Errr, another bug was found.
Probably the constant, XENOPROF_TRACE_BEGIN, should be changed.
Please wait applying this patch series. Sorry for confusion.
On Wed, Dec 12, 2007 at 11:39:06AM +0900, Isaku Yamahata wrote:
> On Tue, Dec 11, 2007 at 10:03:33AM -0700, Alex Williamson wrote:
> >
> > On Tue, 2007-12-11 at 18:02 +0900, Isaku Yamahata wrote:
> > > This patch set addresses the xenorpfo/ia64 issues reported by Xing.
> > >
> > > Xing.
> > > Could you please try this patch set and report the result?
> > >
> > > The fundamental change is
> > > - make psr virtualization code not touch mPSR.pp.
> > > The Xenoprof basic idea is that xenoprof owns PMU so that this is
> > > required.
> > > Anyway the currently code doesn't seem to handle mPSR.pp consistently
> > > so that some kind of clean up is needed.
> > > This clean up might upset a guset that think it's owning PMU.
> > > But at this moment PMU isn't virtualized so that this doesn't an issue.
> > >
> > > Limitations:
> > > Currently perfmon knows only kernel mode (cpl=0) and user mode(cpl=3).
> > > On the other hand Xen/IA64 uses other cpls.
> > > At this moment only generic pmc/pmd's are supported.
> >
> > I haven't tested this, but it looks ok to me. It would be nice to
> > have xenoprof working for 3.2. Please let me know when you have
> > confidence this is ready for inclusion. Thanks,
>
> I'm already confident for inclusion.
> My rough test showed that sampling of xen VMM, PV kernel, PV user process,
> VTi kernel, VTi user process was done as expected.
>
> Probably a document is necessary on how to use and its limitation.
> So I'll write it.
> --
> yamahata
>
> _______________________________________________
> Xen-ia64-devel mailing list
> Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-ia64-devel
>
--
yamahata
_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel
|
Home