Xen devel - Apr 2012 - [PATCH] Fix save/restore of guest PAT table in HAP paging mode.

Hello,

HAP paging mode guests use direct MSR read/write into the VMCS/VMCB
for the guest PAT table, while the current  save/restore code was
accessing only the pat_cr field in hvm_vcpu, used when intercepting
the MSR mostly in shadow mode (the Intel scenario is a bit more
complicated).
This patch fixes this issue creating a new couple of hvm_funcs,
get/set_guest_pat, that access the right PAT table based on the paging
mode and guest configuration.

As a major caveat, I haven''t tested this patch on AMD, for lack of
hardware.

Signed-off-by: Gianluca Guida <gianluca.guida@citrix.com>


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On 04/11/2012 06:04 PM, Gianluca Guida wrote:
> Hello,
>
> HAP paging mode guests use direct MSR read/write into the VMCS/VMCB
> for the guest PAT table, while the current  save/restore code was
> accessing only the pat_cr field in hvm_vcpu, used when intercepting
> the MSR mostly in shadow mode (the Intel scenario is a bit more
> complicated).
> This patch fixes this issue creating a new couple of hvm_funcs,
> get/set_guest_pat, that access the right PAT table based on the paging
> mode and guest configuration.
>
> As a major caveat, I haven''t tested this patch on AMD, for lack of
hardware.
I can test it on my AMD box tomorrow. BTW from my understanding, this 
patch doesn''t have performance implication for nested paging mode, does
it?

-Wei
>
> Signed-off-by: Gianluca Guida<gianluca.guida@citrix.com>

On Wed, Apr 11, 2012 at 11:14 PM, Wei Huang <wei.huang2@amd.com>
wrote:
> On 04/11/2012 06:04 PM, Gianluca Guida wrote:
>> As a major caveat, I haven''t tested this patch on AMD, for
lack of
>> hardware.
>
> I can test it on my AMD box tomorrow. BTW from my understanding, this patch
> doesn''t have performance implication for nested paging mode, does
it?
Thank you! A good but not perfect way is to check that xen-hvmctx
doesn''t return on HAP the reset PAT table after a linux/windows guest
has been running. And that its value is actually preserved across
save/restore.

As for the performances in nested mode, as far as I could see, I think
not -- but I haven''t looked at eventually hidden details of the nested
vm implementation.

Gianluca

I tested it with SLES11 SP1 and Windows 7 guest VM. With nested paging enabled,
I saw XEN saved 0x0007010600070106ULL, instead of default 0x0007040600070406ULL,
to guest disk files. So it behaved as expected.

-Wei

-----Original Message-----
From: Gianluca Guida [mailto:glguida@gmail.com] 
Sent: Thursday, April 12, 2012 2:32 AM
To: Huang2, Wei
Cc: xen-devel@lists.xensource.com; Gianluca Guida
Subject: Re: [Xen-devel] [PATCH] Fix save/restore of guest PAT table in HAP
paging mode.

On Wed, Apr 11, 2012 at 11:14 PM, Wei Huang <wei.huang2@amd.com>
wrote:
> On 04/11/2012 06:04 PM, Gianluca Guida wrote:
>> As a major caveat, I haven''t tested this patch on AMD, for
lack of
>> hardware.
>
> I can test it on my AMD box tomorrow. BTW from my understanding, this patch
> doesn''t have performance implication for nested paging mode, does
it?
Thank you! A good but not perfect way is to check that xen-hvmctx
doesn''t return on HAP the reset PAT table after a linux/windows guest
has been running. And that its value is actually preserved across
save/restore.

As for the performances in nested mode, as far as I could see, I think
not -- but I haven''t looked at eventually hidden details of the nested
vm implementation.

Gianluca

On Fri, Apr 13, 2012 at 9:29 AM, Huang2, Wei <Wei.Huang2@amd.com>
wrote:
> I tested it with SLES11 SP1 and Windows 7 guest VM. With nested paging
enabled, I saw XEN saved 0x0007010600070106ULL, instead of default
0x0007040600070406ULL, to guest disk files. So it behaved as expected.
Yes, that is exactly what I expected to see (WC enabled). Thanks again.

Any change to get this checked in?
Gianluca

At 10:08 -0700 on 13 Apr (1334311697), Gianluca Guida
wrote:
> On Fri, Apr 13, 2012 at 9:29 AM, Huang2, Wei <Wei.Huang2@amd.com>
wrote:
> > I tested it with SLES11 SP1 and Windows 7 guest VM. With nested paging
enabled, I saw XEN saved 0x0007010600070106ULL, instead of default
0x0007040600070406ULL, to guest disk files. So it behaved as expected.
> 
> Yes, that is exactly what I expected to see (WC enabled). Thanks again.
> 
> Any change to get this checked in?
FWIW, Acked-by: Tim Deegan <tim@xen.org>

You can treat my experiments as an ack. But other folks need to ack it too.

-Wei

On 04/13/2012 12:08 PM, Gianluca Guida wrote:
> On Fri, Apr 13, 2012 at 9:29 AM, Huang2, Wei<Wei.Huang2@amd.com> 
wrote:
>> I tested it with SLES11 SP1 and Windows 7 guest VM. With nested paging
enabled, I saw XEN saved 0x0007010600070106ULL, instead of default
0x0007040600070406ULL, to guest disk files. So it behaved as expected.
> Yes, that is exactly what I expected to see (WC enabled). Thanks again.
>
> Any change to get this checked in?
> Gianluca
>