kernel networking problem

Wes Bateman plug-discuss@lists.PLUG.phoenix.az.us
Sat, 22 Dec 2001 01:21:43 -0600 (CST)


On Thu, 20 Dec 2001, Eric wrote:

> 
> Hi,
> 
> I am setting up a laptop firewall, and the pcmcia card services work ok as
> long as not too much data is piped through.  But when the load gets heavier,
> eth1 on the firewall freezes.  And each time I unfreeze it, restart Linux,
> and reinitiate heavy throughput, eth1  freezes, again and again and again,
> always within one minute of the initiation of both streaming and SSH
> transfers.  The concomitant error messages produced when the interface
> freezes are reproduced below.
> 
> The pcmcia cards I am using are the 3ccfe575ct-d and 3ccfe575bt-d, and the
> driver is the 3c59x.  They work fine unless the load gets to heavy.  I have
> upgraded to kernel 2.4.9, but the problem existed on the default RH 7.2
> install (whichever kernel that is, can't remember).  With the latter kernel,
> I first had the problem described here
> http://www.uwsg.iu.edu/hypermail/linux/kernel/0110.3/1030.html, and when I
> added "options 3c59x enable_wol=1 " to /etc/modules.conf, this new problem
> arose.
> 
> Can anyone either
> 1. recommend how to fix this new problem, or
> 2. recommend some pcmcia cards that work with RH 7.1 under
> heavy loads, or else
> 3. recommend a Linux OS that can handle my particular pcmcia cards under
> heavy loads?


Hey Eric:

Hmm, is there anything else you see in
kernel.log/messages/global.log/other pertinent log file?  Also, what does
/proc/pci look like?  Without either NIC, does the notebook function
properly under load?  I wonder if there isn't something broken with either
one of the NICs or the PCI bus.  Possibly even bad memory, but I'm
guessing one of the NICs.  Can you take out the NIC that plays eth1 and
only put the other in?  Does the problem continue?  How about with the
"eth1" NIC all by itself (taking on the role of eth0)?  How about both
NICs in, but PCMCIA card slots swapped?

More verbose log info might help find a resource conflict,
etc.  Additional test results (stuff like the above) might help to nail
it down to a specific piece of hardware (or at least clear the good name
of some pieces of hardware if/when they function properly :) ).

Good luck,

Wes

-- 
Wes Bateman, GCIA
Chief Security Officer
ManISec, Inc. - "Managed Internet Security Services"
http://www.manisec.com
wes@manisec.com