drm/amd/display: fix inputting clk lvl into dml for RN

[Why]
Previous logic is only good for 15W parts. Other configuration
need a smarter logic to match clk levels with pp table in the fuse.

[How]
Cache all 8 DPM level's clock data, find lvl that match each pstate
in the pp table and build input into DML base on that

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

2 files changed, 127 insertions, 41 deletions

diff --git a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c
index 0d506d30d6b6..5f6fb16f16e5 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c
@@ -162,10 +162,10 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_1_soc = {
 	.clock_limits = {
 			{
 				.state = 0,
-				.dcfclk_mhz = 304.0,
-				.fabricclk_mhz = 600.0,
-				.dispclk_mhz = 618.0,
-				.dppclk_mhz = 440.0,
+				.dcfclk_mhz = 400.0,
+				.fabricclk_mhz = 400.0,
+				.dispclk_mhz = 600.0,
+				.dppclk_mhz = 400.00,
 				.phyclk_mhz = 600.0,
 				.socclk_mhz = 278.0,
 				.dscclk_mhz = 205.67,
@@ -173,10 +173,10 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_1_soc = {
 			},
 			{
 				.state = 1,
-				.dcfclk_mhz = 304.0,
-				.fabricclk_mhz = 600.0,
-				.dispclk_mhz = 618.0,
-				.dppclk_mhz = 618.0,
+				.dcfclk_mhz = 464.52,
+				.fabricclk_mhz = 800.0,
+				.dispclk_mhz = 654.55,
+				.dppclk_mhz = 626.09,
 				.phyclk_mhz = 600.0,
 				.socclk_mhz = 278.0,
 				.dscclk_mhz = 205.67,
@@ -184,32 +184,65 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_1_soc = {
 			},
 			{
 				.state = 2,
-				.dcfclk_mhz = 608.0,
-				.fabricclk_mhz = 1066.0,
-				.dispclk_mhz = 888.0,
-				.dppclk_mhz = 888.0,
-				.phyclk_mhz = 810.0,
+				.dcfclk_mhz = 514.29,
+				.fabricclk_mhz = 933.0,
+				.dispclk_mhz = 757.89,
+				.dppclk_mhz = 685.71,
+				.phyclk_mhz = 600.0,
 				.socclk_mhz = 278.0,
 				.dscclk_mhz = 287.67,
-				.dram_speed_mts = 2133.0,
+				.dram_speed_mts = 1866.0,
 			},
 			{
 				.state = 3,
-				.dcfclk_mhz = 676.0,
-				.fabricclk_mhz = 1600.0,
-				.dispclk_mhz = 1015.0,
-				.dppclk_mhz = 1015.0,
-				.phyclk_mhz = 810.0,
+				.dcfclk_mhz = 576.00,
+				.fabricclk_mhz = 1067.0,
+				.dispclk_mhz = 847.06,
+				.dppclk_mhz = 757.89,
+				.phyclk_mhz = 600.0,
 				.socclk_mhz = 715.0,
 				.dscclk_mhz = 318.334,
-				.dram_speed_mts = 4266.0,
+				.dram_speed_mts = 2134.0,
 			},
 			{
 				.state = 4,
-				.dcfclk_mhz = 810.0,
+				.dcfclk_mhz = 626.09,
+				.fabricclk_mhz = 1200.0,
+				.dispclk_mhz = 900.00,
+				.dppclk_mhz = 847.06,
+				.phyclk_mhz = 810.0,
+				.socclk_mhz = 953.0,
+				.dscclk_mhz = 489.0,
+				.dram_speed_mts = 2400.0,
+			},
+			{
+				.state = 5,
+				.dcfclk_mhz = 685.71,
+				.fabricclk_mhz = 1333.0,
+				.dispclk_mhz = 1028.57,
+				.dppclk_mhz = 960.00,
+				.phyclk_mhz = 810.0,
+				.socclk_mhz = 278.0,
+				.dscclk_mhz = 287.67,
+				.dram_speed_mts = 2666.0,
+			},
+			{
+				.state = 6,
+				.dcfclk_mhz = 757.89,
+				.fabricclk_mhz = 1467.0,
+				.dispclk_mhz = 1107.69,
+				.dppclk_mhz = 1028.57,
+				.phyclk_mhz = 810.0,
+				.socclk_mhz = 715.0,
+				.dscclk_mhz = 318.334,
+				.dram_speed_mts = 3200.0,
+			},
+			{
+				.state = 7,
+				.dcfclk_mhz = 847.06,
 				.fabricclk_mhz = 1600.0,
 				.dispclk_mhz = 1395.0,
-				.dppclk_mhz = 1285.0,
+				.dppclk_mhz = 1285.00,
 				.phyclk_mhz = 1325.0,
 				.socclk_mhz = 953.0,
 				.dscclk_mhz = 489.0,
@@ -217,8 +250,8 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_1_soc = {
 			},
 			/*Extra state, no dispclk ramping*/
 			{
-				.state = 5,
-				.dcfclk_mhz = 810.0,
+				.state = 8,
+				.dcfclk_mhz = 847.06,
 				.fabricclk_mhz = 1600.0,
 				.dispclk_mhz = 1395.0,
 				.dppclk_mhz = 1285.0,
@@ -265,7 +298,7 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_1_soc = {
 	.xfc_bus_transport_time_us = 4,
 	.xfc_xbuf_latency_tolerance_us = 4,
 	.use_urgent_burst_bw = 1,
-	.num_states = 5
+	.num_states = 9
 };
 
