Noah,
On 8/19/05, Noah Yan <noah.yan at gmail.com> wrote:
> It is a little bit surprising to me that my last post brought lots of noise
> about kernel loading. And I would like very much to start a new post with a
> summary. I read through all
The list was on the nap-mode for some time and fresh voice awaken it :)
> I also donot fully understand the comments of imp about "having to implement
> complete NFS stack in order to load kernel modules" if using NFS root file
> system, can somebody explain in more detail?
Solaris kernel consists of many separate kernel modules, which are
loaded on demand
from the root (or some other) file system during the boot process.
Naturally the one needs
the file system access methods to be available in order to be able to
load the needed
modules into the memory. In case of NFS root that would be NFS
protocol. Since that
(module loading) would take place when the "normal" IP stack will not
yet be available
(even the network driver itself wouldn't be available at early stages
of booting)
the NFS implementation will have to use OpenFirmware interface. However, Darren
is absolutely right - the very same method is used today in SPARC architecture,
so the port should be relatively straightforward.
>
> Also what is mean "the "official" development platform is the Pegasos"?, or
> any doc about the target ppc system for porting?
It is agreed between participants that the reference platform would be
Pegasos II.
It was clear that having the same hardware among core group will be beneficial
for the project. Support for other platform would be then added incrementally.
Regards,
Cyril
