* periodic bursts
@ 2007-10-01 6:51 Kaya, Sinan
2007-10-01 17:40 ` Jan Kiszka
0 siblings, 1 reply; 2+ messages in thread
From: Kaya, Sinan @ 2007-10-01 6:51 UTC (permalink / raw)
To: linux-rt-users
Hello all,
My realtime system experiences periodic bursts every 4 hours. How can i find the reason for this ? I know the traceit tool but it dumps so much data and it is useless under this case. I need to find what happened at the exact time of burst.
I use rtai_smi module to disable SMI interrupts and set my network interface driver priorities to 99. So it should be something non maskable but what ?
Sinan.
Ömer Sinan Kaya (M.Sc.)
SIEMENS A.S. - SIS SWP TR
Software Engineer
Tel : +90 (216) 459 3720
Mobile : +90 (533) 356 6041
Fax : +90 (216) 459 20 42
Address : Yakacık Yolu No:111, 34870 Kartal, İstanbul , Turkey
Email :sinan.kaya@siemens.com <blocked::mailto:e@siemens.com>
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: periodic bursts
2007-10-01 6:51 periodic bursts Kaya, Sinan
@ 2007-10-01 17:40 ` Jan Kiszka
0 siblings, 0 replies; 2+ messages in thread
From: Jan Kiszka @ 2007-10-01 17:40 UTC (permalink / raw)
To: Kaya, Sinan; +Cc: linux-rt-users
Kaya, Sinan wrote:
> Hello all,
> My realtime system experiences periodic bursts every 4 hours. How can
> i find the reason for this ? I know the traceit tool but it dumps
> so much data and it is useless under this case. I need to find what
> happened at the exact time of burst.
The trace-it tool is a demo, you need to adopt it to your scenario,
specifically make it trigger the stop once you detected some burst.
Means, you need to put its code into your application. Did you do this?
>
> I use rtai_smi module to disable SMI interrupts
Then you will probably like this tool even more:
http://www.rts.uni-hannover.de/rtaddon/#smictrl
> and set my network interface driver priorities to 99. So it should be
> something non maskable but what ?
Out-of-the-box NIC drivers are generally not suited for more than
soft-RT. They allocate memory (skbs) from global pools which takes
varying time (or may even fail if your are short on memory), sometimes
they try to pile up packets first before they raise an IRQ, and they
often contain hardware/link state watchdogs that can inject latencies
right at the wrong time. If you need low latency by design, more work is
required.
In any case, understanding this particular problem comes first, and
collecting the right traces will help.
Jan
--
Siemens AG, Corporate Technology, CT SE 2
Corporate Competence Center Embedded Linux
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2007-10-01 18:12 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-10-01 6:51 periodic bursts Kaya, Sinan
2007-10-01 17:40 ` Jan Kiszka
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.