Re: [PATCH RFC 2/9] xen: Optimize introspection access to guest state

[Razvan Cojocaru <rcojocaru@bitdefender.com>]

On 07/10/2014 02:57 PM, Razvan Cojocaru wrote:
> So this solution would appear to be a bit more involved than the initial
> solution. But you're, of course, right that guest_x86_mode should not be
> VMX-specific.
> 
> Would it be OK if I would replace the call to vmx_guest_x86_mode() to a
> call to hvm_funcs.guest_x86_mode(v) (assuming that's possible)?

Actually, I probably misunderstood this, since "c->guest_x86_mode =
vmx_guest_x86_mode(v);" only happens in vmx.c, and p2m.c does
"req->regs.guest_x86_mode = hvm_guest_x86_mode(current);".

So I think extending this to SVM would simply require a change similar
to the one in vmx.c in svm.c (presumably in svm_vmcb_save(), with the
help of svm_guest_x86_mode()).

How does this tie any SVM extension of our work to VT-x internals?


Thanks,
Razvan Cojocaru