 #ifndef MAX
@@ -1333,26 +1366,78 @@ static void update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_param
 {
 	struct dcn21_resource_pool *pool = TO_DCN21_RES_POOL(dc->res_pool);
 	struct clk_limit_table *clk_table = &bw_params->clk_table;
-	int i;
+	unsigned int i, j, k;
+	int closest_clk_lvl;
+
+	// diags does not retrieve proper values from SMU
+	// cap states to 5 and make state 5 the max state
+	if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment) || IS_DIAG_DC(dc->ctx->dce_environment)) {
+		dcn2_1_soc.num_states = 5;
+
+		dcn2_1_soc.clock_limits[5].state = 5;
+		dcn2_1_soc.clock_limits[5].dcfclk_mhz = 810.0;
+		dcn2_1_soc.clock_limits[5].fabricclk_mhz = 1600.0;
+		dcn2_1_soc.clock_limits[5].dispclk_mhz = 1395.0;
+		dcn2_1_soc.clock_limits[5].dppclk_mhz = 1285.0;
+		dcn2_1_soc.clock_limits[5].phyclk_mhz = 1325.0;
+		dcn2_1_soc.clock_limits[5].socclk_mhz = 953.0;
+		dcn2_1_soc.clock_limits[5].dscclk_mhz = 489.0;
+		dcn2_1_soc.clock_limits[5].dram_speed_mts = 4266.0;
+	} else {
+		dcn2_1_ip.max_num_otg = pool->base.res_cap->num_timing_generator;
+		dcn2_1_ip.max_num_dpp = pool->base.pipe_count;
+		dcn2_1_soc.num_chans = bw_params->num_channels;
+
+		/* Vmin: leave lowest DCN clocks, override with dcfclk, fclk, memclk from fuse */
+		dcn2_1_soc.clock_limits[0].state = 0;
+		dcn2_1_soc.clock_limits[0].dcfclk_mhz = clk_table->entries[0].dcfclk_mhz;
+		dcn2_1_soc.clock_limits[0].fabricclk_mhz = clk_table->entries[0].fclk_mhz;
+		dcn2_1_soc.clock_limits[0].socclk_mhz = clk_table->entries[0].socclk_mhz;
+		dcn2_1_soc.clock_limits[0].dram_speed_mts = clk_table->entries[0].memclk_mhz * 2;
+
+		/*
+		 * Other levels: find cloest DCN clocks that fit the given clock limit using dcfclk
+		 * as indicater
+		 */
 
-	dcn2_1_ip.max_num_otg = pool->base.res_cap->num_timing_generator;
-	dcn2_1_ip.max_num_dpp = pool->base.pipe_count;
-	dcn2_1_soc.num_chans = bw_params->num_channels;
+		closest_clk_lvl = -1;
+		/* index currently being filled */
+		k = 1;
+		for (i = 1; i < clk_table->num_entries; i++) {
+			/* loop backwards, skip duplicate state, +1 because SMU has precision issue */
+			for (j = dcn2_1_soc.num_states - 2; j >= k; j--) {
+				if ((unsigned int) dcn2_1_soc.clock_limits[j].dcfclk_mhz <= clk_table->entries[i].dcfclk_mhz) {
+					closest_clk_lvl = j;
+					break;
+				}
+			}
 
