Re: [PATCH 04/10] perf stat: Add --exclude-guest option

["Liang, Kan" <kan.liang@linux.intel.com>]

On 2024-09-24 4:21 p.m., Namhyung Kim wrote:
> On Mon, Sep 23, 2024 at 09:47:17AM +0100, James Clark wrote:
>>
>> On 06/09/2024 3:33 pm, Liang, Kan wrote:
>>>
>>> On 2024-09-05 4:24 p.m., Namhyung Kim wrote:
>>>> This option is to support the old behavior of setting exclude_guest by
>>>> default.  Now it doesn't set the bit so users want the old behavior can
>>>> use this option.
>>>>
>>>>    $ perf stat true
>>>>
>>>>     Performance counter stats for 'true':
>>>>
>>>>                  0.86 msec task-clock:u                     #    0.443 CPUs utilized
>>>>                     0      context-switches:u               #    0.000 /sec
>>>>                     0      cpu-migrations:u                 #    0.000 /sec
>>>>                    49      page-faults:u                    #   56.889 K/sec
>>>>                   ...
>>>>
>>>>    $ perf stat --exclude-guest true
>>>>
>>>>     Performance counter stats for 'true':
>>>>
>>>>                  0.79 msec task-clock:Hu                    #    0.490 CPUs utilized
>>>>                     0      context-switches:Hu              #    0.000 /sec
>>>>                     0      cpu-migrations:Hu                #    0.000 /sec
>>>>                    49      page-faults:Hu                   #   62.078 K/sec
>>>>                   ...
>>>>
>>>> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
>>>> ---
>>>>   tools/perf/Documentation/perf-stat.txt | 7 +++++++
>>>>   tools/perf/builtin-stat.c              | 2 ++
>>>>   2 files changed, 9 insertions(+)
>>>>
>>>> diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
>>>> index 2bc06367248691dd..d28d8370a856598f 100644
>>>> --- a/tools/perf/Documentation/perf-stat.txt
>>>> +++ b/tools/perf/Documentation/perf-stat.txt
>>>> @@ -382,6 +382,13 @@ color the metric's computed value.
>>>>   Don't print output, warnings or messages. This is useful with perf stat
>>>>   record below to only write data to the perf.data file.
>>>> +--exclude-guest::
>>>> +Don't count event in the guest mode.  It was the old behavior but the
>>>> +default is changed to count guest events also.  Use this option if you
>>>> +want the old behavior (host only).  Note that this option needs to be
>>>> +before other events in case you added -e/--event option in the command
>>>> +line.
>>> I'm not sure if we really need this option. I think it may bring more
>>> trouble than what we get.
>>>
>>> The name of the "--exclude-guest" sounds like a replacement of the event
>>> modifier "H". But in fact, it's not. It should only affect the default.
>>> It doesn't set the "H" for any events.
> Well I think it's tricky but it'd set "H" modifier events after the
> option.  But I have to agree that it can bring more troubles.

I may have miss-read something before. After some simple tests, yes, the
"H" is applied with the option.

Since there is a limit for the "--exclude-guest" option, can we print a
warning if the option becomes invalid because of the order?

> 
>>> Except for the perf kvm user, I don't think there are many users which
>>> care the exclude_guest. The behavior of the perf kvm is not changed. So
>>> the option seems not that important. If we really want an option to
>>> restore the old behavior, it's better to choose a better name and update
>>> the description.
> Personally I don't want to this option but just worried if there's a
> case where exclude_guest is preferred.

The only case I can imagine is that, with the new vPMU passthrough
introduced, some users may want to explicitly set the exclude_guest to
avoid the fallback. But I'm not sure how useful it is for them.

Thanks,
Kan