Doing this at the binary level would be nice, but I see enough problems
just doing it at the source level and at least for my purposes, doing it
on a dtb would be overkill, I think.   =


The main difficulty remains how to deal with cross references between
nodes in a reasonable way where the references cross from one fragment
to another.

Steve

-----Original Message-----
From: Benjamin Herrenschmidt [mailto:benh@kernel.crashing.org]
Sent: Monday, June 30, 2008 11:22 PM
To: Stephen Neuendorffer
Cc: John Williams; grant.likely@secretlab.ca;

linux-arch@vger.kernel.org; Michal Simek;

vapier.adi@gmail.com; arnd@arndb.de; matthew@wil.cx;

microblaze-uclinux@itee.uq.edu.au; linux-

kernel@vger.kernel.org; linuxppc-dev@ozlabs.org;

will.newton@gmail.com; hpa@zytor.com; John Linn;

monstr@seznam.cz; drepper@redhat.com; alan@lxorguk.ukuu.org.uk
Subject: RE: [PATCH 12/60] microblaze_v4: Generic dts file for

platforms

=

=

quoted

As for the copyright, I haven't been able to find much information

on

quoted

whether or not generated files are even copyrightable.  One might
argue that they
don't have sufficient 'creative value' to be copyrightable.  Or
arguably, they are as copyrightable by the generator author as by

the

quoted

author or the .mhs file.
I admit in this case, I've followed the safe route by claiming a
copyright, which at least at Xilinx has significant precedent.

=

Also, thinking about your idea of sticking bits in BRAM etc...
=

what would be nice would be the ability to "merge" trees. We've been
talking about that multiple times, it would be useful at several

levels:

=

 - We could provide pre-made DTs for known CPUs (ie, 440GP, 440GX,
405EX, ...)
 - Boards can then include that, and then "override" some properties
(clocks, PHY wiring, ...)
 - That could be done at the binary level too so that the BRAM

contains

on "overlay" on top of the base ref. platform device-tree that comes
with the kernel for example.
=

This is slightly different between doing that in the .dts source via
some kind of #include vs. doing that by merging blobs but we could

make

it be essentially be the same internally: The #include generates a

blob

that is then "merged in".
=

Just random thoughts...
=

Ben.
=

=

This email and any attachments are intended for the sole use of the named r=
ecipient(s) and contain(s) confidential information that may be proprietary=
, privileged or copyrighted under applicable law. If you are not the intend=
ed recipient, do not read, copy, or forward this email message or any attac=
hments. Delete this email message and any attachments immediately.