drm/amd/display: Add support for zstate during extended vblank

[why]
When we enter FREESYNC_STATE_VIDEO, we want to use the extra vblank
portion to enter zstate if possible.

[how]
When we enter freesync, a full update is triggered and the new vtotal
with extra lines is passed to dml in a stream update. The time gained
from extra vblank lines is calculated in microseconds. We allow zstate
entry if the time gained is greater than 5 ms, which is the current
policy. Furthermore, an optimized value for min_dst_y_next_start is
calculated and written to its register. When exiting freesync, another
full update is triggered and default values are restored.

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Gabe Teeger <gabe.teeger@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: Gabe Teeger

drivers/gpu/drm/amd/display/dc/core/dc.c

@@ -2393,6 +2393,8 @@ static enum surface_update_type check_update_surfaces_for_stream(
 
 		if (stream_update->mst_bw_update)
 			su_flags->bits.mst_bw = 1;
+		if (stream_update->crtc_timing_adjust && dc_extended_blank_supported(dc))
+			su_flags->bits.crtc_timing_adjust = 1;
 
 		if (su_flags->raw != 0)
 			overall_type = UPDATE_TYPE_FULL;
@@ -2654,6 +2656,9 @@ static void copy_stream_update_to_stream(struct dc *dc,
 	if (update->vrr_infopacket)
 		stream->vrr_infopacket = *update->vrr_infopacket;
 
+	if (update->crtc_timing_adjust)
+		stream->adjust = *update->crtc_timing_adjust;
+
 	if (update->dpms_off)
 		stream->dpms_off = *update->dpms_off;
 
@@ -4055,3 +4060,17 @@ void dc_notify_vsync_int_state(struct dc *dc, struct dc_stream_state *stream, bo
 	if (pipe->stream_res.abm && pipe->stream_res.abm->funcs->set_abm_pause)
 		pipe->stream_res.abm->funcs->set_abm_pause(pipe->stream_res.abm, !enable, i, pipe->stream_res.tg->inst);
 }
+/*
+ * dc_extended_blank_supported: Decide whether extended blank is supported
+ *
+ * Extended blank is a freesync optimization feature to be enabled in the future.
+ * During the extra vblank period gained from freesync, we have the ability to enter z9/z10.
+ *
+ * @param [in] dc: Current DC state
+ * @return: Indicate whether extended blank is supported (true or false)
+ */
+bool dc_extended_blank_supported(struct dc *dc)
+{
+	return dc->debug.extended_blank_optimization && !dc->debug.disable_z10
+		&& dc->caps.zstate_support && dc->caps.is_apu;
+}

drivers/gpu/drm/amd/display/dc/dc.h

@@ -188,6 +188,7 @@ struct dc_caps {
 	bool psp_setup_panel_mode;
 	bool extended_aux_timeout_support;
 	bool dmcub_support;
+	bool zstate_support;
 	uint32_t num_of_internal_disp;
 	enum dp_protocol_version max_dp_protocol_version;
 	unsigned int mall_size_per_mem_channel;
@@ -703,13 +704,14 @@ struct dc_debug_options {
 	bool enable_driver_sequence_debug;
 	enum det_size crb_alloc_policy;
 	int crb_alloc_policy_min_disp_count;
-#if defined(CONFIG_DRM_AMD_DC_DCN)
 	bool disable_z10;
+#if defined(CONFIG_DRM_AMD_DC_DCN)
 	bool enable_z9_disable_interface;
 	bool enable_sw_cntl_psr;
 	union dpia_debug_options dpia_debug;
 #endif
 	bool apply_vendor_specific_lttpr_wa;
+	bool extended_blank_optimization;
 	bool ignore_dpref_ss;
 	uint8_t psr_power_use_phy_fsm;
 };
@@ -1369,6 +1371,8 @@ struct dc_sink_init_data {
 	bool converter_disable_audio;
 };
 
+bool dc_extended_blank_supported(struct dc *dc);
+
 struct dc_sink *dc_sink_create(const struct dc_sink_init_data *init_params);
 
 /* Newer interfaces  */

drivers/gpu/drm/amd/display/dc/dc_stream.h

@@ -131,6 +131,7 @@ union stream_update_flags {
 		uint32_t wb_update:1;
 		uint32_t dsc_changed : 1;
 		uint32_t mst_bw : 1;
+		uint32_t crtc_timing_adjust : 1;
 	} bits;
 
 	uint32_t raw;
@@ -289,6 +290,7 @@ struct dc_stream_update {
 	struct dc_3dlut *lut3d_func;
 
 	struct test_pattern *pending_test_pattern;
+	struct dc_crtc_timing_adjust *crtc_timing_adjust;
 };
 
 bool dc_is_stream_unchanged(

drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c

@@ -1857,6 +1857,7 @@ void dcn20_optimize_bandwidth(
 		struct dc_state *context)
 {
 	struct hubbub *hubbub = dc->res_pool->hubbub;
+	int i;
 
