drm/amd/display: Add FPO + VActive support

[Description]
- When determining FPO support, include FPO + VActive support
- Support FPO + VActive if one display meets regular requirements
  for FPO and the second display is able to switch in VACTIVE with
  a given amount of margin
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -875,6 +875,8 @@ struct dc_debug_options {
 	bool override_dispclk_programming;
 	bool disable_fpo_optimizations;
 	bool support_eDP1_5;
+	uint32_t fpo_vactive_margin_us;
+	bool disable_fpo_vactive;
 };
 
 struct gpu_info_soc_bounding_box_v1_0;

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

@@ -327,6 +327,7 @@ bool dc_dmub_srv_p_state_delegate(struct dc *dc, bool should_manage_pstate, stru
 	int i = 0, k = 0;
 	int ramp_up_num_steps = 1; // TODO: Ramp is currently disabled. Reenable it.
 	uint8_t visual_confirm_enabled;
+	int pipe_idx = 0;
 
 	if (dc == NULL)
 		return false;
@@ -339,6 +340,25 @@ bool dc_dmub_srv_p_state_delegate(struct dc *dc, bool should_manage_pstate, stru
 	cmd.fw_assisted_mclk_switch.config_data.fams_enabled = should_manage_pstate;
 	cmd.fw_assisted_mclk_switch.config_data.visual_confirm_enabled = visual_confirm_enabled;
 
+	if (should_manage_pstate) {
+		for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) {
+			struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
+
+			if (!pipe->stream)
+				continue;
+
+			/* If FAMS is being used to support P-State and there is a stream
+			 * that does not use FAMS, we are in an FPO + VActive scenario.
+			 * Assign vactive stretch margin in this case.
+			 */
+			if (!pipe->stream->fpo_in_use) {
+				cmd.fw_assisted_mclk_switch.config_data.vactive_stretch_margin_us = dc->debug.fpo_vactive_margin_us;
+				break;
+			}
+			pipe_idx++;
+		}
+	}
+
 	for (i = 0, k = 0; context && i < dc->res_pool->pipe_count; i++) {
 		struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
 

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

@@ -726,6 +726,8 @@ static const struct dc_debug_options debug_defaults_drv = {
 	.disable_unbounded_requesting = false,
 	.override_dispclk_programming = true,
 	.disable_fpo_optimizations = false,
+	.fpo_vactive_margin_us = 2000, // 2000us
+	.disable_fpo_vactive = true,
 };
 
 static const struct dc_debug_options debug_defaults_diags = {

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

@@ -39,6 +39,7 @@
 #define DCN3_2_MBLK_HEIGHT_8BPE 64
 #define DCN3_2_VMIN_DISPCLK_HZ 717000000
 #define DCN3_2_DCFCLK_DS_INIT_KHZ 10000 // Choose 10Mhz for init DCFCLK DS freq
+#define DCN3_2_MIN_ACTIVE_SWITCH_MARGIN_FPO_US 100 // Only allow FPO + Vactive if active margin >= 100
 
 #define TO_DCN32_RES_POOL(pool)\
 	container_of(pool, struct dcn32_resource_pool, base)
@@ -146,6 +147,8 @@ void dcn32_restore_mall_state(struct dc *dc,
 		struct dc_state *context,
 		struct mall_temp_config *temp_config);
 
+struct dc_stream_state *dcn32_can_support_mclk_switch_using_fw_based_vblank_stretch(struct dc *dc, const struct dc_state *context);
+
 bool dcn32_allow_subvp_with_active_margin(struct pipe_ctx *pipe);
 
 unsigned int dcn32_calc_num_avail_chans_for_mall(struct dc *dc, int num_chans);

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_resource_helpers.c

@@ -27,6 +27,7 @@
 #include "dcn32_resource.h"
 #include "dcn20/dcn20_resource.h"
 #include "dml/dcn32/display_mode_vba_util_32.h"
+#include "dml/dcn32/dcn32_fpu.h"
 
 static bool is_dual_plane(enum surface_pixel_format format)
 {
@@ -500,3 +501,158 @@ void dcn32_restore_mall_state(struct dc *dc,
 			pipe->plane_state->is_phantom = temp_config->is_phantom_plane[i];
 	}
 }
+
+#define MAX_STRETCHED_V_BLANK 1000 // in micro-seconds (must ensure to match value in FW)
+/*
+ * Scaling factor for v_blank stretch calculations considering timing in
+ * micro-seconds and pixel clock in 100hz.
+ * Note: the parenthesis are necessary to ensure the correct order of
+ * operation where V_SCALE is used.
+ */
+#define V_SCALE (10000 / MAX_STRETCHED_V_BLANK)
+
+static int get_frame_rate_at_max_stretch_100hz(
+		struct dc_stream_state *fpo_candidate_stream,
+		uint32_t fpo_vactive_margin_us)
+{
+	struct dc_crtc_timing *timing = NULL;
+	uint32_t sec_per_100_lines;
+	uint32_t max_v_blank;
+	uint32_t curr_v_blank;
+	uint32_t v_stretch_max;
+	uint32_t stretched_frame_pix_cnt;
+	uint32_t scaled_stretched_frame_pix_cnt;
+	uint32_t scaled_refresh_rate;
+	uint32_t v_scale;
+
+	if (fpo_candidate_stream == NULL)
+		return 0;
+
+	/* check if refresh rate at least 120hz */
+	timing = &fpo_candidate_stream->timing;
+	if (timing == NULL)
+		return 0;
+
+	v_scale = 10000 / (MAX_STRETCHED_V_BLANK + fpo_vactive_margin_us);
+
+	sec_per_100_lines = timing->pix_clk_100hz / timing->h_total + 1;
+	max_v_blank = sec_per_100_lines / v_scale + 1;
+	curr_v_blank = timing->v_total - timing->v_addressable;
+	v_stretch_max = (max_v_blank > curr_v_blank) ? (max_v_blank - curr_v_blank) : (0);
+	stretched_frame_pix_cnt = (v_stretch_max + timing->v_total) * timing->h_total;
+	scaled_stretched_frame_pix_cnt = stretched_frame_pix_cnt / 10000;
+	scaled_refresh_rate = (timing->pix_clk_100hz) / scaled_stretched_frame_pix_cnt + 1;
+
+	return scaled_refresh_rate;
+
+}
+
+static bool is_refresh_rate_support_mclk_switch_using_fw_based_vblank_stretch(
+		struct dc_stream_state *fpo_candidate_stream, uint32_t fpo_vactive_margin_us)
+{
+	int refresh_rate_max_stretch_100hz;
+	int min_refresh_100hz;
+
+	if (fpo_candidate_stream == NULL)
+		return false;
+
+	refresh_rate_max_stretch_100hz = get_frame_rate_at_max_stretch_100hz(fpo_candidate_stream, fpo_vactive_margin_us);
+	min_refresh_100hz = fpo_candidate_stream->timing.min_refresh_in_uhz / 10000;
+
+	if (refresh_rate_max_stretch_100hz < min_refresh_100hz)
+		return false;
+
+	return true;
+}
+
+static int get_refresh_rate(struct dc_stream_state *fpo_candidate_stream)
+{
+	int refresh_rate = 0;
+	int h_v_total = 0;
+	struct dc_crtc_timing *timing = NULL;
+
+	if (fpo_candidate_stream == NULL)
+		return 0;
+
+	/* check if refresh rate at least 120hz */
+	timing = &fpo_candidate_stream->timing;
+	if (timing == NULL)
+		return 0;
+
+	h_v_total = timing->h_total * timing->v_total;
+	if (h_v_total == 0)
+		return 0;
+
+	refresh_rate = ((timing->pix_clk_100hz * 100) / (h_v_total)) + 1;
+	return refresh_rate;
+}
+
+/**
+ * dcn32_can_support_mclk_switch_using_fw_based_vblank_stretch - Determines if config can support FPO
+ *
+ * @param [in]: dc - current dc state
+ * @param [in]: context - new dc state
+ *
+ * Return: Pointer to FPO stream candidate if config can support FPO, otherwise NULL
+ */
+struct dc_stream_state *dcn32_can_support_mclk_switch_using_fw_based_vblank_stretch(struct dc *dc, const struct dc_state *context)
+{
+	int refresh_rate = 0;
+	const int minimum_refreshrate_supported = 120;
+	struct dc_stream_state *fpo_candidate_stream = NULL;
+	bool is_fpo_vactive = false;
+	uint32_t fpo_vactive_margin_us = 0;
+
+	if (context == NULL)
+		return NULL;
+
+	if (dc->debug.disable_fams)
+		return NULL;
+
+	if (!dc->caps.dmub_caps.mclk_sw)
+		return NULL;
+
+	if (context->bw_ctx.bw.dcn.clk.fw_based_mclk_switching_shut_down)
+		return NULL;
+
+	/* For FPO we can support up to 2 display configs if:
+	 * - first display uses FPO
+	 * - Second display switches in VACTIVE */
+	if (context->stream_count > 2)
+		return NULL;
+	else if (context->stream_count == 2) {
+		DC_FP_START();
+		dcn32_assign_fpo_vactive_candidate(dc, context, &fpo_candidate_stream);
+		DC_FP_END();
+
+		DC_FP_START();
+		is_fpo_vactive = dcn32_find_vactive_pipe(dc, context, DCN3_2_MIN_ACTIVE_SWITCH_MARGIN_FPO_US);
+		DC_FP_END();
+		if (!is_fpo_vactive || dc->debug.disable_fpo_vactive)
+			return NULL;
+	} else
+		fpo_candidate_stream = context->streams[0];
+
+	if (!fpo_candidate_stream)
+		return NULL;
+
+	if (fpo_candidate_stream->sink->edid_caps.panel_patch.disable_fams)
+		return NULL;
+
+	refresh_rate = get_refresh_rate(fpo_candidate_stream);
+	if (refresh_rate < minimum_refreshrate_supported)
+		return NULL;
+
+	fpo_vactive_margin_us = is_fpo_vactive ? dc->debug.fpo_vactive_margin_us : 0; // For now hardcode the FPO + Vactive stretch margin to be 2000us
+	if (!is_refresh_rate_support_mclk_switch_using_fw_based_vblank_stretch(fpo_candidate_stream, fpo_vactive_margin_us))
+		return NULL;
+
+	// check if freesync enabled
+	if (!fpo_candidate_stream->allow_freesync)
+		return NULL;
+
+	if (fpo_candidate_stream->vrr_active_variable)
+		return NULL;
+
+	return fpo_candidate_stream;
+}

