more NetGear mising ARPs

Neale Banks neale@lowendale.com.au
Fri Jun 11 07:23:58 1999


On Mon, 24 May 1999, jferg wrote:

> Neale Banks wrote:
> 
> > With 2.0.36 and tulip.c:v0.89H on a not-so-slow machine (349.80 bogomips)
> > I appear to be bumping into this problem :-(
[...]
> > I'll try to load and compile v0.91 of tulip.c tomorrow to see how that
> > performs.

v0.91 (as a module on kernel-2.0.36) doesn't appear to have changed
anything :-(

[...]
> Check half/full duplex status of everything connected to hub (when you say hub, I assume
> you mean a PHY-level repeater; they don't handle full-duplex very well)

Yes, hub as in a repeater.

I'm pretty sure that it's not this as:

1) it can talk to one host on the same hub (linux) but not to another
(Cisco 25xx).
2) the other two hosts can talk happily to each other

As others have suggested, it's as though the arp reply is either not seen
by the card or not delivered up by the driver.  Also, whilst the arp from
the tulip-host remains "(incomplete)" the Cisco cannot successfully arp to
the tulip-host - however in the case where there is no arp entry for the
Cisco on the tulip-host then the Cisco can initiate communication with the
tulip-host (which then gets an appropriate arp entry for the Cisco).

Mmmm, an ugly kludge might be to make a static arp entry?  But that won't
solve the underlying problem and will only make for horrible surprises
when hardware gets changed.

Any other ideas?

Thanks,
Neale.