Re: Recent NetBSD-current as DOM0 crashes

[Brad Spencer <brad%anduin.eldar.org@localhost>]

oskar%fessel.org@localhost writes:

> Hej,
>
>> Am 03.08.2023 um 03:24 schrieb Brad Spencer <brad%anduin.eldar.org@localhost>:
>> 
>> 
>> I was testing a recent NetBSD-current, 10.99.6, with a Xen
>> xenkernel415-4.15.3 kernel built from pkgsrc a while back (probably
>> 2022Q3).  
>
> probably not the best combo. I am running dom0:
> NetBSD 10.99.2 (XEN3_DOM0) #0: Tue Dec 27 03:19:40 EET 2022
> with xenkernel415-4.15.4
> and that is rock solid with netbsd 9 and 10 kernels i have installed as domUs up to 10.99.2
>
> probably a good idea to try xen 4.15.4
>
> Cheers
> 	Oskar


Thanks for the info...  After more investigation I filed a PR about what
I saw.  It is likely that Dec 2022 was before the commits that seem to
cause the problem for me.

The commit that caused me trouble was revision 1.109 of
sys/arch/xen/xen/xennetback_xenbus.c which hit the tree in February 2023
and was pulled into NetBSD 10 ...  it may be that 1.110 and 1.111 are
also problems, but I don't know that really...  I had to revert back to
1.108 to get a working DOM0.  It also appears that these were just
reverted in -current, but not NetBSD 10 (yet?).

I also agree that a newer Xen kernel could help, but that is a much more
involved process (at least for me).  I was also concerned about the
surprise factor...  the DOM0 boots up and the guests would begin to
start, and then the DOM0 would crash when the DOMU ran /etc/rc.d/network
(or equiv..) in a way that does not seem to produce a traceback (looks
like pressing the reset button to me).  Being a bit paranoid I started a
DOMU with its console...  had the DOMUs start automatically upon the
boot up of the DOM0, it would have just been a mysterious reboot event
which would happen again on the reboot...





-- 
Brad Spencer - brad%anduin.eldar.org@localhost - KC8VKS - http://anduin.eldar.org