[TCP]: MTU probing

Implementation of packetization layer path mtu discovery for TCP, based on
the internet-draft currently found at
<http://www.ietf.org/internet-drafts/draft-ietf-pmtud-method-05.txt>.

Signed-off-by: John Heffner <jheffner@psc.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 301 insertions, 37 deletions

diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index 16984d4a8a06..ebf2e0b363c4 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -664,6 +664,22 @@ ctl_table ipv4_table[] = {
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,
 	},
+	{
+		.ctl_name	= NET_TCP_MTU_PROBING,
+		.procname	= "tcp_mtu_probing",
+		.data		= &sysctl_tcp_mtu_probing,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+	{
+		.ctl_name	= NET_TCP_BASE_MSS,
+		.procname	= "tcp_base_mss",
+		.data		= &sysctl_tcp_base_mss,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
 
 	{ .ctl_name = 0 }
 };
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index e9a54ae7d690..0ac388e3d01d 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -1891,6 +1891,34 @@ static void tcp_try_to_open(struct sock *sk, struct tcp_sock *tp, int flag)
 	}
 }
 
+static void tcp_mtup_probe_failed(struct sock *sk)
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	icsk->icsk_mtup.search_high = icsk->icsk_mtup.probe_size - 1;
+	icsk->icsk_mtup.probe_size = 0;
+}
+
+static void tcp_mtup_probe_success(struct sock *sk, struct sk_buff *skb)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	/* FIXME: breaks with very large cwnd */
+	tp->prior_ssthresh = tcp_current_ssthresh(sk);
+	tp->snd_cwnd = tp->snd_cwnd *
+		       tcp_mss_to_mtu(sk, tp->mss_cache) /
+		       icsk->icsk_mtup.probe_size;
+	tp->snd_cwnd_cnt = 0;
+	tp->snd_cwnd_stamp = tcp_time_stamp;
+	tp->rcv_ssthresh = tcp_current_ssthresh(sk);
+
+	icsk->icsk_mtup.search_low = icsk->icsk_mtup.probe_size;
+	icsk->icsk_mtup.probe_size = 0;
+	tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
+}
+
+
 /* Process an event, which can update packets-in-flight not trivially.
  * Main goal of this function is to calculate new estimate for left_out,
  * taking into account both packets sitting in receiver's buffer and
@@ -2023,6 +2051,17 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una,
 			return;
 		}
 
+		/* MTU probe failure: don't reduce cwnd */
+		if (icsk->icsk_ca_state < TCP_CA_CWR &&
+		    icsk->icsk_mtup.probe_size &&
+		    tp->snd_una == icsk->icsk_mtup.probe_seq_start) {
+			tcp_mtup_probe_failed(sk);
+			/* Restores the reduction we did in tcp_mtup_probe() */
+			tp->snd_cwnd++;
+			tcp_simple_retransmit(sk);
+			return;
+		}
+
 		/* Otherwise enter Recovery state */
 
 		if (IsReno(tp))
@@ -2243,6 +2282,13 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p)
 			tp->retrans_stamp = 0;
 		}
 
+		/* MTU probing checks */
+		if (icsk->icsk_mtup.probe_size) {
+			if (!after(icsk->icsk_mtup.probe_seq_end, TCP_SKB_CB(skb)->end_seq)) {
+				tcp_mtup_probe_success(sk, skb);
+			}
+		}
+
 		if (sacked) {
 			if (sacked & TCPCB_RETRANS) {
 				if(sacked & TCPCB_SACKED_RETRANS)
@@ -4101,6 +4147,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		if (tp->rx_opt.sack_ok && sysctl_tcp_fack)
 			tp->rx_opt.sack_ok |= 2;
 
+		tcp_mtup_init(sk);
 		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 		tcp_initialize_rcv_mss(sk);
 
@@ -4211,6 +4258,7 @@ discard:
 		if (tp->ecn_flags&TCP_ECN_OK)
 			sock_set_flag(sk, SOCK_NO_LARGESEND);
 
+		tcp_mtup_init(sk);
 		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 		tcp_initialize_rcv_mss(sk);
 
@@ -4399,6 +4447,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
 				 */
 				tp->lsndtime = tcp_time_stamp;
 
