Problems using DWL-650H

[Jan Rychter]

David Gibson on Jan 8, 2003:
> On Sun, Dec 15, 2002 at 10:30:54PM +0100, Jan Rychter wrote:
> > I'm getting errors like:
> > 
> > [...]
> > eth0: Error -110 writing Tx descriptor to BAP
> > eth0: Error -110 writing Tx descriptor to BAP
> > eth0: Error -110 writing Tx descriptor to BAP
> > [...]
> > 
> > which a "cardctl eject; cardctl insert" fixes for a while.
> > 
> > Now, this is a known problem (I've reported it before) -- but I wanted
> > to report some additional data. Perhaps there will be hidden hints
> > there.
> > 
> > I'm still using 0.11b with the double-reset patch:
> > 
> >   orinoco_cs.c 0.11b (David Gibson <hermes at gibson.dropbear.id.au> and others)
> >   eth0: Station identity 001f:0002:0002:0001
> >   eth0: Looks like a Symbol firmware version [V2.51-01] (parsing to 25101)
> >   eth0: Ad-hoc demo mode supported
> >   eth0: IEEE standard IBSS ad-hoc mode supported
> >   eth0: WEP supported, 104-bit key
> >   eth0: MAC address 00:05:5D:D7:3D:BF
> >   eth0: Station name "Prism  I"
> >   eth0: ready
> >   eth0: index 0x01: Vcc 5.0, irq 3, io 0x0100-0x0147
>
> Ok - I suggest you try using 0.13a (I haven't had time to push it to
> Marcelo yet).  It fixes at least one bug that can result in those
> errors (although I suspect there are multiple problems :-( ).
> 
> > I've been running a kernel with software suspend for the last month or
> > so. Now, the networking crashes occur with varying
> > frequencies. Depending on something the card will either freeze every
> > 4-5 minutes, or work for hours without a crash. This is repeatable
> > within a suspend/resume or reboot cycle -- e.g. once I boot the system
> > (or resume), the crashes occur at fairly regular intervals.
> 
> Hmm... that's quite bizarre.  Unfortunately I'm not sure that I can
> make much use of the information beyond stashing it away as another
> datapoint until some pattern emerges.

> It seems that certain "configurations" are more prone to crashes than
> others. Could this be related to the layout of buffers in memory or
> something else that changes at each power-on? Perhaps the initialization
> is not done properly and there is a random factor somewhere?
> 
> There is a relation to actual network load -- the bigger the load, the
> faster the crash occurs -- but it's a weak relation, weaker than the
> above symptoms. Putting a big load on the card might either crash it in
> minutes or keep it running for half an hour, depending on the system
> "mood".
> 
> I don't know whether this helps any. I'm getting desperate, because
> sometimes I get repeatable crashes every 3-4 minutes. It makes the card
> nearly unusable.

To give you another datapoint, the card does not crash when there's
activity on the wire. I've "solved" the problem by running "ping -i 0.2
somehost_on_the_local_network" in the background. That effectively
prevents the dreaded "Error -110" hang from occurring.

Perhaps that'll be useful info.

--J.