usb: xhci-mtk: improve bandwidth scheduling

Mainly improve SuperSpeed ISOC bandwidth in last microframe,
and LowSpeed/FullSpeed IN INT/ISOC bandwidth in split and
idle microframes by introduing a bandwidth budget table;

Signed-off-by: Chunfeng Yun <[e-mail]>
Signed-off-by: Mathias Nyman <[e-mail]>
Signed-off-by: Greg Kroah-Hartman <[e-mail]>

drivers/usb/host/xhci-mtk-sch.c

···
 #define HS_BW_BOUNDARY	6144
 /* usb2 spec section11.18.1: at most 188 FS bytes per microframe */
 #define FS_PAYLOAD_MAX 188
+/*
+ * max number of microframes for split transfer,
+ * for fs isoc in : 1 ss + 1 idle + 7 cs
+ */
+#define TT_MICROFRAMES_MAX 9
 
 /* mtk scheduler bitmasks */
 #define EP_BPKTS(p)	((p) & 0x3f)
···
 	return bw_index;
 }
 
+static u32 get_esit(struct xhci_ep_ctx *ep_ctx)
+{
+	u32 esit;
+
+	esit = 1 << CTX_TO_EP_INTERVAL(le32_to_cpu(ep_ctx->ep_info));
+	if (esit > XHCI_MTK_MAX_ESIT)
+		esit = XHCI_MTK_MAX_ESIT;
+
+	return esit;
+}
+
+static struct mu3h_sch_ep_info *create_sch_ep(struct usb_device *udev,
+	struct usb_host_endpoint *ep, struct xhci_ep_ctx *ep_ctx)
+{
+	struct mu3h_sch_ep_info *sch_ep;
+	u32 len_bw_budget_table;
+	size_t mem_size;
+
+	if (is_fs_or_ls(udev->speed))
+		len_bw_budget_table = TT_MICROFRAMES_MAX;
+	else if ((udev->speed == USB_SPEED_SUPER)
+			&& usb_endpoint_xfer_isoc(&ep->desc))
+		len_bw_budget_table = get_esit(ep_ctx);
+	else
+		len_bw_budget_table = 1;
+
+	mem_size = sizeof(struct mu3h_sch_ep_info) +
+			len_bw_budget_table * sizeof(u32);
+	sch_ep = kzalloc(mem_size, GFP_KERNEL);
+	if (!sch_ep)
+		return ERR_PTR(-ENOMEM);
+
+	sch_ep->ep = ep;
+
+	return sch_ep;
+}
+
 static void setup_sch_info(struct usb_device *udev,
 		struct xhci_ep_ctx *ep_ctx, struct mu3h_sch_ep_info *sch_ep)
 {
 	u32 ep_type;
-	u32 ep_interval;
-	u32 max_packet_size;
+	u32 maxpkt;
 	u32 max_burst;
 	u32 mult;
 	u32 esit_pkts;
 	u32 max_esit_payload;
+	u32 *bwb_table = sch_ep->bw_budget_table;
+	int i;
 
 	ep_type = CTX_TO_EP_TYPE(le32_to_cpu(ep_ctx->ep_info2));
-	ep_interval = CTX_TO_EP_INTERVAL(le32_to_cpu(ep_ctx->ep_info));
-	max_packet_size = MAX_PACKET_DECODED(le32_to_cpu(ep_ctx->ep_info2));
+	maxpkt = MAX_PACKET_DECODED(le32_to_cpu(ep_ctx->ep_info2));
 	max_burst = CTX_TO_MAX_BURST(le32_to_cpu(ep_ctx->ep_info2));
 	mult = CTX_TO_EP_MULT(le32_to_cpu(ep_ctx->ep_info));
 	max_esit_payload =
···
 			le32_to_cpu(ep_ctx->ep_info)) << 16) |
 		 CTX_TO_MAX_ESIT_PAYLOAD(le32_to_cpu(ep_ctx->tx_info));
 
-	sch_ep->esit = 1 << ep_interval;
+	sch_ep->esit = get_esit(ep_ctx);
 	sch_ep->offset = 0;
 	sch_ep->burst_mode = 0;
+	sch_ep->repeat = 0;
 
 	if (udev->speed == USB_SPEED_HIGH) {
 		sch_ep->cs_count = 0;
···
 		 * in a interval
 		 */
 		sch_ep->num_budget_microframes = 1;
-		sch_ep->repeat = 0;
 
