Re: Xen Signature == "NovellShimHv" ???

Re: Xen Signature == "NovellShimHv" ???

[Konrad Rzeszutek Wilk]

On Tue, Sep 15, 2009 at 11:24:43AM +1000, James Harper wrote:
> A user of GPLPV is reporting a problem with the drivers on Windows 2008,
> and it appears that GPLPV is finding a signature of 'NovellShimHv'
> instead of 'XenVMMXenVMM'.
> 
> Can anyone suggest a reason for this?

SLES10 SP2 Xen has the code for this. It is basically an early version of the
Veridian interface.
> 
> Thanks
> 
> James
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Xen Signature == "NovellShimHv" ???

[James Harper]

A user of GPLPV is reporting a problem with the drivers on Windows 2008,
and it appears that GPLPV is finding a signature of 'NovellShimHv'
instead of 'XenVMMXenVMM'.

Can anyone suggest a reason for this?

Thanks

James

RE: Xen Signature == "NovellShimHv" ???

[James Harper]

> 
> On Tue, Sep 15, 2009 at 11:24:43AM +1000, James Harper wrote:
> > A user of GPLPV is reporting a problem with the drivers on Windows
2008,
> > and it appears that GPLPV is finding a signature of 'NovellShimHv'
> > instead of 'XenVMMXenVMM'.
> >
> > Can anyone suggest a reason for this?
> 
> SLES10 SP2 Xen has the code for this. It is basically an early version
of the
> Veridian interface.

Ah. Hmmm. So does that mean that PV drivers on HVM domains (windows or
Linux) aren't going to work at all? Should I check for both
'NovellShimHv' and 'XenVMMXenVMM'? I notice that the unmodified drivers
tree under xen doesn't check for this, and neither does hvmloader.

James

Re: Xen Signature == "NovellShimHv" ???

[Konrad Rzeszutek Wilk]

On Tue, Sep 15, 2009 at 11:41:41AM +1000, James Harper wrote:
> > 
> > On Tue, Sep 15, 2009 at 11:24:43AM +1000, James Harper wrote:
> > > A user of GPLPV is reporting a problem with the drivers on Windows
> 2008,
> > > and it appears that GPLPV is finding a signature of 'NovellShimHv'
> > > instead of 'XenVMMXenVMM'.
> > >
> > > Can anyone suggest a reason for this?
> > 
> > SLES10 SP2 Xen has the code for this. It is basically an early version
> of the
> > Veridian interface.
> 
> Ah. Hmmm. So does that mean that PV drivers on HVM domains (windows or
> Linux) aren't going to work at all? Should I check for both

I don't think you need to worry about Linux. The shim enabled extra functionality
so that under Windows 2008 (and further) you wouldn't hit the BSOD watchdog
failure. I don't think any extra functionaly was provided besides that.

> 'NovellShimHv' and 'XenVMMXenVMM'? I notice that the unmodified drivers
> tree under xen doesn't check for this, and neither does hvmloader.

Correct. SLES version to my recollection never made it to Xen-devel.
> 
> James
> 

Re: Xen Signature == "NovellShimHv" ???

[Keir Fraser]

On 15/09/2009 02:41, "James Harper" <james.harper@bendigoit.com.au> wrote:

>> SLES10 SP2 Xen has the code for this. It is basically an early version
> of the
>> Veridian interface.
> 
> Ah. Hmmm. So does that mean that PV drivers on HVM domains (windows or
> Linux) aren't going to work at all? Should I check for both
> 'NovellShimHv' and 'XenVMMXenVMM'? I notice that the unmodified drivers
> tree under xen doesn't check for this, and neither does hvmloader.

They may have moved XenVMMXenVMM out of the way to another leaf? Like we do
with the checked-in Viridian shim.

 -- Keir

Re: Xen Signature == "NovellShimHv" ???

[Jan Beulich]

>>> Keir Fraser <keir.fraser@eu.citrix.com> 15.09.09 09:30 >>>
>On 15/09/2009 02:41, "James Harper" <james.harper@bendigoit.com.au> wrote:
>
>>> SLES10 SP2 Xen has the code for this. It is basically an early version
>> of the
>>> Veridian interface.
>> 
>> Ah. Hmmm. So does that mean that PV drivers on HVM domains (windows or
>> Linux) aren't going to work at all? Should I check for both
>> 'NovellShimHv' and 'XenVMMXenVMM'? I notice that the unmodified drivers
>> tree under xen doesn't check for this, and neither does hvmloader.
>
>They may have moved XenVMMXenVMM out of the way to another leaf? Like we do
>with the checked-in Viridian shim.

