mptcp: fix splat when incoming connection is never accepted before exit/close
authorFlorian Westphal <fw@strlen.de>
Fri, 17 Apr 2020 07:28:22 +0000 (09:28 +0200)
committerDavid S. Miller <davem@davemloft.net>
Sat, 18 Apr 2020 22:43:20 +0000 (15:43 -0700)
commitdf1036da90108b1a9969721beab34f4c76228bcc
treecb661c0c9eced47d8af51bc080956fa987b5d7b9
parent310660a14b74c380b0ef5c12b66933d6a3d1b59f
mptcp: fix splat when incoming connection is never accepted before exit/close

Following snippet (replicated from syzkaller reproducer) generates
warning: "IPv4: Attempt to release TCP socket in state 1".

int main(void) {
 struct sockaddr_in sin1 = { .sin_family = 2, .sin_port = 0x4e20,
                             .sin_addr.s_addr = 0x010000e0, };
 struct sockaddr_in sin2 = { .sin_family = 2,
                     .sin_addr.s_addr = 0x0100007f, };
 struct sockaddr_in sin3 = { .sin_family = 2, .sin_port = 0x4e20,
                     .sin_addr.s_addr = 0x0100007f, };
 int r0 = socket(0x2, 0x1, 0x106);
 int r1 = socket(0x2, 0x1, 0x106);

 bind(r1, (void *)&sin1, sizeof(sin1));
 connect(r1, (void *)&sin2, sizeof(sin2));
 listen(r1, 3);
 return connect(r0, (void *)&sin3, 0x4d);
}

Reason is that the newly generated mptcp socket is closed via the ulp
release of the tcp listener socket when its accept backlog gets purged.

To fix this, delay setting the ESTABLISHED state until after userspace
calls accept and via mptcp specific destructor.

Fixes: 58b09919626bf ("mptcp: create msk early")
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/9
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/mptcp/protocol.c
net/mptcp/subflow.c