drm/amd/display: Refactor SubVP cursor limiting logic

[WHY]
There are several gaps that can result in SubVP being enabled with
incompatible HW cursor sizes, and unjust restrictions to cursor size due
to wrong predictions on future usage of SubVP.

[HOW]
- remove "prediction" logic in favor of tagging based on previous SubVP
usage
- block SubVP if current HW cursor settings are incompatible
- provide interface for DM to determine if HW cursor should be disabled
due to an attempt to enable SubVP

Reviewed-by: Alvin Lee <[e-mail]>
Signed-off-by: Dillon Varone <[e-mail]>
Signed-off-by: Ray Wu <[e-mail]>
Tested-by: Daniel Wheeler <[e-mail]>
Signed-off-by: Alex Deucher <[e-mail]>

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

···
 #include "dc_state.h"
 #include "dc_state_priv.h"
 #include "dc_plane_priv.h"
+#include "dc_stream_priv.h"
 
 #include "gpio_service_interface.h"
 #include "clk_mgr.h"
···
 	int i;
 	enum surface_update_type overall_type = UPDATE_TYPE_FAST;
 
-	if (dc->idle_optimizations_allowed)
+	if (dc->idle_optimizations_allowed || dc_can_clear_cursor_limit(dc))
 		overall_type = UPDATE_TYPE_FULL;
 
 	if (stream_status == NULL || stream_status->plane_count != surface_count)
···
 		if (dsc_validate_context) {
 			stream->timing.dsc_cfg = *update->dsc_config;
 			stream->timing.flags.DSC = enable_dsc;
-			if (!dc->res_pool->funcs->validate_bandwidth(dc, dsc_validate_context, true)) {
+			if (dc->res_pool->funcs->validate_bandwidth(dc, dsc_validate_context, true) != DC_OK) {
 				stream->timing.dsc_cfg = old_dsc_cfg;
 				stream->timing.flags.DSC = old_dsc_enabled;
 				update->dsc_config = NULL;
···
 	}
 
 	if (update_type == UPDATE_TYPE_FULL) {
-		if (!dc->res_pool->funcs->validate_bandwidth(dc, context, false)) {
+		if (dc->res_pool->funcs->validate_bandwidth(dc, context, false) != DC_OK) {
 			BREAK_TO_DEBUGGER();
 			goto fail;
 		}
···
 
 	backup_and_set_minimal_pipe_split_policy(dc, base_context, policy);
 	/* commit minimal state */
-	if (dc->res_pool->funcs->validate_bandwidth(dc, minimal_transition_context, false)) {
+	if (dc->res_pool->funcs->validate_bandwidth(dc, minimal_transition_context, false) == DC_OK) {
 		/* prevent underflow and corruption when reconfiguring pipes */
 		force_vsync_flip_in_minimal_transition_context(minimal_transition_context);
 	} else {
···
 	if (dc->idle_optimizations_allowed)
 		return true;
 
+	if (dc_can_clear_cursor_limit(dc))
+		return true;
+
 	return false;
 }
 
···
 	copy_stream_update_to_stream(dc, context, stream, stream_update);
 
 	if (update_type >= UPDATE_TYPE_FULL) {
-		if (!dc->res_pool->funcs->validate_bandwidth(dc, context, false)) {
+		if (dc->res_pool->funcs->validate_bandwidth(dc, context, false) != DC_OK) {
 			DC_ERROR("Mode validation failed for stream update!\n");
 			dc_state_release(context);
 			return false;
···
 void dc_query_current_properties(struct dc *dc, struct dc_current_properties *properties)
 {
 	unsigned int i;
-	bool subvp_sw_cursor_req = false;
+	unsigned int max_cursor_size = dc->caps.max_cursor_size;
+	unsigned int stream_cursor_size;
 
-	for (i = 0; i < dc->current_state->stream_count; i++) {
-		if (check_subvp_sw_cursor_fallback_req(dc, dc->current_state->streams[i]) && !dc->current_state->streams[i]->hw_cursor_req) {
-			subvp_sw_cursor_req = true;
-			break;
+	if (dc->debug.allow_sw_cursor_fallback && dc->res_pool->funcs->get_max_hw_cursor_size) {
+		for (i = 0; i < dc->current_state->stream_count; i++) {
+			stream_cursor_size = dc->res_pool->funcs->get_max_hw_cursor_size(dc,
+					dc->current_state,
+					dc->current_state->streams[i]);
+
+			if (stream_cursor_size < max_cursor_size) {
+				max_cursor_size = stream_cursor_size;
+			}
 		}
 	}
-	properties->cursor_size_limit = subvp_sw_cursor_req ? 64 : dc->caps.max_cursor_size;
+
+	properties->cursor_size_limit = max_cursor_size;
 }
 
 /**
···
 		return dc->res_pool->funcs->get_det_buffer_size(context);
 	else
 		return 0;
+}
+
+bool dc_is_cursor_limit_pending(struct dc *dc)
+{
+	uint32_t i;
+
+	for (i = 0; i < dc->current_state->stream_count; i++) {
+		if (dc_stream_is_cursor_limit_pending(dc, dc->current_state->streams[i]))
+			return true;
+	}
+
+	return false;
+}
+
+bool dc_can_clear_cursor_limit(struct dc *dc)
+{
+	uint32_t i;
+
+	for (i = 0; i < dc->current_state->stream_count; i++) {
+		if (dc_state_can_clear_stream_cursor_subvp_limit(dc->current_state->streams[i], dc->current_state))
+			return true;
+	}
+
+	return false;
 }

