[PATCH 1/1] drm/shmem: Dual licence the files as GPL-2 and MIT

Rob Herring robh at kernel.org
Mon Nov 14 18:59:34 UTC 2022


On Sat, Nov 12, 2022 at 1:44 PM Robert Swindells <rjs at fdy2.co.uk> wrote:
>
> Contributors to these files are:
>
> Noralf Trønnes <noralf at tronnes.org>
> Liu Zixian <liuzixian4 at huawei.com>
> Dave Airlie <airlied at redhat.com>
> Thomas Zimmermann <tzimmermann at suse.de>
> Lucas De Marchi <lucas.demarchi at intel.com>
> Gerd Hoffmann <kraxel at redhat.com>
> Rob Herring <robh at kernel.org>

My contributions are related to the madvise functions. That's largely
lifted or inspired from the MSM code which is GPL only. That in turn
looks inspired from i915 which is MIT (though not much more than the
comment):

$ git grep 'Our goal here is to return as much of the memory'
drivers/gpu/drm/drm_gem_shmem_helper.c: /* Our goal here is to return
as much of the memory as
drivers/gpu/drm/i915/gem/i915_gem_shmem.c:       * Our goal here is to
return as much of the memory as
drivers/gpu/drm/msm/msm_gem.c:  /* Our goal here is to return as much
of the memory as


I imagine this is not the only example in this file. In fact, looking
at the introduction of this file, it looks like it originated from V3D
code as that was the first driver to convert over. V3D is licensed
GPL2+. Of course, its code was not written in a vacuum either and came
from ???

This to me is a problem with the dual licensing in DRM drivers. Code
moves around with little attention paid at the time to licensing. I
wouldn't trust anything claiming MIT license is not GPL
'contaminated'.

OTOH, there's really only one way for the madvise code to work, so
maybe not a copyrightable work on its own.

Rob


More information about the dri-devel mailing list