[AG-TECH] Linux networking problems

Kay Gunn gunn at colltech.com
Wed Jul 25 16:06:29 CDT 2001


>    I only know enough about this to be dangerous but are you sure the
> switch port is full duplex?  

Our network admin just verified that it is full-duplex.
This means either mii-diag gave me faulty information or I'm reading it
wrong. Of course, I did use the very same switch port successfully with
Redhat 7.1 forced to do 100MB full duplex, so if the port was set to
half-duplex, I should have seen problems there, too.  
mii-diag output included the following:

 Speed fixed at 100 mbps, full-duplex.  #that's my linux box/card

Your link partner is generating 100baseTx link beat  (no
autonegotiation).   #that's the switch

I'm assuming that's 100BT HD fixed, since card info possibilities
listed 
100baseTx-FD and 100baseTx.

> if it is 100BaseT.  The reason I ask is that I think the full duplex end
> of a theoretical mismatch would never get a collision because it knows
> you cannot get collisions running at full duplex.  But the hdx end would
> not

That makes sense...but I would expect to see errors.  I have never worked
with Linux before, but I have worked alot with Solaris and 100BT duplex
problems.  In the past when I've seen a duplex mismatch on Solaris, I'd 
see *ALOT* of errors (using netstat -i time slices).  

The network admin looked at the port statistics on the switch I'm
connecting to, and he is seeing alot of errors there (but no collisions)
for my grid nodes. 99% of the errors are RUNTS, with a few CRC.  It
doesn't make sense to me that the switch is reporting the errors, but the
linux boxes aren't (they're just performing slowly).  I'll install Redhat
7.1 back on one of machines that's generating the RUNTS to see if they go
away with the different OS.

------
Kay Gunn








More information about the ag-tech mailing list