tcp: don't annotate mark on control socket from tcp_v6_send_response()
authorPablo Neira <pablo@netfilter.org>
Thu, 26 Jan 2017 21:56:21 +0000 (22:56 +0100)
committerDavid S. Miller <davem@davemloft.net>
Fri, 27 Jan 2017 15:33:56 +0000 (10:33 -0500)
commit92e55f412cffd016cc245a74278cb4d7b89bb3bc
treeab90e2e59c82daae60d2c953ef6c780e1f6e5550
parenta47b70ea86bdeb3091341f5ae3ef580f1a1ad822
tcp: don't annotate mark on control socket from tcp_v6_send_response()

Unlike ipv4, this control socket is shared by all cpus so we cannot use
it as scratchpad area to annotate the mark that we pass to ip6_xmit().

Add a new parameter to ip6_xmit() to indicate the mark. The SCTP socket
family caches the flowi6 structure in the sctp_transport structure, so
we cannot use to carry the mark unless we later on reset it back, which
I discarded since it looks ugly to me.

Fixes: bf99b4ded5f8 ("tcp: fix mark propagation with fwmark_reflect enabled")
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ipv6.h
net/dccp/ipv6.c
net/ipv6/inet6_connection_sock.c
net/ipv6/ip6_output.c
net/ipv6/tcp_ipv6.c
net/sctp/ipv6.c