From e37f4e4f738b605fe5ea1030e39da8d723260007 Mon Sep 17 00:00:00 2001 From: chenjiankun Date: Fri, 18 Mar 2022 11:19:28 +0800 Subject: [PATCH] docker: fix rwlayer umountd after container restart if exit event be handled to slow, then the exit event maybe handled again. we need to add a check after the container lock acquired. --- components/engine/daemon/monitor.go | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/components/engine/daemon/monitor.go b/components/engine/daemon/monitor.go index 0aadf33fd..0bf7f0379 100644 --- a/components/engine/daemon/monitor.go +++ b/components/engine/daemon/monitor.go @@ -60,6 +60,17 @@ func (daemon *Daemon) ProcessEvent(id string, e libcontainerd.EventType, ei libc if int(ei.Pid) == c.Pid { logrus.Infof("handle container %s exit event pid=%d", c.ID, c.Pid) c.Lock() + + // ProcessEvent could be called concurrently, and will execute serial + // for c.Lock(), but int(ei.Pid) == c.Pid has already pass. It will cause + // daemon.Cleanup be called twice. This will make rwlayer umount in docker + // restart, get "fork/exec /proc/self/exe: no such file or directory" err. + // Adding this under c.Lock(), could avaid daemon.Cleanup be called again. + if c.Pid == 0 || int(ei.Pid) != c.Pid { + c.Unlock() + return nil + } + _, _, err := daemon.containerd.DeleteTask(context.Background(), c.ID) if err != nil { logrus.WithError(err).Warnf("failed to delete container %s from containerd", c.ID) -- 2.23.0