eepro100.c (v 1.06 and v1.09l) problems

Brian E. Bothwell brian@wisdomtools.com
Thu Sep 9 12:22:26 1999


Hi, I have a Intel PCI EtherExpress Pro100 (Intel 82557), and am having
some problems with it that seem to be very similar to those others have
reported. I am about at wits end, and I am dead meat if I don't get this
server working correctly by tomorrow (we have clients starting to hit one
of our web-based apps).  I'm willing (and leaning towards) replacing the
Intel NIC with another new card if that's what will get me up and running
ASAP.

here are the stats:
---------------------
Redhat Linux 6.0, kernel 2.2.5-22, PPro 200, 100Mb full-duplex
networking connected to a Netgear FS308 switch.  I am forcing 100Mb-FD
in all cases (to avoid negotiation problems I have had before).  I am
loading the module like so: "insmod eepro100 debug=1 options=0x30,0,0"

I was having the infamous "Transmit timed out...trying to restart the
transmitter" with eepro100.c v1.06, which I think I narrowed down to
Netatalk/AppleTalk traffic (when the main AppleTalk router when down, it's
error messages seemed to cause my card to spit out the error messages

Sep  8 21:35:34 gonzo kernel: eth0: Trying to restart the transmitter...
Sep  8 21:36:29 gonzo kernel: eth0: Transmit timed out: status 0050  0000 at 75041/75057 command 00000000.
Sep  8 21:36:29 gonzo kernel: eth0: Trying to restart the transmitter...
Sep  8 21:39:04 gonzo kernel: eth0: Transmit timed out: status 0050  0000 at 75114/75130 command 00000000.

etc....

So I upgraded to eepro100.c v1.091, and now I am getting tons of error
messages (any time there is heavy traffic) like the following:

Sep  9 09:57:08 gonzo kernel: eth0: Transmit timed out: status 0050  0000 at 
  23094/23106 command 000c0000.
Sep  9 09:57:08 gonzo kernel: eth0: Tx ring dump,  Tx queue 23106 / 23094:
Sep  9 09:57:08 gonzo kernel: eth0:   0 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   1 400ca000.
Sep  9 09:57:08 gonzo kernel: eth0:  =2 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   3 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   4 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   5 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   6 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   7 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   8 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   9 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   10 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   11 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   12 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   13 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   14 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   15 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   16 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   17 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   18 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   19 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   20 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   21 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0: * 22 000c0000.
Sep  9 09:57:08 gonzo kernel: eth0:   23 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   24 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   25 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   26 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   27 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   28 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   29 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   30 000ca000.
Sep  9 09:57:08 gonzo kernel: eth0:   31 000ca000.                 



The speeds I am getting on the eepro100.c v1.091 were really slow
(900-1200Kbs) , but giving the vast amount of error messages I wasn't
surprised.  I dropped back to the v1.06 version, and just got 7200Kbs
during a SMB (Samba) file trasfer from the Linux box. (but Netatalk speeds
are still pathetic, around 20-50Kbs from a 10Mbs iMac).  

If I were to go out and buy a new 10/100 Mbs NIC for this Linux box, which
one should I get to avoid any of these problems?  I had a 3Com 3c595
Vortex in this same box before, and it was acting flaky so I decided to
drop in the Intel NIC in its place.  I have an onboard 3C905B in our other
Linux box (a Dual PIII-450 running 2.2.5-22-smp and Redhat 6.0), and
while I have heard bad things about these, once I forced 100Mbs-FD on it,
it works like a charm and has good transfer rates (6-9Mbs).  Maybe I
should go back to a 3Com NIC in the PPro200?  If so, which one?

Thanks much!!

-Brian




--------------
Brian Bothwell                             brian@wisdomtools.com
System & Network Administrator             http://www.wisdomtools.com       
WisdomTools, Inc.                          812-856-4200 #7
--------------