 		/*
 		 * xHCI spec section6.2.3.4
···
 		 * opportunities per microframe
 		 */
 		sch_ep->pkts = max_burst + 1;
-		sch_ep->bw_cost_per_microframe = max_packet_size * sch_ep->pkts;
+		sch_ep->bw_cost_per_microframe = maxpkt * sch_ep->pkts;
+		bwb_table[0] = sch_ep->bw_cost_per_microframe;
 	} else if (udev->speed == USB_SPEED_SUPER) {
 		/* usb3_r1 spec section4.4.7 & 4.4.8 */
 		sch_ep->cs_count = 0;
+		sch_ep->burst_mode = 1;
 		/*
 		 * some device's (d)wBytesPerInterval is set as 0,
 		 * then max_esit_payload is 0, so evaluate esit_pkts from
 		 * mult and burst
 		 */
-		esit_pkts = DIV_ROUND_UP(max_esit_payload, max_packet_size);
+		esit_pkts = DIV_ROUND_UP(max_esit_payload, maxpkt);
 		if (esit_pkts == 0)
 			esit_pkts = (mult + 1) * (max_burst + 1);
 
 		if (ep_type == INT_IN_EP || ep_type == INT_OUT_EP) {
 			sch_ep->pkts = esit_pkts;
 			sch_ep->num_budget_microframes = 1;
-			sch_ep->repeat = 0;
+			bwb_table[0] = maxpkt * sch_ep->pkts;
 		}
 
 		if (ep_type == ISOC_IN_EP || ep_type == ISOC_OUT_EP) {
+			u32 remainder;
+
 			if (sch_ep->esit == 1)
 				sch_ep->pkts = esit_pkts;
 			else if (esit_pkts <= sch_ep->esit)
···
 			sch_ep->num_budget_microframes =
 				DIV_ROUND_UP(esit_pkts, sch_ep->pkts);
 
-			if (sch_ep->num_budget_microframes > 1)
-				sch_ep->repeat = 1;
-			else
-				sch_ep->repeat = 0;
+			sch_ep->repeat = !!(sch_ep->num_budget_microframes > 1);
+			sch_ep->bw_cost_per_microframe = maxpkt * sch_ep->pkts;
+
+			remainder = sch_ep->bw_cost_per_microframe;
+			remainder *= sch_ep->num_budget_microframes;
+			remainder -= (maxpkt * esit_pkts);
+			for (i = 0; i < sch_ep->num_budget_microframes - 1; i++)
+				bwb_table[i] = sch_ep->bw_cost_per_microframe;
+
+			/* last one <= bw_cost_per_microframe */
+			bwb_table[i] = remainder;
 		}
-		sch_ep->bw_cost_per_microframe = max_packet_size * sch_ep->pkts;
 	} else if (is_fs_or_ls(udev->speed)) {
-
-		/*
-		 * usb_20 spec section11.18.4
-		 * assume worst cases
-		 */
-		sch_ep->repeat = 0;
 		sch_ep->pkts = 1; /* at most one packet for each microframe */
-		if (ep_type == INT_IN_EP || ep_type == INT_OUT_EP) {
-			sch_ep->cs_count = 3; /* at most need 3 CS*/
-			/* one for SS and one for budgeted transaction */
-			sch_ep->num_budget_microframes = sch_ep->cs_count + 2;
-			sch_ep->bw_cost_per_microframe = max_packet_size;
-		}
-		if (ep_type == ISOC_OUT_EP) {
+		sch_ep->cs_count = DIV_ROUND_UP(maxpkt, FS_PAYLOAD_MAX);
+		sch_ep->num_budget_microframes = sch_ep->cs_count + 2;
+		sch_ep->bw_cost_per_microframe =
+			(maxpkt < FS_PAYLOAD_MAX) ? maxpkt : FS_PAYLOAD_MAX;
 
-			/*
-			 * the best case FS budget assumes that 188 FS bytes
-			 * occur in each microframe
-			 */
-			sch_ep->num_budget_microframes = DIV_ROUND_UP(
-				max_packet_size, FS_PAYLOAD_MAX);
-			sch_ep->bw_cost_per_microframe = FS_PAYLOAD_MAX;
-			sch_ep->cs_count = sch_ep->num_budget_microframes;
-		}
-		if (ep_type == ISOC_IN_EP) {
-			/* at most need additional two CS. */
-			sch_ep->cs_count = DIV_ROUND_UP(
-				max_packet_size, FS_PAYLOAD_MAX) + 2;
-			sch_ep->num_budget_microframes = sch_ep->cs_count + 2;
-			sch_ep->bw_cost_per_microframe = FS_PAYLOAD_MAX;
+		/* init budget table */
+		if (ep_type == ISOC_OUT_EP) {
+			for (i = 0; i < sch_ep->num_budget_microframes; i++)
+				bwb_table[i] =	sch_ep->bw_cost_per_microframe;
+		} else if (ep_type == INT_OUT_EP) {
+			/* only first one consumes bandwidth, others as zero */
+			bwb_table[0] = sch_ep->bw_cost_per_microframe;
+		} else { /* INT_IN_EP or ISOC_IN_EP */
+			bwb_table[0] = 0; /* start split */
+			bwb_table[1] = 0; /* idle */
+			for (i = 2; i < sch_ep->num_budget_microframes; i++)
+				bwb_table[i] =	sch_ep->bw_cost_per_microframe;
 		}
 	}
 }
···
 {
 	u32 num_esit;
 	u32 max_bw = 0;
+	u32 bw;
 	int i;
 	int j;
 