 	/* program dchubbub watermarks */
 	hubbub->funcs->program_watermarks(hubbub,
@@ -1873,6 +1874,17 @@ void dcn20_optimize_bandwidth(
 			dc->clk_mgr,
 			context,
 			true);
+	if (dc_extended_blank_supported(dc) && context->bw_ctx.bw.dcn.clk.zstate_support == DCN_ZSTATE_SUPPORT_ALLOW) {
+		for (i = 0; i < dc->res_pool->pipe_count; ++i) {
+			struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
+
+			if (pipe_ctx->stream && pipe_ctx->plane_res.hubp->funcs->program_extended_blank
+				&& pipe_ctx->stream->adjust.v_total_min == pipe_ctx->stream->adjust.v_total_max
+				&& pipe_ctx->stream->adjust.v_total_max > pipe_ctx->stream->timing.v_total)
+					pipe_ctx->plane_res.hubp->funcs->program_extended_blank(pipe_ctx->plane_res.hubp,
+						pipe_ctx->dlg_regs.optimized_min_dst_y_next_start);
+		}
+	}
 	/* increase compbuf size */
 	if (hubbub->funcs->program_compbuf_size)
 		hubbub->funcs->program_compbuf_size(hubbub, context->bw_ctx.bw.dcn.compbuf_size_kb, true);

drivers/gpu/drm/amd/display/dc/dcn31/dcn31_hubp.c

@@ -54,6 +54,13 @@ void hubp31_soft_reset(struct hubp *hubp, bool reset)
 	REG_UPDATE(DCHUBP_CNTL, HUBP_SOFT_RESET, reset);
 }
 
+void hubp31_program_extended_blank(struct hubp *hubp, unsigned int min_dst_y_next_start_optimized)
+{
+	struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp);
+
+	REG_SET(BLANK_OFFSET_1, 0, MIN_DST_Y_NEXT_START, min_dst_y_next_start_optimized);
+}
+
 static struct hubp_funcs dcn31_hubp_funcs = {
 	.hubp_enable_tripleBuffer = hubp2_enable_triplebuffer,
 	.hubp_is_triplebuffer_enabled = hubp2_is_triplebuffer_enabled,
@@ -80,6 +87,7 @@ static struct hubp_funcs dcn31_hubp_funcs = {
 	.set_unbounded_requesting = hubp31_set_unbounded_requesting,
 	.hubp_soft_reset = hubp31_soft_reset,
 	.hubp_in_blank = hubp1_in_blank,
+	.program_extended_blank = hubp31_program_extended_blank,
 };
 
 bool hubp31_construct(

drivers/gpu/drm/amd/display/dc/dcn31/dcn31_resource.c

@@ -2232,6 +2232,7 @@ static bool dcn31_resource_construct(
 	dc->caps.extended_aux_timeout_support = true;
 	dc->caps.dmcub_support = true;
 	dc->caps.is_apu = true;
+	dc->caps.zstate_support = true;
 
 	/* Color pipeline capabilities */
 	dc->caps.color.dpp.dcn_arch = 1;

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

@@ -722,8 +722,10 @@ static enum dcn_zstate_support_state  decide_zstate_support(struct dc *dc, struc
 {
 	int plane_count;
 	int i;
+	unsigned int optimized_min_dst_y_next_start_us;
 
 	plane_count = 0;
+	optimized_min_dst_y_next_start_us = 0;
 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
 		if (context->res_ctx.pipe_ctx[i].plane_state)
 			plane_count++;
@@ -744,11 +746,22 @@ static enum dcn_zstate_support_state  decide_zstate_support(struct dc *dc, struc
 		struct dc_link *link = context->streams[0]->sink->link;
 		struct dc_stream_status *stream_status = &context->stream_status[0];
 
+		if (dc_extended_blank_supported(dc)) {
+			for (i = 0; i < dc->res_pool->pipe_count; i++) {
+				if (context->res_ctx.pipe_ctx[i].stream == context->streams[0]
+					&& context->res_ctx.pipe_ctx[i].stream->adjust.v_total_min == context->res_ctx.pipe_ctx[i].stream->adjust.v_total_max
+					&& context->res_ctx.pipe_ctx[i].stream->adjust.v_total_min > context->res_ctx.pipe_ctx[i].stream->timing.v_total) {
+						optimized_min_dst_y_next_start_us =
+							context->res_ctx.pipe_ctx[i].dlg_regs.optimized_min_dst_y_next_start_us;
+						break;
+				}
+			}
+		}
 		/* zstate only supported on PWRSEQ0  and when there's <2 planes*/
 		if (link->link_index != 0 || stream_status->plane_count > 1)
 			return DCN_ZSTATE_SUPPORT_DISALLOW;
 