drivers/gpu/drm/amd/display/dc/dcn321/dcn321_resource.c

@@ -724,6 +724,8 @@ static const struct dc_debug_options debug_defaults_drv = {
 	.disable_unbounded_requesting = false,
 	.override_dispclk_programming = true,
 	.disable_fpo_optimizations = false,
+	.fpo_vactive_margin_us = 2000, // 2000us
+	.disable_fpo_vactive = true,
 };
 
 static const struct dc_debug_options debug_defaults_diags = {

drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.c

@@ -1927,6 +1927,7 @@ void dcn32_calculate_wm_and_dlg_fpu(struct dc *dc, struct dc_state *context,
 	unsigned int min_dram_speed_mts_margin;
 	bool need_fclk_lat_as_dummy = false;
 	bool is_subvp_p_drr = false;
+	struct dc_stream_state *fpo_candidate_stream = NULL;
 
 	dc_assert_fp_enabled();
 
@@ -1968,8 +1969,11 @@ void dcn32_calculate_wm_and_dlg_fpu(struct dc *dc, struct dc_state *context,
 	if (!pstate_en || (!dc->debug.disable_fpo_optimizations &&
 			pstate_en && vlevel != 0)) {
 		/* only when the mclk switch can not be natural, is the fw based vblank stretch attempted */
-		context->bw_ctx.bw.dcn.clk.fw_based_mclk_switching =
-			dcn30_can_support_mclk_switch_using_fw_based_vblank_stretch(dc, context);
+		fpo_candidate_stream = dcn32_can_support_mclk_switch_using_fw_based_vblank_stretch(dc, context);
+		if (fpo_candidate_stream) {
+			fpo_candidate_stream->fpo_in_use = true;
+			context->bw_ctx.bw.dcn.clk.fw_based_mclk_switching = true;
+		}
 
