Problems with the Netgear FA310TX - Broadcasts and TX Lockup

John Stoffel john@casc.com
Tue May 9 11:05:40 2000


>> tulip-0.90q
>> tulip-0.90z
>> tulip-0.91g (from 2.2.15pre??)
>> tulip-0.91g-ppc (from 2.2.15pre13)

Donald> These versions should work the PNIC-I, although I tested the
Donald> original code not the modified versions in the kernel.  ("With
Donald> many eyes, all bugs are shallow.  With many fingers, all bugs
Donald> exist.")

I tried the .92 version of the driver last night and I got the same
problems.  Lots of dropped packets and horrible performance.  I had
three systems up and running at this time:

      SparcClassic, Linux 2.2.15pre13, Sun HME 10/100 card, connected
		    at 100mb/s

      Mac Performa 6100CD with MacOS 7.5.5, 10mb/s connection

      Problem box, Linux 2.2.15pre13, NetGear Card, .92 driver.

The Space and the Mac were very happy talking back and forth with each
other, with each on one side of the NetGear DS108 DualSpeed hub I have
at home.  I had been starting to think it was a hub problem, but since
they worked fine, I still think it's a card/driver issue.

>> Basically, when I ifconfig eth0 up, it auto-negotiates to be
>> 100baseTx, which is great, this is exactly what the [pc] does as
Donald> ...
>> But I can't seem to send packets to either the classic or the pc
>> without lots of timeouts and dropped packets.  The performance is

Donald> What is the timeout message?  I'm especially interested in the
Donald> v92 driver message.

I got and compiled the .92 driver last night, but I had to fight a bit
to make it and pci-scan.c compile properly without tons of warnings.
Ended up commenting out the include for k_compat.h and/or
kern_compat.h to get it to compile.  I also modified the source to set 

	      debug=7 

but default, but since I compiled as a module, I maybe need to
actually use the options with insmod?

Anyway, the message I got in the logs (which is from memory since I
didn't have time to make notes last night after fighting the compile)
from the .92 driver was:

     Advertised 401a, negotiated 001a

which I assume means it wanted 100baseTx-FD, but got 100baseTx-HD
instead, which is fine.  But the wierd thing is that I kept getting
these messages fairly frequently in the logs.  It's as if it was
trying to re-negotiate all the time. 

Tonight I'll try to get more detailed info with tulip-diag and
mii-diag, along with log messages. 

>> eth0: Lite-On 82c168 PNIC rev 32 at 0x6100, 00:A0:CC:57:9A:C1, IRQ 10. 
>> eth0:  MII transceiver #1 config 3000 status 782d advertising 01e1. 

Donald> Looks normal.  I traced this down to a 82c169 board produced
Donald> around mid-'98.  The same board layout was sold under
Donald> Kingston, Linksys and Netgear brands.

Donald> I tried out the Linksys version (thanks to Linksys for
Donald> providing a sample!), and it works fine with tulip.c v092 in a
Donald> BP6 (dual Celeron 440BX) talking to a Linksys 5 port 10/100
Donald> switch (purchased, not a sample :-<).

Donald> ..And it works through a Intel InBusiness 10+100 repeater,
Donald> which should be the same type of hardware you have.

I guess I'll try to get another switch and/or hub and see what my
tests show with this card and driver.  

Donald, a big thanks for all your help here!  I really appreciate the
work you've put into this for me.

John
   John Stoffel - Senior Unix Systems Administrator - Lucent Technologies
	 stoffel@lucent.com - http://www.lucent.com - 978-952-7548
	    john.stoffel@ascend.com - http://www.ascend.com
-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-tulip-request@beowulf.org