Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Dan Streetman <ddstreet@ieee.org>
To:     "David S. Miller" <davem@davemloft.net>
Cc:     Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>,
        Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        Dan Streetman <ddstreet@ieee.org>,
        Dan Streetman <ddstreet@canonical.com>
Subject: [PATCH] net: tcp: close sock if net namespace is exiting
Date:   Thu, 18 Jan 2018 16:14:26 -0500
Message-Id: <20180118211426.24441-1-ddstreet@ieee.org>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

When a tcp socket is closed, if it detects that its net namespace is
exiting, close immediately and do not wait for FIN sequence.

For normal sockets, a reference is taken to their net namespace, so it will
never exit while the socket is open.  However, kernel sockets do not take a
reference to their net namespace, so it may begin exiting while the kernel
socket is still open.  In this case if the kernel socket is a tcp socket,
it will stay open trying to complete its close sequence.  The sock's dst(s)
hold a reference to their interface, which are all transferred to the
namespace's loopback interface when the real interfaces are taken down.
When the namespace tries to take down its loopback interface, it hangs
waiting for all references to the loopback interface to release, which
results in messages like:

unregister_netdevice: waiting for lo to become free. Usage count = 1

These messages continue until the socket finally times out and closes.
Since the net namespace cleanup holds the net_mutex while calling its
registered pernet callbacks, any new net namespace initialization is
blocked until the current net namespace finishes exiting.

After this change, the tcp socket notices the exiting net namespace, and
closes immediately, releasing its dst(s) and their reference to the
loopback interface, which lets the net namespace continue exiting.

Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1711407
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=97811
Signed-off-by: Dan Streetman <ddstreet@canonical.com>
---
 include/net/net_namespace.h | 10 ++++++++++
 net/ipv4/tcp.c              |  3 +++
 net/ipv4/tcp_timer.c        | 15 +++++++++++++++
 3 files changed, 28 insertions(+)

diff --git a/include/net/net_namespace.h b/include/net/net_namespace.h
index f8a84a2c2341..f306b2aa15a4 100644
--- a/include/net/net_namespace.h
+++ b/include/net/net_namespace.h
@@ -223,6 +223,11 @@ int net_eq(const struct net *net1, const struct net *net2)
 	return net1 == net2;
 }
 
+static inline int check_net(const struct net *net)
+{
+	return refcount_read(&net->count) != 0;
+}
+
 void net_drop_ns(void *);
 
 #else
@@ -247,6 +252,11 @@ int net_eq(const struct net *net1, const struct net *net2)
 	return 1;
 }
 
+static inline int check_net(const struct net *net)
+{
+	return 1;
+}
+
 #define net_drop_ns NULL
 #endif
 
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d7cf861bf699..9389193e73f3 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2298,6 +2298,9 @@ void tcp_close(struct sock *sk, long timeout)
 			tcp_send_active_reset(sk, GFP_ATOMIC);
 			__NET_INC_STATS(sock_net(sk),
 					LINUX_MIB_TCPABORTONMEMORY);
+		} else if (!check_net(sock_net(sk))) {
+			/* Not possible to send reset; just close */
+			tcp_set_state(sk, TCP_CLOSE);
 		}
 	}
 
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 6db3124cdbda..41b40b805aa3 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -48,11 +48,19 @@ static void tcp_write_err(struct sock *sk)
  *  to prevent DoS attacks. It is called when a retransmission timeout
  *  or zero probe timeout occurs on orphaned socket.
  *
+ *  Also close if our net namespace is exiting; in that case there is no
+ *  hope of ever communicating again since all netns interfaces are already
+ *  down (or about to be down), and we need to release our dst references,
+ *  which have been moved to the netns loopback interface, so the namespace
+ *  can finish exiting.  This condition is only possible if we are a kernel
+ *  socket, as those do not hold references to the namespace.
+ *
  *  Criteria is still not confirmed experimentally and may change.
  *  We kill the socket, if:
  *  1. If number of orphaned sockets exceeds an administratively configured
  *     limit.
  *  2. If we have strong memory pressure.
+ *  3. If our net namespace is exiting.
  */
 static int tcp_out_of_resources(struct sock *sk, bool do_reset)
 {
@@ -81,6 +89,13 @@ static int tcp_out_of_resources(struct sock *sk, bool do_reset)
 		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONMEMORY);
 		return 1;
 	}
+
+	if (!check_net(sock_net(sk))) {
+		/* Not possible to send reset; just close */
+		tcp_done(sk);
+		return 1;
+	}
+
 	return 0;
 }
 
-- 
2.14.1