Re: [PATCH] tests/functional: Convert the kvm_xen_guest avocado test

[Thomas Huth]

On 19/12/2024 09.49, David Woodhouse wrote:
> On 19 December 2024 09:35:13 CET, Thomas Huth <thuth@redhat.com> wrote:
> > On 18/12/2024 23.14, David Woodhouse wrote:
> > > On Wed, 2024-12-18 at 16:54 +0100, Thomas Huth wrote:
> > > > On 18/12/2024 12.48, David Woodhouse wrote:
> > > > > On 18 December 2024 12:32:49 CET, Thomas Huth <thuth@redhat.com> wrote:
> > > > > > Use the serial console to execute the commands in the guest instead
> > > > > > of using ssh since we don't have ssh support in the functional
> > > > > > framework yet.
> > > > > >
> > > > > > Signed-off-by: Thomas Huth <thuth@redhat.com>
> > > > >
> > > > > Hm, but serial is lossy and experience shows that it leads to flaky tests if 
> > > > > the guest (or host) misses bytes. While SSH would just go slower.
> > > >
> > > > I now noticed some issue with the serial console in this test, too.
> > > > Looks like the "Starting dropbear sshd: OK" is not print in an atomic way by
> > > > the guest, sometimes there are other kernel messages between the ":" and the
> > > > "OK". It works reliable when removing the "OK" from the string.
> > >
> > > Nah, that still isn't atomic; you just got lucky because the race
> > > window is smaller. It's not like serial ports are at a premium; can't
> > > you have a separate port for kernel vs. userspace messages?
> >
> > Maybe easiest solution: Simply add "quiet" to the kernel command line, then it 
> > does not write the kernel messages to the serial console anymore.
>
> Want to resend the bug report about that test failing again? But without the 
> kernel messages this time... :)

With "quiet", the output just looks like this when it hangs:

 Cannot get hvm parameter CONSOLE_EVTCHN (18): -22!
 Spectre V2 : Kernel not compiled with retpoline; no mitigation available!
 kvm_intel: VMX not supported by CPU 0
 Cannot get hvm parameter CONSOLE_EVTCHN (18): -22!
 fail to initialize ptp_kvm

Anyway, to properly track this, I've now created a ticket with the full log:

 https://gitlab.com/qemu-project/qemu/-/issues/2731

 Thomas