fix(policy-smart): avoid canceled poison and run SCI on PR only

Author: MaurUppi

.github/workflows/policy-smart-sci.yml

@@ -1,16 +1,6 @@
 name: Policy Smart SCI
 
 on:
-  push:
-    branches:
-      - feat/policy-smart
-    paths:
-      - "common/**/*.go"
-      - "component/**/*.go"
-      - "control/**/*.go"
-      - "go.mod"
-      - "go.sum"
-      - ".github/workflows/policy-smart-sci.yml"
   pull_request:
     types: [opened, synchronize, reopened, ready_for_review]
     paths:

control/tcp.go

@@ -210,6 +210,9 @@ func (c *ControlPlane) RouteDialTcp(ctx context.Context, p *RouteDialParam) (con
 			c.AddTcpConnectionTotal(networkType.StringWithoutDns(), outbound.Name)
 			return conn, nil
 		}
+		if errors.Is(lastErr, context.Canceled) || errors.Is(lastErr, net.ErrClosed) {
+			return nil, lastErr
+		}
 
 		triedDialer = append(triedDialer, d)
 		d.ReportUnavailable(networkType, lastErr)

control/tcp_route_dial_retry_test.go

@@ -192,3 +192,32 @@ func TestRouteDialTcpRetry_Dedup(t *testing.T) {
 		t.Fatalf("dedup should avoid retrying the same dialer, got call count=%d", impl1.CallCount())
 	}
 }
+
+func TestRouteDialTcpRetry_CanceledDoesNotPoisonDialer(t *testing.T) {
+	impl1 := &scriptedDialer{err: context.Canceled}
+	impl2 := &scriptedDialer{conn: newMockConn(false, nil)}
+	d1 := newTestRouteDialer(t, "d1", impl1)
+	d2 := newTestRouteDialer(t, "d2", impl2)
+
+	cp := newTestControlPlaneWithGroup(consts.DialerSelectionPolicy_MinLastLatency, []*dialer.Dialer{d1, d2})
+	networkType := &dialer.NetworkType{L4Proto: consts.L4ProtoStr_TCP, IpVersion: consts.IpVersionStr_4, IsDns: false}
+	d1.MustGetLatencies10(networkType).AppendLatency(10 * time.Millisecond)
+	d2.MustGetLatencies10(networkType).AppendLatency(20 * time.Millisecond)
+	group := cp.outbounds[consts.OutboundUserDefinedMin]
+	group.MustGetAliveDialerSet(networkType).NotifyLatencyChange(d1, true)
+	group.MustGetAliveDialerSet(networkType).NotifyLatencyChange(d2, true)
+
+	_, err := cp.RouteDialTcp(context.Background(), newTestRouteDialParam())
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context canceled, got: %v", err)
+	}
+	if impl1.CallCount() != 1 {
+		t.Fatalf("unexpected dialer-1 call count: got=%d want=1", impl1.CallCount())
+	}
+	if impl2.CallCount() != 0 {
+		t.Fatalf("unexpected dialer-2 call count: got=%d want=0", impl2.CallCount())
+	}
+	if !d1.MustGetAlive(networkType) {
+		t.Fatal("dialer should not be marked unavailable on context canceled")
+	}
+}