github/containerd
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 2 Wiki Activity

idmapped: use pidfd to avoid pid reuse issue

Browse Source 
It's followup for #5890.

The containerd-shim process depends on the mount package to init rootfs
for container. For the container enable user namespace, the mount
package needs to fork child process to get the brand-new user namespace.
However, there are two reapers in one process (described by the
following list) and there are race-condition cases.

1. mount package
2. sys.Reaper as global one which watch all the SIGCHLD.

=== [kill(2)][kill] the wrong process ===

Currently, we use pipe to ensure that child process is alive. However,
the pide file descriptor can be hold by other process, which the child
process cannot exit by self. We should use [kill(2)][kill] to ensure the
child process. But we might kill the wrong process if the child process
might be reaped by containerd-shim and the PID might be reused by other
process.

=== [waitid(2)][waitid] on the wrong child process ===

```
containerd-shim process:

Goroutine 1(GetUsernsFD):                   Goroutine 2(Reaper)

1. Ready to wait for child process X

                                            2. Received SIGCHLD from X

                                            3. Reaped the zombie child process X

                                            (X has been reused by other child process)

4. Wait on process X

The goroutine 1 will be stuck until the process X has been terminated.
```

=== open `/proc/X/ns/user` on the wrong child process ===

There is also pid-reused risk between opening `/proc/$pid/ns/user` and
writing `/proc/$pid/u[g]id_map`.

```
containerd-shim process:

Goroutine 1(GetUsernsFD):                   Goroutine 2(Reaper)

1. Fork child process X

2. Write /proc/X/uid_map,gid_map

                                            3. Received SIGCHLD from X

                                            4. Reaped the zombie child process X

                                            (X has been reused by other process)

5. Open /proc/X/ns/user file as usernsFD

The usernsFD links to the wrong X!!!
```

In order to fix the race-condition, we should use [CLONE_PIDFD][clone2] (Since
Linux v5.2).

When we fork child process `X`, the kernel will return a process file
descriptor `X_PIDFD` referencing to child process `X`. With the pidfd, we can
use [pidfd_send_signal(2)][pidfd_send_signal] (Since Linux v5.1)
to send signal(0) to ensure the child process `X` is alive. If the `X` has
terminated and its PID has been recycled for another process. The
pidfd_send_signal fails with the error ESRCH.

Therefore, we can open `/proc/X/{ns/user,uid_map,gid_map}` file
descriptors as first and then use pidfd_send_signal to check the process
is still alive. If so, we can ensure the file descriptors are valid and
reference to the child process `X`. Even if the `X` PID has been reused
after pidfd_send_signal call, the file descriptors are still valid.

```code
X, pidfd = clone2(CLONE_PIDFD)

usernsFD = open /proc/X/ns/user
uidmapFD = open /proc/X/uid_map
gidmapFD = open /proc/X/gid_map

pidfd_send_signal pidfd, signal(0)
  return err if no such process

== When we arrive here, we can ensure usernsFD/uidmapFD/gidmapFD are correct
== even if X has been reused after pidfd_send_signal call.

update uid/gid mapping by uidmapFD/gidmapFD
return usernsFD
```

And the [waitid(2)][waitid] also supports pidfd type (Since Linux 5.4).
We can use pidfd type waitid to ensure we are waiting for the correct
process. All the PID related race-condition issues can be resolved by
pidfd.

```bash
➜  mount git:(followup-idmapped) pwd
/home/fuwei/go/src/github.com/containerd/containerd/mount

➜  mount git:(followup-idmapped) sudo go test -test.root -run TestGetUsernsFD -count=1000 -failfast -p 100  ./...
PASS
ok      github.com/containerd/containerd/mount  3.446s
```

[kill]: <https://man7.org/linux/man-pages/man2/kill.2.html>
[clone2]: <https://man7.org/linux/man-pages/man2/clone.2.html>
[pidfd_send_signal]: <https://man7.org/linux/man-pages/man2/pidfd_send_signal.2.html>
[waitid]: <https://man7.org/linux/man-pages/man2/waitid.2.html>

Signed-off-by: Wei Fu <fuweid89@gmail.com>
This commit is contained in:
Wei Fu 2023-10-12 21:58:56 +08:00
parent 420503072e
commit 3742f7f0db
5 changed files with 406 additions and 99 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

