Re: resume from S3 sleep not working in Dom0 - Xen4.2.1

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Tue, Feb 12, 2013 at 10:40:32PM +0100, Milan opath wrote:
> xen-acpi-processor module is loaded and xenpm get-cpufreq-para seems to be
> ok. Unfortunately resume from S3 still doesn't work.

So then the process of elimination starts. Try doing it without having Xorg or any
KMS drivers running (so text-console). Then also try without serial output.
And also as somebody suggested with dom0_pin_vcpus arguments.

> 
> 
> 
> 
> 
> 
> 2013/2/12 Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> 
> > On Sat, Feb 09, 2013 at 06:21:34PM +0100, Milan opath wrote:
> > > I apologize for the patching confusion. I applied Konrad's acpi-s3 patch
> > to
> > > linux 3.7.5-1 and Ben's Xen patches to Xen 4.2.1-3 stable. I've also put
> > > dom0_vcpus_pin to xen kernel line in bootloader but still I cannot
> > modprobe
> > > acpi_cpufreq and S3 sleep resume still not working.
> >
> > The module is xen-acpi-processor and you should see whether it works by
> > running 'xenpm get-cpufreq-para'. If you get values then it worked.
> >
> > >
> > > How can I debug it?
> > > Thanks.
> > >
> > >
> > > 2013/2/8 Konrad Rzeszutek Wilk <konrad@kernel.org>
> > >
> > > > On Fri, Feb 8, 2013 at 5:11 AM, Tomasz Wroblewski
> > > > <tomasz.wroblewski@citrix.com> wrote:
> > > > >
> > > > >> Hm, that is suspect. There should not be any cpuidle_register?
> > Perhaps
> > > > >> you are .. ah yes, you are hitting a bug that should be in the
> > stable
> > > > >> tree fix.
> > > > >>
> > > > >> Here is the git commit b88a634a903d9670aa5f2f785aa890628ce0dece and
> > > > >> 6f8c2e7933679f54b6478945dc72e59ef9a3d5e0
> > > > >>
> > > > >>
> > > > >
> > > > > Thanks Konrad. I've tried your patches, and whilst I have not seen
> > this
> > > > > crash in cpuidle_register anymore, the others are still present (in
> > > > > build_schedule_domains for example).
> > > > >
> > > >
> > > > That looks familiar too. I think it got fixed in the upstream kernel
> > > > and it was a generic bug - but I can't recall which commit it was.
> > > >
> > > > > I've stumbled on a bit of interesting info - using dom0_vcpus_pin on
> > xen
> > > > > commandline stops both the xen scheduler and dom0 kernel crashing
> > and all
> > > > > works fine - made dozens of succesfull s3 attempts. Was wondering if
> > you
> > > > > guys had any thoughts on this? Is the dom0 kernel even supposed to
> > cope
> > > > > during s3 in non dom0 vcpu pin case?
> > > > >
> > > > No that is something new. Is this only an issue on Intel boxes but not
> > AMD?
> > > >
> >