Hi SeongJae,
On Mon, Oct 17, 2022 at 12:27 PM SeongJae Park <sj@xxxxxxxxxx> wrote:
>
> Hello,
>
>
> The commit that this patch is fixing[1] also causes yet another segfault for
> 'perf-script' of tracepoint records. For example:
>
> $ sudo timeout 3 perf record -e exceptions:page_fault_user
> [ perf record: Woken up 1 times to write data ]
> [ perf record: Captured and wrote 0.228 MB perf.data (74 samples) ]
> $ sudo perf script
> Segmentation fault
>
> Reverting this patch and the original bug commit[1] fixes the issue. I haven't
> deep dive yet because I'm not familiar with this area. Anybody has any idea
> about this?
>
> [1] 838d9bb62d13 ("perf: Use sample_flags for raw_data")
Sorry for the trouble. I think you also need to apply the below:
https://lore.kernel.org/r/20221012143857.48198-1-james.clark@xxxxxxx
Thanks,
Namhyung
>
> On Fri, 7 Oct 2022 10:13:27 +0200 Sumanth Korikkar <sumanthk@xxxxxxxxxxxxx> wrote:
>
> > * Raw data is also filled by bpf_perf_event_output.
> > * Add sample_flags to indicate raw data.
> > * This eliminates the segfaults as shown below:
> > Run ./samples/bpf/trace_output
> > BUG pid 9 cookie 1001000000004 sized 4
> > BUG pid 9 cookie 1001000000004 sized 4
> > BUG pid 9 cookie 1001000000004 sized 4
> > Segmentation fault (core dumped)
> >
> > Fixes: 838d9bb62d13 ("perf: Use sample_flags for raw_data")
> > Acked-by: Namhyung Kim <namhyung@xxxxxxxxxx>
> > Signed-off-by: Sumanth Korikkar <sumanthk@xxxxxxxxxxxxx>
> > ---
> > kernel/trace/bpf_trace.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> > index 49fb9ec8366d..1ed08967fb97 100644
> > --- a/kernel/trace/bpf_trace.c
> > +++ b/kernel/trace/bpf_trace.c
> > @@ -687,6 +687,7 @@ BPF_CALL_5(bpf_perf_event_output, struct pt_regs *, regs, struct bpf_map *, map,
> >
> > perf_sample_data_init(sd, 0, 0);
> > sd->raw = &raw;
> > + sd->sample_flags |= PERF_SAMPLE_RAW;
> >
> > err = __bpf_perf_event_output(regs, map, flags, sd);
> >
> > @@ -745,6 +746,7 @@ u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
> > perf_fetch_caller_regs(regs);
> > perf_sample_data_init(sd, 0, 0);
> > sd->raw = &raw;
> > + sd->sample_flags |= PERF_SAMPLE_RAW;
> >
> > ret = __bpf_perf_event_output(regs, map, flags, sd);
> > out:
> > --
> > 2.36.1
