Re: [Xen-devel] [PATCH RESEND v1 1/7] x86: add a flag to enable Intel processor trace

["Kang, Luwei" <luwei.kang@xxxxxxxxx>]

> >>> On 27.04.18 at 10:22, <luwei.kang@xxxxxxxxx> wrote:
> >> > diff --git a/docs/misc/xen-command-line.markdown
> >> > b/docs/misc/xen-command-line.markdown
> >> > index 781110d..95411cf 100644
> >> > --- a/docs/misc/xen-command-line.markdown
> >> > +++ b/docs/misc/xen-command-line.markdown
> >> > @@ -1009,6 +1009,13 @@ debug hypervisor only).
> >> >  ### idle\_latency\_factor
> >> >  > `= <integer>`
> >> >
> >> > +### intel\_pt
> >> > +> `= <boolean>`
> >> > +
> >> > +> Default: `true`
> >> > +
> >>
> >> After reading the manual a bit I think this option needs to be more
> >> sophisticated.
> >>
> >> The series only implements guest-only tracing, while in the future we
> >> might want host-only tracing and system wide tracing.
> >>
> >> Even the other modes aren't implemented yet we should leave room for
> >> them.
> >>
> >
> > Hi Wei,
> >      Thanks for the review. So what about define guest mode like this
> > and make the option as a string. Other mode can be added like 'system
> > | host' in future.
> >
> > ### Intel\_pt
> >> `= guest`
> 
> That's still too simple, as you implement it for HVM only. While it's not 
> clear
> to me whether it could be done properly for PV, that path shouldn't be
> closed unless it's crystal clear that it's impossible to ever happen.

The great change in Intel Processor Trace improvements is CPU will treat PT 
output addresses as Guest Physical Addresses (GPAs) and translate them using 
EPT. So if EPT is not supported or disabled in PV guest, intel PT feature can't 
be exposed to guest.
System wide mode may have a little different, we can disable intel address 
translated by EPT and enable PT in dom0  to record the behavior of hypervisor 
and  guest. Although Intel PT is still enabled(IA32_RTIT_CTL.EN=1) in guest but 
actually guest don't aware this. This mode is not implemented in this patch set 
and may need more discussion in next step.

> 
> Jan
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel