With kernel 4.10-rc1 I get the following error on Carrizo: [ 5.414764] Console: switching to colour frame buffer device 240x67 [ 5.419628] amdgpu 0000:00:01.0: fb0: amdgpudrmfb frame buffer device [ 5.426001] [drm] ib test on ring 0 succeeded [ 5.426315] [drm] ib test on ring 1 succeeded [ 5.426384] [drm] ib test on ring 2 succeeded [ 5.426426] [drm] ib test on ring 3 succeeded [ 5.426464] [drm] ib test on ring 4 succeeded [ 5.426506] [drm] ib test on ring 5 succeeded [ 5.426545] [drm] ib test on ring 6 succeeded [ 5.426583] [drm] ib test on ring 7 succeeded [ 5.426623] [drm] ib test on ring 8 succeeded [ 5.426657] [drm] ib test on ring 9 succeeded [ 5.426688] [drm] ib test on ring 10 succeeded [ 6.453373] [drm] ib test on ring 11 succeeded [ 7.688045] [drm:amdgpu_vce_ring_test_ib] *ERROR* amdgpu: IB test timed out. [ 7.688088] [drm:amdgpu_ib_ring_tests] *ERROR* amdgpu: failed testing IB on ring 12 (-110). [ 7.688122] [drm:amdgpu_device_init] *ERROR* ib ring test failed (-110). [ 7.688268] [ powerplay ] min_core_set_clock not set [ 8.397417] [drm] Initialized amdgpu 3.9.0 20150101 for 0000:00:01.0 on minor 0 Bisecting was a pain in the ass this time cause of three other bugs. But I was able to track this down go: commit ecc2cf7cc8baa1fdb73a7bb9495f6befbcac8cd8 Author: Maruthi Srinivas Bayyavarapu <Maruthi.Bayyavarapu@amd.com> Date: Thu Nov 17 17:29:50 2016 +0530 drm/amdgpu: enable VCE clockgating in Polaris-10/11 VCE clocks are set to be disabled, when not in use. Signed-off-by: Maruthi Bayyavarapu <maruthi.bayyavarapu@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Simply reverting this wasn't possible due to conflicts, but only reverting this part diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 39f03f137a56..6b3293a1c7b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -134,7 +134,7 @@ static void vce_v3_0_set_vce_sw_clock_gating(struct amdgpu_device *adev, accessible but the firmware will throttle the clocks on the fly as necessary. */ - if (gated) { + if (!gated) { data = RREG32(mmVCE_CLOCK_GATING_B); data |= 0x1ff; data &= ~0xef0000; made my system boot as before.
Created attachment 248741 [details] full dmesg output
Created attachment 250711 [details] dmesg output with RX460 I have same errors with Gigabyte RX460 and ~agd5f/linux/log/drivers/gpu/drm/amd?h=drm-next-4.10-wip kernel that I cloned today. Computer seems to work normally, but booting is 3 seconds slower because of this and cpu firmware bug traces.
With amdgpu.dpm=0 this doesn't occur. Also tested with amdgpu.powerplay=0, but it didn't help. I don't know about the meaning of the values applied in vce_v3_0_set_vce_sw_clock_gating(), but just inverting the "if (gated)" looks wrong to me.
I can confirm that https://lists.freedesktop.org/archives/amd-gfx/2017-January/004537.html fixes boot for me. Tested on top of linux-4.10.0-rc3-00029-gbd5d7428f5e5
fixed -> closing