RE: [PATCH 2/3] X86: Add a check to catch Xen emulation of Hyper-V

[KY Srinivasan]

> -----Original Message-----
> From: Michael S. Tsirkin [mailto:mst@xxxxxxxxxx]
> Sent: Wednesday, April 17, 2013 8:49 AM
> To: KY Srinivasan
> Cc: x86@xxxxxxxxxx; gregkh@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> devel@xxxxxxxxxxxxxxxxxxxxxx; olaf@xxxxxxxxx; apw@xxxxxxxxxxxxx;
> jasowang@xxxxxxxxxx; tglx@xxxxxxxxxxxxx; hpa@xxxxxxxxx;
> JBeulich@xxxxxxxx; bp@xxxxxxxxx
> Subject: Re: [PATCH 2/3] X86: Add a check to catch Xen emulation of Hyper-V
> 
> On Wed, Apr 17, 2013 at 01:20:58PM +0000, KY Srinivasan wrote:
> >
> >
> > > -----Original Message-----
> > > From: Michael S. Tsirkin [mailto:mst@xxxxxxxxxx]
> > > Sent: Wednesday, April 17, 2013 3:06 AM
> > > To: KY Srinivasan
> > > Cc: x86@xxxxxxxxxx; gregkh@xxxxxxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx;
> > > devel@xxxxxxxxxxxxxxxxxxxxxx; olaf@xxxxxxxxx; apw@xxxxxxxxxxxxx;
> > > jasowang@xxxxxxxxxx; tglx@xxxxxxxxxxxxx; hpa@xxxxxxxxx;
> > > JBeulich@xxxxxxxx; bp@xxxxxxxxx
> > > Subject: Re: [PATCH 2/3] X86: Add a check to catch Xen emulation of Hyper-V
> > >
> > > On Sun, Feb 03, 2013 at 05:22:38PM -0800, K. Y. Srinivasan wrote:
> > > > Xen emulates Hyper-V to host enlightened Windows. Looks like this
> > > > emulation may be turned on by default even for Linux guests. Check and
> > > > fail Hyper-V detection if we are on Xen.
> > > >
> > > > Signed-off-by: K. Y. Srinivasan <kys@xxxxxxxxxxxxx>
> > >
> > > I'm very curious what's the specific bug that is fixed here?
> > > I went over the original discussion in
> > > https://patchwork.kernel.org/patch/2064331/
> > > and that's still not clear to me. Is there a configuration
> > > that is broken without this patch but starts working with
> > > this patch?
> > >
> > > It seems that one might want to use hyper-v emulation e.g. to test
> > > hyper-v code without using windows, so the functionality
> > > that this patch disables is not completely useless,
> > > so there should be a good reason for disabling it.
> > >
> > > Could you enlighten me please? :)
> >
> > Michael,
> >
> > If I recall correctly, the issue here was that Xen was enabling
> > Hyper-V emulation un-conditionally even for Linux guests.
> 
> Okay so basically some Xen specific strangeness.
> 
> > Clearly,
> > this was not the intention to begin with. This check was added to
> > ensure that while hosting Linux we would not un-intentionally enable
> > Hyper-V emulation in Xen.
> 
> What if you want to intentionally enable Hyper-V emulation in Xen?
> There's no way to do this now, right?
> 
> > Not all Hyper-V features are tagged with a
> > CPUID feature bit and so it was agreed that it would be better to
> > disable Hyper-V emulation when hosting Linux on Xen.
> >
> > Regards,
> >
> > K. Y
> 
> I was just asking whether there are any specific examples of a bug
> or is this change done just in case?

Ideally, if Hyper-V could be emulated in a plug compatible way, in that the guest would not know
the difference as to whether Hyper-V was being emulated or indeed it was Hyper-V underneath, I would agree with you.
Even in this case, I would argue that we would somehow give preference to the native implementation rather
than the emulated environment. Having done the initial emulation of Hyper-V on Xen, I know the current emulation of 
Hyper-V is restricted to emulating a few features advertised via the CPUID bits. If I recall correctly, Xen did not implement
the enlightened TLB flush scheme. Furthermore, none of the Hyper-V emulations implement the I/O related infrastructure.
For Linux to run efficiently on Hyper-V, the I/O paths are significantly more important than what Hyper-V emulation there is
in Xen and I suspect in KVM. So, I don't think we want to run Linux on Xen/KVM that is emulating Hyper-V (cpuids).
The specific case I ran into that required this patch was when I moved the VMBUS interrupts to be delivered via a special
hypervisor vector (to support full interrupt distribution amongst all VCPUs in the guest). This feature is not tagged in the CPUID
feature bits and without this check, I would install an interrupt handler for the hypervisor vector that clearly did not make
any sense in anything other than Hyper-V. I suspect there will be more such situations as we go forward.

Regards,

K. Y 
> 
> > >
> > > > ---
> > > >  arch/x86/kernel/cpu/mshyperv.c |    7 +++++++
> > > >  1 files changed, 7 insertions(+), 0 deletions(-)
> > > >
> > > > diff --git a/arch/x86/kernel/cpu/mshyperv.c
> > > b/arch/x86/kernel/cpu/mshyperv.c
> > > > index 646d192..4dab317 100644
> > > > --- a/arch/x86/kernel/cpu/mshyperv.c
> > > > +++ b/arch/x86/kernel/cpu/mshyperv.c
> > > > @@ -30,6 +30,13 @@ static bool __init ms_hyperv_platform(void)
> > > >  	if (!boot_cpu_has(X86_FEATURE_HYPERVISOR))
> > > >  		return false;
> > > >
> > > > +	/*
> > > > +	 * Xen emulates Hyper-V to support enlightened Windows.
> > > > +	 * Check to see first if we are on a Xen Hypervisor.
> > > > +	 */
> > > > +	if (xen_cpuid_base())
> > > > +		return false;
> > > > +
> > > >  	cpuid(HYPERV_CPUID_VENDOR_AND_MAX_FUNCTIONS,
> > > >  	      &eax, &hyp_signature[0], &hyp_signature[1], &hyp_signature[2]);
> > > >
> > > > --
> > > > 1.7.4.1
> > > >
> > > > --
> > > > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > > > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > > > Please read the FAQ at  http://www.tux.org/lkml/
> > >
> >
> 


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/