Re: Unable to use tools/perf sched and timechart correctly

From: TÃrÃk Edwin
Date: Sun Dec 06 2009 - 07:52:38 EST


On 2009-12-06 12:41, Xiao Guangrong wrote:
> TÃrÃk Edwin wrote:
>
>> Hi,
>>
>> I installed 2.6.32 and tried its tools/perf.
>> I noticed two problems so far, please tell me if I am doing anything
>> wrong here:
>>
>> 1. 'perf sched replay' segfaults on any recorded workload
>> I tested like this:
>> $ sudo ./perf sched record
>> ^C[ perf record: Woken up 1 times to write data ]
>> [ perf record: Captured and wrote 0.295 MB perf.data (~12869 samples) ]
>> $ sudo perf sched replay
>> run measurement overhead: 1258 nsecs
>> sleep measurement overhead: 55232 nsecs
>> the run test took 999778 nsecs
>> the sleep test took 1055791 nsecs
>> nr_run_events: 182
>> nr_sleep_events: 209
>> nr_wakeup_events: 103
>> target-less wakeups: 12
>> [...]
>> task 25 ( icedove-bin: 8757), nr_events: 3
>> task 26 ( events/0: 15), nr_events: 3
>> ------------------------------------------------------------
>> Segmentation fault
>>
>
> You should enable "CONFIG_SCHED_DEBUG", because we use
> "/proc/<pid>/sched" file to get the task's cpu usage.
>

Thanks, that got rid of the segfault, and perf sched replay runs without
crashing now.

Am I missing a config option for 'perf timechart' to work?
I tried running 4 infloop executables (on a 4-core box), the last one as
argument to perf timechart record.
Only one of those processes has shown up, as using CPU for 26.5ms, and
then sleeping (when in fact it was running the entire time).
The CPU graphs are mostly white, with a thicker blue line where a.out
shown up, and some scarce thin blue lines.


Best regards,
--Edwin
--
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/