drm/amd/display: Refactor fast update to use new HWSS build sequence

[Description]
- Refactor HW sequencer to use a build / execute sequence
- Also move gamma updates to become fast

v2: squash in build fix ("drm/amd/display: Fix guarding of 'if (dc->debug.visual_confirm)'")
Acked-by: Stylon Wang <stylon.wang@amd.com>
Signed-off-by: Alvin Lee <alvin.lee2@amd.com>
Reviewed-by: Jun Lei <jun.lei@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

@@ -896,6 +896,7 @@ struct dc_debug_options {
 	bool disable_dp_plus_plus_wa;
 	uint32_t fpo_vactive_min_active_margin_us;
 	uint32_t fpo_vactive_max_blank_us;
+	bool enable_legacy_fast_update;
 };
 
 struct gpu_info_soc_bounding_box_v1_0;

drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c

@@ -401,6 +401,10 @@ static const struct dc_plane_cap plane_cap = {
 	}
 };
 
+static const struct dc_debug_options debug_defaults = {
+		.enable_legacy_fast_update = true,
+};
+
 #define CTX  ctx
 #define REG(reg) mm ## reg
 
@@ -1071,6 +1075,7 @@ static bool dce100_resource_construct(
 	dc->caps.dual_link_dvi = true;
 	dc->caps.disable_dp_clk_share = true;
 	dc->caps.extended_aux_timeout_support = false;
+	dc->debug = debug_defaults;
 
 	for (i = 0; i < pool->base.pipe_count; i++) {
 		pool->base.timing_generators[i] =

drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c

@@ -424,6 +424,10 @@ static const struct dc_plane_cap plane_cap = {
 		64
 };
 
+static const struct dc_debug_options debug_defaults = {
+		.enable_legacy_fast_update = true,
+};
+
 static const struct dc_plane_cap underlay_plane_cap = {
 		.type = DC_PLANE_TYPE_DCE_UNDERLAY,
 		.per_pixel_alpha = 1,
@@ -1368,6 +1372,7 @@ static bool dce110_resource_construct(
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.is_apu = true;
 	dc->caps.extended_aux_timeout_support = false;
+	dc->debug = debug_defaults;
 
 	/*************************************************
 	 *  Create resources                             *

drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c

@@ -429,6 +429,10 @@ static const struct dc_plane_cap plane_cap = {
 	64
 };
 
+static const struct dc_debug_options debug_defaults = {
+		.enable_legacy_fast_update = true,
+};
+
 #define CTX  ctx
 #define REG(reg) mm ## reg
 
@@ -1239,6 +1243,7 @@ static bool dce112_resource_construct(
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.dual_link_dvi = true;
 	dc->caps.extended_aux_timeout_support = false;
+	dc->debug = debug_defaults;
 
 	/*************************************************
 	 *  Create resources                             *

drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c

@@ -526,6 +526,7 @@ static const struct dc_plane_cap plane_cap = {
 
 static const struct dc_debug_options debug_defaults = {
 		.disable_clock_gate = true,
+		.enable_legacy_fast_update = true,
 };
 
 static struct clock_source *dce120_clock_source_create(

drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c

@@ -418,6 +418,10 @@ static const struct dc_plane_cap plane_cap = {
 	}
 };
 
+static const struct dc_debug_options debug_defaults = {
+		.enable_legacy_fast_update = true,
+};
+
 static const struct dce_dmcu_registers dmcu_regs = {
 		DMCU_DCE80_REG_LIST()
 };
@@ -969,6 +973,7 @@ static bool dce80_construct(
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.dual_link_dvi = true;
 	dc->caps.extended_aux_timeout_support = false;
+	dc->debug = debug_defaults;
 
 	/*************************************************
 	 *  Create resources                             *
@@ -1369,6 +1374,7 @@ static bool dce83_construct(
 	dc->caps.max_cursor_size = 128;
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.is_apu = true;
+	dc->debug = debug_defaults;
 