172
mount/mount_idmapped_linux.go
View File

@ -19,15 +19,15 @@ package mount
import (
"fmt"
"os"
"runtime"
"strconv"
"strings"
"sync"
"syscall"
"unsafe"
"golang.org/x/sys/unix"
"github.com/containerd/containerd/sys"
"github.com/sirupsen/logrus"
)
// TODO: Support multiple mappings in future
@ -36,21 +36,26 @@ func parseIDMapping(mapping string) ([]syscall.SysProcIDMap, error) {
if len(parts) != 3 {
return nil, fmt.Errorf("user namespace mappings require the format `container-id:host-id:size`")
}
cID, err := strconv.Atoi(parts[0])
if err != nil {
return nil, fmt.Errorf("invalid container id for user namespace remapping, %w", err)
}
hID, err := strconv.Atoi(parts[1])
if err != nil {
return nil, fmt.Errorf("invalid host id for user namespace remapping, %w", err)
}
size, err := strconv.Atoi(parts[2])
if err != nil {
return nil, fmt.Errorf("invalid size for user namespace remapping, %w", err)
}
if cID != 0 || hID < 0 || size < 0 {
return nil, fmt.Errorf("invalid mapping %s, all IDs and size must be positive integers (container ID of 0 is only supported)", mapping)
if cID < 0 || hID < 0 || size < 0 {
return nil, fmt.Errorf("invalid mapping %s, all IDs and size must be positive integers", mapping)
}
return []syscall.SysProcIDMap{
{
ContainerID: cID,
@ -87,80 +92,121 @@ func IDMapMount(source, target string, usernsFd int) (err error) {
return nil
}
// GetUsernsFD forks the current process and creates a user namespace using the specified
// mappings.
//
// It returns:
// 1. The file descriptor of the /proc/[pid]/ns/user of the newly
// created mapping.
// 2. "Clean up" function that should be called once user namespace
// file descriptor is no longer needed.
// 3. Usual error.
func GetUsernsFD(uidmap, gidmap string) (_ int, _ func(), err error) {
var (
usernsFile *os.File
pipeMap [2]int
pid uintptr
errno syscall.Errno
uidMaps, gidMaps []syscall.SysProcIDMap
)
if uidMaps, err = parseIDMapping(uidmap); err != nil {
return -1, nil, err
}
if gidMaps, err = parseIDMapping(gidmap); err != nil {
return -1, nil, err
// GetUsernsFD forks the current process and creates a user namespace using
// the specified mappings.
func GetUsernsFD(uidmap, gidmap string) (_usernsFD *os.File, _ error) {
uidMaps, err := parseIDMapping(uidmap)
if err != nil {
return nil, err
}
syscall.ForkLock.Lock()
if err = syscall.Pipe2(pipeMap[:], syscall.O_CLOEXEC); err != nil {
syscall.ForkLock.Unlock()
return -1, nil, err
gidMaps, err := parseIDMapping(gidmap)
if err != nil {
return nil, err
}
return getUsernsFD(uidMaps, gidMaps)
}
pid, errno = sys.ForkUserns(pipeMap)
syscall.ForkLock.Unlock()
func getUsernsFD(uidMaps, gidMaps []syscall.SysProcIDMap) (_usernsFD *os.File, retErr error) {
runtime.LockOSThread()
defer runtime.UnlockOSThread()
pid, pidfd, errno := sys.ForkUserns()
if errno != 0 {
syscall.Close(pipeMap[0])
syscall.Close(pipeMap[1])
return -1, nil, errno
return nil, errno
}
syscall.Close(pipeMap[0])
pidFD := os.NewFile(pidfd, "pidfd")
defer func() {
unix.PidfdSendSignal(int(pidFD.Fd()), unix.SIGKILL, nil, 0)
writeMappings := func(fname string, idmap []syscall.SysProcIDMap) error {
pidfdWaitid(pidFD)
pidFD.Close()
}()
// NOTE:
//
// The usernsFD will hold the userns reference in kernel. Even if the
// child process is reaped, the usernsFD is still valid.
