Re: [PATCH] hwlat_detector: Detect hardware-induced latencies

[Daniel Wagner <wagi@monom.org>]

Hi,

On 04/09/2015 04:10 PM, Steven Rostedt wrote:
> On Thu, 09 Apr 2015 15:46:12 +0200
> Carsten Emde <C.Emde@osadl.org> wrote:
> 
>> Hi Daniel,
>>
>>> BTW, are you also taking care of the latency-hist.patch patch?
>> I do.
>>
>> However, the request (and the challenge) is not to simply rewrite the 
>> code. The idea is to provide a general histogram framework that can be 
>> used from anywhere in the kernel and to equip it with everything needed 
>> by the existing histograms. This framework should no longer be part of 
>> the RT patches but submitted directly to mainline Linux. In a second 
>> step, we will then rewrite the existing histogram code to use the new 
>> framework.
>>
>> Any help is greatly appreciated!

I'll start poking around then :)

> Hmm, Currently Tom Zanussi is working on a hist trigger for tracing.
> Not sure it can be used here, but perhaps ideas can be spawned.
> 
> http://marc.info/?l=linux-kernel&m=142807630917288

>From a quick look, it should be possible to add on top of those patches
the two TRACE_EVENTs (preemptirqsoff_hist, hrtimer_interrupt), annotate
start_critical_*() & and friends (from latency-hist.patch). I'll try
that and see what explodes.

cheers,
daniel