 	/*************************************************
 	 *  Create resources                             *

drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c

@@ -553,6 +553,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 		.recovery_enabled = false, /*enable this by default after testing.*/
 		.max_downscale_src_width = 3840,
 		.underflow_assert_delay_us = 0xFFFFFFFF,
+		.enable_legacy_fast_update = true,
 };
 
 static const struct dc_debug_options debug_defaults_diags = {

drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c

@@ -722,6 +722,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 		.scl_reset_length10 = true,
 		.sanity_checks = false,
 		.underflow_assert_delay_us = 0xFFFFFFFF,
+		.enable_legacy_fast_update = true,
 };
 
 void dcn20_dpp_destroy(struct dpp **dpp)

drivers/gpu/drm/amd/display/dc/dcn201/dcn201_resource.c

@@ -613,6 +613,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 		.sanity_checks = false,
 		.underflow_assert_delay_us = 0xFFFFFFFF,
 		.enable_tri_buf = false,
+		.enable_legacy_fast_update = true,
 };
 
 static void dcn201_dpp_destroy(struct dpp **dpp)

drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c

@@ -653,6 +653,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 		.usbc_combo_phy_reset_wa = true,
 		.dmub_command_table = true,
 		.use_max_lb = true,
+		.enable_legacy_fast_update = true,
 };
 
 static const struct dc_panel_config panel_config_defaults = {

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

@@ -887,6 +887,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 		}
 	},
 	.disable_z10 = true,
+	.enable_legacy_fast_update = true,
 	.enable_z9_disable_interface = true, /* Allow support for the PMFW interface for disable Z9*/
 	.dml_hostvm_override = DML_HOSTVM_OVERRIDE_FALSE,
 };

drivers/gpu/drm/amd/display/dc/dcn315/dcn315_resource.c

@@ -887,6 +887,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 			.afmt = true,
 		}
 	},
+	.enable_legacy_fast_update = true,
 	.psr_power_use_phy_fsm = 0,
 };
 

drivers/gpu/drm/amd/display/dc/dcn316/dcn316_resource.c

@@ -884,6 +884,7 @@ static const struct dc_debug_options debug_defaults_drv = {
 			.afmt = true,
 		}
 	},
+	.enable_legacy_fast_update = true,
 };
 
 static const struct dc_panel_config panel_config_defaults = {

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hwseq.c

@@ -409,6 +409,30 @@ void dcn32_subvp_pipe_control_lock(struct dc *dc,
 	}
 }
 
+void dcn32_subvp_pipe_control_lock_fast(union block_sequence_params *params)
+{
+	struct dc *dc = params->subvp_pipe_control_lock_fast_params.dc;
+	bool lock = params->subvp_pipe_control_lock_fast_params.lock;
+	struct pipe_ctx *pipe_ctx = params->subvp_pipe_control_lock_fast_params.pipe_ctx;
+	bool subvp_immediate_flip = false;
+
+	if (pipe_ctx && pipe_ctx->stream && pipe_ctx->plane_state) {
+		if (pipe_ctx->stream->mall_stream_config.type == SUBVP_MAIN &&
+				pipe_ctx->plane_state->flip_immediate)
+			subvp_immediate_flip = true;
+	}
+
+	// Don't need to lock for DRR VSYNC flips -- FW will wait for DRR pending update cleared.
+	if (subvp_immediate_flip) {
+		union dmub_inbox0_cmd_lock_hw hw_lock_cmd = { 0 };
+
+		hw_lock_cmd.bits.command_code = DMUB_INBOX0_CMD__HW_LOCK;
+		hw_lock_cmd.bits.hw_lock_client = HW_LOCK_CLIENT_DRIVER;
+		hw_lock_cmd.bits.lock = lock;
+		hw_lock_cmd.bits.should_release = !lock;
+		dmub_hw_lock_mgr_inbox0_cmd(dc->ctx->dmub_srv, hw_lock_cmd);
+	}
+}
 
 bool dcn32_set_mpc_shaper_3dlut(
 	struct pipe_ctx *pipe_ctx, const struct dc_stream_state *stream)

