Bug#1112535: /usr/lib/systemd/systemd-networkd: segfault after upgrade to 257.8-1~deb13u1, no connectivity after reboot

[Luca Boccassi]

Contron: fixed -1 257.9-1~deb13u1
Control: fixed -1 258~rc3-1
Control: tags -1 pending

On Mon, 8 Sept 2025 at 07:12, Jonathan Wiltshire <jmw at debian.org> wrote:
>
> Hi,
>
> On Sun, Sep 07, 2025 at 11:58:46PM +0100, Luca Boccassi wrote:
> > Control: severity -1 minor
> >
> > On Sun, 07 Sep 2025 23:02:13 +0100 "Adam D. Barratt"
> > <adam at adam-barratt.org.uk> wrote:
> > > On Sun, 2025-09-07 at 22:52 +0100, Tim Small wrote:
> > > > This was reported a good week before Debian 13.1 went out, so I'm a
> > > > bit surprised it still went into that?
> > >
> > > It was reported at a non release critical severity, tagged "moreinfo"
> > > and neither the submitter nor the maintainer (nor anyone else) raised
> > > it to the Release Team's attention.
> >
> > There was absolutely no need to disturb RT and waste your time, as you
> > have much more important things to take care of,
>
> For future reference, a highly impactful issue in something we're about to
> release (even if optional or esoteric) is absolutely something SRMs want to
> know about in the week ahead. It doesn't have to be a drama, just please
> draw our attention to it so we can keep an eye on developments. We would
> rather know and disregard than not know at all.
>
> > as this is just a
> > minor issue with a particular corner case of a custom config of an
> > optional component. Anybody who is unable to deal with that should just
> > stick to the default Debian components.
>
> This doesn't seem right. If the upstream test case is correct, the broken
> configuration looks very much like example #6 in systemd.network(5). There
> are no warnings that this might be experimental or unsupported.
> systemd-networkd is called out as "The modern network configuration without
> GUI" in the Debian reference. VLAN-aware bridges are absolutely a common
> production use in dense environments. A reasonable user would expect this
> to work without issue.
>
> Whether you intended it or not you're now maintaining a critical component
> for a lot of people in a production scenario.

"highly impactful" is in the eye of the beholder. It would have never
crossed my mind to waste your limited time with something like this,
as your plates are already full as is, and this is such a niche corner
case that there wasn't even test coverage for it, despite networkd's
code coverage sitting at ~80%.
Regardless of what one's opinions may be, the fact is that the default
in Debian is network-manager for desktops and ifupdown for the rest,
so anybody demanding nothing less than perfection would do well to
stick to those defaults instead of being adventurous and then throwing
abuse when experimental things don't work.

> > The next stable update in ~2
> > months will contain a fix.
>
> I think we need an accelerated stable-updates fix much sooner than that
> please. The upstream patch is one line.

Last week in Salsa I had already prepared the next LTS update, which
contains this and other more important bug fixes, waiting for p-u to
unlock after the weekend release as usual to upload it. Of course in
the end I didn't upload it today, because while it is expected of this
project to tolerate self-entitled random users lashing out with,
frankly, disgusting messages at volunteers working for them for free
(but never enough), it is jaw-dropping to see project members joining
in and throwing around threats because they *disagree on some bug
triaging*, and incredibly demotivating. One would think project
members would know better than using threats and insults and absurd
hyperboles to try and convince volunteers to do free work for them,
but alas.

Once things calm down to a sensible level I'll find the energy to
upload the already tagged branch to p-u. The 13.2 release is scheduled
for 2 months from now, but of course as SRM you can do as you see fit
with it, if you want to release it sooner from there, it's entirely up
to you.