drm/amdgpu: use sdma_v6 single packet invalidation

This achieves the same result as the sequence used in emit_flush_gpu_tlb
but the invalidation is now a single packet instead of the 3 packets
required to implement reg_write_reg_wait.

Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: peppsac

drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c

@@ -1173,7 +1173,28 @@ static void sdma_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 static void sdma_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					 unsigned vmid, uint64_t pd_addr)
 {
-	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
+	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
+
+	/* Update the PD address for this VMID. */
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 +
+			      (hub->ctx_addr_distance * vmid),
+			      lower_32_bits(pd_addr));
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 +
+			      (hub->ctx_addr_distance * vmid),
+			      upper_32_bits(pd_addr));
+
+	/* Trigger invalidation. */
+	amdgpu_ring_write(ring,
+			  SDMA_PKT_VM_INVALIDATION_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_VM_INVALIDATION_HEADER_SUB_OP(SDMA_SUBOP_VM_INVALIDATION) |
+			  SDMA_PKT_VM_INVALIDATION_HEADER_GFX_ENG_ID(ring->vm_inv_eng) |
+			  SDMA_PKT_VM_INVALIDATION_HEADER_MM_ENG_ID(0x1f));
+	amdgpu_ring_write(ring, req);
+	amdgpu_ring_write(ring, 0xFFFFFFFF);
+	amdgpu_ring_write(ring,
+			  SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_INVALIDATEACK(1 << vmid) |
+			  SDMA_PKT_VM_INVALIDATION_ADDRESSRANGEHI_ADDRESSRANGEHI(0x1F));
 }
 
 static void sdma_v6_0_ring_emit_wreg(struct amdgpu_ring *ring,