Re: [RFC/PATCH 0/4] perf tools: Do not set attr.exclude_guest by default (v1)

[Peter Zijlstra]

On Mon, Sep 02, 2024 at 10:20:41AM -0700, Namhyung Kim wrote:
> Hi Peter,
> 
> On Mon, Sep 2, 2024 at 1:56 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> >
> > On Sun, Sep 01, 2024 at 06:46:17PM -0700, Namhyung Kim wrote:
> > > Hello,
> > >
> > > I found perf tools set exclude_guest bit inconsistently.  It used to
> > > set the bit but now the default event for perf record doesn't.  So I'm
> > > wondering why we want the bit in the first place.
> > >
> > > Actually it's not good for PMUs don't support any exclusion like AMD
> > > IBS because it disables new features after the exclude_guest due to
> > > the missing feature detection logic.
> > >
> > > AFAIK it doesn't matter for the most cases but perf kvm.  If users
> > > need to set the bit, they can still use :H modifier.  Please let me
> > > know if it's ok for you.
> >
> > IIRc the point of setting exclude_guest by default was so that default
> > perf keeps working in the precense of that PMU pass through mess, no?
> 
> Sorry I haven't followed the vPMU pass through threads.  It makes
> sense to keep the exclude_guest for the kernels with it.  But I guess
> it's configurable via kernel config and/or boot option.

Yeah, you need KVM build with the passthrough thing on, and then at
runtime enable it or something -- I too have to still read the latest
posting.

> A question, with PMU pass through would the kernel reject host events
> which don't set the exclude_guest?

Yes.