interfaces: try larger read from /proc/net/route

Work around https://github.com/google/gvisor/issues/5732
by trying to read /proc/net/route with a larger bufsize if
it fails the first time.

Signed-off-by: Denton Gentry <dgentry@tailscale.com>

net/interfaces/interfaces_default_route_test.go

@@ -2,16 +2,85 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build linux darwin,!redo
+// +build linux,!redo
 
 package interfaces
 
-import "testing"
+import (
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"testing"
+)
 
 func TestDefaultRouteInterface(t *testing.T) {
+	// tests /proc/net/route on the local system, cannot make an assertion about
+	// the correct interface name, but good as a sanity check.
 	v, err := DefaultRouteInterface()
 	if err != nil {
 		t.Fatal(err)
 	}
 	t.Logf("got %q", v)
 }
+
+// test the specific /proc/net/route path as found on Google Cloud Run instances
+func TestGoogleCloudRunDefaultRouteInterface(t *testing.T) {
+	dir := t.TempDir()
+	savedProcNetRoutePath := procNetRoutePath
+	defer func() { procNetRoutePath = savedProcNetRoutePath }()
+	procNetRoutePath = filepath.Join(dir, "CloudRun")
+	buf := []byte("Iface\tDestination\tGateway\tFlags\tRefCnt\tUse\tMetric\tMask\tMTU\tWindow\tIRTT\n" +
+		"eth0\t8008FEA9\t00000000\t0001\t0\t0\t0\t01FFFFFF\t0\t0\t0\n" +
+		"eth1\t00000000\t00000000\t0001\t0\t0\t0\t00000000\t0\t0\t0\n")
+	err := ioutil.WriteFile(procNetRoutePath, buf, 0644)
+	if err != nil {
+		t.Fatal(err)
+	}
+	got, err := DefaultRouteInterface()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if got != "eth1" {
+		t.Fatalf("got %s, want eth1", got)
+	}
+}
+
+// we read chunks of /proc/net/route at a time, test that files longer than the chunk
+// size can be handled.
+func TestExtremelyLongProcNetRoute(t *testing.T) {
+	dir := t.TempDir()
+	savedProcNetRoutePath := procNetRoutePath
+	defer func() { procNetRoutePath = savedProcNetRoutePath }()
+	procNetRoutePath = filepath.Join(dir, "VeryLong")
+	f, err := os.Create(procNetRoutePath)
+	if err != nil {
+		t.Fatal(err)
+	}
+	_, err = f.Write([]byte("Iface\tDestination\tGateway\tFlags\tRefCnt\tUse\tMetric\tMask\tMTU\tWindow\tIRTT\n"))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	for n := 0; n <= 1000; n++ {
+		line := fmt.Sprintf("eth%d\t8008FEA9\t00000000\t0001\t0\t0\t0\t01FFFFFF\t0\t0\t0\n", n)
+		_, err := f.Write([]byte(line))
+		if err != nil {
+			t.Fatal(err)
+		}
+	}
+	_, err = f.Write([]byte("tokenring1\t00000000\t00000000\t0001\t0\t0\t0\t00000000\t0\t0\t0\n"))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	got, err := DefaultRouteInterface()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if got != "tokenring1" {
+		t.Fatalf("got %q, want tokenring1", got)
+	}
+}

net/interfaces/interfaces_linux.go

@@ -135,14 +135,16 @@ func DefaultRouteInterface() (string, error) {
 }
 
 var zeroRouteBytes = []byte("00000000")
+var procNetRoutePath = "/proc/net/route"
 
-func defaultRouteInterfaceProcNet() (string, error) {
-	f, err := os.Open("/proc/net/route")
+func defaultRouteInterfaceProcNetInternal(bufsize int) (string, error) {
+	f, err := os.Open(procNetRoutePath)
 	if err != nil {
 		return "", err
 	}
 	defer f.Close()
-	br := bufio.NewReaderSize(f, 128)
+
+	br := bufio.NewReaderSize(f, bufsize)
 	for {
 		line, err := br.ReadSlice('\n')
 		if err == io.EOF {
@@ -170,7 +172,25 @@ func defaultRouteInterfaceProcNet() (string, error) {
 	}
 
 	return "", errors.New("no default routes found")
+}
 
+func defaultRouteInterfaceProcNet() (string, error) {
+	rc, err := defaultRouteInterfaceProcNetInternal(128)
+	if rc == "" && (err == io.EOF || err == nil) {
+		// https://github.com/google/gvisor/issues/5732
+		// On a regular Linux kernel you can read the first 128 bytes of /proc/net/route,
+		// then come back later to read the next 128 bytes and so on.
+		//
+		// In Google Cloud Run, where /proc/net/route comes from gVisor, you have to
+		// read it all at once. If you read only the first few bytes then the second
+		// read returns 0 bytes no matter how much originally appeared to be in the file.
+		//
+		// At the time of this writing (Mar 2021) Google Cloud Run has eth0 and eth1
+		// with a 384 byte /proc/net/route. We allocate a large buffer to ensure we'll
+		// read it all in one call.
+		return defaultRouteInterfaceProcNetInternal(4096)
+	}
+	return rc, err
 }
 
 // defaultRouteInterfaceAndroidIPRoute tries to find the machine's default route interface name