Re: wireguard, was: kernel: Fatal data abort - issue on rpi4

From: <decke_at_freebsd.org>
Date: Thu, 25 Aug 2022 11:10:17 UTC
 ---- On Thu, 25 Aug 2022 11:59:36 +0200    wrote --- 
 > 
 > 
 > > On 25. Aug 2022, at 11:39, Bernhard Froehlich decke@FreeBSD.org> wrote:
 > > 
 > > 
 > > --
 > > Bernhard Froehlich
 > > https://www.bluelife.at/
 > > ---- On Wed, 24 Aug 2022 11:29:58 +0200  Marek Soudny  wrote --- 
 > >> Hello,
 > >>>  I don't use wireguard myself, but if a package includes a kernel
 > >>> driver you have to be really careful that you keep them in sync.  I run
 > >>> into this possibility with video drivers (nvidia, radion, etc) all the time.
 > >> 
 > >> I use GENERIC kernel with freebsd-update and quarterly pkg branch, so 
 > >> everything should be in sync.
 > >> 
 > >> Anyway, I can confirm that loading if_wg kernel module does not do 
 > >> anything wrong, on it's own. Starting the wireguard service does not do 
 > >> anything wrong either, on it's own. BUT :) As soon as I tried to ping 
 > >> the wireguard interface from another box, my rpi stopped responding 
 > >> without stating why (I was tailing 
 > >> /var/log/{messages,debug.log,daemon.log}) and rebooted itself after (a) 
 > >> few minutes.
 > >> 
 > >> No idea how to debug this, thou. Anyway, I can reproduce it anytime.
 > >> 
 > >> BR,
 > >> Marek
 > > 
 > > There is no need to waste any more time on this. We have seen this all
 > > already and it's caused because all ports (latest and quarterly branch) for
 > > FreeBSD 13.x are build on 13.0 currently. But because of an ABI break in
 > > 13.1 the wireguard kernel module will panic when used.
 > Which ABI was changed between 13.0 and 13.1?

Check PR 264114 for more details.

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=264114

--
Bernhard Froehlich
https://www.bluelife.at/