-	for (i = 0; i < clk_table->num_entries; i++) {
+			/* if found a lvl that fits, use the DCN clks from it, if not, go to next clk limit*/
+			if (closest_clk_lvl != -1) {
+				dcn2_1_soc.clock_limits[k].state = i;
+				dcn2_1_soc.clock_limits[k].dcfclk_mhz = clk_table->entries[i].dcfclk_mhz;
+				dcn2_1_soc.clock_limits[k].fabricclk_mhz = clk_table->entries[i].fclk_mhz;
+				dcn2_1_soc.clock_limits[k].socclk_mhz = clk_table->entries[i].socclk_mhz;
+				dcn2_1_soc.clock_limits[k].dram_speed_mts = clk_table->entries[i].memclk_mhz * 2;
+
+				dcn2_1_soc.clock_limits[k].dispclk_mhz = dcn2_1_soc.clock_limits[closest_clk_lvl].dispclk_mhz;
+				dcn2_1_soc.clock_limits[k].dppclk_mhz = dcn2_1_soc.clock_limits[closest_clk_lvl].dppclk_mhz;
+				dcn2_1_soc.clock_limits[k].dram_bw_per_chan_gbps = dcn2_1_soc.clock_limits[closest_clk_lvl].dram_bw_per_chan_gbps;
+				dcn2_1_soc.clock_limits[k].dscclk_mhz = dcn2_1_soc.clock_limits[closest_clk_lvl].dscclk_mhz;
+				dcn2_1_soc.clock_limits[k].dtbclk_mhz = dcn2_1_soc.clock_limits[closest_clk_lvl].dtbclk_mhz;
+				dcn2_1_soc.clock_limits[k].phyclk_d18_mhz = dcn2_1_soc.clock_limits[closest_clk_lvl].phyclk_d18_mhz;
+				dcn2_1_soc.clock_limits[k].phyclk_mhz = dcn2_1_soc.clock_limits[closest_clk_lvl].phyclk_mhz;
+				k++;
+			}
+		}
 
-		dcn2_1_soc.clock_limits[i].state = i;
-		dcn2_1_soc.clock_limits[i].dcfclk_mhz = clk_table->entries[i].dcfclk_mhz;
-		dcn2_1_soc.clock_limits[i].fabricclk_mhz = clk_table->entries[i].fclk_mhz;
-		dcn2_1_soc.clock_limits[i].socclk_mhz = clk_table->entries[i].socclk_mhz;
-		dcn2_1_soc.clock_limits[i].dram_speed_mts = clk_table->entries[i].memclk_mhz * 2;
+		/* duplicate last level */
+		dcn2_1_soc.clock_limits[k] = dcn2_1_soc.clock_limits[k - 1];
+		dcn2_1_soc.clock_limits[k].state = k;
+		dcn2_1_soc.num_states = k + 1;
 	}
-	dcn2_1_soc.clock_limits[i] = dcn2_1_soc.clock_limits[i - 1];
-	dcn2_1_soc.num_states = i;
 
-	// diags does not retrieve proper values from SMU, do not update DML instance for diags
-	if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment) && !IS_DIAG_DC(dc->ctx->dce_environment))
-		dml_init_instance(&dc->dml, &dcn2_1_soc, &dcn2_1_ip, DML_PROJECT_DCN21);
+	dml_init_instance(&dc->dml, &dcn2_1_soc, &dcn2_1_ip, DML_PROJECT_DCN21);
 }
 
 /* Temporary Place holder until we can get them from fuse */
diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h
index 1f5446321112..9bb8bff4cdd9 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h
+++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h
@@ -25,7 +25,7 @@
 #ifndef __DISPLAY_MODE_STRUCTS_H__
 #define __DISPLAY_MODE_STRUCTS_H__
 
-#define MAX_CLOCK_LIMIT_STATES 8
+#define MAX_CLOCK_LIMIT_STATES 9
 
 typedef struct _vcs_dpi_voltage_scaling_st voltage_scaling_st;
 typedef struct _vcs_dpi_soc_bounding_box_st soc_bounding_box_st;
@@ -61,6 +61,7 @@ struct _vcs_dpi_voltage_scaling_st {
 	double dram_speed_mts;
 	double fabricclk_mhz;
 	double dispclk_mhz;
+	double dram_bw_per_chan_gbps;
 	double phyclk_mhz;
 	double dppclk_mhz;
 	double dtbclk_mhz;