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hwseq.h

@@ -84,6 +84,8 @@ void dcn32_subvp_pipe_control_lock(struct dc *dc,
 		struct pipe_ctx *top_pipe_to_program,
 		bool subvp_prev_use);
 
+void dcn32_subvp_pipe_control_lock_fast(union block_sequence_params *params);
+
 void dcn32_unblank_stream(struct pipe_ctx *pipe_ctx,
 		struct dc_link_settings *link_settings);
 

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_init.c

@@ -110,6 +110,7 @@ static const struct hw_sequencer_funcs dcn32_funcs = {
 	.enable_phantom_streams = dcn32_enable_phantom_streams,
 	.subvp_pipe_control_lock = dcn32_subvp_pipe_control_lock,
 	.update_visual_confirm_color = dcn10_update_visual_confirm_color,
+	.subvp_pipe_control_lock_fast = dcn32_subvp_pipe_control_lock_fast,
 	.update_phantom_vp_position = dcn32_update_phantom_vp_position,
 	.update_dsc_pg = dcn32_update_dsc_pg,
 	.apply_update_flags_for_phantom = dcn32_apply_update_flags_for_phantom,

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

@@ -37,6 +37,7 @@
 #include "dwb.h"
 #include "mcif_wb.h"
 #include "panel_cntl.h"
+#include "dmub/inc/dmub_cmd.h"
 
 #define MAX_CLOCK_SOURCES 7
 #define MAX_SVP_PHANTOM_STREAMS 2
@@ -499,6 +500,11 @@ struct bw_context {
 	struct display_mode_lib dml;
 };
 
+struct dc_dmub_cmd {
+	union dmub_rb_cmd dmub_cmd;
+	enum dm_dmub_wait_type wait_type;
+};
+
 /**
  * struct dc_state - The full description of a state requested by users
  */
@@ -547,6 +553,11 @@ struct dc_state {
 	 */
 	struct bw_context bw_ctx;
 
+	struct block_sequence block_sequence[50];
+	unsigned int block_sequence_steps;
+	struct dc_dmub_cmd dc_dmub_cmd[10];
+	unsigned int dmub_cmd_count;
+
 	/**
 	 * @refcount: refcount reference
 	 *

