[Soekris] 4801 Ethernet errors

[Bill Maas]

On Wed, 2007-02-07 at 11:29 -0500, Joshua Coombs wrote:
> Joshua Coombs wrote:
> > (Bah, forgot you have to be subscribed to post, I monitor the list via 
> > the gmane.org archives...)
> > 
> > I just setup a 4801-60 + vpn1411 with m0n0wall 1.23b (FreeBSD 4.11 
> > based).  I've got the 1.5A higher output powersupply.  All three 
> > interfaces are cabled up to HP 6108 gigabit switches, two 5' cables, one 
> > 8' cable.  Ports are hardcoded to 100/Full on both ends.
> > 
> > Even at low sustained data rates (100kbps) I'm racking up a high number 
> > of CRC and other ethernet errors on the 6108s.
> > 
> > Is this a cable length issue, should I not be hardcoding to 100/full? 
> > I'm open to any suggestions on how to get these interfaces running clean.
> > 
> > Joshua Coombs
> 
> Further digging shows my 4501-30s running full FreeBSD 4.11 as ntp 
> servers are showing similar problems.  I checked and the short cable fix 
> is included in the sis driver under 4.11, so now I'm not sure where to look.
> 
> Whats odd is the nics are reporting collisions locally, how can you have 
> a collision if you're running full duplex?

Sounds more like a messed up switch to me. Those things run quite
complex firmware too, which isn't necessarily bug-free. Some cache may
be holding contradicting data or something. My first reaction to such
errors would be to switch off and unplug _everything_, then plug in and
boot the devices one by one. It may turn out that one of the boxes is
messing up the network due to a config error, but more often than not in
my (humble) experience, everything might as well run fine after just
that.

Bill Maas

> Joshua Coombs
> _______________________________________________
> Soekris-tech mailing list
> Soekris-tech at lists.soekris.com
> http://lists.soekris.com/mailman/listinfo/soekris-tech
> 
-- 
"Incompetence is our watchword" - John Peel