Re: [PATCH v2 0/6] DRM: VMA Access Management and Render Nodes

[David Herrmann <dh.herrmann@gmail.com>]

Hi

On Fri, Aug 23, 2013 at 1:28 PM, Christian König
<christian.koenig@amd.com> wrote:
> Hi David,
>
> Am 23.08.2013 13:13, schrieb David Herrmann:
>
>> Hi
>>
>> I reduced the vma access-management patches to a minimum. I now do filp*
>> tracking in gem unconditionally and force drm_gem_mmap() to check this.
>> Hence,
>> all gem drivers are safe now. For TTM drivers, I now use the already
>> available
>> verify_access() callback to get access to the underlying gem-object.
>> Pretty
>> simple.. Why hadn't I thought of that before?
>>
>> Long story short: All drivers using GEM are safe now. This leaves vmwgfx..
>> But
>> they do their own access-management, anyway.
>>
>> The 3 patches on top implement render-nodes. I added a "drm_rnodes" module
>> parameter to core drm. You need to pass "drm.rnodes=1" on the kernel
>> command-line or via sysfs _before_ loading a driver. Otherwise, render
>> nodes
>> will not be created.
>>
>> This allows us to test render-nodes and play with the API. I added FLINK
>> for
>> now so we can better test it. Not sure whether we should allow it in the
>> end,
>> though.
>>
>> Maybe we can get this into 3.11?
>
>
> A bit unlikely, but 3.12 should work fine.

whoops, 3.12 of course.

> I'm working on a project that can make good use of this, so if Alex doesn't
> mind like to add the necessary radeon flags (should be only a few one liners
> anyway).

Feel free to send a patch to dri-devel or just let me know the ioctls
and I will include it in this series.

Regards
David