Matthias Drochner wrote:

> [late followup on an old mail about problems with the 3x90x boot ROM]
> 
> Yesterday I've fixed a stupid bug in the code - botched
> logics in the device recognition made that the io base
> address was not set correctly for the 3c905b (it was correct
> for the 3c900b). This can lead to hangs if a bit is polled of
> course...

Cool!  This is better (I actually get some network activity now!), but
I've still had a problem or two trying to get the kernel via NFS.  Turns
out that:

 a) Using sys/lib/libsa/nfs.c just plain doesn't work (haven't figured
    out why yet).  I had tried both the i386 nfs.c and the libsa one in
    earlier attempts to get this working.

 b) I had to delete the leading slash early on in nfs_open().  The
    bootptab has "bf=netbsd", and it tried to open "/netbsd", but it
    looks like nfs_open() wasn't doing the moral equivalent of "cd / ;
    open netbsd".

> Further, there is a nasty hardware bug in many 3c905b chips:
> It does only decode the first 4k of the ROM properly, the
> rest of the image consists of aliases of that range.
> No boot code will fit into 4k, so these cards are basically
> unusable for boot ROMs.
> The buggy chips have a number ending with -001 on them,
> newer ones ending in -002 are OK.

Mine's a -004, so I guess that's good :-)

Simon.