Re: [RFC PATCH 12/16]: PVH xen: return PVH features during creation, etc...

[Mukesh Rathor <mukesh.rathor@oracle.com>]

On Fri, 18 Jan 2013 09:23:01 +0000
"Jan Beulich" <JBeulich@suse.com> wrote:

> >>> On 18.01.13 at 03:29, Mukesh Rathor <mukesh.rathor@oracle.com>
> >>> wrote:
> > Done. No other callers of prop_page_fault for PVH. So are you OK
> > with something like this:
> > 
> > static noinline int vmxit_invalid_op(struct cpu_user_regs *regs)
> > {
> >     ulong addr=0;
> > 
> >     if ( guest_kernel_mode(current, regs) ||
> >          (addr = emulate_forced_invalid_op(regs)) == 0 )
> 
> Actually, on a second thought that depends on whether you want
> to be able to build kernels that can run both PV and PVH. If so,
> you may need to emulate this even for the guest kernel.

Actually, I changed linux code so that PVH paths will only go 
thru native_cpuid(). So we can leave this as is to discourage
future XEN_EMULATE_PREFIX. Sound good? In PV mode it will not be 
running in HVM container, hence not go thru this path.


> >     if ( is_pvh_vcpu(current) )
> >         regs->eip = eip;
> >     else
> >         instruction_done(regs, eip, 0);
> 
> Why can't you use instruction_done() (or make it fit your needs,
> so that other code wanting to use it wouldn't need similar special
> casing)?

Sure, NP. Done.

Thanks,
Mukesh