Re: [RFC 8/9] RISC-V: KVM: Implement perf support
From: Atish Patra
Date: Wed Dec 07 2022 - 20:11:26 EST
On Wed, Dec 7, 2022 at 8:31 AM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
>
> On Wed, Dec 07, 2022, Atish Patra wrote:
> > On Fri, Dec 2, 2022 at 9:09 AM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
> > >
> > > On Mon, Jul 18, 2022, Atish Patra wrote:
> > > > RISC-V SBI PMU & Sscofpmf ISA extension allows supporting perf in
> > > > the virtualization enviornment as well. KVM implementation
> > > > relies on SBI PMU extension for most of the part while traps
> > > > & emulates the CSRs read for counter access.
> > >
> > > For the benefit of non-RISCV people, the changelog (and documentation?) should
> > > explain why RISC-V doesn't need to tap into kvm_register_perf_callbacks().
> >
> > As per my understanding, kvm_register_perf_callbacks is only useful
> > during event sampling for guests. Please let me know if I missed
> > something.
> > This series doesn't support sampling and guest counter overflow interrupt yet.
> > That's why kvm_register_perf_callbacks support is missing.
>
> Ah, I missed that connection in the cover letter.
>
> In the future, if a patch adds partial support for a thing/feature, it's very
> helpful to call that out in the lack shortlog and changelog, even for RFCs. E.g.
> adding a single word in the shortlog and sentence or two in the changelog doesn't
> take much time on your end, and helps avoid cases like this where drive-by reviewers
> like me from cause a fuss about non-issues.
>
Absolutely. I will update the commit text in v2 accordingly.
> RISC-V: KVM: Implement partial perf support
>
> ...
>
> Counter overflow and interrupts are not supported as the relevant
> architectural specifications are still under discussion.
>
> Thanks!
--
Regards,
Atish