Re: [PATCH] perf: record: Clear BPF headers if not needed

[Namhyung Kim <namhyung@kernel.org>]

On Wed, Jul 01, 2026 at 02:31:47PM +0000, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [Medium] The patch conditionally clears BPF header features based on `rec->opts.no_bpf_event` too early in `record__init_features()`, failing to fix the spurious error messages for default and dynamically evaluated cases when `perf` is compiled without `libbpf`.
> --
> 
> commit 9515d1b27d6dcc9884718d4505a8cf399d665c9e
> Author: Mayuresh Chitale <mayuresh.chitale@oss.qualcomm.com>
> 
> perf: record: Clear BPF headers if not needed
> 
> This patch clears the HEADER_BPF_PROG_INFO and HEADER_BPF_BTF bits from the
> file header feature bitmask when the '--no-bpf-event' flag is passed. This
> aims to prevent spurious error messages about missing libbpf support when
> writing perf data.
> 
> > diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> > index e915390556752..692ab7138d2c3 100644
> > --- a/tools/perf/builtin-record.c
> > +++ b/tools/perf/builtin-record.c
> > @@ -1829,6 +1829,11 @@ static void record__init_features(struct record *rec)
> >  	if (!record__comp_enabled(rec))
> >  		perf_header__clear_feat(&session->header, HEADER_COMPRESSED);
> >  
> > +	if (rec->opts.no_bpf_event) {
> > +		perf_header__clear_feat(&session->header, HEADER_BPF_PROG_INFO);
> > +		perf_header__clear_feat(&session->header, HEADER_BPF_BTF);
> > +	}
> > +
> 
> [Severity: Medium]
> Does this conditionally clear the features too early in the initialization
> process?
> 
> If the user does not explicitly pass the flag (the default case), or if the
> flag is dynamically updated to true later (e.g. if updated after checking
> !evlist__needs_bpf_sb_event(rec->evlist)), this condition will evaluate to
> false.

Probably we can move this check before record__init_features().  It
seems there's nothing can change the evlist between the two points.

Thanks,
Namhyung

> 
> Will this leave the spurious errors unresolved for default and dynamically
> evaluated cases when perf is compiled without libbpf?
> 
> >  	perf_header__clear_feat(&session->header, HEADER_STAT);
> >  }
> 
> -- 
> Sashiko AI review · https://sashiko.dev/#/patchset/20260701141347.4098670-1-mayuresh.chitale@oss.qualcomm.com?part=1