usernsFD, err := os.Open(fmt.Sprintf("/proc/%d/ns/user", pid))
if err != nil {
return nil, fmt.Errorf("failed to get userns file descriptor for /proc/%d/user/ns: %w", pid, err)
}
defer func() {
if retErr != nil {
usernsFD.Close()
}
}()
uidmapFile, err := os.OpenFile(fmt.Sprintf("/proc/%d/%s", pid, "uid_map"), os.O_WRONLY, 0600)
if err != nil {
return nil, fmt.Errorf("failed to open /proc/%d/uid_map: %w", pid, err)
}
defer uidmapFile.Close()
gidmapFile, err := os.OpenFile(fmt.Sprintf("/proc/%d/%s", pid, "gid_map"), os.O_WRONLY, 0600)
if err != nil {
return nil, fmt.Errorf("failed to open /proc/%d/gid_map: %w", pid, err)
}
defer gidmapFile.Close()
testHookKillChildBeforePidfdSendSignal(pid, pidFD)
// Ensure the child process is still alive. If the err is ESRCH, we
// should return error because we can't guarantee the usernsFD and
// u[g]idmapFile are valid. It's safe to return error and retry.
if err := unix.PidfdSendSignal(int(pidFD.Fd()), 0, nil, 0); err != nil {
return nil, fmt.Errorf("failed to ensure child process is alive: %w", err)
}
testHookKillChildAfterPidfdSendSignal(pid, pidFD)
// NOTE:
//
// The u[g]id_map file descriptor is still valid if the child process
// is reaped.
writeMappings := func(f *os.File, idmap []syscall.SysProcIDMap) error {
mappings := ""
for _, m := range idmap {
mappings = fmt.Sprintf("%d %d %d\n", m.ContainerID, m.HostID, m.Size)
}
return os.WriteFile(fmt.Sprintf("/proc/%d/%s", pid, fname), []byte(mappings), 0600)
mappings = fmt.Sprintf("%s%d %d %d\n", mappings, m.ContainerID, m.HostID, m.Size)
}
cleanUpChild := func() {
sync := sys.ProcSyncExit
if _, _, errno := syscall.Syscall6(syscall.SYS_WRITE, uintptr(pipeMap[1]), uintptr(unsafe.Pointer(&sync)), unsafe.Sizeof(sync), 0, 0, 0); errno != 0 {
logrus.WithError(errno).Warnf("failed to sync with child (ProcSyncExit)")
_, err := f.Write([]byte(mappings))
if err1 := f.Close(); err1 != nil && err == nil {
err = err1
}
syscall.Close(pipeMap[1])
if _, err := unix.Wait4(int(pid), nil, 0, nil); err != nil {
logrus.WithError(err).Warnf("failed to wait for child process; the SIGHLD might be received by shim reaper")
}
}
defer cleanUpChild()
if err := writeMappings("uid_map", uidMaps); err != nil {
return -1, nil, err
}
if err := writeMappings("gid_map", gidMaps); err != nil {
return -1, nil, err
return err
}
if usernsFile, err = os.Open(fmt.Sprintf("/proc/%d/ns/user", pid)); err != nil {
return -1, nil, fmt.Errorf("failed to get user ns file descriptor for - /proc/%d/user/ns: %w", pid, err)
if err := writeMappings(uidmapFile, uidMaps); err != nil {
return nil, fmt.Errorf("failed to write uid_map: %w", err)
}
return int(usernsFile.Fd()), func() {
usernsFile.Close()
}, nil
if err := writeMappings(gidmapFile, gidMaps); err != nil {
return nil, fmt.Errorf("failed to write gid_map: %w", err)
}
return usernsFD, nil
}
func pidfdWaitid(pidFD *os.File) error {
// https://elixir.bootlin.com/linux/v5.4.258/source/include/uapi/linux/wait.h#L20
const PPidFD = 3
var e syscall.Errno
for {
_, _, e = syscall.Syscall6(syscall.SYS_WAITID, PPidFD, pidFD.Fd(), 0, syscall.WEXITED, 0, 0)
if e != syscall.EINTR {
break
}
}
return e
}
var (
testHookLock sync.Mutex
testHookKillChildBeforePidfdSendSignal = func(_pid uintptr, _pidFD *os.File) {}
testHookKillChildAfterPidfdSendSignal = func(_pid uintptr, _pidFD *os.File) {}
)

