drm/amdgpu/gfx11: wait for reset done before remap

There is a racing condition that cp firmware modifies
MQD in reset sequence after driver updates it for
remapping. We have to wait till CP_HQD_ACTIVE becoming
false then remap the queue.

Acked-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Signed-off-by: Jiadong Zhu <Jiadong.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: jiadozhu

drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c

@@ -6582,16 +6582,29 @@ static int gfx_v11_0_reset_kgq(struct amdgpu_ring *ring, unsigned int vmid)
 static int gfx_v11_0_reset_kcq(struct amdgpu_ring *ring, unsigned int vmid)
 {
 	struct amdgpu_device *adev = ring->adev;
-	int r;
+	int i, r = 0;
 
 	gfx_v11_0_set_safe_mode(adev, 0);
 	mutex_lock(&adev->srbm_mutex);
 	soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 	WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 0x2);
 	WREG32_SOC15(GC, 0, regSPI_COMPUTE_QUEUE_RESET, 0x1);
+
+	/* make sure dequeue is complete*/
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
 	soc21_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 	gfx_v11_0_unset_safe_mode(adev, 0);
+	if (r) {
+		dev_err(adev->dev, "fail to wait on hqd deactivate\n");
+		return r;
+	}
 
 	r = amdgpu_bo_reserve(ring->mqd_obj, false);
 	if (unlikely(r != 0)) {