[vortex] duplex problems with 3c905

Tim Bell bhat@trinity.unimelb.edu.au
Wed, 24 Oct 2001 15:58:21 +1000


I posted to this list on Mon, 17 Sep, with some duplex problems I was
having:
<http://www.scyld.com/pipermail/vortex/2001-September/001326.html>

I now have some more info available.

When running 10baseT half duplex (as detected), I get a transfer speed
of about 0.9 Mbit/s.  When running 10baseT full duplex (forced), the
transfer speed is about 6 Mbit/s.  The router port at the other end is
supposedly half-duplex only.

As another speed comparison, I put a 10/100 switch between the network
card and the media converter.  In this configuration, the speed was 1.5
Mbit/s, with the switch reporting half duplex operation with plenty of
collisions.

Bogdan Costescu <bogdan.costescu@iwr.uni-heidelberg.de> wrote:

> If you force media settings, you have to do it at both ends of the
> connection (if the partner is manageable) to avoid confusion.

Unfortunately, I don't have ready access to the other end of this
connection; my understanding is that it only does 10 Mbit half duplex,
so this shouldn't be a problem.

> Use the mii-diag tool (from ftp://ftp.scyld.com/pub/diag/) to see what
> the card is thinking about its partner.

Here's the output (the same whether I've forced full duplex or not):

Basic registers of MII PHY #24:  3000 782d 0040 6176 0541 0021 0000 0000.
 Basic mode control register 0x3000: Auto-negotiation enabled.
 You have link beat, and everything is working OK.
 Your link partner is generating 10baseT link beat  (no autonegotiation).
   End of basic transceiver information.

Ruben <ruben@nutz.nl> wrote:

> Start tcpdump on an appropriate machine. Pull the fiber out, and ping
> something on the other site before the arp-cache expires. If you get a
> ICMP-destination-unreachable you have a smart bridge (ie.: a two-port
> switch with a different media on the second port), if you see nothing
> but a timeout you have a dumb one.

There was no destination unreachable response, so I suppose it's a dumb
media converter (at least at this end).

If anyone has any more ideas of things to investigate or try, please let
me know.  (I'll try to get around to them a little sooner this time.)

Thanks,

Tim.
-- 
Tim Bell - bhat@trinity.unimelb.edu.au - System Administrator & Programmer
    Trinity College, Royal Parade, Parkville, Victoria, 3052, Australia