On Tue, 14 Jan 2020 08:43:17 -0800
Adam Williamson <adamwill(a)fedoraproject.org> wrote:
On Tue, 2020-01-14 at 11:58 +0100, Normand wrote:
> Hello Adam,
>
> I am surprised by the number of errors for PowerPC on openqa stg
> since 20200110 compose:
>
https://openqa.stg.fedoraproject.org/group_overview/3
> Is there some pending changes on those machines ?
>
> We do not have such errors in our local openqa server at IBM for
> same compose versions.
Hi Michel!
Long story short, it's:
https://bugzilla.redhat.com/show_bug.cgi?id=1784961
I took out the workaround we had in the templates to set the machine
to pseries-4.0 because I thought we didn't need it any more, but I've
just figured out that's what's causing the problems. We're hitting the
behaviour described in that bug, where the kernel checks something to
do with xive, then triggers a reboot. That breaks all tests where we
want to pass in kernel params, because when the reboot happens, our
params get lost (our test code doesn't know to enter them again, as
it's not something we envisaged when writing the tests).
Kevin was trying to do a backport of the qemu fix, if he's succeeded
with that I'll try it out, otherwise I'll put the pseries-4.0
workaround back in. Sorry for the inconvenience!
things should work again without any workarounds with the new qemu (now
in infra repo, rawhide 4.2 + a patch from master) and SLOF from rawhide
BTW, did you notice I enabled update tests for ppc64le on staging? :)
no, but thank you :-)
Dan