[PATCH 1/2] drm/amdgpu/gfx9: set additional bits on CP halt

Alex Deucher alexander.deucher at amd.com
Thu Sep 19 16:46:59 UTC 2024


Need to set the pipe reset and cache invalidation bits
on halt otherwise we can get stale state if the CP firmware
changes (e.g., on module unload and reload).

Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 23f0573ae47b3..d4f19eb8ff6d3 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -3184,6 +3184,15 @@ static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp = RREG32_SOC15(GC, 0, mmCP_ME_CNTL);
 
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE1_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1);
@@ -3393,7 +3402,15 @@ static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, 0);
 	} else {
 		WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL,
-			(CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
+				 (CP_MEC_CNTL__MEC_INVALIDATE_ICACHE_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE0_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE1_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE2_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE3_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME2_PIPE0_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME2_PIPE1_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_HALT_MASK |
+				  CP_MEC_CNTL__MEC_ME2_HALT_MASK));
 		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
-- 
2.46.0



More information about the amd-gfx mailing list