Experimental Physics and
Industrial Control System

Ralph Lange <[email protected]> · Wed, 02 May 2007 17:19:13 +0200

Dave,
we have been "playing" with these driver settings a lot.
For us, things usually seem to work with both ends (2100 and switch)
being on auto-neg. Note that we only found a severe problem when one
side is fixed to full-duplex (any speed), and the other side is on
auto-neg. In that case, the auto-neg side falls back to half-duplex and
auto-sensing. So it will get the speed right, but not the duplex
configuration. The full-duplex side, when sending, will always force the
half-duplex side into detecting a collision.
So far, we have not seen any 2100s getting "stuck" while both sides are
on auto-neg.
Cheers,
Ralph

Thompson, David H. wrote:

Our dec driver has for some time been afflicted with just going dead
for no apparent reason. It is rate but when it happens we end up
rebooting the IOC. The driver seems to be trying to auto negotiate
and is stuck. Messages show up on the console from CA beacons and
clients when the entries in the arp table age out. The mii flags are
where we get the idea that it is trying to negotiate, that and the
fact that the IOC and the switch port it is connected to are on
different speeds.

Does anyone know if WRS ever fixed this? If so is there a reference
that I can put my tech contact at WRS on to to see if the driver that
is being shipped with vxWorks 6.x includes the fix? I thought that I
had heard that a fix had been done for it, possibly outside the US
office of WRS.

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System