Yes, to 0x40001xxx.

Jan

RE: Xen Signature == "NovellShimHv" ???

[James Harper]

> >They may have moved XenVMMXenVMM out of the way to another leaf? Like
we do
> >with the checked-in Viridian shim.
> 
> Yes, to 0x40001xxx.
> 

I think I only check up to 0x0FFF (I think that's all the HVM Linux
drivers do). What is a reasonable ceiling?

James

Re: Xen Signature == "NovellShimHv" ???

[Ky Srinivasan]

James,

This user must be using a SLES based system. On SLES, the HyperV shim returns this signature.

Regards,

K. Y

>>> On 9/14/2009 at  9:24 PM, in message
<AEC6C66638C05B468B556EA548C1A77D0177D2B0@trantor>, "James Harper"
<james.harper@bendigoit.com.au> wrote: 
> A user of GPLPV is reporting a problem with the drivers on Windows 2008,
> and it appears that GPLPV is finding a signature of 'NovellShimHv'
> instead of 'XenVMMXenVMM'.
> 
> Can anyone suggest a reason for this?
> 
> Thanks
> 
> James
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

RE: Xen Signature == "NovellShimHv" ???

[Ky Srinivasan]

>>> On 9/14/2009 at  9:41 PM, in message
<AEC6C66638C05B468B556EA548C1A77D0177D2B2@trantor>, "James Harper"
<james.harper@bendigoit.com.au> wrote: 
>>  
>> On Tue, Sep 15, 2009 at 11:24:43AM +1000, James Harper wrote:
>> > A user of GPLPV is reporting a problem with the drivers on Windows
> 2008,
>> > and it appears that GPLPV is finding a signature of 'NovellShimHv'
>> > instead of 'XenVMMXenVMM'.
>> >
>> > Can anyone suggest a reason for this?
>> 
>> SLES10 SP2 Xen has the code for this. It is basically an early version
> of the
>> Veridian interface.
> 
> Ah. Hmmm. So does that mean that PV drivers on HVM domains (windows or
> Linux) aren't going to work at all? Should I check for both
> 'NovellShimHv' and 'XenVMMXenVMM'? I notice that the unmodified drivers
> tree under xen doesn't check for this, and neither does hvmloader.
> 
> James
A simple workaround here might be to disable the HyperV shim when you are loading GPL PV drivers for windows on a SLES host; this is done by setting extid = 0 in the VM config file. If you want the shim enabled and run GPL PV , you have more issues than the hypervisor signature. Some of the CPUID leaves and MSRs of Xen and HyerV collide and this needs to be disambiguated.

Regards,

K. Y
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

RE: Xen Signature == "NovellShimHv" ???

[Jan Beulich]

>>> "James Harper" <james.harper@bendigoit.com.au> 15.09.09 11:15 >>>
> >They may have moved XenVMMXenVMM out of the way to another leaf? Like
>we do
>> >with the checked-in Viridian shim.
>> 
>> Yes, to 0x40001xxx.
>> 
>
>I think I only check up to 0x0FFF (I think that's all the HVM Linux
>drivers do). What is a reasonable ceiling?

Perhaps you'll want to scan the whole virtualization reserved space, which
iirc is up to 0x4000ffff.

Keir, perhaps xen-detect should also be adjusted accordingly?

Jan

Re: Xen Signature == "NovellShimHv" ???

[Keir Fraser]

On 15/09/2009 13:54, "Jan Beulich" <JBeulich@novell.com> wrote:

>> I think I only check up to 0x0FFF (I think that's all the HVM Linux
>> drivers do). What is a reasonable ceiling?
> 
> Perhaps you'll want to scan the whole virtualization reserved space, which
> iirc is up to 0x4000ffff.
> 
> Keir, perhaps xen-detect should also be adjusted accordingly?

Can do. Is the step size of 0x100 okay (i.e., searching 0x4000xx00)?

 -- Keir

Re: Xen Signature == "NovellShimHv" ???

[Jan Beulich]

>>> Keir Fraser <keir.fraser@eu.citrix.com> 15.09.09 15:13 >>>
>On 15/09/2009 13:54, "Jan Beulich" <JBeulich@novell.com> wrote:
>
>>> I think I only check up to 0x0FFF (I think that's all the HVM Linux
>>> drivers do). What is a reasonable ceiling?
>> 
>> Perhaps you'll want to scan the whole virtualization reserved space, which
>> iirc is up to 0x4000ffff.
>> 
>> Keir, perhaps xen-detect should also be adjusted accordingly?
>
>Can do. Is the step size of 0x100 okay (i.e., searching 0x4000xx00)?

Yes.