net/ipv4: Set oif in fib_compute_spec_dst
authorDavid Ahern <dsahern@gmail.com>
Sat, 7 Jul 2018 23:15:26 +0000 (16:15 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 25 Jul 2018 09:25:09 +0000 (11:25 +0200)
commit78382d78bb4a99ba312bfeef8b23006eb9f2e368
treec791474038675d296a053f8902769f476cd68334
parent829f4fd66354560fa66c83a3d199453036bef7ed
net/ipv4: Set oif in fib_compute_spec_dst

[ Upstream commit e7372197e15856ec4ee66b668020a662994db103 ]

Xin reported that icmp replies may not use the address on the device the
echo request is received if the destination address is broadcast. Instead
a route lookup is done without considering VRF context. Fix by setting
oif in flow struct to the master device if it is enslaved. That directs
the lookup to the VRF table. If the device is not enslaved, oif is still
0 so no affect.

Fixes: cd2fbe1b6b51 ("net: Use VRF device index for lookups on RX")
Reported-by: Xin Long <lucien.xin@gmail.com>
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/fib_frontend.c