Re: [PATCH v3 5/5] perf/sdt: Add support to perf record to trace SDT events

[Masami Hiramatsu]

(2014/10/23 14:54), Srikar Dronamraju wrote:
> * Masami Hiramatsu <masami.hiramatsu.pt@xxxxxxxxxxx> [2014-10-22 18:41:58]:
> 
>> (2014/10/22 17:20), Hemant Kumar wrote:
>>>>> From "file_sdt_ent" we will find out the file name.
>>>>> Convert this sdt note into a perf event and then write this into uprobe_events
>>>>> file to be able to record the event.
>>>>> Then, corresponding entries are added to uprobe_events file for
>>>>> the SDT events.
>>>>> After recording is done, these events are silently deleted from uprobe_events
>>>>> file. The uprobe_events file is present in debugfs/tracing directory.
>>>>>
>>>>> To support the addition and deletion of SDT events to/from uprobe_events
>>>>> file, a record_sdt struct is maintained which has the event data.
>>>> OK, I have some comments on this.
>>>>
>>>>> An example usage:
>>>>>
>>>>> # ./perf record -e %user_app:fun_start -aR /home/user_app
>>>> At first, I'd like to add SDT support for adding probes too, like below;
>>>>
>>>> ./perf probe -a '%user_app:fun_start $vars'
>>>
>>> But I think, previously we discussed that we won't be having "perf
>>> probe" for SDT events.
>>> We list them and probe/trace them using "perf record" directly.
>>
>> Right, sorry for confusing you. I meant that I'd like to support SDT on both of
>> perf-record and perf-probe :)
>> And even if we'll hide sdt related events via perf, users can access it via ftrace.
>> So, I doubt that we can completely hide them, in that case, honesty is the best way;)
>>
> 
> I am somehow not able to figure out how perf probe comes into the
> current workflow.
> 
> I think the current design was
> 1. perf sdt-cache --add <file> (only once per file)
> 2. perf record -e <sdt-event>
> 
> So what is the additional thing that perf probe does or Is it going to
> replace any of the above steps?

3. perf probe -a <sdt-event>

And this will be done subsequently in this series (without user interface part).
However, current implementation of 2. will do the following steps

s1. get sdt event data from sdt-cache
s2. set up sdt events with suppressing messages
s3. do recording events
(s4. and hiding existing sdt events from perf-probe --list)
s5. remove sdt events

So, what I proposed were ;
- to implement s2., we can introduce --quiet(-q) option and use it
  instead of ->sdt flag checking
- removing s4. and s5.
- and add verification of existing sdt events at s2. if needed.

This will simplify your patch and removing complex part of sdt-specific code.
What would you think about this?

Thank you,

-- 
Masami HIRAMATSU
Software Platform Research Dept. Linux Technology Research Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@xxxxxxxxxxx


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/