Re: Beavis is back and I wanna show him a raw IP dump

["Scott M. Stone" <sstone@example.com>]

On Wed, 6 Sep 2000, Christopher Sekiya wrote:

> On Wed, Sep 06, 2000 at 06:54:58PM +0900, Stephen J. Turnbull wrote:
> 
> > What I'm seeing is lots of errors in the RX line of ifconfig's report
> > (normally >1%, I can push that to almost 20% by doing a high-bandwidth
> > transfer as above).  Most (70--90%) of these are "frame" problems.
> 
> Uh, duplex mismatch?

*possibly*... duplex mismatches usually make the whole thing break, in my
experience.  Besides, you have to deliberately CAUSE one of those, ie, you
have to force a card to use full duplex when it doesn't want to.  modern
NICs will autosense if the hub/switch can do FDX or not and act
accordingly, for the most part.  Every time I've seen the autonegotiation
fail, it's been erring on the side of using HALF duplex on both sides,
which will always work.

Now, there is an error like this that I've seen with Ultra5/10 machines
and Cisco switches - if you force hme0 to be 100mbit FDX, even though the
Cisco switch can support that, it won't bring the link up.  you have to
tell the U5/U10 to autonegotiate and let the SWITCH figure out the speed
and duplex.

(FWIW, the switch picks 100mbit, FDX, just like the U5 wanted in the first
place)...

--------------------------
Scott M. Stone, CCNA <sstone@example.com>
UNIX Systems and Network Engineer
Taos - The SysAdmin Company