[PATCH][HVM][SVM] Fix machine check intercept correctly

[PATCH][HVM][SVM] Fix machine check intercept correctly

[Christoph Egger]

[-- Attachment #1: Type: text/plain, Size: 629 bytes --]


Hi!

Attached patch fixes machine check interception
for HVM guests.

Keir: Please apply this patch to xen-staging *and* to Xen 3.1.1

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>


-- 
AMD Saxony, Dresden, Germany
Operating System Research Center

Legal Information:
AMD Saxony Limited Liability Company & Co. KG
Sitz (Geschäftsanschrift):
   Wilschdorfer Landstr. 101, 01109 Dresden, Deutschland
Registergericht Dresden: HRA 4896
vertretungsberechtigter Komplementär:
   AMD Saxony LLC (Sitz Wilmington, Delaware, USA)
Geschäftsführer der AMD Saxony LLC:
   Dr. Hans-R. Deppe, Thomas McCoy

[-- Attachment #2: xen_svm_mc.diff --]
[-- Type: text/plain, Size: 791 bytes --]

diff -r 5c7afb32df99 xen/arch/x86/hvm/svm/svm.c
--- a/xen/arch/x86/hvm/svm/svm.c	Mon Oct 01 06:39:00 2007 +0100
+++ b/xen/arch/x86/hvm/svm/svm.c	Mon Oct 01 15:29:14 2007 +0200
@@ -2166,7 +2166,6 @@ asmlinkage void svm_vmexit_handler(struc
 
     case VMEXIT_EXCEPTION_MC:
         HVMTRACE_0D(MCE, v);
-        do_machine_check(regs);
         break;
 
     case VMEXIT_VINTR:
diff -r 5c7afb32df99 xen/arch/x86/hvm/svm/vmcb.c
--- a/xen/arch/x86/hvm/svm/vmcb.c	Mon Oct 01 06:39:00 2007 +0100
+++ b/xen/arch/x86/hvm/svm/vmcb.c	Wed Sep 26 13:32:24 2007 +0200
@@ -252,6 +252,7 @@ static int construct_vmcb(struct vcpu *v
     {
         vmcb->exception_intercepts = HVM_TRAP_MASK | (1U << TRAP_page_fault);
     }
+    vmcb->exception_intercepts |= 1U << TRAP_machine_check;
 
     return 0;
 }

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

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

Re: [PATCH][HVM][SVM] Fix machine check intercept correctly

[Keir Fraser]

On 1/10/07 16:33, "Christoph Egger" <Christoph.Egger@amd.com> wrote:

> Hi!
> 
> Attached patch fixes machine check interception
> for HVM guests.
> 
> Keir: Please apply this patch to xen-staging *and* to Xen 3.1.1

First chunk (adding TRAP_machine_check to exception intercepts) is not
needed as that bit is already in HVM_TRAP_MASK. Also why remove the
invocation of do_machine_check()? That is the function that actually handles
the #MC.

 -- Keir

RE: [PATCH][HVM][SVM] Fix machine check intercept correctly

[Woller, Thomas]

In theory the MCE that occurs in SVM partition, will be held pending
until the VMEXIT occurs, and then released when the STGI instruction is
executed, rendering any VMEXIT code non-useful.  Same as the NMI
handling.  I might be missing something here, but the do_machine_check()
will be executed in the context of the hv trap handlers immediately
after STGI instruction, so we don't want to call it twice in the vmexit
handler code.  (?)
tom 

> -----Original Message-----
> From: xen-devel-bounces@lists.xensource.com 
> [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of 
> Keir Fraser
> Sent: Monday, October 01, 2007 11:11 AM
> To: Egger, Christoph; xen-devel@lists.xensource.com
> Subject: Re: [Xen-devel] [PATCH][HVM][SVM] Fix machine check 
> intercept correctly
> 
> On 1/10/07 16:33, "Christoph Egger" <Christoph.Egger@amd.com> wrote:
> 
> > Hi!
> > 
> > Attached patch fixes machine check interception for HVM guests.
> > 
> > Keir: Please apply this patch to xen-staging *and* to Xen 3.1.1
> 
> First chunk (adding TRAP_machine_check to exception 
> intercepts) is not needed as that bit is already in 
> HVM_TRAP_MASK. Also why remove the invocation of 
> do_machine_check()? That is the function that actually 
> handles the #MC.
> 
>  -- Keir
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
> 
> 
> 

Re: [PATCH][HVM][SVM] Fix machine check intercept correctly

[Keir Fraser]

The documentation is unclear on what happens to #MC. Certainly it is held
pending if GIF==0, but the manual does not explicitly say that #MC is held
pending on vmexit. Is it possible to inject a #MC, to test what happens?
That would be very handy, otherwise the code path basically gets no testing.
:-)

 -- Keir

On 1/10/07 17:17, "Woller, Thomas" <thomas.woller@amd.com> wrote:

> In theory the MCE that occurs in SVM partition, will be held pending
> until the VMEXIT occurs, and then released when the STGI instruction is
> executed, rendering any VMEXIT code non-useful.  Same as the NMI
> handling.  I might be missing something here, but the do_machine_check()
> will be executed in the context of the hv trap handlers immediately
> after STGI instruction, so we don't want to call it twice in the vmexit
> handler code.  (?)
> tom 
> 
>> -----Original Message-----
>> From: xen-devel-bounces@lists.xensource.com
>> [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of
>> Keir Fraser
>> Sent: Monday, October 01, 2007 11:11 AM
>> To: Egger, Christoph; xen-devel@lists.xensource.com
>> Subject: Re: [Xen-devel] [PATCH][HVM][SVM] Fix machine check
>> intercept correctly
>> 
>> On 1/10/07 16:33, "Christoph Egger" <Christoph.Egger@amd.com> wrote:
>> 
>>> Hi!
>>> 
>>> Attached patch fixes machine check interception for HVM guests.
>>> 
>>> Keir: Please apply this patch to xen-staging *and* to Xen 3.1.1
>> 
>> First chunk (adding TRAP_machine_check to exception
>> intercepts) is not needed as that bit is already in
>> HVM_TRAP_MASK. Also why remove the invocation of
>> do_machine_check()? That is the function that actually
>> handles the #MC.
>> 
>>  -- Keir
>> 
>> 
>> 
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel
>> 
>> 
>> 
> 
>