[vbox-dev] Arrangement of EPT structures with SMP

[Sander van Leeuwen]

Each VCPU has a root page table pointer that points to the same EPT root 
page. src\vbox\vmm\PGMShw.h, Enter shows how
it's set up. The same applies to AMD-V nested paging and real + 
protected mode without paging.

Tom Birch wrote:
> So this means the value of EPTP on each vcpu is the same?
>
> My question before was: how do they actually share the same page
> table? Is the EPT pointer copied between VCPUs (either directly or
> indirectly)? Can you point to the code that does this?
>
> cheers,
> Tom
>
> On Thu, Oct 29, 2009 at 1:58 PM, Sander van Leeuwen
> <Sander.Vanleeuwen at sun.com> wrote:
>   
>> Each VCPU has its own root page table, but all VCPUs are sharing the same
>> one with EPT (or AMD nested paging for that matter).
>>
>> Tom Birch wrote:
>>     
>>> Cool, thanks, that's what I thought. Is there any reason the EPT is a
>>> per-VCPU data structure? Could you please point me to the bits of code
>>> that do this sharing? I'm getting lost looking at all the assignments
>>> to the pShwPageCR3* variables, they still all seem to be per-VCPU, I
>>> can't find where it hooks into the root VM structure.
>>>
>>> cheers,
>>> Tom
>>>
>>> On Thu, Oct 29, 2009 at 1:19 PM, Sander van Leeuwen
>>> <Sander.Vanleeuwen at sun.com> wrote:
>>>
>>>       
>>>> The EPT structures are shared between all VCPUs as they contain a guest
>>>> physical to host physical mapping, which
>>>> is independent of the VCPU. The host OS is irrelevant.
>>>>
>>>> Tom Birch wrote:
>>>>
>>>>         
>>>>> I'm looking through the page management code, specifically hardware
>>>>> acceleration on Intel with EPT, and I can't quite figure out the
>>>>> high-level relationships between various structures. Both the VMCS and
>>>>> EPTP seem to be per-PVMCPU data structures, does this mean that within
>>>>> a single VM, you can have multiple distinct EPTPs (on different
>>>>> VCpus), or is this arrangement just a consequence of the way shadow
>>>>> paging worked before EPT? Put another way, can I guarantee that all
>>>>> VCpus within a single VM will all share the same EPTP? I'm mainly
>>>>> concerned about how this applies to Darwin if that matters, but I
>>>>> seems like a fairly Host-OS-agnostic feature.
>>>>>
>>>>> cheers,
>>>>> Tom
>>>>>
>>>>> _______________________________________________
>>>>> vbox-dev mailing list
>>>>> vbox-dev at virtualbox.org
>>>>> http://vbox.innotek.de/mailman/listinfo/vbox-dev
>>>>>
>>>>>
>>>>>           
>>>> --
>>>> Kind regards / Mit freundlichen Gruessen / Met vriendelijke groet
>>>>
>>>> --
>>>>
>>>> Sun Microsystems GmbH        Sander van Leeuwen
>>>> Werkstrasse 24               Senior Staff Engineer, VirtualBox
>>>> 71384 Weinstadt, Germany     mailto:Sander.Vanleeuwen at sun.com
>>>>
>>>>
>>>> ================================================
>>>> Sitz der Gesellschaft: Sun Microsystems GmbH,
>>>> Sonnenallee 1, 85551 Kirchheim-Heimstetten
>>>> Amtsgericht Muenchen: HRB 161028
>>>> Geschaeftsfuehrer: Thomas Schroeder,
>>>> Wolfgang Engels, Dr. Roland Boehmer
>>>> Vorsitzender des Aufsichtsrates: Martin Haering
>>>> ================================================
>>>>
>>>>
>>>>
>>>>         
>> --
>> Kind regards / Mit freundlichen Gruessen / Met vriendelijke groet
>>
>> --
>>
>> Sun Microsystems GmbH        Sander van Leeuwen
>> Werkstrasse 24               Senior Staff Engineer, VirtualBox
>> 71384 Weinstadt, Germany     mailto:Sander.Vanleeuwen at sun.com
>>
>>
>> ================================================
>> Sitz der Gesellschaft: Sun Microsystems GmbH,
>> Sonnenallee 1, 85551 Kirchheim-Heimstetten
>> Amtsgericht Muenchen: HRB 161028
>> Geschaeftsfuehrer: Thomas Schroeder,
>> Wolfgang Engels, Dr. Roland Boehmer
>> Vorsitzender des Aufsichtsrates: Martin Haering
>> ================================================
>>
>>
>>     


-- 
Kind regards / Mit freundlichen Gruessen / Met vriendelijke groet

--

Sun Microsystems GmbH        Sander van Leeuwen
Werkstrasse 24               Senior Staff Engineer, VirtualBox
71384 Weinstadt, Germany     mailto:Sander.Vanleeuwen at sun.com


================================================
Sitz der Gesellschaft: Sun Microsystems GmbH,
Sonnenallee 1, 85551 Kirchheim-Heimstetten
Amtsgericht Muenchen: HRB 161028
Geschaeftsfuehrer: Thomas Schroeder,
Wolfgang Engels, Dr. Roland Boehmer
Vorsitzender des Aufsichtsrates: Martin Haering
================================================