On 07/14/2016 05:27 PM, Sebastian Andrzej Siewior wrote:

* Roosen Henri | 2016-06-30 16:05:09 [+0000]:

quoted

Hi,

Hi,

quoted

I just got an info-backtrace while running cyclictest on an iMX28 with kernel 4.1.27-rt30.
This resulted in a latency spike (i'm not sure whether caused by the trace dumping).
Can anyone share some info on this and what to do to prevent the cause of this info-trace?

Did you do anything else besides starting cyclictest?

To put the system under load we're running a "dohell" script, which
boils down to parallel running hackbench and dd, ls, ps, cat, .. commands.

The problem is easily triggered, but only after starting a flood-ping to
the PREEMPT_RT-system under test. This also results in huge latency,
much bigger than on a 'normal' PREEMPT-system, which seems to be
resistant against the flood-ping.

Any suggestions how to best trace this down?

Thanks,
Henri

RCU was somehow blocked and did not make any progress. This might if a
task in an RCU-read section got preempted for long time _or_ is active
in that section over a very long time. More information can be found at
  https://www.kernel.org/doc/Documentation/RCU/stallwarn.txt

Sebastian

________________________________

Ginzinger electronic systems GmbH
Gewerbegebiet Pirath 16
4952 Weng im Innkreis
www.ginzinger.com

Firmenbuchnummer: FN 364958d
Firmenbuchgericht: Ried im Innkreis
UID-Nr.: ATU66521089