podman restart with port bindings results in 'port in use' errors

[baude]

Sometimes when restarting a container with podman restart, it results in a port already in use error. It seems like you need to have an application in the container that takes a little longer to execute to trip it more reliably.

Reproducer

$ sudo podman run -dt --name foo -p 80:80 alpine_nginx
b7fc4ef1bc8a130ec0dec234c59bc11a7d6e88b24e48d04e8afede3b832db3db
$ sudo podman restart foo
Error: cannot listen on the TCP port: listen tcp4 :80: bind: address already in use

split from containers/podman#9465

[rhatdan]

Should restart also imbed a podman wait. Probably does podman stop; podman start. Should do
podman stop; podman wait; podman start. Since podman container cleanup has to happen after the stop completes, and this is where the race probably is.
Have not looked at the code, but that's my guess.

[mheon]

It calls cleanup itself, automatically - we actually SIGKILL conmon so the cleanup process never fires (for unrelated reasons - there was a race where conmon took too long to die and was holding ports we needed).