A remark to rt_task_make_periodic in lxrt (bug report !?)

Nathan Gustavson ngustavson at emacinc.com
Fri Aug 23 18:59:39 CEST 2002


I agree, I have fallen into the same pitfall several times.
It has been my experiance that it's not really rt_task_make_periodic that
causes the problem, but the confusing rt_gettime functions, which return
timer units sometimes, nanoseconds other times, and apparently random
nonsense numbers other times.

If you put the wrong one in your rt_task_make_periodic starttime, you wind
up with lock ups and weird bugs that manifest themselves depending on where
rt_task_make_periodic is called. I have finally stopped having this problem
by always using rt_get_cpu_time_ns to get the time. This function seems to
always return the time in nanoseconds regardless of what mode I'm running
in. Newbies however, should beware the time functions.

I am running in one_shot mode a lot of the time on a non-TSC machine. This
seems to cause a lot of the time functions to return a strange constant
whenever they are called that has nothing to do with the time.

Just an opionion,
 
Nathan Z. Gustavson
Embedded Engineer
EMAC, Inc.
2380 EMAC Way
Carbondale, IL  62902
Phone (618) 529-4525 ext 320
Fax (618) 457-0110
Website: http://www.emacinc.com



-----Original Message-----
From: rtai-admin at rtai.org [mailto:rtai-admin at rtai.org]On Behalf Of
andreas.spieker at daimlerchrysler.com
Sent: Friday, August 23, 2002 7:36 AM
To: rtai at rtai.org
Subject: A remark to rt_task_make_periodic in lxrt (bug report !?)


Hello,

I had serious trouble with rtai/lxrt for months -
a system with 5-10 processes in userspace,
comunicating with systemV IPC (sem, shm, msq).
I use rtai/lxrt mainly to schedule the processes
which are running with periods of 5/20/40 ms,
some are running aperiodic.

I had freezes (20-60 seconds) during the startup of my processes
and also freezes when restarting a process.
I had the problems on uniprocessor and SMP machines,
on a very slow old toshiba laptop, the freezes had a duration
of about 1 hour. But when this freezes had finished, the system
ran fine.

In order to find the error I even tried the LTT (fine).
An then, a few days ago, while trying to understand
the system behaviour with the LLT visualizer
(but even the LTT data was corrupted/data loss with a warning
from the collecting LTT module)
and I couldn't find the reason from the trace data --
I understood that the problem was due to a starting time to
    rt_task_make_periodic(...)
lying a few seconds in the past.

After two hours of work, the error was fixed and I had
no more problems until now. I'm happy with rtai/lxrt.

So I'd like to ask the developers, if they could prevent
other users from making the same experience.

Greetings, Andreas

-----------------------------------------------------
Dr.-Ing. Andreas Spieker
Research Information and Communication
Assisting Systems (RIC/AA)

DaimlerChrysler AG
HPC T728
D-70546 Stuttgart
Telefon: (+49) 711-17-41434
Telefax: (+49) 711-17-47054
email: andreas.spieker at daimlerchrysler.com
Besucheranschrift:
Hedelfinger Straße 4 / 6
Esslingen
-----------------------------------------------------

_______________________________________________
RTAI mailing list
RTAI at rtai.org
https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai





More information about the Rtai mailing list