[eepro100] Second eepro100 recognized "Receiver Lock-Up bug", but nothing gets done about it....

Graham Wooden GWooden@interland.com
Tue, 13 Mar 2001 12:45:01 -0500


I did notice the IRQ sharing in /proc/interrupts of eth0/eth1 on Int 9, and
now, after I rebooted with that noapic statement, dmesg is also showing
Interrupt 9, when before they were on separate IRQs.

Am I wrong to think that the OS can handle the NICs on the same IRQ? It can
right? it just when it comes to Address location is where it will freak.

Anyways, I am not sure what to think right now on this. 
Any help or suggestions would be great and appreciated!

Graham Wooden

-----Original Message-----
From: Donald Becker [mailto:becker@scyld.com]
Sent: Tuesday, March 13, 2001 10:45 AM
To: Graham Wooden
Cc: 'eepro100@scyld.com'; Eric Van Wieren
Subject: Re: [eepro100] Second eepro100 recognized "Receiver Lock-Up
bug", but nothing ge ts done about it....


On Tue, 13 Mar 2001, Graham Wooden wrote:

> I have a HP LT6000 with an on-board Intel Pro 100 NIC and a PCI Intel
> EtherExpress NIC.
> I am having a problem with the PCI one, eth1.
> 
> According to dmesg, for eth0, it's found the receiver bug, and the work
> around gets activated.
> And for eth1, its say it does exist, and stats enabling work around, but
it
> doesn't.

The receive lock-up work-around is pretty rare, and likely unrelated to
your problem.

> Eth1 goes to sleep about 3-4 minutes after I disconnect from the machine.
> The only way I can get Eth1 to respond is to have activity on eth0.

This sounds like an IRQ mapping problem.
Try passing the kernel option 'noapic' on the boot line.

Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993