Re: question regarding cobalt raq2 nic's
----- Original Message -----
From: "Peter Horton" <firstname.lastname@example.org>
Sent: Sunday, October 17, 2004 6:18 PM
On Sun, Oct 17, 2004 at 06:50:34PM +0200, Thiemo Seufer wrote:
vizi0n (debian-mips) wrote:
> hi everyone!
> i had a quick question regarding my cobalt raq2+ (the one with 2 nic
> and a scsi port).
> for no apparrent reason once in a while, one of the nic gets
> "overloaded" if i can say so (most of the time eth0, which is my lan),
> by this i mean that the data doesn't go through it anymore and doing
> a ifdown eth0 ifup eth0 does fix the problem. did anyone ever had
> that problem before? i would be really grateful, since this box is
> now my router and i often have to connect to it and reload eth0.
> i am wondering if this problem is mips specific (drivers problem?)
> or if it is on all platforms.
The problem is known and seems to be bound to some cobalt/mips specific
way to use the NIC driver. Debugging it further failed so far, mostly
due to the lack of test hardware.
I hit the same problems with the Tulip driver in CoLo. If either the PCI
latency registers or the PCI timeout/retry register on Galileo (PCI
bridge) aren't set right the problem occurs all the time. Somehow or
another the PCI bus master transactions fail without an error
indication in the bridge or on the NIC. Could be a problem with the
okay, any hope of seeing someone find a fix for this issue soon? or is this
a serious problem due to non-reversible stuff? did the "real" cobalt linux
have that problem? i've never used the cobalt linux so i dont know much
about it hehe