Commit 27069e7c authored by Matthieu Baerts (NGI0)'s avatar Matthieu Baerts (NGI0) Committed by Jakub Kicinski
Browse files

mptcp: disable active MPTCP in case of blackhole 

An MPTCP firewall blackhole can be detected if the following SYN
retransmission after a fallback to "plain" TCP is accepted.

In case of blackhole, a similar technique to the one in place with TFO
is now used: MPTCP can be disabled for a certain period of time, 1h by
default. This time period will grow exponentially when more blackhole
issues get detected right after MPTCP is re-enabled and will reset to
the initial value when the blackhole issue goes away.

The blackhole period can be modified thanks to a new sysctl knob:
blackhole_timeout. Two new MIB counters help understanding what's
happening:

- 'Blackhole', incremented when a blackhole is detected.
- 'MPCapableSYNTXDisabled', incremented when an MPTCP connection
  directly falls back to TCP during the blackhole period.

Because the technique is inspired by the one used by TFO, an important
part of the new code is similar to what can find in tcp_fastopen.c, with
some adaptations to the MPTCP case.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/57


Signed-off-by: default avatarMatthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20240909-net-next-mptcp-fallback-x-mpc-v1-3-da7ebb4cd2a3@kernel.org


Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
parent 6982826f

Documentation/networking/mptcp-sysctl.rst

+11 −0
Original line number Diff line number Diff line
@@ -34,6 +34,17 @@ available_schedulers - STRING 
	Shows the available schedulers choices that are registered. More packet

	schedulers may be available, but not loaded.



blackhole_timeout - INTEGER (seconds)

	Initial time period in second to disable MPTCP on active MPTCP sockets

	when a MPTCP firewall blackhole issue happens. This time period will

	grow exponentially when more blackhole issues get detected right after

	MPTCP is re-enabled and will reset to the initial value when the

	blackhole issue goes away.



	0 to disable the blackhole detection.



	Default: 3600



checksum_enabled - BOOLEAN

	Control whether DSS checksum can be enabled.

net/mptcp/ctrl.c

+117 −4
Original line number Diff line number Diff line
@@ -28,8 +28,11 @@ struct mptcp_pernet { 
#endif



	unsigned int add_addr_timeout;

	unsigned int blackhole_timeout;

	unsigned int close_timeout;

	unsigned int stale_loss_cnt;

	atomic_t active_disable_times;

	unsigned long active_disable_stamp;

	u8 mptcp_enabled;

	u8 checksum_enabled;

	u8 allow_join_initial_addr_port;
@@ -88,6 +91,8 @@ static void mptcp_pernet_set_defaults(struct mptcp_pernet *pernet) 
{

	pernet->mptcp_enabled = 1;

	pernet->add_addr_timeout = TCP_RTO_MAX;

	pernet->blackhole_timeout = 3600;

	atomic_set(&pernet->active_disable_times, 0);

	pernet->close_timeout = TCP_TIMEWAIT_LEN;

	pernet->checksum_enabled = 0;

	pernet->allow_join_initial_addr_port = 1;
@@ -152,6 +157,20 @@ static int proc_available_schedulers(const struct ctl_table *ctl, 
	return ret;

}



static int proc_blackhole_detect_timeout(const struct ctl_table *table,

					 int write, void *buffer, size_t *lenp,

					 loff_t *ppos)

{

	struct mptcp_pernet *pernet = mptcp_get_pernet(current->nsproxy->net_ns);

	int ret;



	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);

	if (write && ret == 0)

		atomic_set(&pernet->active_disable_times, 0);



	return ret;

}



static struct ctl_table mptcp_sysctl_table[] = {

	{

		.procname = "enabled",
@@ -218,6 +237,13 @@ static struct ctl_table mptcp_sysctl_table[] = { 
		.mode = 0644,

		.proc_handler = proc_dointvec_jiffies,

	},

	{

		.procname = "blackhole_timeout",

		.maxlen = sizeof(unsigned int),

		.mode = 0644,

		.proc_handler = proc_blackhole_detect_timeout,

		.extra1 = SYSCTL_ZERO,

	},

};



static int mptcp_pernet_new_table(struct net *net, struct mptcp_pernet *pernet)
@@ -241,6 +267,7 @@ static int mptcp_pernet_new_table(struct net *net, struct mptcp_pernet *pernet) 
	table[6].data = &pernet->scheduler;

