rtl8139 problems

Donald Becker becker@cesdis1.gsfc.nasa.gov
Thu Sep 3 17:31:37 1998


On Thu, 3 Sep 1998, James Stevens wrote:
> I also have an app called "open_socket" that saturates a line to test
> the throughput speed between two points, reporting bytes of user data
> downloaded per second. The app is also useful for seeing much CPU
> individual ethernet cards require and lets us judge which cards to use
> in what products. 
..
> With me, running "open_socket" to AND from a 8139 based system at the
> same time locks it up within a few seconds, but it can happily client OR
> server to about 1,100,000 bytes/s with no problems. As a client this
> only take 6% cpu (94% idle), as a server it takes about 40% cpu on a
> P75.

I ran this program simultaneously to and from a RTL8139 running v1.01 on a
P6/440FX system for over thirty minutes without a problem.
I only have one rtl8139 card -- the remote end was an epic100 card.

I've put about 50 million packets through the system today trying to
reproduce the problem, without any success.

I have located at least one problem with the card reset code, and several
other problems.  Look for ... (voice trails off)  Damn.  I looked over at
the machine and it just had a Tx-timeout error.  Didn't have one all day,
while running netperf, a bunch of  'ttcp' test, and open_socket
simultaneously.  But it get an error right after I said I couldn't reproduce
the problem.

Anyway, the good news is the reset code immediately restarted the chip and
everything continued running.  So I'll release this driver and wait for
reports.


Donald Becker					  becker@cesdis.gsfc.nasa.gov
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center,  Greenbelt, MD.  20771
301-286-0882	     http://cesdis.gsfc.nasa.gov/people/becker/whoiam.html

 | To unsubscribe, send mail to Majordomo@cesdis.gsfc.nasa.gov, and within the
 |  body of the mail, include only the text:
 |   unsubscribe this-list-name youraddress@wherever.org
 | You will be unsubscribed as speedily as possible.