Re: [Xen-devel] [PATCH v2 10/30] xen/x86: Annotate VM applicability in featureset

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 05.02.16 at 14:42, <andrew.cooper3@xxxxxxxxxx> wrote:
> Use attributes to specify whether a feature is applicable to be exposed to:
>  1) All guests
>  2) HVM guests
>  3) HVM HAP guests

No provisions for PV-only or shadow-only features?

> +#define X86_FEATURE_MTRR          ( 0*32+12) /*S  Memory Type Range 
> Registers */

Why is this being hidden from PV guests (namely Dom0)? Right now
this is being cleared for PHV only.

> +#define X86_FEATURE_VMXE          ( 1*32+ 5) /*S  Virtual Machine Extensions 
> */

Shouldn't this get a "nested-only" class? Also don't we currently
require HAP for nested mode to work?

>  #define X86_FEATURE_OSXSAVE       ( 1*32+27) /*   OSXSAVE */

Leaving this untouched warrants at least a comment in the commit
message I would think.

> +#define X86_FEATURE_RDTSCP        ( 2*32+27) /*S  RDTSCP */

Hmm, I'm confused - on one hand we currently clear this bit for
PV guests, but otoh do_invalid_op() emulates it.

> +#define X86_FEATURE_LM            ( 2*32+29) /*A  Long Mode (x86-64) */
> [...]
> -#define X86_FEATURE_LAHF_LM       ( 3*32+ 0) /*   LAHF/SAHF in long mode */
> +#define X86_FEATURE_LAHF_LM       ( 3*32+ 0) /*A  LAHF/SAHF in long mode */

How do you intend to handle exposing these to 64-bit PV guests,
but not to 32-bit ones?

>  #define X86_FEATURE_EXTAPIC       ( 3*32+ 3) /*   Extended APIC space */

This currently is left untouched for HVM guests, and gets cleared
only when !cpu_has_apic (i.e. effectively never) for PV ones.

>  #define X86_FEATURE_MWAITX        ( 3*32+29) /*   MWAIT extension 
> (MONITORX/MWAITX) */

Why not exposed to HVM (also for _MWAIT as I now notice)?

Jan


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