Re: [Qemu-devel] [PATCH for-3.1] tests/cpu-plug-test: check CPU hotplug on ppc64 with KVM

[Greg Kurz]

On Fri, 27 Jul 2018 13:25:33 +0200
Thomas Huth <address@hidden> wrote:

> On 07/27/2018 11:00 AM, Greg Kurz wrote:
> > On Fri, 27 Jul 2018 10:18:14 +0200
> > Thomas Huth <address@hidden> wrote:
> >   
> >> On 07/27/2018 09:54 AM, Greg Kurz wrote:  
> >>> On Fri, 27 Jul 2018 15:27:24 +1000
> >>> David Gibson <address@hidden> wrote:
> >>>     
> >>>> On Wed, Jul 25, 2018 at 04:45:26PM +0200, Greg Kurz wrote:    
> >>>>> Commit b585395b655 fixed a regression introduced by some recent changes
> >>>>> in the XICS code, that was causing QEMU to crash instantly during CPU
> >>>>> hotplug with KVM. This is typically the kind of bug we'd like our
> >>>>> test suite to detect before it gets merged. Unfortunately, the current
> >>>>> tests run with '-machine accel=qtest' and don't exercise KVM specific
> >>>>> paths in QEMU.
> >>>>>
> >>>>> This patch hence changes add_pseries_test_case() to launch QEMU with
> >>>>> '-machine accel=kvm' if KVM is available.
> >>>>>
> >>>>> A notable consequence is that the guest will execute SLOF, but for some
> >>>>> reasons SLOF sometimes hits a program exception. This causes the guest
> >>>>> to loop forever and the test to be stuck.  Since we don't need the guest
> >>>>> to be truely running, let's pass -S to QEMU to avoid that.
> >>>>>
> >>>>> Also disable machine capabilities that could be unavailable in KVM, eg,
> >>>>> when using PR KVM.
> >>>>>
> >>>>> Signed-off-by: Greg Kurz <address@hidden>      
> >>>>
> >>>> I'm pretty sure trying to change the accelerator on a qtest test just
> >>>> doesn't make sense.  We'd need a different approach for testing cpu
> >>>> hotplug against kvm & tcg backends.
> >>>>    
> >>>
> >>> The test starts QEMU, triggers the CPU hotplug code with a QMP command
> >>> and checks the command didn't fail (or QEMU didn't crash, as it would
> >>> have before commit b585395b655a)... I really don't understand what
> >>> is wrong with that... Please elaborate.    
> >>
> >> For a "real" test, I think you'd need a guest OS that is reacting to the
> >> hot plug events. So maybe this should rather be done in the avocado
> >> framework instead?
> >>  
> > 
> > The intent isn't a "real" test actually, but just to exercise the XICS-KVM
> > paths in QEMU that get called during CPU hotplug. This cannot be achieved
> > with the qtest accelerator.
> > 
> > This patch would have revealed the regression in b585395b655a right away,
> > with the simple 'make check' developers are expected to run before posting. 
> >  
> 
> OK, that's fair. I think what rather bugs me here is that you start QEMU
> with -S to work around something that sounds like real bug in SLOF ...
> any chance that you could fix that bug in SLOF instead of using -S here?
> 

Yeah, -S is a workaround indeed. I've chosen to go that way because I couldn't
find any impact on the hotplug path, but I may have overlooked something, so
I'm not surprised by your request :)

I guess that's fair too and I'll investigate.

Cheers,

--
Greg

>  Thomas