net: stmmac: Fix incorrect IP-csum insertion algo

Currently the IP checksum insertion on Tx is implied to be performed by HW
for any IPv4/IPv6 network protocol types. It's definitely incorrect since
based on [1,2,3] Tx COE supports the TCP/UDP/ICMP payloads only. Moreover
the IPv6 extension headers are limited to be a single Hop-by-Hop and
multiple Destination headers. From that perspective the commit
c2945c435c99 ("net: stmmac: Prevent DSA tags from breaking COE") wasn't
fully correct in providing just IP-type constraint.

Let's fix the problem by providing a comprehensive method which would make
sure that the passed frame can be checksummed by the controller Tx
Checksum Offload Engine.

[1] DesignWare Cores Ethernet MAC Universal Databook, Revision 3.73a,
    October 2013, p. 187.
[2] DesignWare Cores Ethernet Quality-of-Service Databook, Revision 5.20a,
    April 2020, pp. 366.
[3] DesignWare Cores XGMAC - 10G Ethernet MAC Databook, Revision 3.20a,
    September 2022, p.251.

Fixes: c2945c435c99 ("net: stmmac: Prevent DSA tags from breaking COE")
Signed-off-by: Serge Semin <fancer.lancer@gmail.com>

1 files changed, 95 insertions, 42 deletions

diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index d74927fbf148..f0d395b86319 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -18,6 +18,7 @@
 #include <linux/kernel.h>
 #include <linux/interrupt.h>
 #include <linux/ip.h>
+#include <linux/ipv6.h>
 #include <linux/tcp.h>
 #include <linux/skbuff.h>
 #include <linux/ethtool.h>
@@ -41,6 +42,7 @@
 #include <linux/udp.h>
 #include <linux/bpf_trace.h>
 #include <net/devlink.h>
+#include <net/ipv6.h>
 #include <net/page_pool/helpers.h>
 #include <net/pkt_cls.h>
 #include <net/xdp_sock_drv.h>
@@ -4309,6 +4311,91 @@ static bool stmmac_vlan_insert(struct stmmac_priv *priv, struct sk_buff *skb,
 }
 
 /**
+ * stmmac_csum_insert - Insert CSUM into the frame
+ * @priv:  driver private structure
+ * @skb :  the socket buffer
+ * @queue: TX queue index
+ * Description :
+ * This function will either insert all the uninitialized checksum into the
+ * frame or return a request for the Tx COE to do that if it's supported.
+ *
+ * Return: zero if CSUM was inserted into the frame, one if Tx COE is supposed
+ * to do that, negative error otherwise.
+ */
+static int stmmac_csum_insert(struct stmmac_priv *priv, struct sk_buff *skb,
+			      int queue)
+{
+	struct ipv6_opt_hdr *hp;
+	__be16 l3proto;
+	int depth = 0;
+	u8 l4proto;
+
+	/* No checksum required */
+	if (!(skb->ip_summed == CHECKSUM_PARTIAL))
+		return 0;
+
+	/* Tx COE might be available on a limited set of the queues */
+	if (priv->plat->tx_queues_cfg[queue].coe_unsupported)
+		goto sw_checksum;
+
+	/* Tx COE supports Ethernet II frames only (EtherType >= 0x0600) */
+	l3proto = eth_header_parse_protocol(skb);
+	if (!eth_proto_is_802_3(l3proto))
+		goto sw_checksum;
+
+	/* Tx COE supports C-/S-VLAN tagged IP-packets (excl. hwaccel tag?) */
+	l3proto = __vlan_get_protocol(skb, l3proto, &depth);
+	if (depth > (ETH_HLEN + VLAN_HLEN))
+		goto sw_checksum;
+
+	/* Tx COE supports IPv4/IPv6 types with TCP/UDP/ICMP protocols */
+	if ((priv->dev->features & NETIF_F_IP_CSUM) &&
+	    (l3proto == htons(ETH_P_IP))) {
+		l4proto = ip_hdr(skb)->protocol;
+
+		switch(l4proto) {
+		case IPPROTO_ICMP:
+		case IPPROTO_TCP:
+		case IPPROTO_UDP:
+			return 1;
+		default:
+			goto sw_checksum;
+		}
+	} else if ((priv->dev->features & NETIF_F_IPV6_CSUM) &&
+		   (l3proto == htons(ETH_P_IPV6))) {
+		l4proto = ipv6_hdr(skb)->nexthdr;
+		depth = sizeof(struct ipv6hdr);
+
+		/* Tx COE supports single Hop-by-Hop and multiple Dest headers */
+		do {
+			switch(l4proto) {
+			case NEXTHDR_TCP:
+			case NEXTHDR_UDP:
+			case NEXTHDR_ICMP:
+				return 1;
+			case NEXTHDR_HOP:
+				if (depth != sizeof(struct ipv6hdr))
+					goto sw_checksum;
+				break;
+			case NEXTHDR_DEST:
+				break;
+			default:
+				goto sw_checksum;
+			}
+
+			hp = (struct ipv6_opt_hdr *)(skb_network_header(skb) +
+						     depth);
+			l4proto = hp->nexthdr;
+			depth += ipv6_optlen(hp);
+		} while (l4proto != NEXTHDR_NONE);
+	}
+
+sw_checksum:
+	return unlikely(skb_csum_is_sctp(skb)) ?
+	       skb_crc32c_csum_help(skb) : skb_checksum_help(skb);
+}
+
+/**
  *  stmmac_tso_allocator - close entry point of the driver
  *  @priv: driver private structure
  *  @des: buffer start address
@@ -4663,28 +4750,6 @@ flush_ring:
 }
 
