Merge tag 'amd-drm-next-6.8-2024-01-05' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.8-2024-01-05:

amdgpu:
- VRR fixes
- PSR-SU fixes
- SubVP fixes
- DCN 3.5 fixes
- Documentation updates
- DMCUB fixes
- DML2 fixes
- UMC 12.0 updates
- GPUVM fix
- Misc code cleanups and whitespace cleanups
- DP MST fix
- Let KFD sync with GPUVM fences
- GFX11 reset fix
- SMU 13.0.6 fixes
- VSC fix for DP/eDP
- Navi12 display fix
- RN/CZN system aperture fix
- DCN 2.1 bandwidth validation fix
- DCN INIT cleanup

amdkfd:
- SVM fixes
- Revert TBA/TMA location change

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240105220522.4976-1-alexander.deucher@amd.com

10 files changed, 226 insertions, 158 deletions

diff --git a/drivers/gpu/drm/amd/display/dc/link/link_factory.c b/drivers/gpu/drm/amd/display/dc/link/link_factory.c
index 5464d8d26bd3..37d3027c32dc 100644
--- a/drivers/gpu/drm/amd/display/dc/link/link_factory.c
+++ b/drivers/gpu/drm/amd/display/dc/link/link_factory.c
@@ -216,6 +216,7 @@ static void construct_link_service_edp_panel_control(struct link_service *link_s
 	link_srv->edp_send_replay_cmd = edp_send_replay_cmd;
 	link_srv->edp_set_coasting_vtotal = edp_set_coasting_vtotal;
 	link_srv->edp_replay_residency = edp_replay_residency;
+	link_srv->edp_set_replay_power_opt_and_coasting_vtotal = edp_set_replay_power_opt_and_coasting_vtotal;
 
 	link_srv->edp_wait_for_t12 = edp_wait_for_t12;
 	link_srv->edp_is_ilr_optimization_required =
diff --git a/drivers/gpu/drm/amd/display/dc/link/link_validation.h b/drivers/gpu/drm/amd/display/dc/link/link_validation.h
index 4a954317d0da..595fb05946e9 100644
--- a/drivers/gpu/drm/amd/display/dc/link/link_validation.h
+++ b/drivers/gpu/drm/amd/display/dc/link/link_validation.h
@@ -25,6 +25,7 @@
 #ifndef __LINK_VALIDATION_H__
 #define __LINK_VALIDATION_H__
 #include "link.h"
+
 enum dc_status link_validate_mode_timing(
 		const struct dc_stream_state *stream,
 		struct dc_link *link,
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
index 3c5334cdb3fb..289f5d133342 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
@@ -1398,7 +1398,7 @@ static bool get_usbc_cable_id(struct dc_link *link, union dp_cable_id *cable_id)
 	cmd.cable_id.header.payload_bytes = sizeof(cmd.cable_id.data);
 	cmd.cable_id.data.input.phy_inst = resource_transmitter_to_phy_idx(
 			link->dc, link->link_enc->transmitter);
-	if (dm_execute_dmub_cmd(link->dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY) &&
+	if (dc_wake_and_execute_dmub_cmd(link->dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY) &&
 			cmd.cable_id.header.ret_status == 1) {
 		cable_id->raw = cmd.cable_id.data.output_raw;
 		DC_LOG_DC("usbc_cable_id = %d.\n", cable_id->raw);
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia.c
index 0bb749133909..982eda3c46f5 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia.c
@@ -90,7 +90,8 @@ bool dpia_query_hpd_status(struct dc_link *link)
 	cmd.query_hpd.data.ch_type = AUX_CHANNEL_DPIA;
 
 	/* Return HPD status reported by DMUB if query successfully executed. */
-	if (dm_execute_dmub_cmd(dmub_srv->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY) && cmd.query_hpd.data.status == AUX_RET_SUCCESS)
+	if (dc_wake_and_execute_dmub_cmd(dmub_srv->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT_WITH_REPLY) &&
+	    cmd.query_hpd.data.status == AUX_RET_SUCCESS)
 		is_hpd_high = cmd.query_hpd.data.result;
 
 	DC_LOG_DEBUG("%s: link(%d) dpia(%d) cmd_status(%d) result(%d)\n",
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.c
index 7581023daa47..a7aa8c9da868 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.c
@@ -50,6 +50,7 @@ static bool get_bw_alloc_proceed_flag(struct dc_link *tmp)
 			&& tmp->hpd_status
 			&& tmp->dpia_bw_alloc_config.bw_alloc_enabled);
 }
+
 static void reset_bw_alloc_struct(struct dc_link *link)
 {
 	link->dpia_bw_alloc_config.bw_alloc_enabled = false;
@@ -58,7 +59,13 @@ static void reset_bw_alloc_struct(struct dc_link *link)
 	link->dpia_bw_alloc_config.estimated_bw = 0;
 	link->dpia_bw_alloc_config.bw_granularity = 0;
 	link->dpia_bw_alloc_config.response_ready = false;
+	link->dpia_bw_alloc_config.sink_allocated_bw = 0;
 }
+
+#define BW_GRANULARITY_0 4 // 0.25 Gbps
+#define BW_GRANULARITY_1 2 // 0.5 Gbps
+#define BW_GRANULARITY_2 1 // 1 Gbps
+
 static uint8_t get_bw_granularity(struct dc_link *link)
 {
 	uint8_t bw_granularity = 0;
@@ -71,16 +78,20 @@ static uint8_t get_bw_granularity(struct dc_link *link)
 
 	switch (bw_granularity & 0x3) {
 	case 0:
-		bw_granularity = 4;
+		bw_granularity = BW_GRANULARITY_0;
 		break;
 	case 1:
+		bw_granularity = BW_GRANULARITY_1;
+		break;
+	case 2:
 	default:
-		bw_granularity = 2;
+		bw_granularity = BW_GRANULARITY_2;
 		break;
 	}
 
 	return bw_granularity;
 }
+
 static int get_estimated_bw(struct dc_link *link)
 {
 	uint8_t bw_estimated_bw = 0;
@@ -93,31 +104,33 @@ static int get_estimated_bw(struct dc_link *link)
 
 	return bw_estimated_bw * (Kbps_TO_Gbps / link->dpia_bw_alloc_config.bw_granularity);
 }
-static bool allocate_usb4_bw(int *stream_allocated_bw, int bw_needed, struct dc_link *link)
+
+static int get_non_reduced_max_link_rate(struct dc_link *link)
 {
-	if (bw_needed > 0)
-		*stream_allocated_bw += bw_needed;
+	uint8_t nrd_max_link_rate = 0;
 
-	return true;
+	core_link_read_dpcd(
+			link,
+			DP_TUNNELING_MAX_LINK_RATE,
+			&nrd_max_link_rate,
+			sizeof(uint8_t));
+
+	return nrd_max_link_rate;
 }
-static bool deallocate_usb4_bw(int *stream_allocated_bw, int bw_to_dealloc, struct dc_link *link)
-{
-	bool ret = false;
 
-	if (*stream_allocated_bw > 0) {
-		*stream_allocated_bw -= bw_to_dealloc;
-		ret = true;
-	} else {
-		//Do nothing for now
-		ret = true;
-	}
+static int get_non_reduced_max_lane_count(struct dc_link *link)
+{
+	uint8_t nrd_max_lane_count = 0;
 
-	// Unplug so reset values
-	if (!link->hpd_status)
-		reset_bw_alloc_struct(link);
+	core_link_read_dpcd(
+			link,
+			DP_TUNNELING_MAX_LANE_COUNT,
+			&nrd_max_lane_count,
+			sizeof(uint8_t));
 
-	return ret;
+	return nrd_max_lane_count;
 }
+
 /*
  * Read all New BW alloc configuration ex: estimated_bw, allocated_bw,
  * granuality, Driver_ID, CM_Group, & populate the BW allocation structs
@@ -125,10 +138,22 @@ static bool deallocate_usb4_bw(int *stream_allocated_bw, int bw_to_dealloc, stru
  */
 static void init_usb4_bw_struct(struct dc_link *link)
 {
-	// Init the known values
+	reset_bw_alloc_struct(link);
+
+	/* init the known values */
 	link->dpia_bw_alloc_config.bw_granularity = get_bw_granularity(link);
 	link->dpia_bw_alloc_config.estimated_bw = get_estimated_bw(link);
+	link->dpia_bw_alloc_config.nrd_max_link_rate = get_non_reduced_max_link_rate(link);
+	link->dpia_bw_alloc_config.nrd_max_lane_count = get_non_reduced_max_lane_count(link);
+
+	DC_LOG_DEBUG("%s: bw_granularity(%d), estimated_bw(%d)\n",
+		__func__, link->dpia_bw_alloc_config.bw_granularity,
+		link->dpia_bw_alloc_config.estimated_bw);
+	DC_LOG_DEBUG("%s: nrd_max_link_rate(%d), nrd_max_lane_count(%d)\n",
+		__func__, link->dpia_bw_alloc_config.nrd_max_link_rate,
+		link->dpia_bw_alloc_config.nrd_max_lane_count);
 }
+
 static uint8_t get_lowest_dpia_index(struct dc_link *link)
 {
 	const struct dc *dc_struct = link->dc;
@@ -141,51 +166,66 @@ static uint8_t get_lowest_dpia_index(struct dc_link *link)
 				dc_struct->links[i]->ep_type != DISPLAY_ENDPOINT_USB4_DPIA)
 			continue;
 
-		if (idx > dc_struct->links[i]->link_index)
+		if (idx > dc_struct->links[i]->link_index) {
 			idx = dc_struct->links[i]->link_index;
+			break;
+		}
 	}
 
 	return idx;
 }
+
 /*
- * Get the Max Available BW or Max Estimated BW for each Host Router
+ * Get the maximum dp tunnel banwidth of host router
  *
- * @link: pointer to the dc_link struct instance
- * @type: ESTIMATD BW or MAX AVAILABLE BW
+ * @dc: pointer to the dc struct instance
+ * @hr_index: host router index
  *
- * return: response_ready flag from dc_link struct
+ * return: host router maximum dp tunnel bandwidth
  */
-static int get_host_router_total_bw(struct dc_link *link, uint8_t type)
+static int get_host_router_total_dp_tunnel_bw(const struct dc *dc, uint8_t hr_index)
 {
-	const struct dc *dc_struct = link->dc;
-	uint8_t lowest_dpia_index = get_lowest_dpia_index(link);
-	uint8_t idx = (link->link_index - lowest_dpia_index) / 2, idx_temp = 0;
-	struct dc_link *link_temp;
+	uint8_t lowest_dpia_index = get_lowest_dpia_index(dc->links[0]);
+	uint8_t hr_index_temp = 0;
+	struct dc_link *link_dpia_primary, *link_dpia_secondary;
 	int total_bw = 0;
-	int i;
-
-	for (i = 0; i < MAX_PIPES * 2; ++i) {
 
-		if (!dc_struct->links[i] || dc_struct->links[i]->ep_type != DISPLAY_ENDPOINT_USB4_DPIA)
-			continue;
+	for (uint8_t i = 0; i < MAX_PIPES * 2; ++i) {
 
-		link_temp = dc_struct->links[i];
-		if (!link_temp || !link_temp->hpd_status)
+		if (!dc->links[i] || dc->links[i]->ep_type != DISPLAY_ENDPOINT_USB4_DPIA)
 			continue;
 
-		idx_temp = (link_temp->link_index - lowest_dpia_index) / 2;
-
-		if (idx_temp == idx) {
-
-			if (type == HOST_ROUTER_BW_ESTIMATED)
-				total_bw += link_temp->dpia_bw_alloc_config.estimated_bw;
-			else if (type == HOST_ROUTER_BW_ALLOCATED)
-				total_bw += link_temp->dpia_bw_alloc_config.sink_allocated_bw;
+		hr_index_temp = (dc->links[i]->link_index - lowest_dpia_index) / 2;
+
+		if (hr_index_temp == hr_index) {
+			link_dpia_primary = dc->links[i];
+			link_dpia_secondary = dc->links[i + 1];
+
+			/**
+			 * If BW allocation enabled on both DPIAs, then
+			 * HR BW = Estimated(dpia_primary) + Allocated(dpia_secondary)
+			 * otherwise HR BW = Estimated(bw alloc enabled dpia)
+			 */
+			if ((link_dpia_primary->hpd_status &&
+				link_dpia_primary->dpia_bw_alloc_config.bw_alloc_enabled) &&
+				(link_dpia_secondary->hpd_status &&
+				link_dpia_secondary->dpia_bw_alloc_config.bw_alloc_enabled)) {
+				total_bw += link_dpia_primary->dpia_bw_alloc_config.estimated_bw +
+					link_dpia_secondary->dpia_bw_alloc_config.sink_allocated_bw;
+			} else if (link_dpia_primary->hpd_status &&
+					link_dpia_primary->dpia_bw_alloc_config.bw_alloc_enabled) {
+				total_bw = link_dpia_primary->dpia_bw_alloc_config.estimated_bw;
+			} else if (link_dpia_secondary->hpd_status &&
+				link_dpia_secondary->dpia_bw_alloc_config.bw_alloc_enabled) {
+				total_bw += link_dpia_secondary->dpia_bw_alloc_config.estimated_bw;
+			}
+			break;
 		}
 	}
 
 	return total_bw;
 }
+
 /*
  * Cleanup function for when the dpia is unplugged to reset struct
  * and perform any required clean up
@@ -194,42 +234,49 @@ static int get_host_router_total_bw(struct dc_link *link, uint8_t type)
  *
  * return: none
  */
-static bool dpia_bw_alloc_unplug(struct dc_link *link)
+static void dpia_bw_alloc_unplug(struct dc_link *link)
 {
-	if (!link)
-		return true;
-
-	return deallocate_usb4_bw(&link->dpia_bw_alloc_config.sink_allocated_bw,
-			link->dpia_bw_alloc_config.sink_allocated_bw, link);
+	if (link) {
+		DC_LOG_DEBUG("%s: resetting bw alloc config for link(%d)\n",
+			__func__, link->link_index);
+		reset_bw_alloc_struct(link);
+	}
 }
+
 static void set_usb4_req_bw_req(struct dc_link *link, int req_bw)
 {
 	uint8_t requested_bw;
 	uint32_t temp;
 
-	// 1. Add check for this corner case #1
-	if (req_bw > link->dpia_bw_alloc_config.estimated_bw)
+	/* Error check whether request bw greater than allocated */
+	if (req_bw > link->dpia_bw_alloc_config.estimated_bw) {
+		DC_LOG_ERROR("%s: Request bw greater than estimated bw for link(%d)\n",
+			__func__, link->link_index);
 		req_bw = link->dpia_bw_alloc_config.estimated_bw;
+	}
 
 	temp = req_bw * link->dpia_bw_alloc_config.bw_granularity;
 	requested_bw = temp / Kbps_TO_Gbps;
 
-	// Always make sure to add more to account for floating points
+	/* Always make sure to add more to account for floating points */
 	if (temp % Kbps_TO_Gbps)
 		++requested_bw;
 
-	// 2. Add check for this corner case #2
+	/* Error check whether requested and allocated are equal */
 	req_bw = requested_bw * (Kbps_TO_Gbps / link->dpia_bw_alloc_config.bw_granularity);
-	if (req_bw == link->dpia_bw_alloc_config.sink_allocated_bw)
-		return;
+	if (req_bw == link->dpia_bw_alloc_config.sink_allocated_bw) {
+		DC_LOG_ERROR("%s: Request bw equals to allocated bw for link(%d)\n",
+			__func__, link->link_index);
+	}
 
-	if (core_link_write_dpcd(
+	link->dpia_bw_alloc_config.response_ready = false; // Reset flag
+	core_link_write_dpcd(
 		link,
 		REQUESTED_BW,
 		&requested_bw,
-		sizeof(uint8_t)) == DC_OK)
-		link->dpia_bw_alloc_config.response_ready = false; // Reset flag
+		sizeof(uint8_t));
 }
+
 /*
  * Return the response_ready flag from dc_link struct
  *
@@ -241,6 +288,7 @@ static bool get_cm_response_ready_flag(struct dc_link *link)
 {
 	return link->dpia_bw_alloc_config.response_ready;
 }
+
 // ------------------------------------------------------------------
 //					PUBLIC FUNCTIONS
 // ------------------------------------------------------------------
@@ -277,27 +325,27 @@ bool link_dp_dpia_set_dptx_usb4_bw_alloc_support(struct dc_link *link)
 				DPTX_BW_ALLOCATION_MODE_CONTROL,
 				&response,
 				sizeof(uint8_t)) != DC_OK) {
-			DC_LOG_DEBUG("%s: **** FAILURE Enabling DPtx BW Allocation Mode Support ***\n",
-					__func__);
+			DC_LOG_DEBUG("%s: FAILURE Enabling DPtx BW Allocation Mode Support for link(%d)\n",
+				__func__, link->link_index);
 		} else {
 			// SUCCESS Enabled DPtx BW Allocation Mode Support
-			link->dpia_bw_alloc_config.bw_alloc_enabled = true;
-			DC_LOG_DEBUG("%s: **** SUCCESS Enabling DPtx BW Allocation Mode Support ***\n",
-					__func__);
+			DC_LOG_DEBUG("%s: SUCCESS Enabling DPtx BW Allocation Mode Support for link(%d)\n",
+				__func__, link->link_index);
 
 			ret = true;
 			init_usb4_bw_struct(link);
+			link->dpia_bw_alloc_config.bw_alloc_enabled = true;
 		}
 	}
 
 out:
 	return ret;
 }
+
 void dpia_handle_bw_alloc_response(struct dc_link *link, uint8_t bw, uint8_t result)
 {
 	int bw_needed = 0;
 	int estimated = 0;
-	int host_router_total_estimated_bw = 0;
 
 	if (!get_bw_alloc_proceed_flag((link)))
 		return;
@@ -306,14 +354,22 @@ void dpia_handle_bw_alloc_response(struct dc_link *link, uint8_t bw, uint8_t res
 
 	case DPIA_BW_REQ_FAILED:
 
-		DC_LOG_DEBUG("%s: *** *** BW REQ FAILURE for DP-TX Request *** ***\n", __func__);
+		/*
+		 * Ideally, we shouldn't run into this case as we always validate available
+		 * bandwidth and request within that limit
+		 */
+		estimated = bw * (Kbps_TO_Gbps / link->dpia_bw_alloc_config.bw_granularity);
+
+		DC_LOG_ERROR("%s: BW REQ FAILURE for DP-TX Request for link(%d)\n",
+			__func__, link->link_index);
+		DC_LOG_ERROR("%s: current estimated_bw(%d), new estimated_bw(%d)\n",
+			__func__, link->dpia_bw_alloc_config.estimated_bw, estimated);
 
-		// Update the new Estimated BW value updated by CM
-		link->dpia_bw_alloc_config.estimated_bw =
-				bw * (Kbps_TO_Gbps / link->dpia_bw_alloc_config.bw_granularity);
+		/* Update the new Estimated BW value updated by CM */
+		link->dpia_bw_alloc_config.estimated_bw = estimated;
 
+		/* Allocate the previously requested bandwidth */
 		set_usb4_req_bw_req(link, link->dpia_bw_alloc_config.estimated_bw);
-		link->dpia_bw_alloc_config.response_ready = false;
 
 		/*
 		 * If FAIL then it is either:
@@ -326,68 +382,34 @@ void dpia_handle_bw_alloc_response(struct dc_link *link, uint8_t bw, uint8_t res
 
 	case DPIA_BW_REQ_SUCCESS:
 
-		DC_LOG_DEBUG("%s: *** BW REQ SUCCESS for DP-TX Request ***\n", __func__);
-
-		// 1. SUCCESS 1st time before any Pruning is done
-		// 2. SUCCESS after prev. FAIL before any Pruning is done
-		// 3. SUCCESS after Pruning is done but before enabling link
-
 		bw_needed = bw * (Kbps_TO_Gbps / link->dpia_bw_alloc_config.bw_granularity);
 
-		// 1.
-		if (!link->dpia_bw_alloc_config.sink_allocated_bw) {
-
-			allocate_usb4_bw(&link->dpia_bw_alloc_config.sink_allocated_bw, bw_needed, link);
-			link->dpia_bw_alloc_config.sink_verified_bw =
-					link->dpia_bw_alloc_config.sink_allocated_bw;
+		DC_LOG_DEBUG("%s: BW REQ SUCCESS for DP-TX Request for link(%d)\n",
+			__func__, link->link_index);
+		DC_LOG_DEBUG("%s: current allocated_bw(%d), new allocated_bw(%d)\n",
+			__func__, link->dpia_bw_alloc_config.sink_allocated_bw, bw_needed);
 
-			// SUCCESS from first attempt
-			if (link->dpia_bw_alloc_config.sink_allocated_bw >
-			link->dpia_bw_alloc_config.sink_max_bw)
-				link->dpia_bw_alloc_config.sink_verified_bw =
-						link->dpia_bw_alloc_config.sink_max_bw;
-		}
-		// 3.
-		else if (link->dpia_bw_alloc_config.sink_allocated_bw) {
-
-			// Find out how much do we need to de-alloc
-			if (link->dpia_bw_alloc_config.sink_allocated_bw > bw_needed)
-				deallocate_usb4_bw(&link->dpia_bw_alloc_config.sink_allocated_bw,
-						link->dpia_bw_alloc_config.sink_allocated_bw - bw_needed, link);
-			else
-				allocate_usb4_bw(&link->dpia_bw_alloc_config.sink_allocated_bw,
-						bw_needed - link->dpia_bw_alloc_config.sink_allocated_bw, link);
-		}
-
-		// 4. If this is the 2nd sink then any unused bw will be reallocated to master DPIA
-		// => check if estimated_bw changed
+		link->dpia_bw_alloc_config.sink_allocated_bw = bw_needed;
 
 		link->dpia_bw_alloc_config.response_ready = true;
 		break;
 
 	case DPIA_EST_BW_CHANGED:
 
-		DC_LOG_DEBUG("%s: *** ESTIMATED BW CHANGED for DP-TX Request ***\n", __func__);
-
 		estimated = bw * (Kbps_TO_Gbps / link->dpia_bw_alloc_config.bw_granularity);
-		host_router_total_estimated_bw = get_host_router_total_bw(link, HOST_ROUTER_BW_ESTIMATED);
 
-		// 1. If due to unplug of other sink
-		if (estimated == host_router_total_estimated_bw) {
-			// First update the estimated & max_bw fields
-			if (link->dpia_bw_alloc_config.estimated_bw < estimated)
-				link->dpia_bw_alloc_config.estimated_bw = estimated;
-		}
-		// 2. If due to realloc bw btw 2 dpia due to plug OR realloc unused Bw
-		else {
-			// We lost estimated bw usually due to plug event of other dpia
-			link->dpia_bw_alloc_config.estimated_bw = estimated;
-		}
+		DC_LOG_DEBUG("%s: ESTIMATED BW CHANGED for link(%d)\n",
+			__func__, link->link_index);
+		DC_LOG_DEBUG("%s: current estimated_bw(%d), new estimated_bw(%d)\n",
+			__func__, link->dpia_bw_alloc_config.estimated_bw, estimated);
+
+		link->dpia_bw_alloc_config.estimated_bw = estimated;
 		break;
 
 	case DPIA_BW_ALLOC_CAPS_CHANGED:
 
-		DC_LOG_DEBUG("%s: *** BW ALLOC CAPABILITY CHANGED for DP-TX Request ***\n", __func__);
+		DC_LOG_ERROR("%s: BW ALLOC CAPABILITY CHANGED to Disabled for link(%d)\n",
+			__func__, link->link_index);
 		link->dpia_bw_alloc_config.bw_alloc_enabled = false;
 		break;
 	}
@@ -409,17 +431,17 @@ int dpia_handle_usb4_bandwidth_allocation_for_link(struct dc_link *link, int pea
 		set_usb4_req_bw_req(link, link->dpia_bw_alloc_config.sink_max_bw);
 
 		do {
-			if (!(timeout > 0))
+			if (timeout > 0)
 				timeout--;
 			else
 				break;
-			fsleep(10 * 1000);
+			msleep(10);
 		} while (!get_cm_response_ready_flag(link));
 
 		if (!timeout)
 			ret = 0;// ERROR TIMEOUT waiting for response for allocating bw
 		else if (link->dpia_bw_alloc_config.sink_allocated_bw > 0)
-			ret = get_host_router_total_bw(link, HOST_ROUTER_BW_ALLOCATED);
+			ret = link->dpia_bw_alloc_config.sink_allocated_bw;
 	}
 	//2. Cold Unplug
 	else if (!link->hpd_status)
@@ -428,63 +450,74 @@ int dpia_handle_usb4_bandwidth_allocation_for_link(struct dc_link *link, int pea
 out:
 	return ret;
 }
-int link_dp_dpia_allocate_usb4_bandwidth_for_stream(struct dc_link *link, int req_bw)
+
+bool link_dp_dpia_allocate_usb4_bandwidth_for_stream(struct dc_link *link, int req_bw)
 {
-	int ret = 0;
+	bool ret = false;
 	uint8_t timeout = 10;
 
+	DC_LOG_DEBUG("%s: ENTER: link(%d), hpd_status(%d), current allocated_bw(%d), req_bw(%d)\n",
+		__func__, link->link_index, link->hpd_status,
+		link->dpia_bw_alloc_config.sink_allocated_bw, req_bw);
+
 	if (!get_bw_alloc_proceed_flag(link))
 		goto out;
 
-	/*
-	 * Sometimes stream uses same timing parameters as the already
-	 * allocated max sink bw so no need to re-alloc
-	 */
-	if (req_bw != link->dpia_bw_alloc_config.sink_allocated_bw) {
-		set_usb4_req_bw_req(link, req_bw);
-		do {
-			if (!(timeout > 0))
-				timeout--;
-			else
-				break;
-			udelay(10 * 1000);
-		} while (!get_cm_response_ready_flag(link));
+	set_usb4_req_bw_req(link, req_bw);
+	do {
+		if (timeout > 0)
+			timeout--;
+		else
+			break;
+		msleep(10);
+	} while (!get_cm_response_ready_flag(link));
 
-		if (!timeout)
-			ret = 0;// ERROR TIMEOUT waiting for response for allocating bw
-		else if (link->dpia_bw_alloc_config.sink_allocated_bw > 0)
-			ret = get_host_router_total_bw(link, HOST_ROUTER_BW_ALLOCATED);
-	}
+	if (timeout)
+		ret = true;
 
 out:
+	DC_LOG_DEBUG("%s: EXIT: timeout(%d), ret(%d)\n", __func__, timeout, ret);
 	return ret;
 }
+
 bool dpia_validate_usb4_bw(struct dc_link **link, int *bw_needed_per_dpia, const unsigned int num_dpias)
 {
 	bool ret = true;
-	int bw_needed_per_hr[MAX_HR_NUM] = { 0, 0 };
-	uint8_t lowest_dpia_index = 0, dpia_index = 0;
-	uint8_t i;
+	int bw_needed_per_hr[MAX_HR_NUM] = { 0, 0 }, host_router_total_dp_bw = 0;
+	uint8_t lowest_dpia_index, i, hr_index;
 
 	if (!num_dpias || num_dpias > MAX_DPIA_NUM)
 		return ret;
 
-	//Get total Host Router BW & Validate against each Host Router max BW
+	lowest_dpia_index = get_lowest_dpia_index(link[0]);
+
+	/* get total Host Router BW with granularity for the given modes */
 	for (i = 0; i < num_dpias; ++i) {
+		int granularity_Gbps = 0;
+		int bw_granularity = 0;
 
 		if (!link[i]->dpia_bw_alloc_config.bw_alloc_enabled)
 			continue;
 
-		lowest_dpia_index = get_lowest_dpia_index(link[i]);
 		if (link[i]->link_index < lowest_dpia_index)
 			continue;
 
-		dpia_index = (link[i]->link_index - lowest_dpia_index) / 2;
-		bw_needed_per_hr[dpia_index] += bw_needed_per_dpia[i];
-		if (bw_needed_per_hr[dpia_index] > get_host_router_total_bw(link[i], HOST_ROUTER_BW_ALLOCATED)) {
+		granularity_Gbps = (Kbps_TO_Gbps / link[i]->dpia_bw_alloc_config.bw_granularity);
+		bw_granularity = (bw_needed_per_dpia[i] / granularity_Gbps) * granularity_Gbps +
+				((bw_needed_per_dpia[i] % granularity_Gbps) ? granularity_Gbps : 0);
 
-			ret = false;
-			break;
+		hr_index = (link[i]->link_index - lowest_dpia_index) / 2;
+		bw_needed_per_hr[hr_index] += bw_granularity;
+	}
+
+	/* validate against each Host Router max BW */
+	for (hr_index = 0; hr_index < MAX_HR_NUM; ++hr_index) {
+		if (bw_needed_per_hr[hr_index]) {
+			host_router_total_dp_bw = get_host_router_total_dp_tunnel_bw(link[0]->dc, hr_index);
+			if (bw_needed_per_hr[hr_index] > host_router_total_dp_bw) {
+				ret = false;
+				break;
+			}
 		}
 	}
 
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.h b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.h
index 7292690383ae..981bc4eb6120 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.h
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_dpia_bw.h
@@ -59,9 +59,9 @@ bool link_dp_dpia_set_dptx_usb4_bw_alloc_support(struct dc_link *link);
  * @link: pointer to the dc_link struct instance
  * @req_bw: Bw requested by the stream
  *
- * return: allocated bw else return 0
+ * return: true if allocated successfully
  */
-int link_dp_dpia_allocate_usb4_bandwidth_for_stream(struct dc_link *link, int req_bw);
+bool link_dp_dpia_allocate_usb4_bandwidth_for_stream(struct dc_link *link, int req_bw);
 
 /*
  * Handle the USB4 BW Allocation related functionality here:
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.c
index 90339c2dfd84..5a0b04518956 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.c
@@ -807,7 +807,7 @@ void dp_decide_lane_settings(
 		const struct link_training_settings *lt_settings,
 		const union lane_adjust ln_adjust[LANE_COUNT_DP_MAX],
 		struct dc_lane_settings hw_lane_settings[LANE_COUNT_DP_MAX],
-		union dpcd_training_lane dpcd_lane_settings[LANE_COUNT_DP_MAX])
+		union dpcd_training_lane *dpcd_lane_settings)
 {
 	uint32_t lane;
 
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.h b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.h
index 7d027bac8255..851bd17317a0 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.h
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_training.h
@@ -111,7 +111,7 @@ void dp_decide_lane_settings(
 	const struct link_training_settings *lt_settings,
 	const union lane_adjust ln_adjust[LANE_COUNT_DP_MAX],
 	struct dc_lane_settings hw_lane_settings[LANE_COUNT_DP_MAX],
-	union dpcd_training_lane dpcd_lane_settings[LANE_COUNT_DP_MAX]);
+	union dpcd_training_lane *dpcd_lane_settings);
 
 enum dc_dp_training_pattern decide_cr_training_pattern(
 		const struct dc_link_settings *link_settings);
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c
index bf53a86ea817..7f1196528218 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c
@@ -529,6 +529,9 @@ bool edp_set_backlight_level(const struct dc_link *link,
 	if (dc_is_embedded_signal(link->connector_signal)) {
 		struct pipe_ctx *pipe_ctx = get_pipe_from_link(link);
 
+		if (link->panel_cntl)
+			link->panel_cntl->stored_backlight_registers.USER_LEVEL = backlight_pwm_u16_16;
+
 		if (pipe_ctx) {
 			/* Disable brightness ramping when the display is blanked
 			 * as it can hang the DMCU
@@ -1065,6 +1068,33 @@ bool edp_replay_residency(const struct dc_link *link,
 	return true;
 }
 
+bool edp_set_replay_power_opt_and_coasting_vtotal(struct dc_link *link,
+	const unsigned int *power_opts, uint16_t coasting_vtotal)
+{
+	struct dc  *dc = link->ctx->dc;
+	struct dmub_replay *replay = dc->res_pool->replay;
+	unsigned int panel_inst;
+
+	if (!dc_get_edp_link_panel_inst(dc, link, &panel_inst))
+		return false;
+
+	/* Only both power and coasting vtotal changed, this func could return true */
+	if (power_opts && link->replay_settings.replay_power_opt_active != *power_opts &&
+		coasting_vtotal && link->replay_settings.coasting_vtotal != coasting_vtotal) {
+		if (link->replay_settings.replay_feature_enabled &&
+			replay->funcs->replay_set_power_opt_and_coasting_vtotal) {
+			replay->funcs->replay_set_power_opt_and_coasting_vtotal(replay,
+				*power_opts, panel_inst, coasting_vtotal);
+			link->replay_settings.replay_power_opt_active = *power_opts;
+			link->replay_settings.coasting_vtotal = coasting_vtotal;
+		} else
+			return false;
+	} else
+		return false;
+
+	return true;
+}
+
 static struct abm *get_abm_from_stream_res(const struct dc_link *link)
 {
 	int i;
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.h b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.h
index b7493ff4fcee..34e521af7bb4 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.h
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.h
@@ -63,6 +63,8 @@ bool edp_set_coasting_vtotal(struct dc_link *link, uint16_t coasting_vtotal);
 bool edp_replay_residency(const struct dc_link *link,
 	unsigned int *residency, const bool is_start, const bool is_alpm);
 bool edp_get_replay_state(const struct dc_link *link, uint64_t *state);
+bool edp_set_replay_power_opt_and_coasting_vtotal(struct dc_link *link,
+	const unsigned int *power_opts, uint16_t coasting_vtotal);
 bool edp_wait_for_t12(struct dc_link *link);
 bool edp_is_ilr_optimization_required(struct dc_link *link,
        struct dc_crtc_timing *crtc_timing);