Re: [PATCH 6.19 027/844] perf metricgroup: Don't early exit if no CPUID table exists

[Jiri Slaby]

On 09. 03. 26, 15:57, Sasha Levin wrote:
> On Mon, Mar 09, 2026 at 07:53:10AM -0700, Ian Rogers wrote:
> > On Mon, Mar 9, 2026 at 6:40 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> 
> > wrote:
> > > On Mon, Mar 09, 2026 at 08:40:33AM +0100, Jiri Slaby wrote:
> > > > On 28. 02. 26, 18:19, Sasha Levin wrote:
> > > > > From: Ian Rogers <irogers@xxxxxxxxxx>
> > > > >
> > > > > [ Upstream commit cee275edcdb1acfdc8270f80e96f30750b633220 ]
> > > >
> > > > This breaks (userspace) perf:
> > > > $ ./perf stat -a -d -p 1 sleep 5
> > > > PID/TID switch overriding SYSTEM
> > > > Error:
> > > > No supported events found.
> > > >
> > > > Any ideas?
> > >
> > > Is it also broken in 7.0-rc3?  Or is this only a 6.19.y issue?

It's broken in linus/master too.

> > There was a fix:
> > https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools- 
> > next.git/commit/tools/perf/util/metricgroup.c?h=perf-tools- 
> > next&id=c5a244bf17caf2de22f9e100832b75f72b31d3e6
> > was that applied to 6.19.y?
>
> It's not even upstream yet :)

Right. (And it fixes the issue.)

thanks,
--
js
suse labs