\begin{Matthew Sanderson}
> I'm trying to get a SPARCStation SLC working as a diskless x terminal. I'm
> using ARP/RARP to give it its IP address and TFTP to serve it its kernel.
consider using something like bootp/dhcp, since that lets you pass
many more options to the kernel (and bootup scripts)
> boot net linux root=nfs nfsroot=10.0.0.7:/tftpboot/10.0.0.101
> where 10.0.0.7 is the boot server (an intel redhat 6 box) and 10.0.0.101
> is the ip address of the sparc, which it got from arp/rarp from 10.0.0.7 .
>
> The kernel boots, but fails to mount the root filesystem and stops. I
> can't see it attempting to mount it, using 'tcpdump'; I see a whole load
> of UDP traffic which is the kernel being served via tftp, but no TCP
> traffic at all, which I'd expect to see for sun rpc/nfs.
nfs/rpc is udp (unless you're using tcpv3 over tcp, which you aren't)
tcpdump should be able to tell its NFS and print something that
indicates so - check /var/log/, there's bound to be a message from
portmap or nfsd/mountd saying something one way or the other.
> Have I got the right kernel command line?
if /tftpboot/10.0.0.101 is the right export, then yes. usually, you
don't want to put anything under /tftpboot that you don't want
publicly accessable, since tftp doesn't have a whole lot of
security. consider /nfsroot/10.0.0.101 or something.
> Could the problem be that RedHat's SPARC install kernel doesn't have 'root
> on nfs' option compiled in?
could be. i'd expect something like a "panic: couldn't mount vfs root"
if that was the case though. it certainly wouldn't *try*.
i've got some instructions for doing an nfs-root debian install on a
multia ;) most of the file setup, etc should be identical, so it'll
probably give you a head start. see:
http://www.cse.unsw.edu.au/~gusl/multia-howto/
--
- Gus
--
SLUG - Sydney Linux User Group Mailing List - http://slug.org.au/
More Info: http://slug.org.au/lists/listinfo/slug
