[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ZjoJPxVOThaGKTZO@shredder>
Date: Tue, 7 May 2024 13:58:07 +0300
From: Ido Schimmel <idosch@...dia.com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: netdev@...r.kernel.org, davem@...emloft.net, pabeni@...hat.com,
edumazet@...gle.com, razor@...ckwall.org
Subject: Re: [PATCH net] selftests: test_bridge_neigh_suppress.sh: Try to
stabilize test
On Mon, May 06, 2024 at 06:36:13AM -0700, Jakub Kicinski wrote:
> On Sun, 5 May 2024 17:54:12 +0300 Ido Schimmel wrote:
> > I'm unable to reproduce these failures locally.
>
> :(
tl;dr - I was able to reproduce the issue (hopefully the same one as in
the CI) and will post a fix soon.
I woke in the middle of the night and for some reason it occurred to me
that the problem might be related to the MACAddressPolicy=persistent
nonsense that systemd made the default since version 242 [1].
The test creates the veth pairs in the initial network namespace before
moving them to one of the namespaces created by the test. Since they are
created in the same namespace and with the same name, udev will assign
them the same MAC address which breaks bridging.
The reason I couldn't initially reproduce the issue is that long ago I
added the following snippet [2] to the kickstart file with which we
install Fedora on our machines.
[1] https://lore.kernel.org/netdev/20200416095314.0a1dff38@strong.id.au/
[2]
# Prevent systemd from changing MAC addresses
%post
cat > /etc/systemd/network/10-ignore.link << EOL
[Match]
OriginalName=*
[Link]
MACAddressPolicy=none
EOL
%end
Powered by blists - more mailing lists