linux-kernel - Re: [PATCH] perf evsel: Don't set sample_regs_intr/sample_regs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <deb02b38-49eb-088d-6ebf-b5396e50c725@linux.intel.com>
Date:   Fri, 17 Jul 2020 11:33:46 +0800
From:   "Jin, Yao" <yao.jin@...ux.intel.com>
To:     Ian Rogers <irogers@...gle.com>, Jiri Olsa <jolsa@...hat.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>
Cc:     Jiri Olsa <jolsa@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        LKML <Linux-kernel@...r.kernel.org>,
        Andi Kleen <ak@...ux.intel.com>, kan.liang@...el.com,
        "Jin, Yao" <yao.jin@...el.com>,
        Adrian Hunter <adrian.hunter@...el.com>
Subject: Re: [PATCH] perf evsel: Don't set sample_regs_intr/sample_regs_user
 for dummy event

Hi,

On 7/6/2020 8:55 AM, Jin, Yao wrote:
> Hi Ian,
> 
> On 7/6/2020 8:47 AM, Ian Rogers wrote:
>> On Fri, Jul 3, 2020 at 5:31 PM Jin, Yao <yao.jin@...ux.intel.com> wrote:
>>>
>>> Hi Jiri,
>>>
>>> On 7/3/2020 7:00 PM, Jiri Olsa wrote:
>>>> On Fri, Jul 03, 2020 at 08:42:15AM +0800, Jin Yao wrote:
>>>>> Since commit 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis"),
>>>>> a dummy event is added to capture mmaps.
>>>>>
>>>>> But if we run perf-record as,
>>>>>
>>>>>    # perf record -e cycles:p -IXMM0 -a -- sleep 1
>>>>>    Error:
>>>>>    dummy:HG: PMU Hardware doesn't support sampling/overflow-interrupts. Try 'perf stat'
>>>>>
>>
>> Sorry for the breakage caused by modifying the dummy event. Could we
>> add a test to cover the issue? Perhaps in tools/perf/tests/shell/.
>> Trying to reproduce with a register on my skylakex on a 5.6.14 kernel
>> with:
>>
>> $ perf record -e cycles:p -IAX -a -- sleep 1
>>
>> succeeds.
>>
>> Thanks,
>> Ian
>>
> 
> -IAX should be no problem. The issue only occurs on the platform with extended regs supports, such 
> as ICL. So I don't know if it's suitable to add it to perf test suite.
> 
> Thanks
> Jin Yao
> 

Can this fix patch be accepted?

Thanks
Jin Yao

>>>>> The issue is, if we enable the extended regs (-IXMM0), but the
>>>>> pmu->capabilities is not set with PERF_PMU_CAP_EXTENDED_REGS, the kernel
>>>>> will return -EOPNOTSUPP error.
>>>>>
>>>>> See following code pieces.
>>>>>
>>>>> /* in kernel/events/core.c */
>>>>> static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)
>>>>>
>>>>> {
>>>>>       ....
>>>>>       if (!(pmu->capabilities & PERF_PMU_CAP_EXTENDED_REGS) &&
>>>>>           has_extended_regs(event))
>>>>>               ret = -EOPNOTSUPP;
>>>>>       ....
>>>>> }
>>>>>
>>>>> For software dummy event, the PMU should be not set with
>>>>> PERF_PMU_CAP_EXTENDED_REGS. But unfortunately in current code, the dummy
>>>>> event has possibility to be set with PERF_REG_EXTENDED_MASK bit.
>>>>>
>>>>> In evsel__config, /* tools/perf/util/evsel.c */
>>>>>
>>>>> if (opts->sample_intr_regs) {
>>>>>       attr->sample_regs_intr = opts->sample_intr_regs;
>>>>> }
>>>>>
>>>>> If we use -IXMM0, the attr>sample_regs_intr will be set with
>>>>> PERF_REG_EXTENDED_MASK bit.
>>>>>
>>>>> It doesn't make sense to set attr->sample_regs_intr for a
>>>>> software dummy event.
>>>>>
>>>>> This patch adds dummy event checking before setting
>>>>> attr->sample_regs_intr.
>>>>>
>>>>> After:
>>>>>     # ./perf record -e cycles:p -IXMM0 -a -- sleep 1
>>>>>     [ perf record: Woken up 1 times to write data ]
>>>>>     [ perf record: Captured and wrote 0.413 MB perf.data (45 samples) ]
>>>>
>>>> LGTM, Adrian (cc-ed) just added another check to the same place,
>>>> but it looks like both of them should be there:
>>>>
>>>>     https://lore.kernel.org/lkml/20200630133935.11150-2-adrian.hunter@intel.com/
>>>>
>>>> jirka
>>>>
>>>
>>> Thanks Jiri! Yes, it looks like both of checks should be added here.
>>>
>>> So do I post v2 (just rebase) once Adrian's patch gets merged?
>>>
>>> Thanks
>>> Jin Yao
>>>
>>>>>
>>>>> Fixes: 0a892c1c9472 ("perf record: Add dummy event during system wide synthesis")
>>>>> Signed-off-by: Jin Yao <yao.jin@...ux.intel.com>
>>>>> ---
>>>>>    tools/perf/util/evsel.c | 4 ++--
>>>>>    1 file changed, 2 insertions(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
>>>>> index 96e5171dce41..df3315543e86 100644
>>>>> --- a/tools/perf/util/evsel.c
>>>>> +++ b/tools/perf/util/evsel.c
>>>>> @@ -1020,12 +1020,12 @@ void evsel__config(struct evsel *evsel, struct record_opts *opts,
>>>>>       if (callchain && callchain->enabled && !evsel->no_aux_samples)
>>>>>               evsel__config_callchain(evsel, opts, callchain);
>>>>>
>>>>> -    if (opts->sample_intr_regs) {
>>>>> +    if (opts->sample_intr_regs && !is_dummy_event(evsel)) {
>>>>>               attr->sample_regs_intr = opts->sample_intr_regs;
>>>>>               evsel__set_sample_bit(evsel, REGS_INTR);
>>>>>       }
>>>>>
>>>>> -    if (opts->sample_user_regs) {
>>>>> +    if (opts->sample_user_regs && !is_dummy_event(evsel)) {
>>>>>               attr->sample_regs_user |= opts->sample_user_regs;
>>>>>               evsel__set_sample_bit(evsel, REGS_USER);
>>>>>       }
>>>>> -- 
>>>>> 2.17.1
>>>>>
>>>>