On Tue, 29 Nov 2011 14:40:05 -0500 (EST)
Mouse <mo...@rodents-montreal.org> wrote:

> >> And afaik, WLAN problem like yours usually turns out to be driver
> >> issues, [...]
> > Interesting that the very same card works in a different PC with the
> > same the same distribution, the same kernel and hence the same
> > drivers flawlessly....
> 
> "[H]ence the same drivers" is an unwarranted leap.  It may be the same
> driver for the WLAN device itself, but there are numerous other parts
> of the system which probably aren't using the same drivers, such as
> bus-to-bus bridges and controllers and maybe iommus and suchlike.

Uhmm.. do you know what you are saying here?
This would mean, that in all those years where Geode LX systems were
sold (and are still sold), not one of these ten thousands if not
hundred thousands of others who use a Geode LX with linux have 
even once hit a bug in one of the core system drivers, while i can
quite easily make it crash? Sorry, but it would be far easier to believe
that there is a tea pot in an orbit between earth and mars, than such
a bug.


Yes, i have hit many kernel bugs in the past 12 years, since i first
started to use linux, but hardly any of them was in a core component
(just once a bug in the VM), but usually some drivers of devices that
were not so common. And it's very unlikely indeed that a component,
so widely used, and that is well documented as well, should have a bug
in its driver, that only I hit on my system. 

The pattern i see is, the bug is only aparent if and only if the wlan
interface is in use. Having it sitting idle, doesn't trigger it.
There is not an sure way to trigger it, but it is more likely to happen if:
* Runnig wlan on 5GHz instead of 2.4GHz
* High CPU load
* During high disk i/o
* During high network traffic (with low CPU load)

The easiest explenation here is, that there is a small and short drop of
the voltage on one of the power rails. Too short to trigger the reset
circuit but long enough to let the CPU crash. This could be caused, by
the power supplies on the board being brought to their limit by two
or more devices simultanously using a little bit more power than they
usually do.

And as i have already said, the easiest and surest way to verify this is
to solder a few capacitors into the system and see whether this fixes it.

But if it makes you happy, could try to get one of the alix boards with
an Geode LX and try it in there. Though i'm quite sure i wont be able to
trigger it.


                        Attila Kinali

-- 
Why does it take years to find the answers to
the questions one should have asked long ago?
_______________________________________________
Soekris-tech mailing list
Soekris-tech@lists.soekris.com
http://lists.soekris.com/mailman/listinfo/soekris-tech

Reply via email to