235
mount/mount_idmapped_linux_test.go Normal file
View File

@ -0,0 +1,235 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package mount
import (
"fmt"
"os"
"path/filepath"
"syscall"
"testing"
"github.com/containerd/continuity/testutil"
"github.com/stretchr/testify/require"
"golang.org/x/sys/unix"
)
var (
testUIDMaps = []syscall.SysProcIDMap{
{ContainerID: 1000, HostID: 0, Size: 100},
{ContainerID: 5000, HostID: 2000, Size: 100},
{ContainerID: 10000, HostID: 3000, Size: 100},
}
testGIDMaps = []syscall.SysProcIDMap{
{ContainerID: 1000, HostID: 0, Size: 100},
{ContainerID: 5000, HostID: 2000, Size: 100},
{ContainerID: 10000, HostID: 3000, Size: 100},
}
)
func TestGetUsernsFD(t *testing.T) {
testutil.RequiresRoot(t)
t.Run("basic", testGetUsernsFDBasic)
t.Run("when kill child process before write u[g]id maps", testGetUsernsFDKillChildWhenWriteUGIDMaps)
t.Run("when kill child process after open u[g]id_map file", testGetUsernsFDKillChildAfterOpenUGIDMapFiles)
}
func testGetUsernsFDBasic(t *testing.T) {
for idx, tc := range []struct {
uidMaps string
gidMaps string
hasErr bool
}{
{
uidMaps: "0:1000:100",
gidMaps: "0:1000:100",
hasErr: false,
},
{
uidMaps: "100:1000:100",
gidMaps: "0:-1:100",
hasErr: true,
},
{
uidMaps: "100:1000:100",
gidMaps: "-1:1000:100",
hasErr: true,
},
{
uidMaps: "100:1000:100",
gidMaps: "0:1000:-1",
hasErr: true,
},
} {
t.Run(fmt.Sprintf("#%v", idx), func(t *testing.T) {
_, err := GetUsernsFD(tc.uidMaps, tc.gidMaps)
if tc.hasErr {
require.Error(t, err)
} else {
require.NoError(t, err)
}
})
}
}
func testGetUsernsFDKillChildWhenWriteUGIDMaps(t *testing.T) {
hookFunc := func(reap bool) func(uintptr, *os.File) {
return func(_pid uintptr, pidFD *os.File) {
err := unix.PidfdSendSignal(int(pidFD.Fd()), unix.SIGKILL, nil, 0)
require.NoError(t, err)
if reap {
pidfdWaitid(pidFD)
}
}
}
for _, tcReap := range []bool{true, false} {
t.Run(fmt.Sprintf("#reap=%v", tcReap), func(t *testing.T) {
updateTestHookKillForGetUsernsFD(t, nil, hookFunc(tcReap))
usernsFD, err := getUsernsFD(testUIDMaps, testGIDMaps)
require.NoError(t, err)
defer usernsFD.Close()
srcDir, checkFunc := initIDMappedChecker(t, testUIDMaps, testGIDMaps)
destDir := t.TempDir()
defer func() {
require.NoError(t, UnmountAll(destDir, 0))
}()
err = IDMapMount(srcDir, destDir, int(usernsFD.Fd()))
usernsFD.Close()
require.NoError(t, err)
checkFunc(destDir)
})
}
}
func testGetUsernsFDKillChildAfterOpenUGIDMapFiles(t *testing.T) {
hookFunc := func(reap bool) func(uintptr, *os.File) {
return func(_pid uintptr, pidFD *os.File) {
err := unix.PidfdSendSignal(int(pidFD.Fd()), unix.SIGKILL, nil, 0)
require.NoError(t, err)
if reap {
pidfdWaitid(pidFD)
}
}
}
for _, tc := range []struct {
reap bool
expected error
}{
{
reap: false,
expected: nil,
},
{
reap: true,
expected: syscall.ESRCH,
},
} {
t.Run(fmt.Sprintf("#reap=%v", tc.reap), func(t *testing.T) {
updateTestHookKillForGetUsernsFD(t, hookFunc(tc.reap), nil)
usernsFD, err := getUsernsFD(testUIDMaps, testGIDMaps)
if tc.expected != nil {
require.Error(t, tc.expected, err)
return
}
require.NoError(t, err)
defer usernsFD.Close()
srcDir, checkFunc := initIDMappedChecker(t, testUIDMaps, testGIDMaps)
destDir := t.TempDir()
defer func() {
require.NoError(t, UnmountAll(destDir, 0))
}()
err = IDMapMount(srcDir, destDir, int(usernsFD.Fd()))
usernsFD.Close()
require.NoError(t, err)
checkFunc(destDir)
})
}
}
func updateTestHookKillForGetUsernsFD(t *testing.T, newBeforeFunc, newAfterFunc func(uintptr, *os.File)) {
testHookLock.Lock()
oldBefore := testHookKillChildBeforePidfdSendSignal
oldAfter := testHookKillChildAfterPidfdSendSignal
t.Cleanup(func() {
testHookKillChildBeforePidfdSendSignal = oldBefore
testHookKillChildAfterPidfdSendSignal = oldAfter
testHookLock.Unlock()
})
if newBeforeFunc != nil {
testHookKillChildBeforePidfdSendSignal = newBeforeFunc
}
if newAfterFunc != nil {
testHookKillChildAfterPidfdSendSignal = newAfterFunc
}
}
func initIDMappedChecker(t *testing.T, uidMaps, gidMaps []syscall.SysProcIDMap) (_srcDir string, _verifyFunc func(destDir string)) {
testutil.RequiresRoot(t)
srcDir := t.TempDir()
require.Equal(t, len(uidMaps), len(gidMaps))
for idx := range uidMaps {
file := filepath.Join(srcDir, fmt.Sprintf("%v", idx))
f, err := os.Create(file)
require.NoError(t, err, fmt.Sprintf("create file %s", file))
defer f.Close()
uid, gid := uidMaps[idx].ContainerID, gidMaps[idx].ContainerID
err = f.Chown(uid, gid)
require.NoError(t, err, fmt.Sprintf("chown %v:%v for file %s", uid, gid, file))
}
return srcDir, func(destDir string) {
for idx := range uidMaps {
file := filepath.Join(destDir, fmt.Sprintf("%v", idx))
f, err := os.Open(file)
require.NoError(t, err, fmt.Sprintf("open file %s", file))
defer f.Close()
stat, err := f.Stat()
require.NoError(t, err, fmt.Sprintf("stat file %s", file))
sysStat := stat.Sys().(*syscall.Stat_t)
uid, gid := uidMaps[idx].HostID, gidMaps[idx].HostID
require.Equal(t, uint32(uid), sysStat.Uid, fmt.Sprintf("check file %s uid", file))
require.Equal(t, uint32(gid), sysStat.Gid, fmt.Sprintf("check file %s gid", file))
t.Logf("IDMapped File %s uid=%v, gid=%v", file, uid, gid)
}
}
}

