Re: Network booting issues

[Andrew Sharp <andy@netfall.com>]

Nils Ackermann wrote:
> 
> Andrew Sharp <andy@netfall.com> writes:
> 
> > The sun4m boxes are 32 bit processor boxes while you are running
> > your ultra1 in 64 bit land.  But you say that you built the kernel
> > on the client, so it should be good for that arch, as long as you
> > configured the kernel correctly.  I'm not familiar with mknbi.
> 
> > You said above that you built it on the client.  Why would you want
> > to build it on the server for the client?  Besides the fact that the
> > server is faster, I mean.  ~:^)
> 
> I'm planning to remove all disks from the clients to use them for
> /home on the server (buying new disks is not an option at the moment,
> given that the budget of our department doesn't even cover costs for
> periodicals in the library).  It would thus be nice to compile kernels
> for the clients on the server.

That depends on your definition of `nice.'  I don't see what having
the clients be diskless has to do with compiling the kernel on
them.  I run my SS20 with NFS root and everthing else, most of the
time, even though I have two disks, because a 10/100 network card +
cheap x86 NFS server is actually a faster disk combo than the local
disks.  I compile kernels on it all the time.  I also run netscape. 
~:^)

> 
> Andreas Jaehnigen <jaehnias@ims.uni-stuttgart.de> writes:
> 
> >       elftoaout -o vmlinuz-2.2.14.net vmlinuz-2.2.14
> 
> That bit was new to me.  So to make a network bootable image on sparc
> one doesn't use mknbi-linux?  How is the functionality of mknbi-linux
> achieved here (i.e. -d rom -i rom and initrd (not that I need the last
> one, just curious))?

I don't know why that would be necessary, but I load my kernel off
the disk, then use an NFS root.  However, I've got a completely
diskless x86 laptop that I netboot and I didn't have to do anything
special to the kernel in that case.  Perhaps that's apples-2-oranges
however.

a