-		if (context->bw_ctx.dml.vba.StutterPeriod > 5000.0)
+		if (context->bw_ctx.dml.vba.StutterPeriod > 5000.0 || optimized_min_dst_y_next_start_us > 5000)
 			return DCN_ZSTATE_SUPPORT_ALLOW;
 		else if (link->psr_settings.psr_version == DC_PSR_VERSION_1 && !dc->debug.disable_psr)
 			return DCN_ZSTATE_SUPPORT_ALLOW_Z10_ONLY;
@@ -786,8 +799,6 @@ void dcn20_calculate_dlg_params(
 							!= dm_dram_clock_change_unsupported;
 	context->bw_ctx.bw.dcn.clk.dppclk_khz = 0;
 
-	context->bw_ctx.bw.dcn.clk.zstate_support = decide_zstate_support(dc, context);
-
 	context->bw_ctx.bw.dcn.clk.dtbclk_en = is_dtbclk_required(dc, context);
 
 	if (context->bw_ctx.bw.dcn.clk.dispclk_khz < dc->debug.min_disp_clk_khz)
@@ -843,6 +854,7 @@ void dcn20_calculate_dlg_params(
 				&pipes[pipe_idx].pipe);
 		pipe_idx++;
 	}
+	context->bw_ctx.bw.dcn.clk.zstate_support = decide_zstate_support(dc, context);
 }
 
 static void swizzle_to_dml_params(

drivers/gpu/drm/amd/display/dc/dml/dcn31/display_rq_dlg_calc_31.c

@@ -1055,6 +1055,7 @@ static void dml_rq_dlg_get_dlg_params(
 
 	float vba__refcyc_per_req_delivery_pre_l = get_refcyc_per_req_delivery_pre_l_in_us(mode_lib, e2e_pipe_param, num_pipes, pipe_idx) * refclk_freq_in_mhz;  // From VBA
 	float vba__refcyc_per_req_delivery_l = get_refcyc_per_req_delivery_l_in_us(mode_lib, e2e_pipe_param, num_pipes, pipe_idx) * refclk_freq_in_mhz;  // From VBA
+	int blank_lines;
 
 	memset(disp_dlg_regs, 0, sizeof(*disp_dlg_regs));
 	memset(disp_ttu_regs, 0, sizeof(*disp_ttu_regs));
@@ -1080,6 +1081,18 @@ static void dml_rq_dlg_get_dlg_params(
 	dlg_vblank_start = interlaced ? (vblank_start / 2) : vblank_start;
 
 	disp_dlg_regs->min_dst_y_next_start = (unsigned int) (((double) dlg_vblank_start) * dml_pow(2, 2));
+	blank_lines = (dst->vblank_end + dst->vtotal_min - dst->vblank_start - dst->vstartup_start - 1);
+	if (blank_lines < 0)
+		blank_lines = 0;
+	if (blank_lines != 0) {
+		disp_dlg_regs->optimized_min_dst_y_next_start_us =
+			((unsigned int) blank_lines * dst->hactive) / (unsigned int) dst->pixel_rate_mhz;
+		disp_dlg_regs->optimized_min_dst_y_next_start =
+			(unsigned int)(((double) (dlg_vblank_start + blank_lines)) * dml_pow(2, 2));
+	} else {
+		// use unoptimized value
+		disp_dlg_regs->optimized_min_dst_y_next_start = disp_dlg_regs->min_dst_y_next_start;
+	}
 	ASSERT(disp_dlg_regs->min_dst_y_next_start < (unsigned int)dml_pow(2, 18));
 
 	dml_print("DML_DLG: %s: min_ttu_vblank (us)         = %3.2f\n", __func__, min_ttu_vblank);

drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h

@@ -446,6 +446,8 @@ struct _vcs_dpi_display_dlg_regs_st {
 	unsigned int refcyc_h_blank_end;
 	unsigned int dlg_vblank_end;
 	unsigned int min_dst_y_next_start;
+	unsigned int optimized_min_dst_y_next_start;
+	unsigned int optimized_min_dst_y_next_start_us;
 	unsigned int refcyc_per_htotal;
 	unsigned int refcyc_x_after_scaler;
 	unsigned int dst_y_after_scaler;

drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h

@@ -195,6 +195,9 @@ struct hubp_funcs {
 
 	void (*hubp_set_flip_int)(struct hubp *hubp);
 
+	void (*program_extended_blank)(struct hubp *hubp,
+			unsigned int min_dst_y_next_start_optimized);
+
 	void (*hubp_wait_pipe_read_start)(struct hubp *hubp);
 };