smc: move unhash as early as possible in smc_release()
authorCong Wang <xiyou.wangcong@gmail.com>
Sun, 6 Jan 2019 07:45:26 +0000 (23:45 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 22 Jan 2019 20:40:31 +0000 (21:40 +0100)
[ Upstream commit 26d92e951fe0a44ee4aec157cabb65a818cc8151 ]

In smc_release() we release smc->clcsock before unhash the smc
sock, but a parallel smc_diag_dump() may be still reading
smc->clcsock, therefore this could cause a use-after-free as
reported by syzbot.

Reported-and-tested-by: syzbot+fbd1e5476e4c94c7b34e@syzkaller.appspotmail.com
Fixes: 51f1de79ad8e ("net/smc: replace sock_put worker by socket refcounting")
Cc: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reported-by: syzbot+0bf2e01269f1274b4b03@syzkaller.appspotmail.com
Reported-by: syzbot+e3132895630f957306bc@syzkaller.appspotmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/smc/af_smc.c

index 2b8f952..e6e506b 100644 (file)
@@ -144,6 +144,9 @@ static int smc_release(struct socket *sock)
                sock_set_flag(sk, SOCK_DEAD);
                sk->sk_shutdown |= SHUTDOWN_MASK;
        }
+
+       sk->sk_prot->unhash(sk);
+
        if (smc->clcsock) {
                if (smc->use_fallback && sk->sk_state == SMC_LISTEN) {
                        /* wake up clcsock accept */
@@ -168,7 +171,6 @@ static int smc_release(struct socket *sock)
                smc_conn_free(&smc->conn);
        release_sock(sk);
 
-       sk->sk_prot->unhash(sk);
        sock_put(sk); /* final sock_put */
 out:
        return rc;