Bug#984495: systemd segfault soon after daemon-reload

[Benjamin Poirier]

On 2021-03-04 13:38 +0100, Michael Biebl wrote:
[...]
> 
> This does look like a reasonable candidate for a buster backport.
> Have you verified, that this commit fixes the crash you are seeing?

I have not been able to reproduce the issue myself.

According to the discussion that led to the upstream commit, this
problem can be triggered by running `systemctl daemon-reload`
concurrently with `systemctl restart <service>`:
https://github.com/systemd/systemd/issues/15356#issue-595874386
https://github.com/systemd/systemd/issues/15356#issuecomment-610405337

According to the logs and the core file that I have, in this case it
seems that it was `systemctl daemon-reload` and `systemctl reload frr`
running around the same time. Somewhat similarly to the example unit
file given on Github, frr.service defines Exec* directives which are
bash scripts that send signals to other processes, but I don't know if
that is really relevant to the problem.

Scott (cc'ed) tried simultaneous restarts of frr and daemon reloads in
their own loops for an hour but wasn't able to trigger the crash. I
tried simultaneous restarts of a simple service with ExecStart, ExecStop
commands similar to what's shown on Github and `systemctl daemon-reload`
but also didn't reproduce a problem.