Thanks Dan & Tom,

quoted

There are problems as this port was based on an experimental

and unstable

quoted

kernel.
There are patches (tons) that fix some of the problems we've

been seeing,

Really?  Although not suitable for a product, there shouldn't be "tons"
of patches required to make it useful in a development environment.
There may be patches required to work around the different revisions
of the Sandpoint hardware.

We do now need something suitable for a product.  And to use initrd as root.

I was attempting to summarize the following conclusions of a co-worker:

"By the way, last night I found info on the one problem we were having with
the linux kernel booting.  The people at Montavista originally ported
linux 2.3.16 to the PPC 8240 chip. all linux 2.3.* versions are designated
as experimental and unstable.  Since then (feb 2000) there have been over
100 sets of patches to the linux kernel.  I'll take a look to see what is
best to do: apply the changes to the stable 2.2.18 kernel, or apply the
changes to the almost stable 2.4.0-test12, the totally latest kernel
version as of last week.  The changes themselves have to change depending
on what is chosen.  The problem with the initial ram disk is not a ram
disk specific bug, the bug is in the MMU paging/caching system and I don't
think it was fixed until 2.3.47 or so. It can cause other subtle problems,
such as crashing during ftp of large files."

That is where all of the resources seem to be going right now, to
custom hardware.  The 8240 is just a 603 with 106/107 OpenPIC PCI
bridge.  The Sandpoint is what required all of the code changes.  If
your hardware is like a Sandpoint, then there are still changes
required to suit your needs.  Most people aren't building Sandpoint-like
hardware that I know about.

quoted

Eventually we'll make our patches for 8240 available.

Do it quickly, as there will soon be 824x updates in the 2.4 kernel.

We'll probably wait to see what those are, so as to avoid duplication.
Yeah, our board is not very sandpoint-like either.  But changing
sandpoint_setup.c (and sandpoint_pci.c) to suit our board seemed the
easiest.    Now that I'm finished debugging the hardware, the final kernel
changes we needed were minor.

Ron


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/