xdp: fix race on generic receive path
authorIlya Maximets <i.maximets@samsung.com>
Wed, 3 Jul 2019 12:09:16 +0000 (15:09 +0300)
committerDaniel Borkmann <daniel@iogearbox.net>
Mon, 8 Jul 2019 23:43:26 +0000 (01:43 +0200)
commitbf0bdd1343efbbf65b4d53aef1fce14acbd79d50
tree596c5002c0217bb4c9ab8e1ea2fbf49cf0e78f75
parentbc2d8afecbec33bd4549ce1ef904383fde21d385
xdp: fix race on generic receive path

Unlike driver mode, generic xdp receive could be triggered
by different threads on different CPU cores at the same time
leading to the fill and rx queue breakage. For example, this
could happen while sending packets from two processes to the
first interface of veth pair while the second part of it is
open with AF_XDP socket.

Need to take a lock for each generic receive to avoid race.

Fixes: c497176cb2e4 ("xsk: add Rx receive functions and poll support")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Tested-by: William Tu <u9012063@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
include/net/xdp_sock.h
net/xdp/xsk.c