Re: [Xen-devel] [PATCH v2 for-4.9 6/6] x86/emul: Require callers to provide LMA in the emulation context

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 05.04.17 at 19:33, <andrew.cooper3@xxxxxxxxxx> wrote:
> --- a/xen/arch/x86/mm.c
> +++ b/xen/arch/x86/mm.c
> @@ -5412,6 +5412,7 @@ int ptwr_do_page_fault(struct vcpu *v, unsigned long 
> addr,
>              .vendor = d->arch.cpuid->x86_vendor,
>              .addr_size = is_pv_32bit_domain(d) ? 32 : BITS_PER_LONG,
>              .sp_size   = is_pv_32bit_domain(d) ? 32 : BITS_PER_LONG,
> +            .lma = true,
>          },
>      };
>      int rc;
> @@ -5566,6 +5567,7 @@ int mmio_ro_do_page_fault(struct vcpu *v, unsigned long 
> addr,
>          .vendor = v->domain->arch.cpuid->x86_vendor,
>          .addr_size = addr_size,
>          .sp_size = addr_size,
> +        .lma = true,

As mentioned elsewhere already, I continue to consider this wrong
for 32-bit PV guests. I don't think there is any requirement for them
to be meaningfully aware of possibly running in long mode, at least
as far as segmentation is concerned. While likely benign right now,
this would become an active issue if any of the paths into
x86_emulate() wanted to have call gate use emulated (once the
function supports that).

> --- a/xen/arch/x86/x86_emulate/x86_emulate.c
> +++ b/xen/arch/x86/x86_emulate/x86_emulate.c

Can x86_emulate_wrapper() please gain

    ASSERT(!mode_64bit() || ctxt->lma);

or some equivalent?

Jan


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