Subject: Re: Inquiry on WM behaviour on NetBSD 3.0 Beta
To: Joseph A. Dacuma <jadacuma@ched.gov.ph>
From: Manuel Bouyer <bouyer@antioche.eu.org>
List: netbsd-users
Date: 11/21/2005 22:41:45
On Sun, Nov 20, 2005 at 09:20:17PM +0800, Joseph A. Dacuma wrote:
> Hi Mr. Bouyer!
> 
> My current setup is a client PC, IBM R51 with an Intel Pro/1000 MT mobile
> connection and a Dell 2800 with i82541GI 1000 BASET Ethernet rev. 5 card.
> 
> To be sure I tried several possible scenarios. Ifconfig.wm0 was configured
> with media auto select and tried transferring 2 iso files with sizes of
> 369MB and 582MB respectively:
> 
> 1) Both units connected to a 3com Superstack II Dual Speed Hub. Showing
> and ifconfig -a result with (100base TX).
> 
> 2) Both units connected to a D-Link Fast Ethernet switch (DES-1016D).
> Showing an ifcofig -a result with (100 baseTX full-duplex, flow control,rx
> pause, tx pause).
> 
> 3) Both units connected directly using a cross connect cable. Showing an
> ifcofig -a result with (1000 baseT full-duplex, flow control,rx pause, tx
> pause).
> 
> To be sure, I crimped fresh CAT5e cables using T568A and T568B wiring
> standards. Also, vendor supplied CAT5 cables both straight-through and
> cross connect cables. Further, I isolated the network cables from any
> power cords to rule out the possibility of EMI.
> 
> All three setups yeilded the same results. With error messages appearing
> on the test server's console "wm0: discarding oversize frame (len=-4)".
> 
> I do get these messages often. Specially, on large file transfers more
> than 300 MB of data.
> 
> I tried to do md5 and sha1 checksums and had no inconsistencies. I haven't
> experienced any errors yet but I'm not sure if this is safe and be
> deployed on production with 200+ plus connected clients being served on a
> 24x7x365 basis.

So the errors are properly reported. It's probably an issue with this
specific adapter model, because I've never seen this with others Intel
adapters (and you seem to imply you didn't see it on the client PC either).

Now, I don't know if this specific model of chip has know flaws that would
have to be worked around in the driver ...

As long as your tests didn't show problems with the file transfers, I think
it's safe to use it on production.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--