	/* table[7] is for available_schedulers which is read-only info */

	table[8].data = &pernet->close_timeout;

	table[9].data = &pernet->blackhole_timeout;



	hdr = register_net_sysctl_sz(net, MPTCP_SYSCTL_PATH, table,

				     ARRAY_SIZE(mptcp_sysctl_table));
@@ -278,6 +305,88 @@ static void mptcp_pernet_del_table(struct mptcp_pernet *pernet) {} 


#endif /* CONFIG_SYSCTL */



/* The following code block is to deal with middle box issues with MPTCP,

 * similar to what is done with TFO.

 * The proposed solution is to disable active MPTCP globally when SYN+MPC are

 * dropped, while SYN without MPC aren't. In this case, active side MPTCP is

 * disabled globally for 1hr at first. Then if it happens again, it is disabled

 * for 2h, then 4h, 8h, ...

 * The timeout is reset back to 1hr when a successful active MPTCP connection is

 * fully established.

 */



/* Disable active MPTCP and record current jiffies and active_disable_times */

void mptcp_active_disable(struct sock *sk)

{

	struct net *net = sock_net(sk);

	struct mptcp_pernet *pernet;



	pernet = mptcp_get_pernet(net);



	if (!READ_ONCE(pernet->blackhole_timeout))

		return;



	/* Paired with READ_ONCE() in mptcp_active_should_disable() */

	WRITE_ONCE(pernet->active_disable_stamp, jiffies);



	/* Paired with smp_rmb() in mptcp_active_should_disable().

	 * We want pernet->active_disable_stamp to be updated first.

	 */

	smp_mb__before_atomic();

	atomic_inc(&pernet->active_disable_times);



	MPTCP_INC_STATS(net, MPTCP_MIB_BLACKHOLE);

}



/* Calculate timeout for MPTCP active disable

 * Return true if we are still in the active MPTCP disable period

 * Return false if timeout already expired and we should use active MPTCP

 */

bool mptcp_active_should_disable(struct sock *ssk)

{

	struct net *net = sock_net(ssk);

	unsigned int blackhole_timeout;

	struct mptcp_pernet *pernet;

	unsigned long timeout;

	int disable_times;

	int multiplier;



	pernet = mptcp_get_pernet(net);

	blackhole_timeout = READ_ONCE(pernet->blackhole_timeout);



	if (!blackhole_timeout)

		return false;



	disable_times = atomic_read(&pernet->active_disable_times);

	if (!disable_times)

		return false;



	/* Paired with smp_mb__before_atomic() in mptcp_active_disable() */

	smp_rmb();



	/* Limit timeout to max: 2^6 * initial timeout */

	multiplier = 1 << min(disable_times - 1, 6);



	/* Paired with the WRITE_ONCE() in mptcp_active_disable(). */

	timeout = READ_ONCE(pernet->active_disable_stamp) +

		  multiplier * blackhole_timeout * HZ;



	return time_before(jiffies, timeout);

}



/* Enable active MPTCP and reset active_disable_times if needed */

void mptcp_active_enable(struct sock *sk)

{

	struct mptcp_pernet *pernet = mptcp_get_pernet(sock_net(sk));



	if (atomic_read(&pernet->active_disable_times)) {

		struct dst_entry *dst = sk_dst_get(sk);



		if (dst && dst->dev && (dst->dev->flags & IFF_LOOPBACK))

			atomic_set(&pernet->active_disable_times, 0);

	}

}



/* Check the number of retransmissions, and fallback to TCP if needed */

void mptcp_active_detect_blackhole(struct sock *ssk, bool expired)

{
@@ -290,10 +399,14 @@ void mptcp_active_detect_blackhole(struct sock *ssk, bool expired) 
	timeouts = inet_csk(ssk)->icsk_retransmits;

	subflow = mptcp_subflow_ctx(ssk);



	if (subflow->request_mptcp && ssk->sk_state == TCP_SYN_SENT &&

	    (timeouts == 2 || (timeouts < 2 && expired))) {

	if (subflow->request_mptcp && ssk->sk_state == TCP_SYN_SENT) {

		if (timeouts == 2 || (timeouts < 2 && expired)) {

			MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_MPCAPABLEACTIVEDROP);

			subflow->mpc_drop = 1;

			mptcp_subflow_early_fallback(mptcp_sk(subflow->conn), subflow);

