This patch fixes a couple of bugs that fell out from the porting
effort.
Signed-off-by: Tom St Denis <tom.stdenis at amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
index 440e347c617b..9423501bbaad 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
@@ -1607,10 +1607,10 @@ err1:
static void gfx_v6_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
{
int i;
- if (enable)
+ if (enable) {
WREG32(mmCP_ME_CNTL, 0);
- else {
- WREG32(mmCP_ME_CNTL, (CP_ME_CNTL__CE_HALT_MASK |
+ } else {
+ WREG32(mmCP_ME_CNTL, (CP_ME_CNTL__ME_HALT_MASK |
CP_ME_CNTL__PFP_HALT_MASK |
CP_ME_CNTL__CE_HALT_MASK));
WREG32(mmSCRATCH_UMSK, 0);
@@ -1899,6 +1899,7 @@ static int gfx_v6_0_cp_compute_resume(struct amdgpu_device *adev)
r = amdgpu_ring_test_ring(&adev->gfx.compute_ring[i]);
if (r)
return r;
+ adev->gfx.compute_ring[i].ready = true;
}
return 0;
--
2.10.0
