One of my WRAP with 3 NICs is servering as a traffic shaper running
1.2b8 in filtering bridge mode. Its OPT1 port talks to a manageable
switch Nortel Baystack 450-24T. As long as I leave the NIC of OPT1 in
the default auto negotiation mode (which correctly negotiates
100/full) everything works fine. The bridge was nicely shaping a 4
Mbit/s symmetrical ISP connection for about 4 months without a single
For some reason, I have now to force the manual setting 100/full onto
OPT1's configuration, which I did using the documented way of
inserting those options into config.xml. At the same time, I also
manually force the corresponding port on the switch to 100/full to
avoid any mismatch. However, I observe many Frame Errors as soon as I
manually force the peer ports of the WRAP and the switch. As soon as I
go back to auto-negotiation, the Frame Errors vanish.
Interestingly, this WRAP's WAN port is connected to the CPE (Cisco
7xxx router) of our ISP and it is manually forced to 100/full, but
there are no Frame Errors!
Now I started to dig mail list archives and found discussions dating
from 2002 that there were/are issues with the cable length &
attenuation and the National DP82xxx NICs, see for ex. thread
The cable between the WRAP and the switch is only 0.5 m long (where
Frame Errors occur in manual 100/full), whereas the cable between the
WRAP and the ISP's router is some 5 m long (never observing any Frame
Error there in manual 100/full).
Would it be possible that the fixes made in the sis driver regarding
cable length maybe apply only to the auto negotiation mode, but for
whatever reason do not work around the hardware bug in manual 100/full
I'll go and replace the 0.5m cable by a 10 m cable, just to check if I
can get rid of this weird problem. Also, I might have a look at the
sources of the sis driver.
Thanks in advance for any hints and other suggestions what else I should try.