Re: Rtl8139 pcmcia?

On Tue, 18 Mar 2003, Andreas Johansson wrote:

> I have a pile of errors in interface statistics. This information
> is collected after 7 days of usage connected to a hub with half duplex. 
> I have'nt saved information when connected to a switch with full duplex
> but the problem persists there aswell.

Hmmm, there might be two different problems.

One known issues is that the different rtl8139 chip revisions required
different recovery from a too-many-collisions error.  That problem
should only occur in half duplex mode with a repeater.

I'm guessing another part of the problem is that you don't have enough
PCI bandwidth to reliably receive packet.  That's usually caused by a
video card that over-consumes PCI bus bandwidth.

> "eth0 UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1          
> RX packets:24983558 errors:5195 dropped:5885 overruns:5190 frame:13     
> TX packets:25457394 errors:0 dropped:0 overruns:0 carrier:0         
> collisions:255090 txqueuelen:100
> RX bytes:2636611853 (2.4 GiB) TX bytes:1838210793 (1.7 GiB)
> Interrupt:11 Base address:0x4000"
> 
> Surely looks like an recieving problem?
> 
> Recieved messages from "/var/messages".
> 
> "Mar 16 11:13:39 Master kernel: eth0: Oversized Ethernet frame, status
> 7e880eb!"

Do you get this only on the repeater, or also with the switch?

> I have a couple more that looks the same but with different status
> messages. Looks like I don't get these messeges every time the NIC
> hangs, perhaps they are'nt related.

This message is followed by a reset of the chip.
The driver expects the reset to work.  If the reset fails, the driver
will fail as well.

-- 
Donald Becker				[email protected]
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Scyld Beowulf cluster system
Annapolis MD 21403			410-990-9993