Do you think this receiver going deaf is a problem with the kernel, its
architecture, or the driver and/or its architecture?

I've been suffering with this problem all day, and I need to fix it.

Thanks -Mike



On Wed, 6 Jun 2001, David Burgess wrote:

> Date: Wed, 06 Jun 2001 08:56:54 -0500
> From: David Burgess <burgess@mitre.org>
> To: Mike Cheponis <mac@Wireless.Com>
> Cc: netbsd-help@netbsd.org, port-i386@netbsd.org
> Subject: Re: Ethernet receiver goes catatonic
>
> I'm not sure if it's related, but I have one of these cards in my NFS
> server, and every once in a while I will get a "Out of buffers" error (I
> don't have the exact error message, one of my techs cleared it when he
> rebooted).  After that, the card is catatonic and the system (and all of
> the systems that connect to it) need to be rebooted.
>
> The similarities are:
> 1)  They are both rtk cards
> 2)  It only happens when lots of file activity is occurring.
>
> Dave
>
> Mike Cheponis wrote:
> >
> > Hi, I'm running 1.5T on a K6-2/350 w/128MB.
> >
> > With a RealTEk RTL8139A chip, using the rtk0 driver, and FTPing into the
> > box from another machine on a 100 Mb/s LAN causes the interface to stop
> > receiving; I'm sending lots of big files at this machine.
> >
> > If I do:  ifconfig rtk0 down   then   ifconfig rtk0 up    then it comes back
> > to life, and I can use the network again. (The previous ftp session is
> > hosed, however).
> >
> > It only seems to be affected when there are lots of big files going into
> > this machine; just an ssh session or so never causes a problem.
> >
> > Today, I was able to try a D-Link DFE-570TX card (4-ports, tulip chips) uses
> > the de driver.
> >
> > This time, the FTP succeeded for about 10% of the files before having the
> > exact same problem as the rtk0  (the rtk0 only got about 0.1% of the files
> > before croaking).
> >
> > Has anybody seen this before?
> >
> > Thanks -Mike
> >
> > p.s. FWIW, the machine sending to the NetBSD box is Windows 2k, using ftp
> > in a command window, with "MPUT *" in binary mode.  I did try a NetBSD 1.5
> > box as the source and got a similar result, however.
>