[PATCH 01/13] drm/amdgpu: introduce and honour DRM_FORCE_AUTH workaround

Michel Dänzer michel at daenzer.net
Fri Jun 14 14:16:17 UTC 2019


On 2019-06-14 2:55 p.m., Koenig, Christian wrote:
> Am 14.06.19 um 14:09 schrieb Emil Velikov:
> 
>> That said, the proposal will not conflict with the DRM_AUTH removal. If
>> anything it is step 0.5 of the grand master plan.
> 
> That's the point I strongly disagree on.
> 
> By lowering the DRM_AUTH restriction you are encouraging userspace to 
> use the shortcut and use the primary node for rendering instead of 
> fixing the code and using the render node.
> 
> So at the end of the day userspace will most likely completely drop 
> support for the render node, simply for the reason that it became 
> superfluous. You can just open up the primary node and get the same 
> functionality.
> 
> I absolutely can't understand how you can argue that this won't make it 
> harder to cleanly separate rendering and primary node functionality.

FWIW, I agree with Christian on this.


Also FWIW, the solution I'm working on for
https://bugs.freedesktop.org/110903 should allow making libdrm_amdgpu
always use a render node for rendering. For backwards compatibility
it'll probably require adding a new variant of amdgpu_device_initialize
though, since existing users may rely on the first amdgpu_device for a
GPU using the DRM file description passed to amdgpu_device_initialize
for rendering.


-- 
Earthling Michel Dänzer               |              https://www.amd.com
Libre software enthusiast             |             Mesa and X developer


More information about the amd-gfx mailing list