Subject: Re: 3c509 but maybe really intel pci chipset woes
To: Ron Minnich <rminnich@mini.sarnoff.com>
From: Herb Peyerl <hpeyerl@novatel.ca>
List: netbsd-help
Date: 01/09/1995 02:43:44
Ron Minnich <rminnich@mini.sarnoff.com>  wrote:
 > some and not others. the symptom is that i can read the id information of 
 > the 509 just fine -- this is I/O address 0x100. I can read the enet 
 > hardware address and other info fine. What doesn't work is that 
 > the 509 never comes back after a reset has happened. The result is that 
 > the product id at 0x30a always reads back as 0xffff. I can swap 509s 
 > around and the problem stays with the box. 

This sounds like a problem that was just identified recently.  It appears
to be that newer versions of the 3c509 ASIC are "plug-n-play compatible"
and this causes them to exhibit the above behaviour.  The current work
around is to go into the 3c5x9cfg program and disable plug-n-play 
compatibility and then all will work.

I've contacted 3Com and am expecting a new set of documentation which
should help me address the problem more directly.

 > any ideas welcome. This problem is observed with both netbsd and freebsd, 
 > although the if_ep drivers are quite different.

Yes, observe that the FreeBSD if_ep is based on NetBSD's 1.6 version of the
driver and I believe we're up to 1.64 at this point.