syscall: add GidMappingsEnableSetgroups to Linux SysProcAttr

Linux 3.19 made a change in the handling of setgroups and the 'gid_map' file to
address a security issue.
The upshot of the 3.19 changes is that in order to update the 'gid_maps' file,
use of the setgroups() system call in this user namespace must first be disabled
by writing "deny" to one of the /proc/PID/setgroups files for this namespace.

Also added tests for remapping uid_map and gid_map inside new user
namespace.

Fixes #10626

Change-Id: I4d2539acbab741a37092d277e10f31fc39a8feb7
Reviewed-on: https://go-review.googlesource.com/10670
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

src/syscall/exec_linux.go

@@ -33,6 +33,11 @@ type SysProcAttr struct {
 	Cloneflags  uintptr        // Flags for clone calls (Linux only)
 	UidMappings []SysProcIDMap // User ID mappings for user namespaces.
 	GidMappings []SysProcIDMap // Group ID mappings for user namespaces.
+	// GidMappingsEnableSetgroups enabling setgroups syscall.
+	// If false, then setgroups syscall will be disabled for the child process.
+	// This parameter is no-op if GidMappings == nil. Otherwise for unprivileged
+	// users this should be set to false for mappings work.
+	GidMappingsEnableSetgroups bool
 }
 
 // Implemented in runtime package.
@@ -366,6 +371,32 @@ func writeIDMappings(path string, idMap []SysProcIDMap) error {
 	return nil
 }
 
+// writeSetgroups writes to /proc/PID/setgroups "deny" if enable is false
+// and "allow" if enable is true.
+// This is needed since kernel 3.19, because you can't write gid_map without
+// disabling setgroups() system call.
+func writeSetgroups(pid int, enable bool) error {
+	sgf := "/proc/" + itoa(pid) + "/setgroups"
+	fd, err := Open(sgf, O_RDWR, 0)
+	if err != nil {
+		return err
+	}
+
+	var data []byte
+	if enable {
+		data = []byte("allow")
+	} else {
+		data = []byte("deny")
+	}
+
+	if _, err := Write(fd, data); err != nil {
+		Close(fd)
+		return err
+	}
+
+	return Close(fd)
+}
+
 // writeUidGidMappings writes User ID and Group ID mappings for user namespaces
 // for a process and it is called from the parent process.
 func writeUidGidMappings(pid int, sys *SysProcAttr) error {
@@ -377,6 +408,10 @@ func writeUidGidMappings(pid int, sys *SysProcAttr) error {
 	}
 
 	if sys.GidMappings != nil {
+		// If the kernel is too old to support /proc/PID/setgroups, writeSetGroups will return ENOENT; this is OK.
+		if err := writeSetgroups(pid, sys.GidMappingsEnableSetgroups); err != nil && err != ENOENT {
+			return err
+		}
 		gidf := "/proc/" + itoa(pid) + "/gid_map"
 		if err := writeIDMappings(gidf, sys.GidMappings); err != nil {
 			return err

src/syscall/exec_linux_test.go

@@ -0,0 +1,84 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build linux
+
+package syscall_test
+
+import (
+	"os"
+	"os/exec"
+	"strings"
+	"syscall"
+	"testing"
+)
+
+func whoamiCmd(t *testing.T, uid int, setgroups bool) *exec.Cmd {
+	if _, err := os.Stat("/proc/self/ns/user"); err != nil {
+		if os.IsNotExist(err) {
+			t.Skip("kernel doesn't support user namespaces")
+		}
+		t.Fatalf("Failed to stat /proc/self/ns/user: %v", err)
+	}
+	cmd := exec.Command("whoami")
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		Cloneflags: syscall.CLONE_NEWUSER,
+		UidMappings: []syscall.SysProcIDMap{
+			{ContainerID: 0, HostID: uid, Size: 1},
+		},
+		GidMappings: []syscall.SysProcIDMap{
+			{ContainerID: 0, HostID: uid, Size: 1},
+		},
+		GidMappingsEnableSetgroups: setgroups,
+	}
+	return cmd
+}
+
+func testNEWUSERRemap(t *testing.T, uid int, setgroups bool) {
+	cmd := whoamiCmd(t, uid, setgroups)
+	out, err := cmd.CombinedOutput()
+	if err != nil {
+		t.Fatalf("Cmd failed with err %v, output: %s", err, out)
+	}
+	sout := strings.TrimSpace(string(out))
+	want := "root"
+	if sout != want {
+		t.Fatalf("whoami = %q; want %q", out, want)
+	}
+}
+
+func TestCloneNEWUSERAndRemapRootDisableSetgroups(t *testing.T) {
+	if os.Getuid() != 0 {
+		t.Skip("skipping root only test")
+	}
+	testNEWUSERRemap(t, 0, false)
+}
+
+func TestCloneNEWUSERAndRemapRootEnableSetgroups(t *testing.T) {
+	if os.Getuid() != 0 {
+		t.Skip("skipping root only test")
+	}
+	testNEWUSERRemap(t, 0, false)
+}
+
+func TestCloneNEWUSERAndRemapNoRootDisableSetgroups(t *testing.T) {
+	if os.Getuid() == 0 {
+		t.Skip("skipping unprivileged user only test")
+	}
+	testNEWUSERRemap(t, os.Getuid(), false)
+}
+
+func TestCloneNEWUSERAndRemapNoRootSetgroupsEnableSetgroups(t *testing.T) {
+	if os.Getuid() == 0 {
+		t.Skip("skipping unprivileged user only test")
+	}
+	cmd := whoamiCmd(t, os.Getuid(), true)
+	err := cmd.Run()
+	if err == nil {
+		t.Skip("probably old kernel without security fix")
+	}
+	if !strings.Contains(err.Error(), "operation not permitted") {
+		t.Fatalf("Unprivileged gid_map rewriting with GidMappingsEnableSetgroups must fail")
+	}
+}