plugin/cache: Add refresh mode setting to serve_stale (#5131)

This PR adds an optional REFRESH_MODE parameter on the serve_stale configuration directive of the
cache plugin, which verifies that the upstream is still unavailable before returning stale entries.

Signed-off-by: Antoine Tollenaere <atollena@gmail.com>

plugin/cache/cache_test.go

@@ -266,7 +266,7 @@ func TestServeFromStaleCache(t *testing.T) {
 	req.SetQuestion("cached.org.", dns.TypeA)
 	ctx := context.TODO()
 
-	// Cache example.org.
+	// Cache cached.org. with 60s TTL
 	rec := dnstest.NewRecorder(&test.ResponseWriter{})
 	c.staleUpTo = 1 * time.Hour
 	c.ServeDNS(ctx, rec, req)
@@ -304,6 +304,80 @@ func TestServeFromStaleCache(t *testing.T) {
 	}
 }
 
+func TestServeFromStaleCacheFetchVerify(t *testing.T) {
+	c := New()
+	c.Next = ttlBackend(120)
+
+	req := new(dns.Msg)
+	req.SetQuestion("cached.org.", dns.TypeA)
+	ctx := context.TODO()
+
+	// Cache cached.org. with 120s TTL
+	rec := dnstest.NewRecorder(&test.ResponseWriter{})
+	c.staleUpTo = 1 * time.Hour
+	c.verifyStale = true
+	c.ServeDNS(ctx, rec, req)
+	if c.pcache.Len() != 1 {
+		t.Fatalf("Msg with > 0 TTL should have been cached")
+	}
+
+	tests := []struct {
+		name          string
+		upstreamRCode int
+		upstreamTtl   int
+		futureMinutes int
+		expectedRCode int
+		expectedTtl   int
+	}{
+		// After 1 minutes of initial TTL, we should see a cached response
+		{"cached.org.", dns.RcodeSuccess, 200, 1, dns.RcodeSuccess, 60}, // ttl = 120 - 60 -- not refreshed
+
+		// After the 2 more minutes, we should see upstream responses because upstream is available
+		{"cached.org.", dns.RcodeSuccess, 200, 3, dns.RcodeSuccess, 200},
+
+		// After the TTL expired, if the server fails we should get the cached entry
+		{"cached.org.", dns.RcodeServerFailure, 200, 7, dns.RcodeSuccess, 0},
+
+		// After 1 more minutes, if the server serves nxdomain we should see them (despite being within the serve stale period)
+		{"cached.org.", dns.RcodeNameError, 150, 8, dns.RcodeNameError, 150},
+	}
+
+	for i, tt := range tests {
+		rec := dnstest.NewRecorder(&test.ResponseWriter{})
+		c.now = func() time.Time { return time.Now().Add(time.Duration(tt.futureMinutes) * time.Minute) }
+
+		if tt.upstreamRCode == dns.RcodeSuccess {
+			c.Next = ttlBackend(tt.upstreamTtl)
+		} else if tt.upstreamRCode == dns.RcodeServerFailure {
+			// Make upstream fail, should now rely on cache during the c.staleUpTo period
+			c.Next = servFailBackend(tt.upstreamTtl)
+		} else if tt.upstreamRCode == dns.RcodeNameError {
+			c.Next = nxDomainBackend(tt.upstreamTtl)
+		} else {
+			t.Fatal("upstream code not implemented")
+		}
+
+		r := req.Copy()
+		r.SetQuestion(tt.name, dns.TypeA)
+		ret, _ := c.ServeDNS(ctx, rec, r)
+		if ret != tt.expectedRCode {
+			t.Errorf("Test %d: expected rcode=%v, got rcode=%v", i, tt.expectedRCode, ret)
+			continue
+		}
+		if ret == dns.RcodeSuccess {
+			recTtl := rec.Msg.Answer[0].Header().Ttl
+			if tt.expectedTtl != int(recTtl) {
+				t.Errorf("Test %d: expected TTL=%d, got TTL=%d", i, tt.expectedTtl, recTtl)
+			}
+		} else if ret == dns.RcodeNameError {
+			soaTtl := rec.Msg.Ns[0].Header().Ttl
+			if tt.expectedTtl != int(soaTtl) {
+				t.Errorf("Test %d: expected TTL=%d, got TTL=%d", i, tt.expectedTtl, soaTtl)
+			}
+		}
+	}
+}
+
 func TestNegativeStaleMaskingPositiveCache(t *testing.T) {
 	c := New()
 	c.staleUpTo = time.Minute * 10
@@ -454,6 +528,20 @@ func ttlBackend(ttl int) plugin.Handler {
 	})
 }
 
+func servFailBackend(ttl int) plugin.Handler {
+	return plugin.HandlerFunc(func(ctx context.Context, w dns.ResponseWriter, r *dns.Msg) (int, error) {
+		m := new(dns.Msg)
+		m.SetReply(r)
+		m.Response, m.RecursionAvailable = true, true
+
+		m.Ns = []dns.RR{test.SOA(fmt.Sprintf("example.org. %d IN	SOA	sns.dns.icann.org. noc.dns.icann.org. 2016082540 7200 3600 1209600 3600", ttl))}
+
+		m.MsgHdr.Rcode = dns.RcodeServerFailure
+		w.WriteMsg(m)
+		return dns.RcodeServerFailure, nil
+	})
+}
+
 func TestComputeTTL(t *testing.T) {
 	tests := []struct {
 		msgTTL      time.Duration