		} else {

			subflow->mpc_drop = 0;

		}

	}

}

net/mptcp/mib.c

+2 −0
Original line number Diff line number Diff line
@@ -16,6 +16,7 @@ static const struct snmp_mib mptcp_snmp_list[] = { 
	SNMP_MIB_ITEM("MPCapableFallbackACK", MPTCP_MIB_MPCAPABLEPASSIVEFALLBACK),

	SNMP_MIB_ITEM("MPCapableFallbackSYNACK", MPTCP_MIB_MPCAPABLEACTIVEFALLBACK),

	SNMP_MIB_ITEM("MPCapableSYNTXDrop", MPTCP_MIB_MPCAPABLEACTIVEDROP),

	SNMP_MIB_ITEM("MPCapableSYNTXDisabled", MPTCP_MIB_MPCAPABLEACTIVEDISABLED),

	SNMP_MIB_ITEM("MPFallbackTokenInit", MPTCP_MIB_TOKENFALLBACKINIT),

	SNMP_MIB_ITEM("MPTCPRetrans", MPTCP_MIB_RETRANSSEGS),

	SNMP_MIB_ITEM("MPJoinNoTokenFound", MPTCP_MIB_JOINNOTOKEN),
@@ -74,6 +75,7 @@ static const struct snmp_mib mptcp_snmp_list[] = { 
	SNMP_MIB_ITEM("RcvWndConflictUpdate", MPTCP_MIB_RCVWNDCONFLICTUPDATE),

	SNMP_MIB_ITEM("RcvWndConflict", MPTCP_MIB_RCVWNDCONFLICT),

	SNMP_MIB_ITEM("MPCurrEstab", MPTCP_MIB_CURRESTAB),

	SNMP_MIB_ITEM("Blackhole", MPTCP_MIB_BLACKHOLE),

	SNMP_MIB_SENTINEL

};

net/mptcp/mib.h

+2 −0
Original line number Diff line number Diff line
@@ -11,6 +11,7 @@ enum linux_mptcp_mib_field { 
	MPTCP_MIB_MPCAPABLEPASSIVEFALLBACK,/* Server-side fallback during 3-way handshake */

	MPTCP_MIB_MPCAPABLEACTIVEFALLBACK, /* Client-side fallback during 3-way handshake */

	MPTCP_MIB_MPCAPABLEACTIVEDROP,	/* Client-side fallback due to a MPC drop */

	MPTCP_MIB_MPCAPABLEACTIVEDISABLED, /* Client-side disabled due to past issues */

	MPTCP_MIB_TOKENFALLBACKINIT,	/* Could not init/allocate token */

	MPTCP_MIB_RETRANSSEGS,		/* Segments retransmitted at the MPTCP-level */

	MPTCP_MIB_JOINNOTOKEN,		/* Received MP_JOIN but the token was not found */
@@ -75,6 +76,7 @@ enum linux_mptcp_mib_field { 
					 */

	MPTCP_MIB_RCVWNDCONFLICT,	/* Conflict with while updating msk rcv wnd */

	MPTCP_MIB_CURRESTAB,		/* Current established MPTCP connections */

	MPTCP_MIB_BLACKHOLE,		/* A blackhole has been detected */

	__MPTCP_MIB_MAX

};

net/mptcp/protocol.c

+8 −3
Original line number Diff line number Diff line
@@ -3737,10 +3737,15 @@ static int mptcp_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) 
	if (rcu_access_pointer(tcp_sk(ssk)->md5sig_info))

		mptcp_subflow_early_fallback(msk, subflow);

#endif

	if (subflow->request_mptcp && mptcp_token_new_connect(ssk)) {

	if (subflow->request_mptcp) {

		if (mptcp_active_should_disable(sk)) {

			MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_MPCAPABLEACTIVEDISABLED);

			mptcp_subflow_early_fallback(msk, subflow);

		} else if (mptcp_token_new_connect(ssk) < 0) {

			MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_TOKENFALLBACKINIT);

			mptcp_subflow_early_fallback(msk, subflow);

		}

	}



	WRITE_ONCE(msk->write_seq, subflow->idsn);

	WRITE_ONCE(msk->snd_nxt, subflow->idsn);