>> So I set about to do a Netboot.  In the end I have had to;
>>
>> set use-nvramrc? to false
>> strip out non-mac components
>> use a serial console
>> set `arp` by hand on the server.
>>
>> It now boots and gets the loader but freaks out getting the kernel.  
>> It
>> begins flooding the server with DHCP requests (about 30/sec) until 
>> some
>> buffer overflows and the server freezes.
>>
> Did you provide the client a hostname with the dhcp server based on 
> the client MAC adress? and did you make it easy for the OF to fetch a 
> kernel eg. "netbsd.GENERIC_MD.gz" from a top level directory(you can 
> fake this one when its deeper in your FS structure) with NFS for 
> example /netbsd.GENERIC_MD.gz
> you can also try to unpack the kernel on the server and then call it 
> from the client.
> example: boot enet:,ofwboot.xcf netbsd.GENERIC_MD

I've been using boot enet:,ofwboot.xcf enet:,/netbsd

netbsd is uncompressed

>
>>
>> The Host is set up as follows;
>>
>>    MacPPC 9600 w/ G3
>>    NetBSD 1.5.2Y  (Custom kernel which has IPNAT turned ON)
>>
>>    2 ethernet cards
>>        mc0: 10BaseT - set to real-world static IP (a.b.c.d)
>>        ep0: 100BaseT - set to 192.168.1.1
>>    DHCP listens on both cards.  It hands out a limited number of
>> real-world IPs to known MAC addresses and 192 subnet addresses to
>> unknown clients and has a fixed address (192.168.1.99) for this
>> netboot client.
>>    IP Forwarding IS ON:  ep0 forwards to mc0
>>
>> Client IP is fixed at 192.168.1.99 by dhcp through ep0.  arp is set to
>> bind this address to the MAC address.
>>
>> After the bootloader is received, I noticed a lot of collisions on the
>> switch between mc0 and the client.  And, by tailing the messages log 
>> on
>> the host I can watch the client flood the host with dhcp requestst.
>>
> The only thing i can see is ceep the kernel path so simple as possible 
> and try what happens when you completely disable the firewall or 
> disable firewall rules for tftp, bootps, dhcp and NFS

Ok, I turned off ep0 in dhcpd and changed the ip to a real-world 
address.  Otherwise Ieft everything alone.  I'm hoping that I don't 
have to completely reconfigure the host to do this.

1) The collisions stopped.
2) However when I try to boot, I get one of the following floods in the 
log;

Jun 26 18:26:14 hawk dhcpd: BOOTREQUEST from 00:05:02:35:8c:0c via mc0
Jun 26 18:26:14 hawk dhcpd: BOOTREPLY for a.b.c.45 to hawk2.domain.dom 
(00:05:02:35:8c:0c) via mc0

-- or --

Jun 26 18:20:01 hawk dhcpd: DHCPDISCOVER from 00:05:02:35:8c:0c via mc0
Jun 26 18:20:01 hawk dhcpd: DHCPOFFER on a.b.c.45 to 00:05:02:35:8c:0c 
via mc0

-- or --

Jun 26 18:16:44 hawk dhcpd: DHCPREQUEST for a.b.c.45 (a.b.c.42) from 
00:05:02:35:8c:0c via mc0
Jun 26 18:16:44 hawk dhcpd: DHCPACK on a.b.c.45 to 00:05:02:35:8c:0c 
via mc0

Sometimes they cause buffer underflow or full buffers and knock the NIC 
off the network.  I have to restart it with ifconfig.

Is the client malfunctioning or is the server misconfigured.

Also, I have another machine of the same type (probably OF2.01).  Since 
the OF is so buggy, might there be value in trying the other machine?


Thanks for the assist.

Robert Schmid