mirror of
https://github.com/opencontainers/runc.git
synced 2026-07-11 06:03:57 +08:00
38f7404ba7
When signal installation was moved to a goroutine for performance,
containers that exited quickly could complete before SIGCHLD was
registered, causing runc to hang waiting for the signal.
This fix ensures SIGCHLD is registered immediately in the main thread
before other signals are handled in the goroutine, maintaining performance
while guaranteeing no missed SIGCHLD notifications for fast-exiting
containers.
Reported-by: Ayato Tokubi <atokubi@redhat.com>
Signed-off-by: lifubang <lifubang@acmcoder.com>
(cherry picked from commit 404181e4cc)
Signed-off-by: lifubang <lifubang@acmcoder.com>
152 lines
4.1 KiB
Go
152 lines
4.1 KiB
Go
package main
|
|
|
|
import (
|
|
"os"
|
|
"os/signal"
|
|
|
|
"github.com/opencontainers/runc/libcontainer"
|
|
"github.com/opencontainers/runc/libcontainer/system"
|
|
"github.com/opencontainers/runc/libcontainer/utils"
|
|
|
|
"github.com/sirupsen/logrus"
|
|
"golang.org/x/sys/unix"
|
|
)
|
|
|
|
const signalBufferSize = 2048
|
|
|
|
// newSignalHandler returns a signal handler for processing SIGCHLD and SIGWINCH signals
|
|
// while still forwarding all other signals to the process.
|
|
// If notifySocket is present, use it to read systemd notifications from the container and
|
|
// forward them to notifySocketHost.
|
|
func newSignalHandler(enableSubreaper bool, notifySocket *notifySocket) chan *signalHandler {
|
|
if enableSubreaper {
|
|
// set us as the subreaper before registering the signal handler for the container
|
|
if err := system.SetSubreaper(1); err != nil {
|
|
logrus.Warn(err)
|
|
}
|
|
}
|
|
handler := make(chan *signalHandler)
|
|
|
|
// Ensure that we have a large buffer size so that we do not miss any
|
|
// signals in case we are not processing them fast enough.
|
|
s := make(chan os.Signal, signalBufferSize)
|
|
|
|
// signal.Notify is actually quite expensive, as it has to configure the
|
|
// signal mask and add signal handlers for all signals (all ~65 of them).
|
|
// So, defer this to a background thread while doing the rest of the io/tty
|
|
// setup, except for SIGCHLD which is very important (see #5208).
|
|
signal.Notify(s, unix.SIGCHLD)
|
|
go func() {
|
|
// handle all signals for the process.
|
|
signal.Notify(s)
|
|
handler <- &signalHandler{
|
|
signals: s,
|
|
notifySocket: notifySocket,
|
|
}
|
|
}()
|
|
return handler
|
|
}
|
|
|
|
// exit models a process exit status with the pid and
|
|
// exit status.
|
|
type exit struct {
|
|
pid int
|
|
status int
|
|
}
|
|
|
|
type signalHandler struct {
|
|
signals chan os.Signal
|
|
notifySocket *notifySocket
|
|
}
|
|
|
|
// forward handles the main signal event loop forwarding, resizing, or reaping depending
|
|
// on the signal received.
|
|
func (h *signalHandler) forward(process *libcontainer.Process, tty *tty, detach bool) (int, error) {
|
|
// make sure we know the pid of our main process so that we can return
|
|
// after it dies.
|
|
if detach && h.notifySocket == nil {
|
|
return 0, nil
|
|
}
|
|
|
|
pid1, err := process.Pid()
|
|
if err != nil {
|
|
return -1, err
|
|
}
|
|
|
|
if h.notifySocket != nil {
|
|
if detach {
|
|
_ = h.notifySocket.run(pid1)
|
|
return 0, nil
|
|
}
|
|
_ = h.notifySocket.run(os.Getpid())
|
|
go func() { _ = h.notifySocket.run(0) }()
|
|
}
|
|
|
|
// Perform the initial tty resize. Always ignore errors resizing because
|
|
// stdout might have disappeared (due to races with when SIGHUP is sent).
|
|
_ = tty.resize()
|
|
// Handle and forward signals.
|
|
for s := range h.signals {
|
|
switch s {
|
|
case unix.SIGWINCH:
|
|
// Ignore errors resizing, as above.
|
|
_ = tty.resize()
|
|
case unix.SIGCHLD:
|
|
exits, err := h.reap()
|
|
if err != nil {
|
|
logrus.Error(err)
|
|
}
|
|
for _, e := range exits {
|
|
logrus.WithFields(logrus.Fields{
|
|
"pid": e.pid,
|
|
"status": e.status,
|
|
}).Debug("process exited")
|
|
if e.pid == pid1 {
|
|
// call Wait() on the process even though we already have the exit
|
|
// status because we must ensure that any of the go specific process
|
|
// fun such as flushing pipes are complete before we return.
|
|
_, _ = process.Wait()
|
|
return e.status, nil
|
|
}
|
|
}
|
|
case unix.SIGURG:
|
|
// SIGURG is used by go runtime for async preemptive
|
|
// scheduling, so runc receives it from time to time,
|
|
// and it should not be forwarded to the container.
|
|
// Do nothing.
|
|
default:
|
|
us := s.(unix.Signal)
|
|
logrus.Debugf("forwarding signal %d (%s) to %d", int(us), unix.SignalName(us), pid1)
|
|
if err := process.Signal(s); err != nil {
|
|
logrus.Error(err)
|
|
}
|
|
}
|
|
}
|
|
return -1, nil
|
|
}
|
|
|
|
// reap runs wait4 in a loop until we have finished processing any existing exits
|
|
// then returns all exits to the main event loop for further processing.
|
|
func (h *signalHandler) reap() (exits []exit, err error) {
|
|
var (
|
|
ws unix.WaitStatus
|
|
rus unix.Rusage
|
|
)
|
|
for {
|
|
pid, err := unix.Wait4(-1, &ws, unix.WNOHANG, &rus)
|
|
if err != nil {
|
|
if err == unix.ECHILD {
|
|
return exits, nil
|
|
}
|
|
return nil, err
|
|
}
|
|
if pid <= 0 {
|
|
return exits, nil
|
|
}
|
|
exits = append(exits, exit{
|
|
pid: pid,
|
|
status: utils.ExitStatus(ws),
|
|
})
|
|
}
|
|
}
|