+				tcp_mtup_init(sk);
 				tcp_initialize_rcv_mss(sk);
 				tcp_init_buffer_space(sk);
 				tcp_fast_path_on(tp);
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 233bdf259965..57e7a26e8213 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -900,6 +900,7 @@ struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
 		inet_csk(newsk)->icsk_ext_hdr_len = newinet->opt->optlen;
 	newinet->id = newtp->write_seq ^ jiffies;
 
+	tcp_mtup_init(newsk);
 	tcp_sync_mss(newsk, dst_mtu(dst));
 	newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
 	tcp_initialize_rcv_mss(newsk);
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 9f498a6c8895..8197b5e12f1f 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -51,6 +51,12 @@ int sysctl_tcp_retrans_collapse = 1;
  */
 int sysctl_tcp_tso_win_divisor = 3;
 
+int sysctl_tcp_mtu_probing = 0;
+int sysctl_tcp_base_mss = 512;
+
+EXPORT_SYMBOL(sysctl_tcp_mtu_probing);
+EXPORT_SYMBOL(sysctl_tcp_base_mss);
+
 static void update_send_head(struct sock *sk, struct tcp_sock *tp,
 			     struct sk_buff *skb)
 {
@@ -681,6 +687,62 @@ int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
 	return 0;
 }
 
+/* Not accounting for SACKs here. */
+int tcp_mtu_to_mss(struct sock *sk, int pmtu)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	int mss_now;
+
+	/* Calculate base mss without TCP options:
+	   It is MMS_S - sizeof(tcphdr) of rfc1122
+	 */
+	mss_now = pmtu - icsk->icsk_af_ops->net_header_len - sizeof(struct tcphdr);
+
+	/* Clamp it (mss_clamp does not include tcp options) */
+	if (mss_now > tp->rx_opt.mss_clamp)
+		mss_now = tp->rx_opt.mss_clamp;
+
+	/* Now subtract optional transport overhead */
+	mss_now -= icsk->icsk_ext_hdr_len;
+
+	/* Then reserve room for full set of TCP options and 8 bytes of data */
+	if (mss_now < 48)
+		mss_now = 48;
+
+	/* Now subtract TCP options size, not including SACKs */
+	mss_now -= tp->tcp_header_len - sizeof(struct tcphdr);
+
+	return mss_now;
+}
+
+/* Inverse of above */
+int tcp_mss_to_mtu(struct sock *sk, int mss)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	int mtu;
+
+	mtu = mss +
+	      tp->tcp_header_len +
+	      icsk->icsk_ext_hdr_len +
+	      icsk->icsk_af_ops->net_header_len;
+
+	return mtu;
+}
+
+void tcp_mtup_init(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	icsk->icsk_mtup.enabled = sysctl_tcp_mtu_probing > 1;
+	icsk->icsk_mtup.search_high = tp->rx_opt.mss_clamp + sizeof(struct tcphdr) +
+	                       icsk->icsk_af_ops->net_header_len;
+	icsk->icsk_mtup.search_low = tcp_mss_to_mtu(sk, sysctl_tcp_base_mss);
+	icsk->icsk_mtup.probe_size = 0;
+}
+
 /* This function synchronize snd mss to current pmtu/exthdr set.
 
    tp->rx_opt.user_mss is mss set by user by TCP_MAXSEG. It does NOT counts
@@ -708,25 +770,12 @@ unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct inet_connection_sock *icsk = inet_csk(sk);
-	/* Calculate base mss without TCP options:
-	   It is MMS_S - sizeof(tcphdr) of rfc1122
-	 */
-	int mss_now = (pmtu - icsk->icsk_af_ops->net_header_len -
-		       sizeof(struct tcphdr));
+	int mss_now;
 
-	/* Clamp it (mss_clamp does not include tcp options) */
-	if (mss_now > tp->rx_opt.mss_clamp)
-		mss_now = tp->rx_opt.mss_clamp;
+	if (icsk->icsk_mtup.search_high > pmtu)
+		icsk->icsk_mtup.search_high = pmtu;
 
