Hi Joerg.
On Fri, 7 May 1999, Joerg Reuter wrote:
Q> [NON-Text Body part not included]
Not sure what happenned, but your message was listed as an attachment
to a binary message, hence the above. Here's your comments extracted
(and reformatted) after saving them to disk, just in case there were
any problems caused by that header. Nothing deleted, and I've added my
comments only where I have something to say...
Q> > It would seem buffers get overloaded, or something to that
Q> > effect, ie, the 9k6 interface works twice as fast as the data
Q> > can be retransimited on the 4k8 interface causing buffer/memory
Q> > problems.
Q> Assuming that you are using either z8530drv-2.4c in its network
Q> driver mode or z8530drv-3.0 it is most likely not a buffer
Q> problem.
Q> What seems to happen is that the driver loses not only one TX
Q> interrupt but also the "exception" interrupt that notifies the
Q> driver of the TX underrun. Even worse, it seems that in some cases
Q> (probably depending on the mask version) the Z8530 needs a
Q> complete re-initialization of that channel to be able to transmit
Q> again.
Many years back, I was involved in a project using a Z8530 essentially
connected through a parallel port adapter, and I remember having some
problems getting that to work. There was some problem concerned with
having to acknowledge ALL interrupts, so if one was missed, the thing
stopped working until it saw as many acknowledgements as it had issued
interrupts.
I don't know whether this is connected with the problem under
discussion, but you may wish to investigate something along those
lines just in case...
Q> > There is no need to reboot to get over the problem a small
Q> > script i have running via cron, checks the interface with pings
Q> > every 15 minutes, no pings returned will cause ifconfig to
Q> > restart the interface and all is well again.
Q> I'd recommend a DMA capable SCC card instead.
So would I...
Q> > There are never any signs or messages anywhere which could give
Q> > even the smallest hint of where to start looking for the
Q> > problem.
Q> Seems that I'll have to add yet another timer to monitor transmit
Q> interrupts, sigh.
Maybe, maybe not...
Q> > I have explaned this problem to this list at least TWICE in the
Q> > past, it is NOT only me that has this problem, a look at the
Q> > archives will produce quite a few callsigns who all say they
Q> > have the same symptoms in one way or another.
Q> Yes, but messages with little more information than "it does not
Q> work under high load" don't help much when trying to find the
Q> exact condition causing it.
Q> > Joreg DK1BKE has never responded to any of the comments, hints
Q> > or kinks, made about this problem,
Q> This is not true. I've responded to several of your mails, asked
Q> for more information and never got any answers or replies
Q> afterwards. Seems that the path to your mailbox is not very
Q> reliable.
If your messages all start with that binary header, it's quite
possible that some mail agent en route killed the mail because of it,
so this may not be his fault.
Q> > i have tryed many things different timings, MTU's, window sizes,
Q> > buffsize in /etc/z8530drv.conf, parms settings, realy you name
Q> > it i've tryed it, all to no avail, there is a problem with the
Q> > SCC driver but as finding out what it is, is like banging your
Q> > head against a wall.
Q> Has it ever occured to you that the fault may not be a more or
Q> less "buggy" SCC driver but rather broken SCC board design equiped
Q> with broken second-source versions of Zilog's Z8530 on a broken PC
Q> architecture?
Q> > If there are others with this problem please speak now.
Q> ...and perhaps pass me copy when this happens, with the version of
Q> the z8530drv you are using and the output of cat
Q> /proc/net/z8530drv.
Best wishes from Riley.
+----------------------------------------------------------------------+
| There is something frustrating about the quality and speed of Linux |
| development, ie., the quality is too high and the speed is too high, |
| in other words, I can implement this XXXX feature, but I bet someone |
| else has already done so and is just about to release their patch. |
+----------------------------------------------------------------------+
* ftp://ftp.MemAlpha.cx/pub/rhw/Linux
* http://www.MemAlpha.cx/kernel.versions.html
