Xen devel - Mar 2010 - Intel: Overly restrictive test for availablility of CPUID masking MSRs?

Hi,

http://xenbits.xensource.com/xen-unstable.hg?rev/aab9fbd6ffa0 from
http://lists.xensource.com/archives/html/xen-devel/2008-07/msg00518.html
restricts the CPUID masking feature to model 0x1d or model 0x17 with
stepping >=4:
        if (!((model == 0x1d) || ((model == 0x17) && ((eax & 0xf)
>        4)))) {
        	/* fail ... */
        
A Xeon E5520 which is supposed to have FlexMigration has family 6, model
26 (0x1a) and stepping 5 fails the existing test and we think it should
be allowed.

Is there a more precise way of detecting the presence of this
capability? I''ve seen:
        family > 0x6 || (model > 0x17 || (model = 0x17 && stepping
>= 4))
suggested but this looks like it matches exactly the same set of
processors as the "eax < 0x00010674" in the code before
aab9fbd6ffa0
which was apparently too lenient?

Ian.



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

Correcting Jun''s email which is back-to-front in the Signed-off-by.

On Tue, 2010-03-02 at 14:01 +0000, Ian Campbell wrote:
> Hi,
> 
> http://xenbits.xensource.com/xen-unstable.hg?rev/aab9fbd6ffa0 from
> http://lists.xensource.com/archives/html/xen-devel/2008-07/msg00518.html
> restricts the CPUID masking feature to model 0x1d or model 0x17 with
> stepping >=4:
>         if (!((model == 0x1d) || ((model == 0x17) && ((eax &
0xf) >>         4)))) {
>         	/* fail ... */
>         
> A Xeon E5520 which is supposed to have FlexMigration has family 6, model
> 26 (0x1a) and stepping 5 fails the existing test and we think it should
> be allowed.
> 
> Is there a more precise way of detecting the presence of this
> capability? I''ve seen:
>         family > 0x6 || (model > 0x17 || (model = 0x17 &&
stepping >= 4))
> suggested but this looks like it matches exactly the same set of
> processors as the "eax < 0x00010674" in the code before
aab9fbd6ffa0
> which was apparently too lenient?
> 
> Ian.
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel


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

On Tue, 2010-03-02 at 14:11 +0000, Ian Campbell wrote:
> > Is there a more precise way of detecting the presence of this
> > capability? I''ve seen:
> >         family > 0x6 || (model > 0x17 || (model = 0x17
&& stepping >= 4))
> > suggested but this looks like it matches exactly the same set of
> > processors as the "eax < 0x00010674" in the code before
aab9fbd6ffa0
> > which was apparently too lenient?
Or alternatively, is there any reason we can''t just use wrmsr_safe
instead to gracefully handle processors which can''t cope?

Ian.


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