xen 3.0 for AMD Pacifica

xen 3.0 for AMD Pacifica

[Dave Feustel]

> We intend to add support for x86 h/w virtualization extensions 
> as information about them becomes available.

What information is needed beyond the Pacifica documentation already
provided by AMD and the Vanderpool docs provided by Intel?

Is there any possibility of setting up a mailing list directed specifically at
xen 3.0 hardware supported virtualization?

Thanks,
Dave Feustel


-- 
Tired of having to defend against Malware?
You know: trojans, viruses, SPYWARE, ADWARE, 
KEYLOGGERS, rootkits, worms and popups.
Then Switch to OpenBSD with a KDE desktop!!!

Re: xen 3.0 for AMD Pacifica

[Mark Williamson]

> > We intend to add support for x86 h/w virtualization extensions
> > as information about them becomes available.
>
> What information is needed beyond the Pacifica documentation already
> provided by AMD and the Vanderpool docs provided by Intel?

In principle, nothing, but the code and manpower coming from those guys is a 
*huge* help.  The other information - stuff like, when hardware is being 
released, how work is progressing, which optional features of the specs will 
be implemented - is rather hush for business reasons.

> Is there any possibility of setting up a mailing list directed specifically
> at xen 3.0 hardware supported virtualization?

I guess there's an argument for this - h/w virtualisation posts give a 
substantial amount of the traffic on xen-devel these days.  Would it really 
be much of an advantage?

Cheers,
Mark

Re: xen 3.0 for AMD Pacifica

[Anthony Liguori]

Dave Feustel wrote:

>>We intend to add support for x86 h/w virtualization extensions 
>>as information about them becomes available.
>>    
>>
>
>What information is needed beyond the Pacifica documentation already
>provided by AMD and the Vanderpool docs provided by Intel?
>  
>
I don't know where you're quoting from, but both specs are out publicly.

As Mark mentions, there's quite a bit of work to go from specs to 
implementation though.

Both VT and Pacifica are pretty similiar in the basic idea that they 
allow the processor to be put in a mode where 30 or so events that 
normally don't trap actually do trap.

Just having the processor trap though doesn't get you virtualization 
:-)  Each of those events has to be handled, and often emulated.  Throw 
in shadow paging and IO emulation and it's all rather complicated.

Plus, there's not much that can be done without having the hardware at 
this point.

>Is there any possibility of setting up a mailing list directed specifically at
>xen 3.0 hardware supported virtualization?
>  
>
There are an awful lot of Xen mailing lists at this point.  What would 
be *really* useful is having some content on the wiki that went into 
more depth about VT/Pacifica.

Regards,

Anthony Liguori

>Thanks,
>Dave Feustel
>
>
>  
>

Re: xen 3.0 for AMD Pacifica

[Dave Feustel]

On Tuesday 25 October 2005 09:53, Anthony Liguori wrote:
> Dave Feustel wrote:
> 
> >>We intend to add support for x86 h/w virtualization extensions 
> >>as information about them becomes available.
> >>    
> >>
> >
> >What information is needed beyond the Pacifica documentation already
> >provided by AMD and the Vanderpool docs provided by Intel?
> >  
> >
> I don't know where you're quoting from, but both specs are out publicly.

I know. I have both. What I was asking is whether the info in those documents
is sufficient to make Xen work with  the hardware of both chips, or is additional 
info needed (and still not available)?
 
> As Mark mentions, there's quite a bit of work to go from specs to 
> implementation though.
> 
> Both VT and Pacifica are pretty similiar in the basic idea that they 
> allow the processor to be put in a mode where 30 or so events that 
> normally don't trap actually do trap.

Pacifica has a new paging mode (paged real mode) made possible by the 
Opteron on-chip memory controller. This feature is not present in VT.

> Just having the processor trap though doesn't get you virtualization 
> :-)  Each of those events has to be handled, and often emulated.  Throw 
> in shadow paging and IO emulation and it's all rather complicated.
> 
> Plus, there's not much that can be done without having the hardware at 
> this point.
> 
> >Is there any possibility of setting up a mailing list directed specifically at
> >xen 3.0 hardware supported virtualization?
> >  
> >
> There are an awful lot of Xen mailing lists at this point.  What would 
> be *really* useful is having some content on the wiki that went into 
> more depth about VT/Pacifica.
> 
> Regards,
> 
> Anthony Liguori
> 
> >Thanks,
> >Dave Feustel
> >
> >
> >  
> >
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
> 

-- 
Tired of having to defend against Malware?
You know: trojans, viruses, SPYWARE, ADWARE, 
KEYLOGGERS, rootkits, worms and popups.
Then Switch to OpenBSD with a KDE desktop!!!

