Re: [PATCH 1/2] tracing: Add sysctl to enable/disable tracing on oops

[Li Zefan <lizf@cn.fujitsu.com>]

>>>>>>> Currently we always disable tracing on oops, and this patch
>>>>>>> adds a sysctl so one can choose to enable it.
>>>>>> Hmm, we already have a way to enable it.
>>>>>>
>>>>>> # echo 1 > /debug/tracing/tracing_on
>>>>>>
>>>>> What I want is a way to not disable it when an oops happened. :)
>>>>>
>>>> Ah, I misunderstood. May I ask a silly question?
>>>>
>>>> Why?
>>>>
>>> Otherwise we won't get trace output from trae_crash_kexec if
>>> crash_kexec() is not called by panic(). For example:
>>>
>>>   oops_begin()
>>> ->trace_off()
>>> ->panic_on_oops
>>> ->kexec_should_crash()
>>> ->crash_kexec()
>> OK, but I'm not exactly sure what you final goal is here. To have a
>> something to search for in the ring buffer after the crash? Maybe
>> instead we can add a "trace_oops" event? Just put it before the
>> tracing_off call.
> 
> I have another silly question.
> Why should we call tracing_off() in oops_enter()?
> 

I guess it's because trace outputs generated during oops can
overwrite/mess up those generated before oops?

It was added by this commit, but I can't find trace_printk_on_oops.

commit bdff78707f3ce47e891f3201c9666122a70556ce
Author: Thomas Gleixner <tglx@linutronix.de>
Date:   Fri Jul 24 15:30:45 2009 -0400

    trace: stop tracer in oops_enter()

    If trace_printk_on_oops is set we lose interesting trace information
    when the tracer is enabled across oops handling and printing. We want
    the trace which might give us information _WHY_ we oopsed.