tcp: do not release socket ownership in tcp_close()
authorEric Dumazet <edumazet@google.com>
Tue, 2 Oct 2018 06:24:26 +0000 (23:24 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 1 Dec 2018 08:37:28 +0000 (09:37 +0100)
commitcddcc9959a32999ce2d38018900595a5ab453eb2
tree9b32d515c35c86a1d78d460eb5006d675a927dd8
parent5999609a93a4eb4956b6548159ce69295b1557b2
tcp: do not release socket ownership in tcp_close()

commit 8873c064d1de579ea23412a6d3eee972593f142b upstream.

syzkaller was able to hit the WARN_ON(sock_owned_by_user(sk));
in tcp_close()

While a socket is being closed, it is very possible other
threads find it in rtnetlink dump.

tcp_get_info() will acquire the socket lock for a short amount
of time (slow = lock_sock_fast(sk)/unlock_sock_fast(sk, slow);),
enough to trigger the warning.

Fixes: 67db3e4bfbc9 ("tcp: no longer hold ehash lock while calling tcp_get_info()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
include/net/sock.h
net/core/sock.c
net/ipv4/tcp.c