix_edge *edge)
 	return edge->successor->vertex;
 }
 
+static bool unix_graph_maybe_cyclic;
+
+static void unix_update_graph(struct unix_vertex *vertex)
+{
+	/* If the receiver socket is not inflight, no cyclic
+	 * reference could be formed.
+	 */
+	if (!vertex)
+		return;
+
+	unix_graph_maybe_cyclic = true;
+}
+
 static LIST_HEAD(unix_unvisited_vertices);
 
 enum unix_vertex_index {
@@ -138,12 +151,16 @@ static void unix_add_edge(struct scm_fp_list *fpl, struct unix_edge *edge)
 
 	vertex->out_degree++;
 	list_add_tail(&edge->vertex_entry, &vertex->edges);
+
+	unix_update_graph(unix_edge_successor(edge));
 }
 
 static void unix_del_edge(struct scm_fp_list *fpl, struct unix_edge *edge)
 {
 	struct unix_vertex *vertex = edge->predecessor->vertex;
 
+	unix_update_graph(unix_edge_successor(edge));
+
 	list_del(&edge->vertex_entry);
 	vertex->out_degree--;
 
@@ -227,6 +244,7 @@ out:
 void unix_update_edges(struct unix_sock *receiver)
 {
 	spin_lock(&unix_gc_lock);
+	unix_update_graph(unix_sk(receiver->listener)->vertex);
 	receiver->listener = NULL;
 	spin_unlock(&unix_gc_lock);
 }
@@ -268,6 +286,26 @@ void unix_destroy_fpl(struct scm_fp_list *fpl)
 	unix_free_vertices(fpl);
 }
 
+static bool unix_scc_cyclic(struct list_head *scc)
+{
+	struct unix_vertex *vertex;
+	struct unix_edge *edge;
+
+	/* SCC containing multiple vertices ? */
+	if (!list_is_singular(scc))
+		return true;
+
+	vertex = list_first_entry(scc, typeof(*vertex), scc_entry);
+
+	/* Self-reference or a embryo-listener circle ? */
+	list_for_each_entry(edge, &vertex->edges, vertex_entry) {
+		if (unix_edge_successor(edge) == vertex)
+			return true;
+	}
+
+	return false;
+}
+
 static LIST_HEAD(unix_visited_vertices);
 static unsigned long unix_vertex_grouped_index = UNIX_VERTEX_INDEX_MARK2;
 
@@ -353,6 +391,9 @@ prev_vertex:
 			vertex->index = unix_vertex_grouped_index;
 		}
 
+		if (!unix_graph_maybe_cyclic)
+			unix_graph_maybe_cyclic = unix_scc_cyclic(&scc);
+
 		list_del(&scc);
 	}
 
@@ -363,6 +404,8 @@ prev_vertex:
 
 static void unix_walk_scc(void)
 {
+	unix_graph_maybe_cyclic = false;
+
 	/* Visit every vertex exactly once.
 	 * __unix_walk_scc() moves visited vertices to unix_visited_vertices.
 	 */
@@ -524,6 +567,9 @@ static void __unix_gc(struct work_struct *work)
 
 	spin_lock(&unix_gc_lock);
 
+	if (!unix_graph_maybe_cyclic)
+		goto skip_gc;
+
 	unix_walk_scc();
 
 	/* First, select candidates for garbage collection.  Only
@@ -617,7 +663,7 @@ static void __unix_gc(struct work_struct *work)
 
 	/* All candidates should have been detached by now. */
 	WARN_ON_ONCE(!list_empty(&gc_candidates));
-
+skip_gc:
 	/* Paired with READ_ONCE() in wait_for_unix_gc(). */
 	WRITE_ONCE(gc_in_progress, false);