Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755120AbdLLNNh (ORCPT ); Tue, 12 Dec 2017 08:13:37 -0500 Received: from mail.linuxfoundation.org ([140.211.169.12]:34328 "EHLO mail.linuxfoundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754673AbdLLM72 (ORCPT ); Tue, 12 Dec 2017 07:59:28 -0500 From: Greg Kroah-Hartman To: linux-kernel@vger.kernel.org Cc: Greg Kroah-Hartman , stable@vger.kernel.org, Sowmini Varadhan , "David S. Miller" , Sasha Levin Subject: [PATCH 4.9 103/148] rds: tcp: Sequence teardown of listen and acceptor sockets to avoid races Date: Tue, 12 Dec 2017 13:45:13 +0100 Message-Id: <20171212124436.834478712@linuxfoundation.org> X-Mailer: git-send-email 2.15.1 In-Reply-To: <20171212124431.207182779@linuxfoundation.org> References: <20171212124431.207182779@linuxfoundation.org> User-Agent: quilt/0.65 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4767 Lines: 140 4.9-stable review patch. If anyone has any objections, please let me know. ------------------ From: Sowmini Varadhan [ Upstream commit b21dd4506b71bdb9c5a20e759255cd2513ea7ebe ] Commit a93d01f5777e ("RDS: TCP: avoid bad page reference in rds_tcp_listen_data_ready") added the function rds_tcp_listen_sock_def_readable() to handle the case when a partially set-up acceptor socket drops into rds_tcp_listen_data_ready(). However, if the listen socket (rtn->rds_tcp_listen_sock) is itself going through a tear-down via rds_tcp_listen_stop(), the (*ready)() will be null and we would hit a panic of the form BUG: unable to handle kernel NULL pointer dereference at (null) IP: (null) : ? rds_tcp_listen_data_ready+0x59/0xb0 [rds_tcp] tcp_data_queue+0x39d/0x5b0 tcp_rcv_established+0x2e5/0x660 tcp_v4_do_rcv+0x122/0x220 tcp_v4_rcv+0x8b7/0x980 : In the above case, it is not fatal to encounter a NULL value for ready- we should just drop the packet and let the flush of the acceptor thread finish gracefully. In general, the tear-down sequence for listen() and accept() socket that is ensured by this commit is: rtn->rds_tcp_listen_sock = NULL; /* prevent any new accepts */ In rds_tcp_listen_stop(): serialize with, and prevent, further callbacks using lock_sock() flush rds_wq flush acceptor workq sock_release(listen socket) Signed-off-by: Sowmini Varadhan Signed-off-by: David S. Miller Signed-off-by: Sasha Levin Signed-off-by: Greg Kroah-Hartman --- net/rds/tcp.c | 15 ++++++++++----- net/rds/tcp.h | 2 +- net/rds/tcp_listen.c | 9 +++++++-- 3 files changed, 18 insertions(+), 8 deletions(-) --- a/net/rds/tcp.c +++ b/net/rds/tcp.c @@ -478,9 +478,10 @@ static void __net_exit rds_tcp_exit_net( * we do need to clean up the listen socket here. */ if (rtn->rds_tcp_listen_sock) { - rds_tcp_listen_stop(rtn->rds_tcp_listen_sock); + struct socket *lsock = rtn->rds_tcp_listen_sock; + rtn->rds_tcp_listen_sock = NULL; - flush_work(&rtn->rds_tcp_accept_w); + rds_tcp_listen_stop(lsock, &rtn->rds_tcp_accept_w); } } @@ -517,10 +518,10 @@ static void rds_tcp_kill_sock(struct net struct rds_tcp_connection *tc, *_tc; LIST_HEAD(tmp_list); struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid); + struct socket *lsock = rtn->rds_tcp_listen_sock; - rds_tcp_listen_stop(rtn->rds_tcp_listen_sock); rtn->rds_tcp_listen_sock = NULL; - flush_work(&rtn->rds_tcp_accept_w); + rds_tcp_listen_stop(lsock, &rtn->rds_tcp_accept_w); spin_lock_irq(&rds_tcp_conn_lock); list_for_each_entry_safe(tc, _tc, &rds_tcp_conn_list, t_tcp_node) { struct net *c_net = read_pnet(&tc->t_cpath->cp_conn->c_net); @@ -540,8 +541,12 @@ static void rds_tcp_kill_sock(struct net void *rds_tcp_listen_sock_def_readable(struct net *net) { struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid); + struct socket *lsock = rtn->rds_tcp_listen_sock; + + if (!lsock) + return NULL; - return rtn->rds_tcp_listen_sock->sk->sk_user_data; + return lsock->sk->sk_user_data; } static int rds_tcp_dev_event(struct notifier_block *this, --- a/net/rds/tcp.h +++ b/net/rds/tcp.h @@ -66,7 +66,7 @@ void rds_tcp_state_change(struct sock *s /* tcp_listen.c */ struct socket *rds_tcp_listen_init(struct net *); -void rds_tcp_listen_stop(struct socket *); +void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor); void rds_tcp_listen_data_ready(struct sock *sk); int rds_tcp_accept_one(struct socket *sock); int rds_tcp_keepalive(struct socket *sock); --- a/net/rds/tcp_listen.c +++ b/net/rds/tcp_listen.c @@ -227,6 +227,9 @@ void rds_tcp_listen_data_ready(struct so * before it has been accepted and the accepter has set up their * data_ready.. we only want to queue listen work for our listening * socket + * + * (*ready)() may be null if we are racing with netns delete, and + * the listen socket is being torn down. */ if (sk->sk_state == TCP_LISTEN) rds_tcp_accept_work(sk); @@ -235,7 +238,8 @@ void rds_tcp_listen_data_ready(struct so out: read_unlock_bh(&sk->sk_callback_lock); - ready(sk); + if (ready) + ready(sk); } struct socket *rds_tcp_listen_init(struct net *net) @@ -275,7 +279,7 @@ out: return NULL; } -void rds_tcp_listen_stop(struct socket *sock) +void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor) { struct sock *sk; @@ -296,5 +300,6 @@ void rds_tcp_listen_stop(struct socket * /* wait for accepts to stop and close the socket */ flush_workqueue(rds_wq); + flush_work(acceptor); sock_release(sock); }