Re: [PATCH RFC] perf/x86/amd/lbr: Don't freeze the LBR on PMI for user-only filters

[Andrii Nakryiko]

On Thu, Nov 12, 2026 at 3:09 AM Ravi Bangoria <ravi.bangoria@xxxxxxx> wrote:
>
> Puranjay, Andrii,
>
> >> heh, so we basically either
> >>
> >> a) opt-in freeze with *new config*, but do it *only for non-kernel LBR*
> >>
> >> or
> >>
> >> b) opt-out freeze *only for non-kernel LBR* _unconditionally_, but
> >> without adding any new config
> >>
> >> In both cases we have a behavior that makes sense in practice, but a)
> >> adds some new config, while b) doesn't add anything. End result is the
> >> same, just more configuration and need to change user space code to
> >> opt-out the freeze.
> >>
> >> Is a) worth it, or we should stick to b) and keep it simple, Sandipan?
>
> Have you run a PGO experiment with the patch, and does it show a
> benefit? Also, have you seen any regression in any benchmarks? If
> no-freeze regresses some of them, we may need to give the user an
> option to keep the freeze enabled. No?

One of Meta teams internally reported that disabling freezing
unconditionally through the kernel command line showed a lot of
benefit in terms of data consistency, and I don't believe any sort of
regression was reported.

If anything, that same team is asking to unconditionally disable
freezing fleet-wide, and this work by Puranjay is an attempt to have
some middle ground where we don't have to do this all-or-nothing
disablement through the kernel command line.

>
> Thanks,
> Ravi