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

···
 		return "Fail dp payload allocation";
 	case DC_FAIL_DP_LINK_BANDWIDTH:
 		return "Insufficient DP link bandwidth";
+	case DC_FAIL_HW_CURSOR_SUPPORT:
+		return "HW Cursor not supported";
 	case DC_ERROR_UNEXPECTED:
 		return "Unexpected error";
 	}

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

···
 	data->viewport_c.y += src.y / vpc_div;
 }
 
-static bool is_subvp_high_refresh_candidate(struct dc_stream_state *stream)
-{
-	uint32_t refresh_rate;
-	struct dc *dc = stream->ctx->dc;
-
-	refresh_rate = (stream->timing.pix_clk_100hz * (uint64_t)100 +
-		stream->timing.v_total * stream->timing.h_total - (uint64_t)1);
-	refresh_rate = div_u64(refresh_rate, stream->timing.v_total);
-	refresh_rate = div_u64(refresh_rate, stream->timing.h_total);
-
-	/* If there's any stream that fits the SubVP high refresh criteria,
-	 * we must return true. This is because cursor updates are asynchronous
-	 * with full updates, so we could transition into a SubVP config and
-	 * remain in HW cursor mode if there's no cursor update which will
-	 * then cause corruption.
-	 */
-	if ((refresh_rate >= 120 && refresh_rate <= 175 &&
-			stream->timing.v_addressable >= 1080 &&
-			stream->timing.v_addressable <= 2160) &&
-			(dc->current_state->stream_count > 1 ||
-			(dc->current_state->stream_count == 1 && !stream->allow_freesync)))
-		return true;
-
-	return false;
-}
-
 static enum controller_dp_test_pattern convert_dp_to_controller_test_pattern(
 				enum dp_test_pattern test_pattern)
 {
···
 		}
 	}
 
+	/* clear subvp cursor limitations */
+	for (i = 0; i < context->stream_count; i++) {
+		dc_state_set_stream_subvp_cursor_limit(context->streams[i], context, false);
+	}
+
 	res = dc_validate_global_state(dc, context, fast_validate);
 
 	/* calculate pixel rate divider after deciding pxiel clock & odm combine  */
···
 	result = resource_build_scaling_params_for_context(dc, new_ctx);
 
 	if (result == DC_OK)
-		if (!dc->res_pool->funcs->validate_bandwidth(dc, new_ctx, fast_validate))
-			result = DC_FAIL_BANDWIDTH_VALIDATE;
+		result = dc->res_pool->funcs->validate_bandwidth(dc, new_ctx, fast_validate);
 
 	return result;
 }
···
 			return DC_NO_LINK_ENC_RESOURCE;
 
 	return DC_OK;
-}
-
-bool check_subvp_sw_cursor_fallback_req(const struct dc *dc, struct dc_stream_state *stream)
-{
-	if (!dc->debug.disable_subvp_high_refresh && is_subvp_high_refresh_candidate(stream))
-		return true;
-	if (dc->current_state->stream_count == 1 && stream->timing.v_addressable >= 2880 &&
-			((stream->timing.pix_clk_100hz * 100) / stream->timing.v_total / stream->timing.h_total) < 120)
-		return true;
-	else if (dc->current_state->stream_count > 1 && stream->timing.v_addressable >= 1080 &&
-			((stream->timing.pix_clk_100hz * 100) / stream->timing.v_total / stream->timing.h_total) < 120)
-		return true;
-
-	return false;
 }
 
 struct dscl_prog_data *resource_get_dscl_prog_data(struct pipe_ctx *pipe_ctx)

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

···
  * Authors: AMD
  *
  */
+#include "dc_types.h"
 #include "core_types.h"
 #include "core_status.h"
 #include "dc_state.h"
···
 	if (phantom_stream_status) {
 		phantom_stream_status->mall_stream_config.type = SUBVP_PHANTOM;
 		phantom_stream_status->mall_stream_config.paired_stream = main_stream;
+		phantom_stream_status->mall_stream_config.subvp_limit_cursor_size = false;
+		phantom_stream_status->mall_stream_config.cursor_size_limit_subvp = false;
 	}
+
+	dc_state_set_stream_subvp_cursor_limit(main_stream, state, true);
 
 	return res;
 }
···
 		const struct dc *dc,
 		struct dc_state *state)
 {
+	unsigned int phantom_count;
+	struct dc_stream_state *phantom_streams[MAX_PHANTOM_PIPES];
+	struct dc_plane_state *phantom_planes[MAX_PHANTOM_PIPES];
 	int i;
 
-	for (i = 0; i < state->phantom_stream_count; i++)
-		dc_state_release_phantom_stream(dc, state, state->phantom_streams[i]);
+	phantom_count = state->phantom_stream_count;
+	memcpy(phantom_streams, state->phantom_streams, sizeof(struct dc_stream_state *) * MAX_PHANTOM_PIPES);
+	for (i = 0; i < phantom_count; i++)
+		dc_state_release_phantom_stream(dc, state, phantom_streams[i]);
 
