net/tcp: Fix socket lookups with SO_BINDTODEVICE
authorDavid Ahern <dsahern@gmail.com>
Mon, 18 Jun 2018 19:30:37 +0000 (12:30 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 22 Jul 2018 12:28:46 +0000 (14:28 +0200)
commit35e324ebeee0fed7d7d30b6ed9873a3f06f86e94
tree6cee1f8a95e312f8ad8e5022184ecc9c5ad0ac86
parentb3c66b54d8fef1d032580d3f7660279cdd277576
net/tcp: Fix socket lookups with SO_BINDTODEVICE

[ Upstream commit 8c43bd1706885ba1acfa88da02bc60a2ec16f68c ]

Similar to 69678bcd4d2d ("udp: fix SO_BINDTODEVICE"), TCP socket lookups
need to fail if dev_match is not true. Currently, a packet to a given port
can match a socket bound to device when it should not. In the VRF case,
this causes the lookup to hit a VRF socket and not a global socket
resulting in a response trying to go through the VRF when it should not.

Fixes: 3fa6f616a7a4d ("net: ipv4: add second dif to inet socket lookups")
Fixes: 4297a0ef08572 ("net: ipv6: add second dif to inet6 socket lookups")
Reported-by: Lou Berger <lberger@labn.net>
Diagnosed-by: Renato Westphal <renato@opensourcerouting.org>
Tested-by: Renato Westphal <renato@opensourcerouting.org>
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/ipv4/inet_hashtables.c
net/ipv6/inet6_hashtables.c