net: stmmac: Implement UDP Segmentation Offload

Implement the UDP Segmentation Offload feature in stmmac. This is only
available in GMAC4+ cores.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c

@@ -36,6 +36,7 @@
 #endif /* CONFIG_DEBUG_FS */
 #include <linux/net_tstamp.h>
 #include <linux/phylink.h>
+#include <linux/udp.h>
 #include <net/pkt_cls.h>
 #include "stmmac_ptp.h"
 #include "stmmac.h"
@@ -2916,9 +2917,9 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
 	u32 queue = skb_get_queue_mapping(skb);
 	struct stmmac_tx_queue *tx_q;
 	unsigned int first_entry;
+	u8 proto_hdr_len, hdr;
 	int tmp_pay_len = 0;
 	u32 pay_len, mss;
-	u8 proto_hdr_len;
 	dma_addr_t des;
 	bool has_vlan;
 	int i;
@@ -2926,7 +2927,13 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
 	tx_q = &priv->tx_queue[queue];
 
 	/* Compute header lengths */
+	if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) {
+		proto_hdr_len = skb_transport_offset(skb) + sizeof(struct udphdr);
+		hdr = sizeof(struct udphdr);
+	} else {
 		proto_hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
+		hdr = tcp_hdrlen(skb);
+	}
 
 	/* Desc availability based on threshold should be enough safe */
 	if (unlikely(stmmac_tx_avail(priv, queue) <
@@ -2956,8 +2963,8 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
 	}
 
 	if (netif_msg_tx_queued(priv)) {
-		pr_info("%s: tcphdrlen %d, hdr_len %d, pay_len %d, mss %d\n",
-			__func__, tcp_hdrlen(skb), proto_hdr_len, pay_len, mss);
+		pr_info("%s: hdrlen %d, hdr_len %d, pay_len %d, mss %d\n",
+			__func__, hdr, proto_hdr_len, pay_len, mss);
 		pr_info("\tskb->len %d, skb->data_len %d\n", skb->len,
 			skb->data_len);
 	}
@@ -3071,7 +3078,7 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
 			proto_hdr_len,
 			pay_len,
 			1, tx_q->tx_skbuff_dma[first_entry].last_segment,
-			tcp_hdrlen(skb) / 4, (skb->len - proto_hdr_len));
+			hdr / 4, (skb->len - proto_hdr_len));
 
 	/* If context desc is used to change MSS */
 	if (mss_desc) {
@@ -3130,6 +3137,7 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
 	int i, csum_insertion = 0, is_jumbo = 0;
 	u32 queue = skb_get_queue_mapping(skb);
 	int nfrags = skb_shinfo(skb)->nr_frags;
+	int gso = skb_shinfo(skb)->gso_type;
 	struct dma_desc *desc, *first;
 	struct stmmac_tx_queue *tx_q;
 	unsigned int first_entry;
@@ -3145,7 +3153,9 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	/* Manage oversized TCP frames for GMAC4 device */
 	if (skb_is_gso(skb) && priv->tso) {
-		if (skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))
+		if (gso & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))
+			return stmmac_tso_xmit(skb, dev);
+		if (priv->plat->has_gmac4 && (gso & SKB_GSO_UDP_L4))
 			return stmmac_tso_xmit(skb, dev);
 	}
 
@@ -4036,11 +4046,13 @@ static int stmmac_setup_tc(struct net_device *ndev, enum tc_setup_type type,
 static u16 stmmac_select_queue(struct net_device *dev, struct sk_buff *skb,
 			       struct net_device *sb_dev)
 {
-	if (skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
+	int gso = skb_shinfo(skb)->gso_type;
+
+	if (gso & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6 | SKB_GSO_UDP_L4)) {
 		/*
-		 * There is no way to determine the number of TSO
+		 * There is no way to determine the number of TSO/USO
 		 * capable Queues. Let's use always the Queue 0
-		 * because if TSO is supported then at least this
+		 * because if TSO/USO is supported then at least this
 		 * one will be capable.
 		 */
 		return 0;
@@ -4555,6 +4567,8 @@ int stmmac_dvr_probe(struct device *device,
 
 	if ((priv->plat->tso_en) && (priv->dma_cap.tsoen)) {
 		ndev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
+		if (priv->plat->has_gmac4)
+			ndev->hw_features |= NETIF_F_GSO_UDP_L4;
 		priv->tso = true;
 		dev_info(priv->device, "TSO feature enabled\n");
 	}