drm/amdgpu/vcn3.0: add wptr/rptr reset/update for share memory

Because of dpg, the rptr/wptr need to be saved on fw shared memory,
and restore them back in RBC_RB_RPTR/WPTR in kernel at power up.
Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h

@@ -155,6 +155,7 @@
 		}										\
 	} while (0)
 
+#define AMDGPU_VCN_FW_SHARED_FLAG_0_RB	(1 << 6)
 #define AMDGPU_VCN_MULTI_QUEUE_FLAG	(1 << 8)
 #define AMDGPU_VCN_SW_RING_FLAG		(1 << 9)
 
@@ -243,6 +244,12 @@ struct amdgpu_vcn {
 		int inst_idx, struct dpg_pause_state *new_state);
 };
 
+struct amdgpu_fw_shared_rb_ptrs_struct {
+	/* to WA DPG R/W ptr issues.*/
+	uint32_t  rptr;
+	uint32_t  wptr;
+};
+
 struct amdgpu_fw_shared_multi_queue {
 	uint8_t decode_queue_mode;
 	uint8_t encode_generalpurpose_queue_mode;
@@ -258,10 +265,12 @@ struct amdgpu_fw_shared_sw_ring {
 
 struct amdgpu_fw_shared {
 	uint32_t present_flag_0;
-	uint8_t pad[53];
+	uint8_t pad[44];
+	struct amdgpu_fw_shared_rb_ptrs_struct rb;
+	uint8_t pad1[1];
 	struct amdgpu_fw_shared_multi_queue multi_queue;
 	struct amdgpu_fw_shared_sw_ring sw_ring;
-} __attribute__((__packed__));
+};
 
 struct amdgpu_vcn_decode_buffer {
 	uint32_t valid_buf_flag;

drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c

@@ -238,7 +238,8 @@ static int vcn_v3_0_sw_init(void *handle)
 
 		fw_shared = adev->vcn.inst[i].fw_shared_cpu_addr;
 		fw_shared->present_flag_0 |= cpu_to_le32(AMDGPU_VCN_SW_RING_FLAG) |
-					     cpu_to_le32(AMDGPU_VCN_MULTI_QUEUE_FLAG);
+					     cpu_to_le32(AMDGPU_VCN_MULTI_QUEUE_FLAG) |
+					     cpu_to_le32(AMDGPU_VCN_FW_SHARED_FLAG_0_RB);
 		fw_shared->sw_ring.is_enabled = cpu_to_le32(DEC_SW_RING_ENABLED);
 	}
 
@@ -1074,7 +1075,13 @@ static int vcn_v3_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo
 	WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_WPTR,
 		lower_32_bits(ring->wptr));
 
+	/* Reset FW shared memory RBC WPTR/RPTR */
+	fw_shared->rb.rptr = 0;
+	fw_shared->rb.wptr = lower_32_bits(ring->wptr);
+
+	/*resetting done, fw can check RB ring */
 	fw_shared->multi_queue.decode_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
+
 	/* Unstall DPG */
 	WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS),
 		0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK);
@@ -1239,9 +1246,11 @@ static int vcn_v3_0_start(struct amdgpu_device *adev)
 		/* Initialize the ring buffer's read and write pointers */
 		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR, 0);
 
+		WREG32_SOC15(VCN, i, mmUVD_SCRATCH2, 0);
 		ring->wptr = RREG32_SOC15(VCN, i, mmUVD_RBC_RB_RPTR);
 		WREG32_SOC15(VCN, i, mmUVD_RBC_RB_WPTR,
 			lower_32_bits(ring->wptr));
+		fw_shared->rb.wptr = lower_32_bits(ring->wptr);
 		fw_shared->multi_queue.decode_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
 
 		fw_shared->multi_queue.encode_generalpurpose_queue_mode |= cpu_to_le32(FW_QUEUE_RING_RESET);
@@ -1662,6 +1671,10 @@ static int vcn_v3_0_pause_dpg_mode(struct amdgpu_device *adev,
 				WREG32_SOC15(VCN, inst_idx, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
 				fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET);
 
+				/* restore wptr/rptr with pointers saved in FW shared memory*/
+				WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_RPTR, fw_shared->rb.rptr);
+				WREG32_SOC15(VCN, inst_idx, mmUVD_RBC_RB_WPTR, fw_shared->rb.wptr);
+
 				/* Unstall DPG */
 				WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS),
 					0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK);
@@ -1721,6 +1734,15 @@ static uint64_t vcn_v3_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
 static void vcn_v3_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
+	volatile struct amdgpu_fw_shared *fw_shared;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
+		/*whenever update RBC_RB_WPTR, we save the wptr in shared rb.wptr and scratch2 */
+		fw_shared = adev->vcn.inst[ring->me].fw_shared_cpu_addr;
+		fw_shared->rb.wptr = lower_32_bits(ring->wptr);
+		WREG32_SOC15(VCN, ring->me, mmUVD_SCRATCH2,
+			lower_32_bits(ring->wptr));
+	}
 
 	if (ring->use_doorbell) {
 		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);