Description of problem: Users of AMD graphic processing units leveraging GCN architecture from before the GCN 1.2 revision cannot use amdgpu driver that supports Vulkan (current one doesn't). Version-Release number of selected component (if applicable): 4.14.8 (current) How reproducible: 100% on my GCN 1.0 hardware Steps to Reproduce: 1. Use GCN1.0 or likely GCN 1.1 amd-gpu as well; 2. Add "radeon.si_support=0 radeon.cik_support=0 amdgpu.si_support=1 amdgpu.cik_support=1" to boot parameters 3. Reboot Actual results: no gpu acceleration is present (low resolution etc.) Expected results: Fully working driver that supports Vulkan and is on pair with currently used driver in other areas. Additional info: This driver is considered experimental BUT it will not make it worse for anyone as it still has to be enabled by the kernel command line parameters and it brings the possibility of using Vulkan API plus is a little bit faster at opengl and makes the previous driver redundant in the long run. I can only see benefits of enabling it.
Moving this to the graphics team for appropriate tracking.
I might have not stated it clear enough: - kernel is compiled without CONFIG_DRM_AMDGPU_SI=Y One might say that this driver is experimental (since 4.9) but it's not the default even when you enable it at compile time, users still get radeon driver. Advanced users who would like to try out vulkan, or test the driver etc. still have to set the kernel command line with parameters (radeon.si_support=0 amdgpu.si_support=1) and probably blacklist radeon module elsewhere.
This has been fixed with kernel 4.17 in F28 for me. AMD Oland is working fine when booted with radeon.si_support=0 amdgpu.si_support=1 kernel parameters.
Ups, it was 4.16 whet it started working, not 4.17.
Ah, thanks for catching this. Yes, it was turned on with the rebase to 4.16.