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

@@ -44,6 +44,112 @@ struct dc_virtual_addr_space_config;
 struct dpp;
 struct dce_hwseq;
 struct link_resource;
+struct dc_dmub_cmd;
+
+struct subvp_pipe_control_lock_fast_params {
+	struct dc *dc;
+	bool lock;
+	struct pipe_ctx *pipe_ctx;
+};
+
+struct pipe_control_lock_params {
+	struct dc *dc;
+	struct pipe_ctx *pipe_ctx;
+	bool lock;
+};
+
+struct set_flip_control_gsl_params {
+	struct pipe_ctx *pipe_ctx;
+	bool flip_immediate;
+};
+
+struct program_triplebuffer_params {
+	const struct dc *dc;
+	struct pipe_ctx *pipe_ctx;
+	bool enableTripleBuffer;
+};
+
+struct update_plane_addr_params {
+	struct dc *dc;
+	struct pipe_ctx *pipe_ctx;
+};
+
+struct set_input_transfer_func_params {
+	struct dc *dc;
+	struct pipe_ctx *pipe_ctx;
+	struct dc_plane_state *plane_state;
+};
+
+struct program_gamut_remap_params {
+	struct pipe_ctx *pipe_ctx;
+};
+
+struct program_manual_trigger_params {
+	struct pipe_ctx *pipe_ctx;
+};
+
+struct send_dmcub_cmd_params {
+	struct dc_context *ctx;
+	union dmub_rb_cmd *cmd;
+	enum dm_dmub_wait_type wait_type;
+};
+
+struct setup_dpp_params {
+	struct pipe_ctx *pipe_ctx;
+};
+
+struct program_bias_and_scale_params {
+	struct pipe_ctx *pipe_ctx;
+};
+
+struct set_output_transfer_func_params {
+	struct dc *dc;
+	struct pipe_ctx *pipe_ctx;
+	const struct dc_stream_state *stream;
+};
+
+struct update_visual_confirm_params {
+	struct dc *dc;
+	struct pipe_ctx *pipe_ctx;
+	int mpcc_id;
+};
+
+union block_sequence_params {
+	struct update_plane_addr_params update_plane_addr_params;
+	struct subvp_pipe_control_lock_fast_params subvp_pipe_control_lock_fast_params;
+	struct pipe_control_lock_params pipe_control_lock_params;
+	struct set_flip_control_gsl_params set_flip_control_gsl_params;
+	struct program_triplebuffer_params program_triplebuffer_params;
+	struct set_input_transfer_func_params set_input_transfer_func_params;
+	struct program_gamut_remap_params program_gamut_remap_params;
+	struct program_manual_trigger_params program_manual_trigger_params;
+	struct send_dmcub_cmd_params send_dmcub_cmd_params;
+	struct setup_dpp_params setup_dpp_params;
+	struct program_bias_and_scale_params program_bias_and_scale_params;
+	struct set_output_transfer_func_params set_output_transfer_func_params;
+	struct update_visual_confirm_params update_visual_confirm_params;
+};
+
+enum block_sequence_func {
+	DMUB_SUBVP_PIPE_CONTROL_LOCK_FAST = 0,
+	OPTC_PIPE_CONTROL_LOCK,
+	HUBP_SET_FLIP_CONTROL_GSL,
+	HUBP_PROGRAM_TRIPLEBUFFER,
+	HUBP_UPDATE_PLANE_ADDR,
+	DPP_SET_INPUT_TRANSFER_FUNC,
+	DPP_PROGRAM_GAMUT_REMAP,
+	OPTC_PROGRAM_MANUAL_TRIGGER,
+	DMUB_SEND_DMCUB_CMD,
+	DPP_SETUP_DPP,
+	DPP_PROGRAM_BIAS_AND_SCALE,
+	DPP_SET_OUTPUT_TRANSFER_FUNC,
+	MPC_UPDATE_VISUAL_CONFIRM,
+};
+
+struct block_sequence {
+	union block_sequence_params params;
+	enum block_sequence_func func;
+};
 
 struct hw_sequencer_funcs {
 	void (*hardware_release)(struct dc *dc);
@@ -252,6 +358,7 @@ struct hw_sequencer_funcs {
 			const struct tg_color *solid_color,
 			int width, int height, int offset);
 
+	void (*subvp_pipe_control_lock_fast)(union block_sequence_params *params);
 	void (*z10_restore)(const struct dc *dc);
 	void (*z10_save_init)(struct dc *dc);
 
@@ -313,4 +420,23 @@ void get_mclk_switch_visual_confirm_color(
 		struct pipe_ctx *pipe_ctx,
 		struct tg_color *color);
 
+void hwss_execute_sequence(struct dc *dc,
+		struct block_sequence block_sequence[],
+		int num_steps);
+
+void hwss_build_fast_sequence(struct dc *dc,
+		struct dc_dmub_cmd *dc_dmub_cmd,
+		unsigned int dmub_cmd_count,
+		struct block_sequence block_sequence[],
+		int *num_steps,
+		struct pipe_ctx *pipe_ctx);
+
+void hwss_send_dmcub_cmd(union block_sequence_params *params);
+
+void hwss_program_manual_trigger(union block_sequence_params *params);
+
+void hwss_setup_dpp(union block_sequence_params *params);
+
+void hwss_program_bias_and_scale(union block_sequence_params *params);
+
 #endif /* __DC_HW_SEQUENCER_H__ */