On Tue, 2006-06-20 at 03:32 -0400, Amos Waterland wrote:
> On Mon, Jun 19, 2006 at 11:13:35AM -0400, Jimi Xenidis wrote:
> > On Jun 17, 2006, at 12:06 AM, Amos Waterland wrote:
> > > Since prom_init is not executed by Linux when running as a guest in Xen,
>
> This was bogus: prom_init is indeed executed. Please see the new patch below.
>
> > BTW: the Xen passes the string after "--" in bootargs to Dom0
>
> Yes, I am aware of that. But doesn't that mean that if I have a cluster
> composed of hundreds of blades I would have to configure the NFS root
> mount of their DOM0 via the firmware of each blade? That is, wouldn't I
> have to get physical access to the machine and manually run this?
>
> => setenv bootargs
> 'xenarg=foo -- root=/dev/nfs rw nfsroot=${serverip}:${rootpath}
> ip=${ipaddr}:${serverip}:${gatewayip}:${netmask}:${hostname}::off'
>
> If I needed to temporarily change the NFS path for fifty of the blades,
> it would be really nice to be able to do that in an automated way.
>
> A patch just went into Paulus' tree that allows you to have a single
> zImage for an arbitrary number of compatible machines: you just run a
> script to write new boot arguments into a buffer in the zImage. A
> similar patch could certainly be constructed for Xen. This is nice for
> ten machines, but since you really should copy the kernel before editing
> it, you end up with hundreds of binaries sitting around whose only
> difference is boot arguments. That is, the same situation you would
> have been in had you just written a script to edit the .config and
> compiled a hundred times.
>
I'd hack up a tftp daemon to do the substitution on the fly.
Or, perhaps a bit more generically, to have the tftp daemon recognize
that certain paths are to be fetched from the output of certain
commands.
--
Michal Ostrowski <mostrows@xxxxxxxxxxxxxx>
_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ppc-devel
|
Home