13
mount/mount_linux.go
View File

@ -93,26 +93,23 @@ func (m *Mount) mount(target string) (err error) {
var (
chdir string
recalcOpt bool
usernsFd int
usernsFd *os.File
options = m.Options
)
opt := parseMountOptions(options)
// The only remapping of both GID and UID is supported
if opt.uidmap != "" && opt.gidmap != "" {
var (
childProcCleanUp func()
)
if usernsFd, childProcCleanUp, err = GetUsernsFD(opt.uidmap, opt.gidmap); err != nil {
if usernsFd, err = GetUsernsFD(opt.uidmap, opt.gidmap); err != nil {
return err
}
defer childProcCleanUp()
defer usernsFd.Close()
// overlay expects lowerdir's to be remapped instead
if m.Type == "overlay" {
var (
userNsCleanUp func()
)
options, userNsCleanUp, err = prepareIDMappedOverlay(usernsFd, options)
options, userNsCleanUp, err = prepareIDMappedOverlay(int(usernsFd.Fd()), options)
defer userNsCleanUp()
if err != nil {
@ -196,7 +193,7 @@ func (m *Mount) mount(target string) (err error) {
// remap non-overlay mount point
if opt.uidmap != "" && opt.gidmap != "" && m.Type != "overlay" {
if err := IDMapMount(target, target, usernsFd); err != nil {
if err := IDMapMount(target, target, int(usernsFd.Fd())); err != nil {
return err
}
}

6
snapshots/overlay/overlayutils/check.go
View File

@ -255,13 +255,13 @@ func SupportsIDMappedMounts() (bool, error) {
uidmap := fmt.Sprintf("%d:%d:%d", uidMap.ContainerID, uidMap.HostID, uidMap.Size)
gidmap := fmt.Sprintf("%d:%d:%d", gidMap.ContainerID, gidMap.HostID, gidMap.Size)
usernsFd, childProcCleanUp, err := mount.GetUsernsFD(uidmap, gidmap)
usernsFd, err := mount.GetUsernsFD(uidmap, gidmap)
if err != nil {
return false, err
}
defer childProcCleanUp()
defer usernsFd.Close()
if err = mount.IDMapMount(lowerDir, lowerDir, usernsFd); err != nil {
if err = mount.IDMapMount(lowerDir, lowerDir, int(usernsFd.Fd())); err != nil {
return false, fmt.Errorf("failed to remap lowerdir %s: %w", lowerDir, err)
}
defer func() {

