From: Andy Gospodarek <gospo@redhat.com>
Date: Wed, 13 Feb 2008 21:44:51 -0500
Subject: [net] ixgbe: obtain correct protocol info on xmit
Message-id: 20080214024451.GW31161@gospo.usersys.redhat.com
O-Subject: [RHEL5.2 PATCH] ixgbe: correctly obtain protocol information on transmit
Bugzilla: 428230

Mark Wagner discovered that Xen guests that send traffic through ixgbe
interfaces would cause a panic unless tx checksumming was disabled on
the interface.  This severely limits performance and since the default
has it enabled you will quickly panic the the box.  This isn't ideal, so
I dug around a vmcore he provided and discovered that the driver
incorrectly gathered packet protocol information from socket information
stored in the skb.

This works well on hosts that are generating the traffic, but when you
put the ixgbe interface in a bridge, there is no longer any socket info
for outgoing packets (since they came from the bridging interface).  I
proposed a smaller fix for this upstream, but intel already had one in
their internal driver, so Auke posted this patch.

I've tested it and it works well.  This will resolve BZ 428230.

Acked-by: "David S. Miller" <davem@redhat.com>

diff --git a/drivers/net/ixgbe/ixgbe_main.c b/drivers/net/ixgbe/ixgbe_main.c
index 70775af..c3533df 100644
--- a/drivers/net/ixgbe/ixgbe_main.c
+++ b/drivers/net/ixgbe/ixgbe_main.c
@@ -2206,11 +2206,29 @@ static bool ixgbe_tx_csum(struct ixgbe_adapter *adapter,
 				    IXGBE_ADVTXD_DTYP_CTXT);
 
 		if (skb->ip_summed == CHECKSUM_PARTIAL) {
-			if (skb->protocol == ntohs(ETH_P_IP))
+			switch (skb->protocol) {
+			case __constant_htons(ETH_P_IP):
 				type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4;
+				if (ip_hdr(skb)->protocol == IPPROTO_TCP)
+					type_tucmd_mlhl |=
+						IXGBE_ADVTXD_TUCMD_L4T_TCP;
+				break;
+
+			case __constant_htons(ETH_P_IPV6):
+				/* XXX what about other V6 headers?? */
+				if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
+					type_tucmd_mlhl |=
+						IXGBE_ADVTXD_TUCMD_L4T_TCP;
+				break;
 
-			if (skb->sk->sk_protocol == IPPROTO_TCP)
-				type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
+			default:
+				if (unlikely(net_ratelimit())) {
+					DPRINTK(PROBE, WARNING,
+					 "partial checksum but proto=%x!\n",
+					 skb->protocol);
+				}
+				break;
+			}
 		}
 
 		context_desc->type_tucmd_mlhl = cpu_to_le32(type_tucmd_mlhl);