-	for (i = 0; i < state->phantom_plane_count; i++)
-		dc_state_release_phantom_plane(dc, state, state->phantom_planes[i]);
+	phantom_count = state->phantom_plane_count;
+	memcpy(phantom_planes, state->phantom_planes, sizeof(struct dc_plane_state *) * MAX_PHANTOM_PIPES);
+	for (i = 0; i < phantom_count; i++)
+		dc_state_release_phantom_plane(dc, state, phantom_planes[i]);
 }
 
 struct dc_stream_state *dc_state_get_stream_from_id(const struct dc_state *state, unsigned int id)
···
 		is_fams2_in_use |= dc->current_state->bw_ctx.bw.dcn.fams2_global_config.features.bits.enable;
 
 	return is_fams2_in_use;
+}
+
+void dc_state_set_stream_subvp_cursor_limit(const struct dc_stream_state *stream,
+		struct dc_state *state,
+		bool limit)
+{
+	struct dc_stream_status *stream_status;
+
+	stream_status = dc_state_get_stream_status(state, stream);
+
+	if (stream_status) {
+		stream_status->mall_stream_config.subvp_limit_cursor_size = limit;
+	}
+}
+
+bool dc_state_get_stream_subvp_cursor_limit(const struct dc_stream_state *stream,
+		struct dc_state *state)
+{
+	bool limit = false;
+
+	struct dc_stream_status *stream_status;
+
+	stream_status = dc_state_get_stream_status(state, stream);
+
+	if (stream_status) {
+		limit = stream_status->mall_stream_config.subvp_limit_cursor_size;
+	}
+
+	return limit;
+}
+
+void dc_state_set_stream_cursor_subvp_limit(const struct dc_stream_state *stream,
+		struct dc_state *state,
+		bool limit)
+{
+	struct dc_stream_status *stream_status;
+
+	stream_status = dc_state_get_stream_status(state, stream);
+
+	if (stream_status) {
+		stream_status->mall_stream_config.cursor_size_limit_subvp = limit;
+	}
+}
+
+bool dc_state_get_stream_cursor_subvp_limit(const struct dc_stream_state *stream,
+		struct dc_state *state)
+{
+	bool limit = false;
+
+	struct dc_stream_status *stream_status;
+
+	stream_status = dc_state_get_stream_status(state, stream);
+
+	if (stream_status) {
+		limit = stream_status->mall_stream_config.cursor_size_limit_subvp;
+	}
+
+	return limit;
+}
+
+bool dc_state_can_clear_stream_cursor_subvp_limit(const struct dc_stream_state *stream,
+		struct dc_state *state)
+{
+	bool can_clear_limit = false;
+
+	struct dc_stream_status *stream_status;
+
+	stream_status = dc_state_get_stream_status(state, stream);
+
+	if (stream_status) {
+		can_clear_limit = dc_state_get_stream_cursor_subvp_limit(stream, state) &&
+				(stream_status->mall_stream_config.type == SUBVP_PHANTOM ||
+				stream->hw_cursor_req ||
+				!stream_status->mall_stream_config.subvp_limit_cursor_size ||
+				!stream->cursor_position.enable ||
+				dc_stream_check_cursor_attributes(stream, state, &stream->cursor_attributes));
+	}
+
+	return can_clear_limit;
+}
+
+bool dc_state_is_subvp_in_use(struct dc_state *state)
+{
+	uint32_t i;
+
+	for (i = 0; i < state->stream_count; i++) {
+		if (dc_state_get_stream_subvp_type(state, state->streams[i]) != SUBVP_NONE)
+			return true;
+	}
+
+	return false;
 }

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

···
 }
 
 /*
- * dc_stream_set_cursor_attributes() - Update cursor attributes and set cursor surface address
+ * dc_stream_check_cursor_attributes() - Check validitity of cursor attributes and surface address
  */
-bool dc_stream_set_cursor_attributes(
-	struct dc_stream_state *stream,
+bool dc_stream_check_cursor_attributes(
+	const struct dc_stream_state *stream,
+	struct dc_state *state,
 	const struct dc_cursor_attributes *attributes)
 {
-	struct dc  *dc;
+	const struct dc *dc;
+
+	unsigned int max_cursor_size;
 
 	if (NULL == stream) {
 		dm_error("DC: dc_stream is NULL!\n");
···
 
 	dc = stream->ctx->dc;
 
-	/* SubVP is not compatible with HW cursor larger than 64 x 64 x 4.
-	 * Therefore, if cursor is greater than 64 x 64 x 4, fallback to SW cursor in the following case:
-	 * 1. If the config is a candidate for SubVP high refresh (both single an dual display configs)
-	 * 2. If not subvp high refresh, for single display cases, if resolution is >= 5K and refresh rate < 120hz
-	 * 3. If not subvp high refresh, for multi display cases, if resolution is >= 4K and refresh rate < 120hz
+	/* SubVP is not compatible with HW cursor larger than what can fit in cursor SRAM.
+	 * Therefore, if cursor is greater than this, fallback to SW cursor.
 	 */
-	if (dc->debug.allow_sw_cursor_fallback &&
-		attributes->height * attributes->width * 4 > 16384 &&
-		!stream->hw_cursor_req) {
-		if (check_subvp_sw_cursor_fallback_req(dc, stream))
+	if (dc->debug.allow_sw_cursor_fallback && dc->res_pool->funcs->get_max_hw_cursor_size) {
+		max_cursor_size = dc->res_pool->funcs->get_max_hw_cursor_size(dc, state, stream);
+		max_cursor_size = max_cursor_size * max_cursor_size * 4;
+
+		if (attributes->height * attributes->width * 4 > max_cursor_size) {
 			return false;
+		}
 	}
 
-	stream->cursor_attributes = *attributes;
-
 	return true;
+}
+
+/*
+ * dc_stream_set_cursor_attributes() - Update cursor attributes and set cursor surface address
+ */
+bool dc_stream_set_cursor_attributes(
+	struct dc_stream_state *stream,
+	const struct dc_cursor_attributes *attributes)
+{
+	bool result = false;
+
+	if (dc_stream_check_cursor_attributes(stream, stream->ctx->dc->current_state, attributes)) {
+		stream->cursor_attributes = *attributes;
+		result = true;
+	}
+
+	return result;
 }
 
 bool dc_stream_program_cursor_attributes(
···
 		return 0;
 
 	return dc_stream_get_max_flickerless_instant_vtotal_delta(stream, is_gaming, false);
+}
+
+bool dc_stream_is_cursor_limit_pending(struct dc *dc, struct dc_stream_state *stream)
+{
+	bool is_limit_pending = false;
+
+	if (dc->current_state)
+		is_limit_pending = dc_state_get_stream_cursor_subvp_limit(stream, dc->current_state);
+
+	return is_limit_pending;
+}
+
+bool dc_stream_can_clear_cursor_limit(struct dc *dc, struct dc_stream_state *stream)
+{
+	bool can_clear_limit = false;
+
+	if (dc->current_state)
+		can_clear_limit = dc_state_get_stream_cursor_subvp_limit(stream, dc->current_state) &&
+				(stream->hw_cursor_req ||
+				!stream->cursor_position.enable ||
+				dc_stream_check_cursor_attributes(stream, dc->current_state, &stream->cursor_attributes));
+
+	return can_clear_limit;
 }

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

···
 	uint32_t i2c_speed_in_khz_hdcp;
 	uint32_t dmdata_alloc_size;
 	unsigned int max_cursor_size;
+	unsigned int max_buffered_cursor_size;
 	unsigned int max_video_width;
 	/*
 	 * max video plane width that can be safely assumed to be always
···
 
 bool dc_is_timing_changed(struct dc_stream_state *cur_stream,
 		       struct dc_stream_state *new_stream);
+
+bool dc_is_cursor_limit_pending(struct dc *dc);
+bool dc_can_clear_cursor_limit(struct dc *dc);
 
 #endif /* DC_INTERFACE_H_ */

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

···
 		const struct dc *dc,
 		const struct dc_state *state);
 
+
+void dc_state_set_stream_subvp_cursor_limit(const struct dc_stream_state *stream,
+		struct dc_state *state,
+		bool limit);
+
+bool dc_state_get_stream_subvp_cursor_limit(const struct dc_stream_state *stream,
+		struct dc_state *state);
+
+void dc_state_set_stream_cursor_subvp_limit(const struct dc_stream_state *stream,
+		struct dc_state *state,
+		bool limit);
+
+bool dc_state_get_stream_cursor_subvp_limit(const struct dc_stream_state *stream,
+		struct dc_state *state);
+
+bool dc_state_can_clear_stream_cursor_subvp_limit(const struct dc_stream_state *stream,
+		struct dc_state *state);
+
+bool dc_state_is_subvp_in_use(struct dc_state *state);
+
 #endif /* _DC_STATE_PRIV_H_ */

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

···
 	 */
 	enum mall_stream_type type;
 	struct dc_stream_state *paired_stream;	// master / slave stream
+	bool subvp_limit_cursor_size; /* stream has/is using subvp limiting hw cursor support */
+	bool cursor_size_limit_subvp; /* stream is using hw cursor config preventing subvp */
 };
 
 struct dc_stream_status {
···
 	struct dc *dc,
 	struct dc_stream_state *stream);
 
+bool dc_stream_check_cursor_attributes(
+	const struct dc_stream_state *stream,
+	struct dc_state *state,
+	const struct dc_cursor_attributes *attributes);
+
 bool dc_stream_set_cursor_attributes(
 	struct dc_stream_state *stream,
 	const struct dc_cursor_attributes *attributes);
···
 			       struct dc_stream_state *stream,
 			       struct dc_surface_update *srf_updates,
 			       struct dc_state *context);
+
+bool dc_stream_is_cursor_limit_pending(struct dc *dc, struct dc_stream_state *stream);
+bool dc_stream_can_clear_cursor_limit(struct dc *dc, struct dc_stream_state *stream);
 
 #endif /* DC_STREAM_H_ */

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

···
 		 */
 		if (pipe->plane_state && !pipe->top_pipe && !pipe->prev_odm_pipe && !dcn32_is_center_timing(pipe) &&
 				!pipe->stream->hw_cursor_req &&
+				!dc_state_get_stream_cursor_subvp_limit(pipe->stream, context) &&
 				!(pipe->stream->timing.pix_clk_100hz / 10000 > DCN3_2_MAX_SUBVP_PIXEL_RATE_MHZ) &&
 				(!dcn32_is_psr_capable(pipe) || (context->stream_count == 1 && dc->caps.dmub_caps.subvp_psr)) &&
 				dc_state_get_pipe_subvp_type(context, pipe) == SUBVP_NONE &&

drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c

···
 
 static void populate_dml21_stream_overrides_from_stream_state(
 		struct dml2_stream_parameters *stream_desc,
-		struct dc_stream_state *stream)
+		struct dc_stream_state *stream,
+		struct dc_stream_status *stream_status)
 {
 	switch (stream->debug.force_odm_combine_segments) {
 	case 0:
···
 	if (!stream->ctx->dc->debug.enable_single_display_2to1_odm_policy ||
 			stream->debug.force_odm_combine_segments > 0)
 		stream_desc->overrides.disable_dynamic_odm = true;
-	stream_desc->overrides.disable_subvp = stream->ctx->dc->debug.force_disable_subvp || stream->hw_cursor_req;
+	stream_desc->overrides.disable_subvp = stream->ctx->dc->debug.force_disable_subvp ||
+			stream->hw_cursor_req ||
+			stream_status->mall_stream_config.cursor_size_limit_subvp;
 }
 
 static enum dml2_swizzle_mode gfx_addr3_to_dml2_swizzle_mode(enum swizzle_mode_addr3_values addr3_mode)
···
 		populate_dml21_timing_config_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location].timing, context->streams[stream_index], dml_ctx);
 		adjust_dml21_hblank_timing_config_from_pipe_ctx(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location].timing, &context->res_ctx.pipe_ctx[stream_index]);
 		populate_dml21_output_config_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location].output, context->streams[stream_index], &context->res_ctx.pipe_ctx[stream_index]);
-		populate_dml21_stream_overrides_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location], context->streams[stream_index]);
+		populate_dml21_stream_overrides_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location], context->streams[stream_index], &context->stream_status[stream_index]);
 
 		dml_dispcfg->stream_descriptors[disp_cfg_stream_location].overrides.hw.twait_budgeting.fclk_pstate = dml2_twait_budgeting_setting_if_needed;
 		dml_dispcfg->stream_descriptors[disp_cfg_stream_location].overrides.hw.twait_budgeting.uclk_pstate = dml2_twait_budgeting_setting_if_needed;

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

···
 	struct dce_hwseq *hws = dc->hwseq;
 
 	/* recalculate DML parameters */
-	if (!dc->res_pool->funcs->validate_bandwidth(dc, context, false))
+	if (dc->res_pool->funcs->validate_bandwidth(dc, context, false) != DC_OK)
 		return false;
 
 	/* apply updated bandwidth parameters */

drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c

···
 	struct dce_hwseq *hws = dc->hwseq;
 
 	/* recalculate DML parameters */
-	if (!dc->res_pool->funcs->validate_bandwidth(dc, context, false))
+	if (dc->res_pool->funcs->validate_bandwidth(dc, context, false) != DC_OK)
 		return false;
 
 	/* apply updated bandwidth parameters */

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

···
 #ifndef _CORE_STATUS_H_
 #define _CORE_STATUS_H_
 
+#include "dc_hw_types.h"
+
 enum dc_status {
 	DC_OK = 1,
 
···
 	DC_NO_LINK_ENC_RESOURCE = 26,
 	DC_FAIL_DP_PAYLOAD_ALLOCATION = 27,
 	DC_FAIL_DP_LINK_BANDWIDTH = 28,
+	DC_FAIL_HW_CURSOR_SUPPORT = 29,
 	DC_ERROR_UNEXPECTED = -1
 };
 

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

···
 	/* Create a minimal link encoder object with no dc_link object
 	 * associated with it. */
 	struct link_encoder *(*link_enc_create_minimal)(struct dc_context *ctx, enum engine_id eng_id);
-	bool (*validate_bandwidth)(
+	enum dc_status (*validate_bandwidth)(
 					struct dc *dc,
 					struct dc_state *context,
 					bool fast_validate);
···
 	int (*get_power_profile)(const struct dc_state *context);
 	unsigned int (*get_det_buffer_size)(const struct dc_state *context);
 	unsigned int (*get_vstartup_for_pipe)(struct pipe_ctx *pipe_ctx);
+	unsigned int (*get_max_hw_cursor_size)(const struct dc *dc,
+			struct dc_state *state,
+			const struct dc_stream_state *stream);
 };
 
 struct audio_support{

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

···
 		struct dc_state *context,
 		struct pipe_ctx *pipe_ctx);
 
-bool check_subvp_sw_cursor_fallback_req(const struct dc *dc, struct dc_stream_state *stream);
-
 /* Get hw programming parameters container from pipe context
  * @pipe_ctx: pipe context
  * @dscl_prog_data: struct to hold programmable hw reg values

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

···
 	return DC_OK;
 }
 
-static bool dce100_validate_bandwidth(
+static enum dc_status dce100_validate_bandwidth(
 	struct dc  *dc,
 	struct dc_state *context,
 	bool fast_validate)
···
 		context->bw_ctx.bw.dce.yclk_khz = 0;
 	}
 
-	return true;
+	return DC_OK;
 }
 
 static bool dce100_validate_surface_sets(

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

···
 	return DC_OK;
 }
 
-static bool dce110_validate_bandwidth(
+static enum dc_status dce110_validate_bandwidth(
 	struct dc *dc,
 	struct dc_state *context,
 	bool fast_validate)
···
 			context->bw_ctx.bw.dce.yclk_khz,
 			context->bw_ctx.bw.dce.blackout_recovery_time_us);
 	}
-	return result;
+	return result ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 static enum dc_status dce110_validate_plane(const struct dc_plane_state *plane_state,

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

···
 	return DC_OK;
 }
 
-bool dce112_validate_bandwidth(
+enum dc_status dce112_validate_bandwidth(
 	struct dc *dc,
 	struct dc_state *context,
 	bool fast_validate)
···
 			context->bw_ctx.bw.dce.yclk_khz,
 			context->bw_ctx.bw.dce.blackout_recovery_time_us);
 	}
-	return result;
+	return result ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 enum dc_status resource_map_phy_clock_resources(

drivers/gpu/drm/amd/display/dc/resource/dce112/dce112_resource.h

···
 		struct dc_state *context,
 		struct dc_state *old_context);
 
-bool dce112_validate_bandwidth(
+enum dc_status dce112_validate_bandwidth(
 	struct dc *dc,
 	struct dc_state *context,
 	bool fast_validate);

drivers/gpu/drm/amd/display/dc/resource/dce60/dce60_resource.c

···
 	}
 }
 
-static bool dce60_validate_bandwidth(
+static enum dc_status dce60_validate_bandwidth(
 	struct dc *dc,
 	struct dc_state *context,
 	bool fast_validate)
···
 		context->bw_ctx.bw.dce.yclk_khz = 0;
 	}
 
-	return true;
+	return DC_OK;
 }
 
 static bool dce60_validate_surface_sets(

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

···
 	}
 }
 
-static bool dce80_validate_bandwidth(
+static enum dc_status dce80_validate_bandwidth(
 	struct dc *dc,
 	struct dc_state *context,
 	bool fast_validate)
···
 		context->bw_ctx.bw.dce.yclk_khz = 0;
 	}
 
-	return true;
+	return DC_OK;
 }
 
 static bool dce80_validate_surface_sets(

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

···
  *
  */
 
+#include "core_status.h"
 #include "dm_services.h"
 #include "dc.h"
 
···
 	*pool = NULL;
 }
 
-static bool dcn10_validate_bandwidth(
+static enum dc_status dcn10_validate_bandwidth(
 		struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
···
 	voltage_supported = dcn_validate_bandwidth(dc, context, fast_validate);
 	DC_FP_END();
 
-	return voltage_supported;
+	return voltage_supported ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 static enum dc_status dcn10_validate_plane(const struct dc_plane_state *plane_state, struct dc_caps *caps)

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

···
 	return out;
 }
 
-bool dcn20_validate_bandwidth(struct dc *dc, struct dc_state *context,
+enum dc_status dcn20_validate_bandwidth(struct dc *dc, struct dc_state *context,
 		bool fast_validate)
 {
 	bool voltage_supported;
···
 
 	pipes = kcalloc(dc->res_pool->pipe_count, sizeof(display_e2e_pipe_params_st), GFP_KERNEL);
 	if (!pipes)
-		return false;
+		return DC_FAIL_BANDWIDTH_VALIDATE;
 
 	DC_FP_START();
 	voltage_supported = dcn20_validate_bandwidth_fp(dc, context, fast_validate, pipes);
 	DC_FP_END();
 
 	kfree(pipes);
-	return voltage_supported;
+	return voltage_supported ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 struct pipe_ctx *dcn20_acquire_free_pipe_for_layer(

drivers/gpu/drm/amd/display/dc/resource/dcn20/dcn20_resource.h

···
 		struct dc_state *context,
 		display_e2e_pipe_params_st *pipes,
 		int pipe_cnt);
-bool dcn20_validate_bandwidth(struct dc *dc, struct dc_state *context, bool fast_validate);
+enum dc_status dcn20_validate_bandwidth(struct dc *dc, struct dc_state *context, bool fast_validate);
 void dcn20_merge_pipes_for_validate(
 		struct dc *dc,
 		struct dc_state *context);

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

···
  * with DC_FP_START()/DC_FP_END(). Use the same approach as for
  * dcn20_validate_bandwidth in dcn20_resource.c.
  */
-static bool dcn21_validate_bandwidth(struct dc *dc, struct dc_state *context,
+static enum dc_status dcn21_validate_bandwidth(struct dc *dc, struct dc_state *context,
 		bool fast_validate)
 {
 	bool voltage_supported;
···
 
 	pipes = kcalloc(dc->res_pool->pipe_count, sizeof(display_e2e_pipe_params_st), GFP_KERNEL);
 	if (!pipes)
-		return false;
+		return DC_FAIL_BANDWIDTH_VALIDATE;
 
 	DC_FP_START();
 	voltage_supported = dcn21_validate_bandwidth_fp(dc, context, fast_validate, pipes);
 	DC_FP_END();
 
 	kfree(pipes);
-	return voltage_supported;
+	return voltage_supported ? DC_OK : DC_NOT_SUPPORTED;
 }
 
 static void dcn21_destroy_resource_pool(struct resource_pool **pool)

drivers/gpu/drm/amd/display/dc/resource/dcn30/dcn30_resource.c

···
 	DC_FP_END();
 }
 
-bool dcn30_validate_bandwidth(struct dc *dc,
+enum dc_status dcn30_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
···
 
 	BW_VAL_TRACE_FINISH();
 
-	return out;
+	return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 void dcn30_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_params)

drivers/gpu/drm/amd/display/dc/resource/dcn30/dcn30_resource.h

···
 		enum mmhubbub_wbif_mode mode,
 		unsigned int urgent_watermark);
 
-bool dcn30_validate_bandwidth(struct dc *dc, struct dc_state *context,
+enum dc_status dcn30_validate_bandwidth(struct dc *dc, struct dc_state *context,
 		bool fast_validate);
 bool dcn30_internal_validate_bw(
 		struct dc *dc,

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

···
 	DC_FP_END();
 }
 
-bool dcn31_validate_bandwidth(struct dc *dc,
+enum dc_status dcn31_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
···
 
 	BW_VAL_TRACE_FINISH();
 
-	return out;
+	return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 static void dcn31_get_panel_config_defaults(struct dc_panel_config *panel_config)

drivers/gpu/drm/amd/display/dc/resource/dcn31/dcn31_resource.h

···
 	struct resource_pool base;
 };
 
-bool dcn31_validate_bandwidth(struct dc *dc,
+enum dc_status dcn31_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate);
 void dcn31_calculate_wm_and_dlg(

drivers/gpu/drm/amd/display/dc/resource/dcn314/dcn314_resource.c

···
 	*panel_config = panel_config_defaults;
 }
 
-bool dcn314_validate_bandwidth(struct dc *dc,
+enum dc_status dcn314_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
···
 
 	BW_VAL_TRACE_FINISH();
 
-	return out;
+	return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 static struct resource_funcs dcn314_res_pool_funcs = {

drivers/gpu/drm/amd/display/dc/resource/dcn314/dcn314_resource.h

···
 	struct resource_pool base;
 };
 
-bool dcn314_validate_bandwidth(struct dc *dc,
+enum dc_status dcn314_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate);
 

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

···
  *
  */
 
+#include "dc_types.h"
 #include "dm_services.h"
 #include "dc.h"
 
···
 	return out;
 }
 
-bool dcn32_validate_bandwidth(struct dc *dc,
+enum dc_status dcn32_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
-	bool out = false;
+	unsigned int i;
+	enum dc_status status;
+	const struct dc_stream_state *stream;
+
+	/* reset cursor limitations on subvp */
+	for (i = 0; i < context->stream_count; i++) {
+		stream = context->streams[i];
+
+		if (dc_state_can_clear_stream_cursor_subvp_limit(stream, context)) {
+			dc_state_set_stream_cursor_subvp_limit(stream, context, false);
+		}
+	}
 
 	if (dc->debug.using_dml2)
-		out = dml2_validate(dc, context,
+		status = dml2_validate(dc, context,
 				context->power_source == DC_POWER_SOURCE_DC ? context->bw_ctx.dml2_dc_power_source : context->bw_ctx.dml2,
-				fast_validate);
+				fast_validate) ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 	else
-		out = dml1_validate(dc, context, fast_validate);
-	return out;
+		status = dml1_validate(dc, context, fast_validate) ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
+
+	if (!fast_validate && status == DC_OK && dc_state_is_subvp_in_use(context)) {
+		/* check new stream configuration still supports cursor if subvp used */
+		for (i = 0; i < context->stream_count; i++) {
+			stream = context->streams[i];
+
+			if (dc_state_get_stream_subvp_type(context, stream) != SUBVP_PHANTOM &&
+					stream->cursor_position.enable &&
+					!dc_stream_check_cursor_attributes(stream, context, &stream->cursor_attributes)) {
+				/* hw cursor cannot be supported with subvp active, so disable subvp for now */
+				dc_state_set_stream_cursor_subvp_limit(stream, context, true);
+				status = DC_FAIL_HW_CURSOR_SUPPORT;
+			}
+		};
+	}
+
+	if (!fast_validate && status == DC_FAIL_HW_CURSOR_SUPPORT) {
+		/* attempt to validate again with subvp disabled due to cursor */
+		if (dc->debug.using_dml2)
+			status = dml2_validate(dc, context,
+					context->power_source == DC_POWER_SOURCE_DC ? context->bw_ctx.dml2_dc_power_source : context->bw_ctx.dml2,
+					fast_validate) ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
+		else
+			status = dml1_validate(dc, context, fast_validate) ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
+	}
+
+	return status;
 }
 
 int dcn32_populate_dml_pipes_from_context(
···
 	DC_FP_END();
 }
 
+unsigned int dcn32_get_max_hw_cursor_size(const struct dc *dc,
+			struct dc_state *state,
+			const struct dc_stream_state *stream)
+{
+	bool limit_cur_to_buf;
+
+	limit_cur_to_buf = dc_state_get_stream_subvp_cursor_limit(stream, state) &&
+			!stream->hw_cursor_req;
+
+	return limit_cur_to_buf ? dc->caps.max_buffered_cursor_size : dc->caps.max_cursor_size;
+}
+
 static struct resource_funcs dcn32_res_pool_funcs = {
 	.destroy = dcn32_destroy_resource_pool,
 	.link_enc_create = dcn32_link_encoder_create,
···
 	.add_phantom_pipes = dcn32_add_phantom_pipes,
 	.build_pipe_pix_clk_params = dcn20_build_pipe_pix_clk_params,
 	.calculate_mall_ways_from_bytes = dcn32_calculate_mall_ways_from_bytes,
-	.get_vstartup_for_pipe = dcn10_get_vstartup_for_pipe
+	.get_vstartup_for_pipe = dcn10_get_vstartup_for_pipe,
+	.get_max_hw_cursor_size = dcn32_get_max_hw_cursor_size,
 };
 
 static uint32_t read_pipe_fuses(struct dc_context *ctx)
···
 	dc->caps.i2c_speed_in_khz_hdcp = 100; /*1.4 w/a applied by default*/
 	/* TODO: Bring max_cursor_size back to 256 after subvp cursor corruption is fixed*/
 	dc->caps.max_cursor_size = 64;
+	dc->caps.max_buffered_cursor_size = 64; // sqrt(16 * 1024 / 4)
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.dmdata_alloc_size = 2048;
 	dc->caps.mall_size_per_mem_channel = 4;

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

···
 		unsigned int pipe_cnt,
 		unsigned int index);
 
-bool dcn32_validate_bandwidth(struct dc *dc,
+enum dc_status dcn32_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate);
 
···
 void dcn32_override_min_req_dcfclk(struct dc *dc, struct dc_state *context);
 
 unsigned int dcn32_calculate_mall_ways_from_bytes(const struct dc *dc, unsigned int total_size_in_mall_bytes);
+
+unsigned int dcn32_get_max_hw_cursor_size(const struct dc *dc,
+			struct dc_state *state,
+			const struct dc_stream_state *stream);
 
 /* definitions for run time init of reg offsets */
 

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

···
 	.add_phantom_pipes = dcn32_add_phantom_pipes,
 	.build_pipe_pix_clk_params = dcn20_build_pipe_pix_clk_params,
 	.calculate_mall_ways_from_bytes = dcn32_calculate_mall_ways_from_bytes,
-	.get_vstartup_for_pipe = dcn10_get_vstartup_for_pipe
+	.get_vstartup_for_pipe = dcn10_get_vstartup_for_pipe,
+	.get_max_hw_cursor_size = dcn32_get_max_hw_cursor_size,
 };
 
 static uint32_t read_pipe_fuses(struct dc_context *ctx)
···
 	dc->caps.i2c_speed_in_khz_hdcp = 100; /*1.4 w/a applied by default*/
 	/* TODO: Bring max cursor size back to 256 after subvp cursor corruption is fixed*/
 	dc->caps.max_cursor_size = 64;
+	dc->caps.max_buffered_cursor_size = 64; // sqrt(16 * 1024 / 4)
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.dmdata_alloc_size = 2048;
 	dc->caps.mall_size_per_mem_channel = 4;

drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c

···
 }
 
 
-static bool dcn35_validate_bandwidth(struct dc *dc,
+static enum dc_status dcn35_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
···
 			fast_validate);
 
 	if (fast_validate)
-		return out;
+		return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 
 	DC_FP_START();
 	dcn35_decide_zstate_support(dc, context);
 	DC_FP_END();
 
-	return out;
+	return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 enum dc_status dcn35_patch_unknown_plane_state(struct dc_plane_state *plane_state)

drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c

···
 }
 
 
-static bool dcn351_validate_bandwidth(struct dc *dc,
+static enum dc_status dcn351_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
···
 			fast_validate);
 
 	if (fast_validate)
-		return out;
+		return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 
 	DC_FP_START();
 	dcn35_decide_zstate_support(dc, context);
 	DC_FP_END();
 
-	return out;
+	return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 static struct resource_funcs dcn351_res_pool_funcs = {

drivers/gpu/drm/amd/display/dc/resource/dcn36/dcn36_resource.c

···
 }
 
 
-static bool dcn35_validate_bandwidth(struct dc *dc,
+static enum dc_status dcn35_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
···
 			fast_validate);
 
 	if (fast_validate)
-		return out;
+		return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 
 	DC_FP_START();
 	dcn35_decide_zstate_support(dc, context);
 	DC_FP_END();
 
-	return out;
+	return out ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
 }
 
 

drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c

···
 	return DC_OK;
 }
 
-bool dcn401_validate_bandwidth(struct dc *dc,
+enum dc_status dcn401_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate)
 {
-	bool out = false;
+	unsigned int i;
+	enum dc_status status = DC_OK;
+	const struct dc_stream_state *stream;
+
+	/* reset cursor limitations on subvp */
+	for (i = 0; i < context->stream_count; i++) {
+		stream = context->streams[i];
+
+		if (dc_state_can_clear_stream_cursor_subvp_limit(stream, context)) {
+			dc_state_set_stream_cursor_subvp_limit(stream, context, false);
+		}
+	}
+
 	if (dc->debug.using_dml2)
-		out = dml2_validate(dc, context,
+		status = dml2_validate(dc, context,
 				context->power_source == DC_POWER_SOURCE_DC ? context->bw_ctx.dml2_dc_power_source : context->bw_ctx.dml2,
-				fast_validate);
-	return out;
+				fast_validate) ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
+
+	if (!fast_validate && status == DC_OK && dc_state_is_subvp_in_use(context)) {
+		/* check new stream configuration still supports cursor if subvp used */
+		for (i = 0; i < context->stream_count; i++) {
+			stream = context->streams[i];
+
+			if (dc_state_get_stream_subvp_type(context, stream) != SUBVP_PHANTOM &&
+					stream->cursor_position.enable &&
+					!dc_stream_check_cursor_attributes(stream, context, &stream->cursor_attributes))	{
+				/* hw cursor cannot be supported with subvp active, so disable subvp for now */
+				dc_state_set_stream_cursor_subvp_limit(stream, context, true);
+				status = DC_FAIL_HW_CURSOR_SUPPORT;
+			}
+		};
+	}
+
+	if (!fast_validate && status == DC_FAIL_HW_CURSOR_SUPPORT) {
+		/* attempt to validate again with subvp disabled due to cursor */
+		if (dc->debug.using_dml2)
+			status = dml2_validate(dc, context,
+					context->power_source == DC_POWER_SOURCE_DC ? context->bw_ctx.dml2_dc_power_source : context->bw_ctx.dml2,
+					fast_validate) ? DC_OK : DC_FAIL_BANDWIDTH_VALIDATE;
+	}
+
+	return status;
 }
 
 void dcn401_prepare_mcache_programming(struct dc *dc,
···
 	.build_pipe_pix_clk_params = dcn401_build_pipe_pix_clk_params,
 	.calculate_mall_ways_from_bytes = dcn32_calculate_mall_ways_from_bytes,
 	.get_power_profile = dcn401_get_power_profile,
-	.get_vstartup_for_pipe = dcn401_get_vstartup_for_pipe
+	.get_vstartup_for_pipe = dcn401_get_vstartup_for_pipe,
+	.get_max_hw_cursor_size = dcn32_get_max_hw_cursor_size
 };
 
 static uint32_t read_pipe_fuses(struct dc_context *ctx)
···
 	dc->caps.max_downscale_ratio = 600;
 	dc->caps.i2c_speed_in_khz = 95;
 	dc->caps.i2c_speed_in_khz_hdcp = 95; /*1.4 w/a applied by default*/
-	/* TODO: Bring max cursor size back to 256 after subvp cursor corruption is fixed*/
+	/* used to set cursor pitch, so must be aligned to power of 2 (HW actually supported 78x78) */
 	dc->caps.max_cursor_size = 64;
+	dc->caps.max_buffered_cursor_size = 64;
 	dc->caps.cursor_not_scaled = true;
 	dc->caps.min_horizontal_blanking_period = 80;
 	dc->caps.dmdata_alloc_size = 2048;

drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.h

···
 
 enum dc_status dcn401_patch_unknown_plane_state(struct dc_plane_state *plane_state);
 
-bool dcn401_validate_bandwidth(struct dc *dc,
+enum dc_status dcn401_validate_bandwidth(struct dc *dc,
 		struct dc_state *context,
 		bool fast_validate);