[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netbsd-tech-net
Subject:    Re: LANCE_REVC_BUG
From:       David Laight <david () l8s ! co ! uk>
Date:       2002-08-10 18:07:56
[Download RAW message or body]

On Sat, Aug 10, 2002 at 03:27:32PM +0200, der Mouse wrote:
> I have some questions about the lance Rev. C bug.

Interesting - that isn't one of the lance errata I'd heard about!

> Also, why does it preclude multicast support?

Well if you tested unicast and multicast addresses separately,
and maybe allowed through valid multicast (or all multicast
- since the user must check the destination address of a
multicast packet before processing it) it would be ok.

The 'usual' lance errata (which I didn't spot anything paging
through the code to get around) concern unreported memory
timeouts during ring lookahead, memory timeouts that happen
between 126 and 128 clocks, and glitches on the ALE line.
(The latter being particularly painful because the C-lance
glitches it when configured for the other polarity).

A heavily loaded multi-cpu mbus/sbus sparc system WILL
show every bus timeout cycle you could possibly imagine :-(

	David

-- 
David Laight: david@l8s.co.uk
[prev in list] [next in list] [prev in thread] [next in thread]