 		if (context->bw_ctx.bw.dcn.clk.fw_based_mclk_switching) {
 			dummy_latency_index = dcn32_find_dummy_latency_index_for_fw_based_mclk_switch(dc,
@@ -2161,7 +2165,13 @@ void dcn32_calculate_wm_and_dlg_fpu(struct dc *dc, struct dc_state *context,
 		 * DCFCLK: Min, as reported by PM FW, when available
 		 * UCLK: Min, as reported by PM FW, when available
 		 */
-		dc->res_pool->funcs->update_soc_for_wm_a(dc, context);
+
+		/* For set A set the correct latency values (i.e. non-dummy values) unconditionally
+		 */
+		context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.pstate_latency_us;
+		context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.sr_enter_plus_exit_time_us;
+		context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_A].dml_input.sr_exit_time_us;
+
 		context->bw_ctx.bw.dcn.watermarks.a.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
 		context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
 		context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
@@ -2796,3 +2806,68 @@ double dcn32_determine_max_vratio_prefetch(struct dc *dc, struct dc_state *conte
 	}
 	return max_vratio_pre;
 }
+
+/**
+ * dcn32_assign_fpo_vactive_candidate - Assign the FPO stream candidate for FPO + VActive case
+ *
+ * This function chooses the FPO candidate stream for FPO + VActive cases (2 stream config).
+ * For FPO + VAtive cases, the assumption is that one display has ActiveMargin > 0, and the
+ * other display has ActiveMargin <= 0. This function will choose the pipe/stream that has
+ * ActiveMargin <= 0 to be the FPO stream candidate if found.
+ *
+ *
+ * @param [in]: dc - current dc state
+ * @param [in]: context - new dc state
+ * @param [out]: fpo_candidate_stream - pointer to FPO stream candidate if one is found
+ *
+ * Return: void
+ */
+void dcn32_assign_fpo_vactive_candidate(struct dc *dc, const struct dc_state *context, struct dc_stream_state **fpo_candidate_stream)
+{
+	unsigned int i, pipe_idx;
+	const struct vba_vars_st *vba = &context->bw_ctx.dml.vba;
+
+	for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) {
+		const struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
+
+		if (!pipe->stream)
+			continue;
+
+		if (vba->ActiveDRAMClockChangeLatencyMarginPerState[vba->VoltageLevel][vba->maxMpcComb][vba->pipe_plane[pipe_idx]] <= 0) {
+			*fpo_candidate_stream = pipe->stream;
+			break;
+		}
+		pipe_idx++;
+	}
+}
+
+/**
+ * dcn32_find_vactive_pipe - Determines if the config has a pipe that can switch in VACTIVE
+ *
+ * @param [in]: dc - current dc state
+ * @param [in]: context - new dc state
+ * @param [in]: vactive_margin_req_us - The vactive marign required for a vactive pipe to be
+ *                                      considered "found"
+ *
+ * Return: True if VACTIVE display is found, false otherwise
+ */
+bool dcn32_find_vactive_pipe(struct dc *dc, const struct dc_state *context, uint32_t vactive_margin_req_us)
+{
+	unsigned int i, pipe_idx;
+	const struct vba_vars_st *vba = &context->bw_ctx.dml.vba;
+	bool vactive_found = false;
+
+	for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) {
+		const struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
+
+		if (!pipe->stream)
+			continue;
+
+		if (vba->ActiveDRAMClockChangeLatencyMarginPerState[vba->VoltageLevel][vba->maxMpcComb][vba->pipe_plane[pipe_idx]] >= vactive_margin_req_us) {
+			vactive_found = true;
+			break;
+		}
+		pipe_idx++;
+	}
+	return vactive_found;
+}

drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.h

@@ -76,4 +76,8 @@ void dcn32_patch_dpm_table(struct clk_bw_params *bw_params);
 void dcn32_zero_pipe_dcc_fraction(display_e2e_pipe_params_st *pipes,
 				  int pipe_cnt);
 
+void dcn32_assign_fpo_vactive_candidate(struct dc *dc, const struct dc_state *context, struct dc_stream_state **fpo_candidate_stream);
+
+bool dcn32_find_vactive_pipe(struct dc *dc, const struct dc_state *context, uint32_t vactive_margin_req);
+
 #endif