drm/amd/display: Various DML fixes to enable higher timings

Fixes to enable higher rate timings for DCN3.2.x.
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Chaitanya Dhere <chaitanya.dhere@amd.com>
Signed-off-by: Nevenko Stupar <Nevenko.Stupar@amd.com>
Acked-by: Jerry Zuo <jerry.zuo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hubbub.c

@@ -65,7 +65,7 @@ static void dcn32_init_crb(struct hubbub *hubbub)
 	REG_SET_2(COMPBUF_RESERVED_SPACE, 0,
 			COMPBUF_RESERVED_SPACE_64B, hubbub2->pixel_chunk_size / 32,
 			COMPBUF_RESERVED_SPACE_ZS, hubbub2->pixel_chunk_size / 128);
-	REG_UPDATE(DCHUBBUB_DEBUG_CTRL_0, DET_DEPTH, 0x17F);
+	REG_UPDATE(DCHUBBUB_DEBUG_CTRL_0, DET_DEPTH, 0x47F);
 }
 
 static void dcn32_program_det_size(struct hubbub *hubbub, int hubp_inst, unsigned int det_buffer_size_in_kbyte)
@@ -119,8 +119,8 @@ static void dcn32_program_compbuf_size(struct hubbub *hubbub, unsigned int compb
 		ASSERT(hubbub2->det0_size + hubbub2->det1_size + hubbub2->det2_size
 				+ hubbub2->det3_size + compbuf_size_segments <= hubbub2->crb_size_segs);
 		REG_UPDATE(DCHUBBUB_COMPBUF_CTRL, COMPBUF_SIZE, compbuf_size_segments);
-		REG_WAIT(DCHUBBUB_COMPBUF_CTRL, COMPBUF_SIZE_CURRENT, compbuf_size_segments, 1, 100);
 		hubbub2->compbuf_size_segments = compbuf_size_segments;
+		ASSERT(REG_GET(DCHUBBUB_COMPBUF_CTRL, CONFIG_ERROR, &compbuf_size_segments) && !compbuf_size_segments);
 	}
 }
 

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_resource.c

@@ -3015,13 +3015,30 @@ int dcn32_populate_dml_pipes_from_context(
 		}
 		pipe_cnt++;
 	}
-	context->bw_ctx.dml.ip.det_buffer_size_kbytes = DCN3_2_DEFAULT_DET_SIZE;
 
-	if (pipe_cnt == 1 && pipe->plane_state && !dc->debug.disable_z9_mpc) {
-		if (!is_dual_plane(pipe->plane_state->format)) {
-			context->bw_ctx.dml.ip.det_buffer_size_kbytes = 192;
-			pipes[0].pipe.src.unbounded_req_mode = true;
+	switch (pipe_cnt) {
+	case 1:
+		context->bw_ctx.dml.ip.det_buffer_size_kbytes = DCN3_2_MAX_DET_SIZE;
+		if (pipe->plane_state && !dc->debug.disable_z9_mpc) {
+			if (!is_dual_plane(pipe->plane_state->format)) {
+				context->bw_ctx.dml.ip.det_buffer_size_kbytes = DCN3_2_DEFAULT_DET_SIZE;
+				pipes[0].pipe.src.unbounded_req_mode = true;
+				if (pipe->plane_state->src_rect.width >= 5120 &&
+					pipe->plane_state->src_rect.height >= 2880)
+					context->bw_ctx.dml.ip.det_buffer_size_kbytes = 320; // 5K or higher
+			}
 		}
+		break;
+	case 2:
+		context->bw_ctx.dml.ip.det_buffer_size_kbytes = DCN3_2_MAX_DET_SIZE / 2; // 576 KB (9 segments)
+		break;
+	case 3:
+		context->bw_ctx.dml.ip.det_buffer_size_kbytes = DCN3_2_MAX_DET_SIZE / 3; // 384 KB (6 segments)
+		break;
+	case 4:
+	default:
+		context->bw_ctx.dml.ip.det_buffer_size_kbytes = DCN3_2_DEFAULT_DET_SIZE; // 256 KB (4 segments)
+		break;
 	}
 
 	return pipe_cnt;
@@ -3283,7 +3300,6 @@ void dcn32_calculate_dlg_params(struct dc *dc, struct dc_state *context, display
 	 * possible with firmware driven vertical blank stretching.
 	 */
 	// context->bw_ctx.bw.dcn.clk.p_state_change_support |= context->bw_ctx.bw.dcn.clk.fw_based_mclk_switching;
-
 	context->bw_ctx.bw.dcn.clk.dppclk_khz = 0;
 	context->bw_ctx.bw.dcn.clk.dtbclk_en = is_dtbclk_required(dc, context);
 	if (context->bw_ctx.dml.vba.FCLKChangeSupport[vlevel][context->bw_ctx.dml.vba.maxMpcComb] == dm_fclock_change_unsupported)
@@ -3339,11 +3355,6 @@ void dcn32_calculate_dlg_params(struct dc *dc, struct dc_state *context, display
 		if (!context->res_ctx.pipe_ctx[i].stream)
 			continue;
 
-		/* cstate disabled on 201 */
-//		if (dc->ctx->dce_version == DCN_VERSION_2_01)
-//			cstate_en = false;
-
-
 		context->bw_ctx.dml.funcs.rq_dlg_get_dlg_reg_v2(&context->bw_ctx.dml,
 				&context->res_ctx.pipe_ctx[i].dlg_regs, &context->res_ctx.pipe_ctx[i].ttu_regs, pipes,
 				pipe_cnt, pipe_idx);

drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c

@@ -1018,31 +1018,8 @@ int dcn20_populate_dml_pipes_from_context(
 			pipes[pipe_cnt].pipe.dest.pixel_rate_mhz *= 2;
 		pipes[pipe_cnt].pipe.dest.otg_inst = res_ctx->pipe_ctx[i].stream_res.tg->inst;
 		pipes[pipe_cnt].dout.dp_lanes = 4;
-		if (res_ctx->pipe_ctx[i].stream->link) {
-			switch (res_ctx->pipe_ctx[i].stream->link->cur_link_settings.link_rate) {
-			case LINK_RATE_HIGH:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_hbr;
-				break;
-			case LINK_RATE_HIGH2:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_hbr2;
-				break;
-			case LINK_RATE_HIGH3:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_hbr3;
-				break;
-			case LINK_RATE_UHBR10:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_uhbr10;
-				break;
-			case LINK_RATE_UHBR13_5:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_uhbr13p5;
-				break;
-			case LINK_RATE_UHBR20:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_uhbr20;
-				break;
-			default:
-				pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_na;
-				break;
-			}
-		}
+		if (res_ctx->pipe_ctx[i].stream->link)
+			pipes[pipe_cnt].dout.dp_rate = dm_dp_rate_na;
 		pipes[pipe_cnt].dout.is_virtual = 0;
 		pipes[pipe_cnt].pipe.dest.vtotal_min = res_ctx->pipe_ctx[i].stream->adjust.v_total_min;
 		pipes[pipe_cnt].pipe.dest.vtotal_max = res_ctx->pipe_ctx[i].stream->adjust.v_total_max;