Re: xen 3.0 for AMD Pacifica

[Mark Williamson]

> I know. I have both. What I was asking is whether the info in those
> documents is sufficient to make Xen work with  the hardware of both chips,
> or is additional info needed (and still not available)?

Well, apart from the code from Intel we also have had VT SDVs in Cambridge for 
quite some time and so are able to test stuff for ourselves.  I'm not sure 
what the status is for AMD: but there's always SimNow for development / 
testing.  So in both cases there's some sort of reference implementation to 
work with.

> Pacifica has a new paging mode (paged real mode) made possible by the
> Opteron on-chip memory controller. This feature is not present in VT.

My initial impression of Pacifica was "much like VT, but with more 
goodies" ;-)

Cheers,
Mark

> > Just having the processor trap though doesn't get you virtualization
> >
> > :-)  Each of those events has to be handled, and often emulated.  Throw
> >
> > in shadow paging and IO emulation and it's all rather complicated.
> >
> > Plus, there's not much that can be done without having the hardware at
> > this point.
> >
> > >Is there any possibility of setting up a mailing list directed
> > > specifically at xen 3.0 hardware supported virtualization?
> >
> > There are an awful lot of Xen mailing lists at this point.  What would
> > be *really* useful is having some content on the wiki that went into
> > more depth about VT/Pacifica.
> >
> > Regards,
> >
> > Anthony Liguori
> >
> > >Thanks,
> > >Dave Feustel
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel

Re: xen 3.0 for AMD Pacifica

[Anthony Liguori]

Dave Feustel wrote:

>On Tuesday 25 October 2005 09:53, Anthony Liguori wrote:
>  
>
>>Dave Feustel wrote:
>>
>>    
>>
>>>>We intend to add support for x86 h/w virtualization extensions 
>>>>as information about them becomes available.
>>>>   
>>>>
>>>>        
>>>>
>>>What information is needed beyond the Pacifica documentation already
>>>provided by AMD and the Vanderpool docs provided by Intel?
>>> 
>>>
>>>      
>>>
>>I don't know where you're quoting from, but both specs are out publicly.
>>    
>>
>
>I know. I have both. What I was asking is whether the info in those documents
>is sufficient to make Xen work with  the hardware of both chips, or is additional 
>info needed (and still not available)?
>
Sorry, then yes, theoritically, all you should need are those specs.

>>Both VT and Pacifica are pretty similiar in the basic idea that they 
>>allow the processor to be put in a mode where 30 or so events that 
>>normally don't trap actually do trap.
>>    
>>
>
>Pacifica has a new paging mode (paged real mode) made possible by the 
>Opteron on-chip memory controller. This feature is not present in VT.
>  
>
Yes, there are considerable differences.  Pacifica has paged real mode 
whereas VT requires vm86 mode to be used for real mode emulation.  
Pacifica also has (optional) support for hardware shadow page tables, 
TLB tagging, and DMA memory protection.  They also have different models 
for dealing with the VMCS/VMCB.  I would categorize this stuff mostly as 
optimizations though since none of it's required to meet the 
Popek/Goldberg requirements.

Just enough differences to make things interesting :-)

Regards,

Anthony Liguori

>>Just having the processor trap though doesn't get you virtualization 
>>:-)  Each of those events has to be handled, and often emulated.  Throw 
>>in shadow paging and IO emulation and it's all rather complicated.
>>
>>Plus, there's not much that can be done without having the hardware at 
>>this point.
>>
>>    
>>
>>>Is there any possibility of setting up a mailing list directed specifically at
>>>xen 3.0 hardware supported virtualization?
>>> 
>>>
>>>      
>>>
>>There are an awful lot of Xen mailing lists at this point.  What would 
>>be *really* useful is having some content on the wiki that went into 
>>more depth about VT/Pacifica.
>>
>>Regards,
>>
>>Anthony Liguori
>>
>>    
>>
>>>Thanks,
>>>Dave Feustel
>>>
>>>
>>> 
>>>
>>>      
>>>
>>_______________________________________________
>>Xen-devel mailing list
>>Xen-devel@lists.xensource.com
>>http://lists.xensource.com/xen-devel
>>
>>    
>>
>
>  
>