bnx2x: Look inside vlan when determining checksum proto.
authorHao Zheng <hzheng@nicira.com>
Thu, 11 Nov 2010 13:47:58 +0000 (13:47 +0000)
committerDavid S. Miller <davem@davemloft.net>
Fri, 12 Nov 2010 20:30:57 +0000 (12:30 -0800)
Currently the skb->protocol field is used to setup checksum
offloading on transmit for the correct protocol.  However, if
vlan offloading is disabled or otherwise not used, the protocol
field will be ETH_P_8021Q, not the actual protocol.  This will
cause the checksum to be not computed correctly, even though the
hardware is capable of looking inside vlan tags.  Instead,
look inside the header if necessary to determine the correct
protocol type.

To some extent this fixes a regression from 2.6.36 because it
was previously not possible to disable vlan offloading and this
error case was not exposed.

Signed-off-by: Hao Zheng <hzheng@nicira.com>
CC: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/bnx2x/bnx2x_cmn.c

index 459614d..94d5f59 100644 (file)
@@ -1680,7 +1680,7 @@ static inline u32 bnx2x_xmit_type(struct bnx2x *bp, struct sk_buff *skb)
                rc = XMIT_PLAIN;
 
        else {
-               if (skb->protocol == htons(ETH_P_IPV6)) {
+               if (vlan_get_protocol(skb) == htons(ETH_P_IPV6)) {
                        rc = XMIT_CSUM_V6;
                        if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
                                rc |= XMIT_CSUM_TCP;