···
 		u32 base = offset + i * sch_ep->esit;
 
 		for (j = 0; j < sch_ep->num_budget_microframes; j++) {
-			if (sch_bw->bus_bw[base + j] > max_bw)
-				max_bw = sch_bw->bus_bw[base + j];
+			bw = sch_bw->bus_bw[base + j] +
+					sch_ep->bw_budget_table[j];
+			if (bw > max_bw)
+				max_bw = bw;
 		}
 	}
 	return max_bw;
 }
 
 static void update_bus_bw(struct mu3h_sch_bw_info *sch_bw,
-	struct mu3h_sch_ep_info *sch_ep, int bw_cost)
+	struct mu3h_sch_ep_info *sch_ep, bool used)
 {
 	u32 num_esit;
 	u32 base;
···
 	num_esit = XHCI_MTK_MAX_ESIT / sch_ep->esit;
 	for (i = 0; i < num_esit; i++) {
 		base = sch_ep->offset + i * sch_ep->esit;
-		for (j = 0; j < sch_ep->num_budget_microframes; j++)
-			sch_bw->bus_bw[base + j] += bw_cost;
+		for (j = 0; j < sch_ep->num_budget_microframes; j++) {
+			if (used)
+				sch_bw->bus_bw[base + j] +=
+					sch_ep->bw_budget_table[j];
+			else
+				sch_bw->bus_bw[base + j] -=
+					sch_ep->bw_budget_table[j];
+		}
 	}
 }
 
···
 {
 	u32 offset;
 	u32 esit;
-	u32 num_budget_microframes;
 	u32 min_bw;
 	u32 min_index;
 	u32 worst_bw;
 	u32 bw_boundary;
 
-	if (sch_ep->esit > XHCI_MTK_MAX_ESIT)
-		sch_ep->esit = XHCI_MTK_MAX_ESIT;
-
 	esit = sch_ep->esit;
-	num_budget_microframes = sch_ep->num_budget_microframes;
 
 	/*
 	 * Search through all possible schedule microframes.
···
 	min_bw = ~0;
 	min_index = 0;
 	for (offset = 0; offset < esit; offset++) {
-		if ((offset + num_budget_microframes) > sch_ep->esit)
+		if ((offset + sch_ep->num_budget_microframes) > sch_ep->esit)
 			break;
 
 		/*
···
 				? SS_BW_BOUNDARY : HS_BW_BOUNDARY;
 
 	/* check bandwidth */
-	if (min_bw + sch_ep->bw_cost_per_microframe > bw_boundary)
+	if (min_bw > bw_boundary)
 		return -ERANGE;
 
 	/* update bus bandwidth info */
-	update_bus_bw(sch_bw, sch_ep, sch_ep->bw_cost_per_microframe);
+	update_bus_bw(sch_bw, sch_ep, 1);
 
 	return 0;
 }
···
 	bw_index = get_bw_index(xhci, udev, ep);
 	sch_bw = &sch_array[bw_index];
 
-	sch_ep = kzalloc(sizeof(struct mu3h_sch_ep_info), GFP_NOIO);
-	if (!sch_ep)
+	sch_ep = create_sch_ep(udev, ep, ep_ctx);
+	if (IS_ERR_OR_NULL(sch_ep))
 		return -ENOMEM;
 
 	setup_sch_info(udev, ep_ctx, sch_ep);
···
 	}
 
 	list_add_tail(&sch_ep->endpoint, &sch_bw->bw_ep_list);
-	sch_ep->ep = ep;
 
 	ep_ctx->reserved[0] |= cpu_to_le32(EP_BPKTS(sch_ep->pkts)
 		| EP_BCSCOUNT(sch_ep->cs_count) | EP_BBM(sch_ep->burst_mode));
···
 
 	list_for_each_entry(sch_ep, &sch_bw->bw_ep_list, endpoint) {
 		if (sch_ep->ep == ep) {
-			update_bus_bw(sch_bw, sch_ep,
-				-sch_ep->bw_cost_per_microframe);
+			update_bus_bw(sch_bw, sch_ep, 0);
 			list_del(&sch_ep->endpoint);
 			kfree(sch_ep);
 			break;

drivers/usb/host/xhci-mtk.h

···
  *		times; 1: distribute the (bMaxBurst+1)*(Mult+1) packets
  *		according to @pkts and @repeat. normal mode is used by
  *		default
+ * @bw_budget_table: table to record bandwidth budget per microframe
  */
 struct mu3h_sch_ep_info {
 	u32 esit;
···
 	u32 pkts;
 	u32 cs_count;
 	u32 burst_mode;
+	u32 bw_budget_table[0];
 };
 
 #define MU3C_U3_PORT_MAX 4