[PATCH 1/3] drm/radeon: Make CIK support in Radeon conditional (v2)

Michel Dänzer michel at daenzer.net
Tue Apr 11 01:39:24 UTC 2017


On 11/04/17 03:13 AM, Felix Kuehling wrote:
> On 17-04-09 09:28 PM, Michel Dänzer wrote:
>> On 10/04/17 09:56 AM, Michel Dänzer wrote:
>>> On 08/04/17 05:15 AM, Felix Kuehling wrote:
>>>> Advertise CIK PCI IDs only when they are not supported by amdgpu.
>>>> Use the CONFIG_DRM_AMDGPU_CIK to check so that a single option in
>>>> the kernel config keeps both drivers in sync.
>>>>
>>>> This is the simplest possible change. A more complete solution
>>>> may want to conditionally disable more CIK-specific code in the
>>>> Radeon driver.
>>>>
>>>> v2: Add CONFIG_DRM_RADEON_FORCE_CIK option
>>>>
>>>> Signed-off-by: Felix Kuehling <Felix.Kuehling at amd.com>
>>> I don't think this is good enough a solution.
>> To elaborate, my main objection is that people who still want to be able
>> to use the radeon driver for CIK/SI GPUs with CONFIG_DRM_AMDGPU_CIK
>> enabled have to enable the new CONFIG_DRM_RADEON_FORCE_CIK/SI option,
>> and then they're merely back to square one, in the current situation
>> where it's not clearly defined which driver will end up being used for
>> such GPUs.
> 
> So what's the usage model here? I think the situation that both AMDGPU
> and Radeon support CIK is for developers only. "Normal end users" would
> never be in this situation.

I agree in the long term, but for now, there's still too many bugs which
force users to use one or the other driver. Since this patch would
require distros to choose one driver for all of their users, with no way
to override at runtime, it would probably result in bad user experiences
and/or a support nightmare at least in the short term.


> For developers, Christian is OK with having just a compile time option,
> and probably choosing the driver by loading the right module manually.
> You want to have a boot time kernel option.

Note that I'm okay with an inconvenient mechanism for choosing the
driver at runtime, as long as there is one. The important thing is to
fix the current situation where it's essentially up to chance which
driver ends up driving a GPU supported by both.


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


More information about the amd-gfx mailing list