 /**
- * stmmac_has_ip_ethertype() - Check if packet has IP ethertype
- * @skb: socket buffer to check
- *
- * Check if a packet has an ethertype that will trigger the IP header checks
- * and IP/TCP checksum engine of the stmmac core.
- *
- * Return: true if the ethertype can trigger the checksum engine, false
- * otherwise
- */
-static bool stmmac_has_ip_ethertype(struct sk_buff *skb)
-{
-	int depth = 0;
-	__be16 proto;
-
-	proto = __vlan_get_protocol(skb, eth_header_parse_protocol(skb),
-				    &depth);
-
-	return (depth <= ETH_HLEN) &&
-		(proto == htons(ETH_P_IP) || proto == htons(ETH_P_IPV6));
-}
-
-/**
  *  stmmac_xmit - Tx entry point of the driver
  *  @skb : the socket buffer
  *  @dev : device pointer
@@ -4755,22 +4820,11 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
 	first_entry = entry;
 	WARN_ON(tx_q->tx_skbuff[first_entry]);
 
-	csum_insertion = (skb->ip_summed == CHECKSUM_PARTIAL);
-	/* DWMAC IPs can be synthesized to support tx coe only for a few tx
-	 * queues. In that case, checksum offloading for those queues that don't
-	 * support tx coe needs to fallback to software checksum calculation.
-	 *
-	 * Packets that won't trigger the COE e.g. most DSA-tagged packets will
-	 * also have to be checksummed in software.
-	 */
-	if (csum_insertion &&
-	    (priv->plat->tx_queues_cfg[queue].coe_unsupported ||
-	     !stmmac_has_ip_ethertype(skb))) {
-		if (unlikely(skb_checksum_help(skb))) {
-			drop = true;
-			goto flush_ring;
-		}
-		csum_insertion = !csum_insertion;
+	/* Check if CSUM should be inserted by HW */
+	csum_insertion = stmmac_csum_insert(priv, skb, queue);
+	if (csum_insertion < 0) {
+		drop = true;
+		goto flush_ring;
 	}
 
 	if (likely(priv->extend_desc))
@@ -6178,9 +6232,6 @@ static netdev_features_t stmmac_fix_features(struct net_device *dev,
 {
 	struct stmmac_priv *priv = netdev_priv(dev);
 
-	if (!priv->plat->tx_coe)
-		features &= ~NETIF_F_CSUM_MASK;
-
 	/* Some GMAC devices have a bugged Jumbo frame support that
 	 * needs to have the Tx COE disabled for oversized frames
 	 * (due to limited buffer sizes). In this case we disable
@@ -8258,10 +8309,12 @@ static int __stmmac_dvr_probe(struct device *device,
 	ndev->xdp_metadata_ops = &stmmac_xdp_metadata_ops;
 	ndev->xsk_tx_metadata_ops = &stmmac_xsk_tx_metadata_ops;
 
-	ndev->hw_features = NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
 	ndev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
 			     NETDEV_XDP_ACT_XSK_ZEROCOPY;
 
+	if (priv->plat->tx_coe)
+		ndev->hw_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+
 	if (priv->plat->rx_coe)
 		ndev->hw_features |= NETIF_F_RXCSUM;