Re: Re: [PATCH][v4] PV extension of HVM(hybrid) support in Xen

[Sheng Yang <sheng@linux.intel.com>]

On Tuesday 02 March 2010 12:39:34 Jeremy Fitzhardinge wrote:
> On 03/01/2010 07:36 PM, Sheng Yang wrote:
> >> static u64 pvclock_get_nsec_offset(struct pvclock_shadow_time *shadow)
> >> {
> >>         u64 delta = native_read_tsc() - shadow->tsc_timestamp;
> >>         return scale_delta(delta, shadow->tsc_to_nsec_mul, shadow-
> >> tsc_shift);
> >> }
> >
> > tsc_timestamp take the vcpu beginning at 0, so that's the assumption.
> 
> Why would it be 0?  Xen sets tsc_timestamp to the current tsc when it
> updates the time parameters, which is whenever the vcpu is scheduled on
> a pcpu (and other times).  There's no expectation that the tsc starts
> from 0, since that won't ever be the case.
> 
Sorry, I misunderstood. HVM assume it start from 0... PV is following the 
native. We set the offset to 0, so that PV tsc is the same as native.

-- 
regards
Yang, Sheng