[ previous ] [ next ] [ threads ]
 
 From:  Josh Coombs <twig at gwi dot net>
 To:  m0n0wall at lists dot m0n0 dot ch
 Subject:  Re: [m0n0wall] Soekris 4801 + m0n0wall + HP 6108 == errors?
 Date:  Wed, 7 Feb 2007 11:38:39 -0500
On Feb 7, 2007, at 9:54 AM, Josh Coombs wrote:

> I just swapped in a Soekris 4801-60 running 1.23b3 for a PC with a  
> known wonky nic.  The dual port nic in the PC is DEC Tulip based,  
> and doesn't quite behave properly.  You HAVE to specify the media,  
> 10 or 100 for it to get link, it won't autoneg on it's own.  Once  
> you've done that, it ignores duplex settings, always running at  
> half duplex even though it reports back that it's running full  
> duplex.  We've got two of these duds in house, they do the same  
> trick under every OS I've tried, including Windows with the mfg's  
> drivers.
>
> So in any case, I swapped machines as 100/half was causing problems  
> at moderate load thanks to collisions.  The Soekris I pressed into  
> service appears to honor both media and duplex settings... but  
> instead it's racking up CRC errors on the switch end even at low  
> data rates.  I've traded one problem for another it seems.

Bit more detail... and this is where I start pulling hair out.   
According to system_advanced.php, polling is disabled.   
status_interfaces.php shows media as 100baseTX, and lots of  
collisions.  ifconfig via exec.php shows the nics as polling enabled,  
100baseTX full-duplex.  netstat -i confirms collisions...

I'm going to downrev to 1.22 as suggested offlist and see what I can  
sort from there.

(Further digging shows my 4500s are doing the same thing, 100/Full,  
reporting collisions and taking errors on the switch, albeit at a far  
lower rate than the 4800... those are running pure FreeBSD 4.11.   
Maybe this hardware IS bad.)

Joshua Coombs