some ftrace events are missed

some ftrace events are missed

[song]

hi guys,

I'm working on an issue with ftrace events, the call stack is like that:

__cobalt_clock_nanosleep
     xnthread_suspend
         xntimer_start
             xnclock_core_ns_to_ticks
                 xntimer_enqueue_and_program
                     xnclock_program_shot()
                         xnclock_core_local_shot
                             ipipe_timer_set


I tried to enable ftrace event with "-f" in cyclictest,however, i can 
only find:

cobalt_clock_entry: clock_irq=4
cobalt_timer_expire: timer=ffffffc07ffd5800
cobalt_clock_exit: clock_irq=4
cobalt_timer_start ....

event traces like trace_cobalt_clock_nanosleep or 
trace_cobalt_thread_suspend are missed, they are all defined in 
cobalt_core.h/cobalt_posix.h and visible in 
/sys/kernel/debugfs/tracing/events/cobalt-core(cobalt-posix).

Does anyone have the same problem before and i would appreciate it if 
you could let me know the way to solve it,many thanks

BR

Song

Re: some ftrace events are missed

[Jan Kiszka]

On 02.02.20 13:22, song via Xenomai wrote:
> hi guys,
> 
> I'm working on an issue with ftrace events, the call stack is like that:
> 
> __cobalt_clock_nanosleep
>      xnthread_suspend
>          xntimer_start
>              xnclock_core_ns_to_ticks
>                  xntimer_enqueue_and_program
>                      xnclock_program_shot()
>                          xnclock_core_local_shot
>                              ipipe_timer_set
> 
> 
> I tried to enable ftrace event with "-f" in cyclictest,however, i can 
> only find:
> 
> cobalt_clock_entry: clock_irq=4
> cobalt_timer_expire: timer=ffffffc07ffd5800
> cobalt_clock_exit: clock_irq=4
> cobalt_timer_start ....
> 
> event traces like trace_cobalt_clock_nanosleep or 
> trace_cobalt_thread_suspend are missed, they are all defined in 
> cobalt_core.h/cobalt_posix.h and visible in 
> /sys/kernel/debugfs/tracing/events/cobalt-core(cobalt-posix).

Are those events also enabled when you do the tracing? Try catching them 
via "trace-cmd record -e 'cobalt*' if unsure.

Jan

> 
> Does anyone have the same problem before and i would appreciate it if 
> you could let me know the way to solve it,many thanks
> 
> BR
> 
> Song
> 
> 
> 
> 
> 
> 
> 
> 
-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

Re: some ftrace events are missed

[song]

I was pretty sure that all events were enabled before testing with "echo 
1 > events/enable", somehow, the events i cared about were missed.I used 
trace-cmd as your suggestion, it took effect, i saw what i wanted.

I haven't figured out how different they are, but at least i can 
continue my work in my hand, many thanks.

BR

/Song

On 2020/2/3 下午3:00, Jan Kiszka wrote:
> On 02.02.20 13:22, song via Xenomai wrote:
>> hi guys,
>>
>> I'm working on an issue with ftrace events, the call stack is like that:
>>
>> __cobalt_clock_nanosleep
>>      xnthread_suspend
>>          xntimer_start
>>              xnclock_core_ns_to_ticks
>>                  xntimer_enqueue_and_program
>>                      xnclock_program_shot()
>>                          xnclock_core_local_shot
>>                              ipipe_timer_set
>>
>>
>> I tried to enable ftrace event with "-f" in cyclictest,however, i can 
>> only find:
>>
>> cobalt_clock_entry: clock_irq=4
>> cobalt_timer_expire: timer=ffffffc07ffd5800
>> cobalt_clock_exit: clock_irq=4
>> cobalt_timer_start ....
>>
>> event traces like trace_cobalt_clock_nanosleep or 
>> trace_cobalt_thread_suspend are missed, they are all defined in 
>> cobalt_core.h/cobalt_posix.h and visible in 
>> /sys/kernel/debugfs/tracing/events/cobalt-core(cobalt-posix).
>
> Are those events also enabled when you do the tracing? Try catching 
> them via "trace-cmd record -e 'cobalt*' if unsure.
>
> Jan
>
>>
>> Does anyone have the same problem before and i would appreciate it if 
>> you could let me know the way to solve it,many thanks
>>
>> BR
>>
>> Song
>>
>>
>>
>>
>>
>>
>>
>>