79
sys/userns_unsafe_linux.go
View File

@ -22,44 +22,73 @@ import (
"unsafe"
)
// ProcSyncType is used for synchronization
// between parent and child processes.
type ProcSyncType uint8
const (
// ProcSyncExit tells child "it's time to exit".
ProcSyncExit ProcSyncType = 0x1
)
// ForkUserns is to fork child process with user namespace. It returns child
// process's pid and pidfd reference to the child process.
//
// Precondition: The runtime OS thread must be locked, which is GO runtime
// requirement.
//
// Beside this, the child process sets PR_SET_PDEATHSIG with SIGKILL so that
// the parent process's OS thread must be locked. Otherwise, the exit event of
// parent process's OS thread will send kill signal to the child process,
// even if parent process is still running.
//
//go:norace
//go:noinline
func ForkUserns(pipeMap [2]int) (pid uintptr, errno syscall.Errno) {
var sync ProcSyncType
func ForkUserns() (_pid uintptr, _pidfd uintptr, _ syscall.Errno) {
var (
pidfd uintptr
pid, ppid uintptr
err syscall.Errno
)
ppid, _, err = syscall.RawSyscall(uintptr(syscall.SYS_GETPID), 0, 0, 0)
if err != 0 {
return 0, 0, err
}
beforeFork()
if runtime.GOARCH == "s390x" {
pid, _, errno = syscall.RawSyscall6(uintptr(syscall.SYS_CLONE), 0, syscall.CLONE_NEWUSER|uintptr(syscall.SIGCHLD), 0, 0, 0, 0)
// NOTE:
//
// On the s390 architectures, the order of the first two
// arguments is reversed.
//
// REF: https://man7.org/linux/man-pages/man2/clone.2.html
pid, _, err = syscall.RawSyscall(syscall.SYS_CLONE,
0,
uintptr(syscall.CLONE_NEWUSER|syscall.SIGCHLD|syscall.CLONE_PIDFD),
uintptr(unsafe.Pointer(&pidfd)),
)
} else {
pid, _, errno = syscall.RawSyscall6(uintptr(syscall.SYS_CLONE), syscall.CLONE_NEWUSER|uintptr(syscall.SIGCHLD), 0, 0, 0, 0, 0)
pid, _, err = syscall.RawSyscall(syscall.SYS_CLONE,
uintptr(syscall.CLONE_NEWUSER|syscall.SIGCHLD|syscall.CLONE_PIDFD),
0,
uintptr(unsafe.Pointer(&pidfd)),
)
}
if errno != 0 || pid != 0 {
if err != 0 || pid != 0 {
afterFork()
return pid, errno
return pid, pidfd, err
}
afterForkInChild()
if _, _, errno = syscall.RawSyscall(syscall.SYS_CLOSE, uintptr(pipeMap[1]), 0, 0); errno != 0 {
goto err
}
if _, _, errno = syscall.RawSyscall6(syscall.SYS_PRCTL, syscall.PR_SET_PDEATHSIG, uintptr(syscall.SIGKILL), 0, 0, 0, 0); errno != 0 {
goto err
}
// wait for parent's signal
if _, _, errno = syscall.RawSyscall6(syscall.SYS_READ, uintptr(pipeMap[0]), uintptr(unsafe.Pointer(&sync)), unsafe.Sizeof(sync), 0, 0, 0); errno != 0 || sync != ProcSyncExit {
if _, _, err = syscall.RawSyscall(syscall.SYS_PRCTL, syscall.PR_SET_PDEATHSIG, uintptr(syscall.SIGKILL), 0); err != 0 {
goto err
}
pid, _, err = syscall.RawSyscall(syscall.SYS_GETPPID, 0, 0, 0)
if err != 0 {
goto err
}
// exit if re-parent
if pid != ppid {
goto err
}
_, _, err = syscall.RawSyscall(syscall.SYS_PPOLL, 0, 0, 0)
err:
syscall.RawSyscall6(syscall.SYS_EXIT, uintptr(errno), 0, 0, 0, 0, 0)
syscall.RawSyscall(syscall.SYS_EXIT, uintptr(err), 0, 0)
panic("unreachable")
}