Re: [Xen-ia64-devel][PATCH]Build new infrastructure for fast fault handling path.

[Isaku Yamahata <yamahata@xxxxxxxxxxxxx>]

On Fri, May 09, 2008 at 03:15:02PM +0800, Xu, Anthony wrote:

> >> diff -r f2457c7aff8d xen/arch/ia64/vmx/vmx_vcpu.c
> >> --- a/xen/arch/ia64/vmx/vmx_vcpu.c Fri Apr 25 20:13:52 2008 +0900
> >> +++ b/xen/arch/ia64/vmx/vmx_vcpu.c Fri May 09 10:58:37 2008 +0800
> >> @@ -172,11 +172,6 @@ IA64FAULT vmx_vcpu_set_rr(VCPU *vcpu, u6  {
> >>      u64 rrval;
> >> 
> >> -    if (unlikely(is_reserved_rr_rid(vcpu, val))) {
> >> -        gdprintk(XENLOG_DEBUG, "use of invalid rrval %lx\n", val);
> >> -        return IA64_RSVDREG_FAULT;
> >> -    }
> >> -
> >>      VMX(vcpu,vrr[reg>>VRN_SHIFT]) = val;
> >>      switch((u64)(reg>>VRN_SHIFT)) {
> >>      case VRN7:
> > 
> > Without the check, VTi domain guest may access other domain's page
> > or xen's page. So it can't be dropped.
> > Do you find that calling is_reserved_rr_rid() is heavy?
> > If so, please make it an inline function.
> > 
> 
> int is_reserved_rr_rid(VCPU* vcpu, u64 reg_value)
> {
>     ia64_rr rr = { .rrval = reg_value };
> 
>     if (rr.rid >= (1UL << vcpu->domain->arch.rid_bits))
>         return 1;
> 
>     return 0;
> }
> 
> 
> Because vcpu->domain is not mapped by TR, this function can not be
> called in fast path.

vcpu->arch.rid_bits can be used.
I'm worried that someone (possibly me!) may accidentaly touch
those C functions to break fast path. Probably it may be difficult
to debug it.
Could you please add warning comments to each C functions on fast path?

-- 
yamahata

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel