Merge tag 'amd-drm-fixes-6.8-2024-01-18' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-fixes-6.8-2024-01-18:

amdgpu:
- DSC fixes
- DC resource pool fixes
- OTG fix
- DML2 fixes
- Aux fix
- GFX10 RLC firmware handling fix
- Revert a broken workaround for SMU 13.0.2
- DC writeback fix
- Enable gfxoff when ROCm apps are active on gfx11 with the proper FW version

amdkfd:
- Fix dma-buf exports using GEM handles

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

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240118223127.4904-1-alexander.deucher@amd.com

6 files changed, 173 insertions, 53 deletions

diff --git a/drivers/gpu/drm/amd/display/dc/link/link_dpms.c b/drivers/gpu/drm/amd/display/dc/link/link_dpms.c
index 5fe8b4871c77..3cbfbf8d107e 100644
--- a/drivers/gpu/drm/amd/display/dc/link/link_dpms.c
+++ b/drivers/gpu/drm/amd/display/dc/link/link_dpms.c
@@ -900,11 +900,15 @@ bool link_set_dsc_pps_packet(struct pipe_ctx *pipe_ctx, bool enable, bool immedi
 {
 	struct display_stream_compressor *dsc = pipe_ctx->stream_res.dsc;
 	struct dc_stream_state *stream = pipe_ctx->stream;
-	DC_LOGGER_INIT(dsc->ctx->logger);
 
-	if (!pipe_ctx->stream->timing.flags.DSC || !dsc)
+	if (!pipe_ctx->stream->timing.flags.DSC)
 		return false;
 
+	if (!dsc)
+		return false;
+
+	DC_LOGGER_INIT(dsc->ctx->logger);
+
 	if (enable) {
 		struct dsc_config dsc_cfg;
 		uint8_t dsc_packed_pps[128];
@@ -2005,17 +2009,11 @@ static enum dc_status enable_link_dp(struct dc_state *state,
 		}
 	}
 
-	/*
-	 * If the link is DP-over-USB4 do the following:
-	 * - Train with fallback when enabling DPIA link. Conventional links are
+	/* Train with fallback when enabling DPIA link. Conventional links are
 	 * trained with fallback during sink detection.
-	 * - Allocate only what the stream needs for bw in Gbps. Inform the CM
-	 * in case stream needs more or less bw from what has been allocated
-	 * earlier at plug time.
 	 */
-	if (link->ep_type == DISPLAY_ENDPOINT_USB4_DPIA) {
+	if (link->ep_type == DISPLAY_ENDPOINT_USB4_DPIA)
 		do_fallback = true;
-	}
 
 	/*
 	 * Temporary w/a to get DP2.0 link rates to work with SST.
@@ -2197,6 +2195,32 @@ static enum dc_status enable_link(
 	return status;
 }
 
+static bool allocate_usb4_bandwidth_for_stream(struct dc_stream_state *stream, int bw)
+{
+	return true;
+}
+
+static bool allocate_usb4_bandwidth(struct dc_stream_state *stream)
+{
+	bool ret;
+
+	int bw = dc_bandwidth_in_kbps_from_timing(&stream->timing,
+			dc_link_get_highest_encoding_format(stream->sink->link));
+
+	ret = allocate_usb4_bandwidth_for_stream(stream, bw);
+
+	return ret;
+}
+
+static bool deallocate_usb4_bandwidth(struct dc_stream_state *stream)
+{
+	bool ret;
+
+	ret = allocate_usb4_bandwidth_for_stream(stream, 0);
+
+	return ret;
+}
+
 void link_set_dpms_off(struct pipe_ctx *pipe_ctx)
 {
 	struct dc  *dc = pipe_ctx->stream->ctx->dc;
@@ -2232,6 +2256,9 @@ void link_set_dpms_off(struct pipe_ctx *pipe_ctx)
 	update_psp_stream_config(pipe_ctx, true);
 	dc->hwss.blank_stream(pipe_ctx);
 
+	if (pipe_ctx->stream->link->ep_type == DISPLAY_ENDPOINT_USB4_DPIA)
+		deallocate_usb4_bandwidth(pipe_ctx->stream);
+
 	if (pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST)
 		deallocate_mst_payload(pipe_ctx);
 	else if (pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT &&
@@ -2474,6 +2501,9 @@ void link_set_dpms_on(
 		}
 	}
 
+	if (pipe_ctx->stream->link->ep_type == DISPLAY_ENDPOINT_USB4_DPIA)
+		allocate_usb4_bandwidth(pipe_ctx->stream);
+
 	if (pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST)
 		allocate_mst_payload(pipe_ctx);
 	else if (pipe_ctx->stream->signal == SIGNAL_TYPE_DISPLAY_PORT &&
diff --git a/drivers/gpu/drm/amd/display/dc/link/link_validation.c b/drivers/gpu/drm/amd/display/dc/link/link_validation.c
index b45fda96eaf6..8fe66c367850 100644
--- a/drivers/gpu/drm/amd/display/dc/link/link_validation.c
+++ b/drivers/gpu/drm/amd/display/dc/link/link_validation.c
@@ -346,23 +346,61 @@ enum dc_status link_validate_mode_timing(
 	return DC_OK;
 }
 
+/*
+ * This function calculates the bandwidth required for the stream timing
+ * and aggregates the stream bandwidth for the respective dpia link
+ *
+ * @stream: pointer to the dc_stream_state struct instance
+ * @num_streams: number of streams to be validated
+ *
+ * return: true if validation is succeeded
+ */
 bool link_validate_dpia_bandwidth(const struct dc_stream_state *stream, const unsigned int num_streams)
 {
-	bool ret = true;
-	int bw_needed[MAX_DPIA_NUM];
-	struct dc_link *link[MAX_DPIA_NUM];
-
-	if (!num_streams || num_streams > MAX_DPIA_NUM)
-		return ret;
+	int bw_needed[MAX_DPIA_NUM] = {0};
+	struct dc_link *dpia_link[MAX_DPIA_NUM] = {0};
+	int num_dpias = 0;
 
 	for (uint8_t i = 0; i < num_streams; ++i) {
+		if (stream[i].signal == SIGNAL_TYPE_DISPLAY_PORT) {
+			/* new dpia sst stream, check whether it exceeds max dpia */
+			if (num_dpias >= MAX_DPIA_NUM)
+				return false;
 
-		link[i] = stream[i].link;
-		bw_needed[i] = dc_bandwidth_in_kbps_from_timing(&stream[i].timing,
-				dc_link_get_highest_encoding_format(link[i]));
+			dpia_link[num_dpias] = stream[i].link;
+			bw_needed[num_dpias] = dc_bandwidth_in_kbps_from_timing(&stream[i].timing,
+					dc_link_get_highest_encoding_format(dpia_link[num_dpias]));
+			num_dpias++;
+		} else if (stream[i].signal == SIGNAL_TYPE_DISPLAY_PORT_MST) {
+			uint8_t j = 0;
+			/* check whether its a known dpia link */
+			for (; j < num_dpias; ++j) {
+				if (dpia_link[j] == stream[i].link)
+					break;
+			}
+
+			if (j == num_dpias) {
+				/* new dpia mst stream, check whether it exceeds max dpia */
+				if (num_dpias >= MAX_DPIA_NUM)
+					return false;
+				else {
+					dpia_link[j] = stream[i].link;
+					num_dpias++;
+				}
+			}
+
+			bw_needed[j] += dc_bandwidth_in_kbps_from_timing(&stream[i].timing,
+				dc_link_get_highest_encoding_format(dpia_link[j]));
+		}
 	}
 
-	ret = dpia_validate_usb4_bw(link, bw_needed, num_streams);
+	/* Include dp overheads */
+	for (uint8_t i = 0; i < num_dpias; ++i) {
+		int dp_overhead = 0;
+
+		dp_overhead = link_dp_dpia_get_dp_overhead_in_dp_tunneling(dpia_link[i]);
+		bw_needed[i] += dp_overhead;
+	}
 
-	return ret;
+	return dpia_validate_usb4_bw(dpia_link, bw_needed, num_dpias);
 }
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 982eda3c46f5..6af42ba9885c 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
@@ -82,25 +82,33 @@ bool dpia_query_hpd_status(struct dc_link *link)
 {
 	union dmub_rb_cmd cmd = {0};
 	struct dc_dmub_srv *dmub_srv = link->ctx->dmub_srv;
-	bool is_hpd_high = false;
 
 	/* prepare QUERY_HPD command */
 	cmd.query_hpd.header.type = DMUB_CMD__QUERY_HPD_STATE;
 	cmd.query_hpd.data.instance = link->link_id.enum_id - ENUM_ID_1;
 	cmd.query_hpd.data.ch_type = AUX_CHANNEL_DPIA;
 
-	/* Return HPD status reported by DMUB if query successfully executed. */
-	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",
-		__func__,
-		link->link_index,
-		link->link_id.enum_id - ENUM_ID_1,
-		cmd.query_hpd.data.status,
-		cmd.query_hpd.data.result);
-
-	return is_hpd_high;
+	/* Query dpia hpd status from dmub */
+	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) {
+		DC_LOG_DEBUG("%s: for link(%d) dpia(%d) success, current_hpd_status(%d) new_hpd_status(%d)\n",
+			__func__,
+			link->link_index,
+			link->link_id.enum_id - ENUM_ID_1,
+			link->hpd_status,
+			cmd.query_hpd.data.result);
+		link->hpd_status = cmd.query_hpd.data.result;
+	} else {
+		DC_LOG_ERROR("%s: for link(%d) dpia(%d) failed with status(%d), current_hpd_status(%d) new_hpd_status(0)\n",
+			__func__,
+			link->link_index,
+			link->link_id.enum_id - ENUM_ID_1,
+			cmd.query_hpd.data.status,
+			link->hpd_status);
+		link->hpd_status = false;
+	}
+
+	return link->hpd_status;
 }
 
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 a7aa8c9da868..dd0d2b206462 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
@@ -54,12 +54,18 @@ static bool get_bw_alloc_proceed_flag(struct dc_link *tmp)
 static void reset_bw_alloc_struct(struct dc_link *link)
 {
 	link->dpia_bw_alloc_config.bw_alloc_enabled = false;
-	link->dpia_bw_alloc_config.sink_verified_bw = 0;
-	link->dpia_bw_alloc_config.sink_max_bw = 0;
+	link->dpia_bw_alloc_config.link_verified_bw = 0;
+	link->dpia_bw_alloc_config.link_max_bw = 0;
+	link->dpia_bw_alloc_config.allocated_bw = 0;
 	link->dpia_bw_alloc_config.estimated_bw = 0;
 	link->dpia_bw_alloc_config.bw_granularity = 0;
+	link->dpia_bw_alloc_config.dp_overhead = 0;
 	link->dpia_bw_alloc_config.response_ready = false;
-	link->dpia_bw_alloc_config.sink_allocated_bw = 0;
+	link->dpia_bw_alloc_config.nrd_max_lane_count = 0;
+	link->dpia_bw_alloc_config.nrd_max_link_rate = 0;
+	for (int i = 0; i < MAX_SINKS_PER_LINK; i++)
+		link->dpia_bw_alloc_config.remote_sink_req_bw[i] = 0;
+	DC_LOG_DEBUG("reset usb4 bw alloc of link(%d)\n", link->link_index);
 }
 
 #define BW_GRANULARITY_0 4 // 0.25 Gbps
@@ -210,8 +216,8 @@ static int get_host_router_total_dp_tunnel_bw(const struct dc *dc, uint8_t hr_in
 				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;
+					total_bw += link_dpia_primary->dpia_bw_alloc_config.estimated_bw +
+						link_dpia_secondary->dpia_bw_alloc_config.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;
@@ -264,7 +270,7 @@ static void set_usb4_req_bw_req(struct dc_link *link, int req_bw)
 
 	/* 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) {
+	if (req_bw == link->dpia_bw_alloc_config.allocated_bw) {
 		DC_LOG_ERROR("%s: Request bw equals to allocated bw for link(%d)\n",
 			__func__, link->link_index);
 	}
@@ -387,9 +393,9 @@ void dpia_handle_bw_alloc_response(struct dc_link *link, uint8_t bw, uint8_t res
 		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);
+			__func__, link->dpia_bw_alloc_config.allocated_bw, bw_needed);
 
-		link->dpia_bw_alloc_config.sink_allocated_bw = bw_needed;
+		link->dpia_bw_alloc_config.allocated_bw = bw_needed;
 
 		link->dpia_bw_alloc_config.response_ready = true;
 		break;
@@ -427,8 +433,8 @@ int dpia_handle_usb4_bandwidth_allocation_for_link(struct dc_link *link, int pea
 	if (link->hpd_status && peak_bw > 0) {
 
 		// If DP over USB4 then we need to check BW allocation
-		link->dpia_bw_alloc_config.sink_max_bw = peak_bw;
-		set_usb4_req_bw_req(link, link->dpia_bw_alloc_config.sink_max_bw);
+		link->dpia_bw_alloc_config.link_max_bw = peak_bw;
+		set_usb4_req_bw_req(link, link->dpia_bw_alloc_config.link_max_bw);
 
 		do {
 			if (timeout > 0)
@@ -440,8 +446,8 @@ int dpia_handle_usb4_bandwidth_allocation_for_link(struct dc_link *link, int pea
 
 		if (!timeout)
 			ret = 0;// ERROR TIMEOUT waiting for response for allocating bw
-		else if (link->dpia_bw_alloc_config.sink_allocated_bw > 0)
-			ret = link->dpia_bw_alloc_config.sink_allocated_bw;
+		else if (link->dpia_bw_alloc_config.allocated_bw > 0)
+			ret = link->dpia_bw_alloc_config.allocated_bw;
 	}
 	//2. Cold Unplug
 	else if (!link->hpd_status)
@@ -450,7 +456,6 @@ int dpia_handle_usb4_bandwidth_allocation_for_link(struct dc_link *link, int pea
 out:
 	return ret;
 }
-
 bool link_dp_dpia_allocate_usb4_bandwidth_for_stream(struct dc_link *link, int req_bw)
 {
 	bool ret = false;
@@ -458,7 +463,7 @@ bool link_dp_dpia_allocate_usb4_bandwidth_for_stream(struct dc_link *link, int r
 
 	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);
+		link->dpia_bw_alloc_config.allocated_bw, req_bw);
 
 	if (!get_bw_alloc_proceed_flag(link))
 		goto out;
@@ -523,3 +528,30 @@ bool dpia_validate_usb4_bw(struct dc_link **link, int *bw_needed_per_dpia, const
 
 	return ret;
 }
+
+int link_dp_dpia_get_dp_overhead_in_dp_tunneling(struct dc_link *link)
+{
+	int dp_overhead = 0, link_mst_overhead = 0;
+
+	if (!get_bw_alloc_proceed_flag((link)))
+		return dp_overhead;
+
+	/* if its mst link, add MTPH overhead */
+	if ((link->type == dc_connection_mst_branch) &&
+		!link->dpcd_caps.channel_coding_cap.bits.DP_128b_132b_SUPPORTED) {
+		/* For 8b/10b encoding: MTP is 64 time slots long, slot 0 is used for MTPH
+		 * MST overhead is 1/64 of link bandwidth (excluding any overhead)
+		 */
+		const struct dc_link_settings *link_cap =
+			dc_link_get_link_cap(link);
+		uint32_t link_bw_in_kbps = (uint32_t)link_cap->link_rate *
+					   (uint32_t)link_cap->lane_count *
+					   LINK_RATE_REF_FREQ_IN_KHZ * 8;
+		link_mst_overhead = (link_bw_in_kbps / 64) + ((link_bw_in_kbps % 64) ? 1 : 0);
+	}
+
+	/* add all the overheads */
+	dp_overhead = link_mst_overhead;
+
+	return dp_overhead;
+}
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 981bc4eb6120..3b6d8494f9d5 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
@@ -99,4 +99,13 @@ void dpia_handle_bw_alloc_response(struct dc_link *link, uint8_t bw, uint8_t res
  */
 bool dpia_validate_usb4_bw(struct dc_link **link, int *bw_needed, const unsigned int num_dpias);
 
+/*
+ * Obtain all the DP overheads in dp tunneling for the dpia link
+ *
+ * @link: pointer to the dc_link struct instance
+ *
+ * return: DP overheads in DP tunneling
+ */
+int link_dp_dpia_get_dp_overhead_in_dp_tunneling(struct dc_link *link);
+
 #endif /* DC_INC_LINK_DP_DPIA_BW_H_ */
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 7f1196528218..046d3e205415 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
@@ -930,8 +930,8 @@ bool edp_get_replay_state(const struct dc_link *link, uint64_t *state)
 bool edp_setup_replay(struct dc_link *link, const struct dc_stream_state *stream)
 {
 	/* To-do: Setup Replay */
-	struct dc *dc = link->ctx->dc;
-	struct dmub_replay *replay = dc->res_pool->replay;
+	struct dc *dc;
+	struct dmub_replay *replay;
 	int i;
 	unsigned int panel_inst;
 	struct replay_context replay_context = { 0 };
@@ -947,6 +947,10 @@ bool edp_setup_replay(struct dc_link *link, const struct dc_stream_state *stream
 	if (!link)
 		return false;
 
+	dc = link->ctx->dc;
+
+	replay = dc->res_pool->replay;
+
 	if (!replay)
 		return false;
 
@@ -975,8 +979,7 @@ bool edp_setup_replay(struct dc_link *link, const struct dc_stream_state *stream
 
 	replay_context.line_time_in_ns = lineTimeInNs;
 
-	if (replay)
-		link->replay_settings.replay_feature_enabled =
+	link->replay_settings.replay_feature_enabled =
 			replay->funcs->replay_copy_settings(replay, link, &replay_context, panel_inst);
 	if (link->replay_settings.replay_feature_enabled) {