Re: [PATCH] perf script: Skip dummy event attr check

[Jiri Olsa <olsajiri@gmail.com>]

On Wed, Sep 07, 2022 at 01:14:52AM -0700, Namhyung Kim wrote:
> On Wed, Sep 7, 2022 at 1:09 AM Jiri Olsa <olsajiri@gmail.com> wrote:
> >
> > On Wed, Sep 07, 2022 at 01:19:32PM +0800, Xing Zhengjun wrote:
> > >
> > >
> > > On 9/7/2022 12:50 PM, Namhyung Kim wrote:
> > > > Hello,
> > > >
> > > > On Tue, Sep 6, 2022 at 7:49 PM Xing Zhengjun
> > > > <zhengjun.xing@linux.intel.com> wrote:
> > > > >
> > > > > Hi,
> > > > >
> > > > > On 9/1/2022 12:23 AM, Arnaldo Carvalho de Melo wrote:
> > > > > > Em Wed, Aug 31, 2022 at 09:02:46AM -0700, Ian Rogers escreveu:
> > > > > > > On Wed, Aug 31, 2022 at 5:40 AM Jiri Olsa <jolsa@kernel.org> wrote:
> > > > > > > >
> > > > > > > > Hongtao Yu reported problem when displaying uregs in perf script
> > > > > > > > for system wide perf.data:
> > > > > > > >
> > > > > > > >     # perf script -F uregs | head -10
> > > > > > > >     Samples for 'dummy:HG' event do not have UREGS attribute set. Cannot print 'uregs' field.
> > > > > > > >
> > > > > > > > The problem is the extra dummy event added for system wide,
> > > > > > > > which does not have proper sample_type setup.
> > > > > > > >
> > > > > > > > Skipping attr check completely for dummy event as suggested
> > > > > > > > by Namhyung, because it does not have any samples anyway.
> > > > > > > >
> > > > > > > > Reported-by: Hongtao Yu <hoy@fb.com>
> > > > > > > > Suggested-by: Namhyung Kim <namhyung@kernel.org>
> > > > > > > > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > > > > > >
> > > > > > > Acked-by: Ian Rogers <irogers@google.com>
> > > > > >
> > > > > > Thanks, applied to perf/urgent.
> > > > > >
> > > > > > - Arnaldo
> > > > >
> > > > > I have met a similar issue on hybrid systems such as ADL, I apply the
> > > > > patch, and it works OK for the normal mode.
> > > > >    # ./perf record  --intr-regs=di,r8,dx,cx -e
> > > > > br_inst_retired.near_call:p -c 1000 --per-thread true
> > > > > [ perf record: Woken up 1 times to write data ]
> > > > > [ perf record: Captured and wrote 0.024 MB perf.data (25 samples) ]
> > > > >
> > > > > # ./perf script -F iregs |head -5
> > > > >    ABI:2    CX:0xffff90e19d024ed8    DX:0x1    DI:0xffff90e19d024ed8
> > > > > R8:0xffffba5e437e7b30
> > > > >    ABI:2    CX:0x7f5239783000    DX:0x80000000    DI:0xffff90e1801eea00
> > > > >    R8:0x71
> > > > >    ABI:2    CX:0x40    DX:0x60    DI:0xffffffff9fde5ab8    R8:0x40
> > > > >    ABI:2    CX:0x0    DX:0xffffffffffffffff    DI:0xffff90e1877408e8
> > > > > R8:0x1
> > > > >    ABI:2    CX:0xcc0    DX:0x1    DI:0xffff90e187d17c60    R8:0x40
> > > > >
> > > > > But the issue still happened when running the test in the pipe mode. In
> > > > > the pipe mode, it calls process_attr() which still checks the attr for
> > > > > the dummy event, so the issue happened.
> > > > >
> > > > >    # ./perf record -o - --intr-regs=di,r8,dx,cx -e
> > > > > br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf
> > > > > script -F iregs |head -5
> > > > > Samples for 'dummy:HG' event do not have IREGS attribute set. Cannot
> > > > > print 'iregs' field.
> > > > > 0x120 [0x90]: failed to process type: 64
> > > > >
> > > > > I have one test patch which can fix the pipe mode issue.
> > > > >
> > > > >    ./perf record -o - --intr-regs=di,r8,dx,cx -e
> > > > > br_inst_retired.near_call:p -c 1000 --per-thread true 2>/dev/null|./perf
> > > > > script -F iregs |head -5
> > > > >    ABI:2    CX:0xffff90e18119e278    DX:0x0    DI:0xffff90e18119f858
> > > > > R8:0xffff90e18119e278
> > > > >    ABI:2    CX:0x0    DX:0x1    DI:0xfffffa2844a91580    R8:0x402
> > > > >    ABI:2    CX:0x0    DX:0x0    DI:0x100cca    R8:0x0
> > > > >    ABI:2    CX:0x0    DX:0x0    DI:0xffffffff9e997ca5    R8:0x0
> > > > >    ABI:2    CX:0x113ce8000    DX:0xffff90e198f46600
> > > > > DI:0xffff90e189de8000    R8:0x290
> > > > >
> > > > >
> > > > > Fixes: b91e5492f9d7 ("perf record: Add a dummy event on hybrid systems
> > > > > to collect metadata records")
> > > > > Signed-off-by: Zhengjun Xing <zhengjun.xing@linux.intel.com>
> > > > > ---
> > > > >    tools/perf/builtin-script.c | 2 ++
> > > > >    1 file changed, 2 insertions(+)
> > > > >
> > > > > diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
> > > > > index 9152e3e45b69..2d863cdb19fe 100644
> > > > > --- a/tools/perf/builtin-script.c
> > > > > +++ b/tools/perf/builtin-script.c
> > > > > @@ -2429,6 +2429,8 @@ static int process_attr(struct perf_tool *tool,
> > > > > union perf_event *event,
> > > > >           }
> > > > >
> > > > >           if (evsel->core.attr.sample_type) {
> > > > > +               if (evsel__is_dummy_event(evsel))
> > > > > +                       return 0;
> > > >
> > > > Maybe we can move this into evsel__check_attr().
> > > >
> > > > Thanks,
> > > > Namhyung
> > >
> > > Yes, the following changes in evsel__check_attr() can fix both normal and
> > > pipe mode issues, Otherwise, we have to patch everywhere when
> > > evsel__check_attr() is called.
> > >
> > > diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
> > > index 13580a9c50b8..fb76e3191858 100644
> > > --- a/tools/perf/builtin-script.c
> > > +++ b/tools/perf/builtin-script.c
> > > @@ -451,6 +451,9 @@ static int evsel__check_attr(struct evsel *evsel, struct
> > > perf_session *session)
> > >         allow_user_set = perf_header__has_feat(&session->header,
> > >                                                HEADER_AUXTRACE);
> > >
> > > +       if (evsel__is_dummy_event(evsel))
> > > +               allow_user_set = true;
> > > +
> >
> > hm, do you need to pass allow_user_set to UREGS check then?
> 
> Well.. actually I thought we can simply return 0 for a dummy event.

true :-)

jirka