| Message ID | 20180804204127.GC15082@ZenIV.linux.org.uk |
|---|---|
| State | Accepted, archived |
| Delegated to: | David Miller |
| Headers | show |
| Series | mellanox: fix the dport endianness in call of __inet6_lookup_established() | expand |
From: Al Viro <viro@ZenIV.linux.org.uk> Date: Sat, 4 Aug 2018 21:41:27 +0100 > __inet6_lookup_established() expect th->dport passed in host-endian, > not net-endian. The reason is microoptimization in __inet6_lookup(), > but if you use the lower-level helpers, you have to play by their > rules... > > Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Mellanox folks, please review.
On 8/6/2018 3:27 AM, David Miller wrote: > From: Al Viro <viro@ZenIV.linux.org.uk> > Date: Sat, 4 Aug 2018 21:41:27 +0100 > >> __inet6_lookup_established() expect th->dport passed in host-endian, >> not net-endian. The reason is microoptimization in __inet6_lookup(), >> but if you use the lower-level helpers, you have to play by their >> rules... >> >> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> > > Mellanox folks, please review. > Looks good to me. Thanks Al.
From: Al Viro <viro@ZenIV.linux.org.uk> Date: Sat, 4 Aug 2018 21:41:27 +0100 > __inet6_lookup_established() expect th->dport passed in host-endian, > not net-endian. The reason is microoptimization in __inet6_lookup(), > but if you use the lower-level helpers, you have to play by their > rules... > > Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Applied to net-next, thanks Al.
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/tls_rxtx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/tls_rxtx.c index 92d37459850e..be137d4a9169 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/tls_rxtx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/tls_rxtx.c @@ -328,7 +328,7 @@ static int tls_update_resync_sn(struct net_device *netdev, sk = __inet6_lookup_established(dev_net(netdev), &tcp_hashinfo, &ipv6h->saddr, th->source, - &ipv6h->daddr, th->dest, + &ipv6h->daddr, ntohs(th->dest), netdev->ifindex, 0); #endif }
__inet6_lookup_established() expect th->dport passed in host-endian, not net-endian. The reason is microoptimization in __inet6_lookup(), but if you use the lower-level helpers, you have to play by their rules... Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> ---