-	/* Now subtract optional transport overhead */
-	mss_now -= icsk->icsk_ext_hdr_len;
-
-	/* Then reserve room for full set of TCP options and 8 bytes of data */
-	if (mss_now < 48)
-		mss_now = 48;
-
-	/* Now subtract TCP options size, not including SACKs */
-	mss_now -= tp->tcp_header_len - sizeof(struct tcphdr);
+	mss_now = tcp_mtu_to_mss(sk, pmtu);
 
 	/* Bound mss with half of window */
 	if (tp->max_window && mss_now > (tp->max_window>>1))
@@ -734,6 +783,8 @@ unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
 
 	/* And store cached results */
 	icsk->icsk_pmtu_cookie = pmtu;
+	if (icsk->icsk_mtup.enabled)
+		mss_now = min(mss_now, tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low));
 	tp->mss_cache = mss_now;
 
 	return mss_now;
@@ -1063,6 +1114,140 @@ static int tcp_tso_should_defer(struct sock *sk, struct tcp_sock *tp, struct sk_
 	return 1;
 }
 
+/* Create a new MTU probe if we are ready.
+ * Returns 0 if we should wait to probe (no cwnd available),
+ *         1 if a probe was sent,
+ *         -1 otherwise */
+static int tcp_mtu_probe(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	struct sk_buff *skb, *nskb, *next;
+	int len;
+	int probe_size;
+	unsigned int pif;
+	int copy;
+	int mss_now;
+
+	/* Not currently probing/verifying,
+	 * not in recovery,
+	 * have enough cwnd, and
+	 * not SACKing (the variable headers throw things off) */
+	if (!icsk->icsk_mtup.enabled ||
+	    icsk->icsk_mtup.probe_size ||
+	    inet_csk(sk)->icsk_ca_state != TCP_CA_Open ||
+	    tp->snd_cwnd < 11 ||
+	    tp->rx_opt.eff_sacks)
+		return -1;
+
+	/* Very simple search strategy: just double the MSS. */
+	mss_now = tcp_current_mss(sk, 0);
+	probe_size = 2*tp->mss_cache;
+	if (probe_size > tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_high)) {
+		/* TODO: set timer for probe_converge_event */
+		return -1;
+	}
+
+	/* Have enough data in the send queue to probe? */
+	len = 0;
+	if ((skb = sk->sk_send_head) == NULL)
+		return -1;
+	while ((len += skb->len) < probe_size && !tcp_skb_is_last(sk, skb))
+		skb = skb->next;
+	if (len < probe_size)
+		return -1;
+
+	/* Receive window check. */
+	if (after(TCP_SKB_CB(skb)->seq + probe_size, tp->snd_una + tp->snd_wnd)) {
+		if (tp->snd_wnd < probe_size)
+			return -1;
+		else
+			return 0;
+	}
+
+	/* Do we need to wait to drain cwnd? */
+	pif = tcp_packets_in_flight(tp);
+	if (pif + 2 > tp->snd_cwnd) {
+		/* With no packets in flight, don't stall. */
+		if (pif == 0)
+			return -1;
+		else
+			return 0;
+	}
+
+	/* We're allowed to probe.  Build it now. */
+	if ((nskb = sk_stream_alloc_skb(sk, probe_size, GFP_ATOMIC)) == NULL)
+		return -1;
+	sk_charge_skb(sk, nskb);
+
+	skb = sk->sk_send_head;
+	__skb_insert(nskb, skb->prev, skb, &sk->sk_write_queue);
+	sk->sk_send_head = nskb;
+
+	TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(skb)->seq;
+	TCP_SKB_CB(nskb)->end_seq = TCP_SKB_CB(skb)->seq + probe_size;
+	TCP_SKB_CB(nskb)->flags = TCPCB_FLAG_ACK;
+	TCP_SKB_CB(nskb)->sacked = 0;
+	nskb->csum = 0;
+	if (skb->ip_summed == CHECKSUM_HW)
+		nskb->ip_summed = CHECKSUM_HW;
+
+	len = 0;
+	while (len < probe_size) {
+		next = skb->next;
+
+		copy = min_t(int, skb->len, probe_size - len);
+		if (nskb->ip_summed)
+			skb_copy_bits(skb, 0, skb_put(nskb, copy), copy);
+		else
+			nskb->csum = skb_copy_and_csum_bits(skb, 0,
+			                 skb_put(nskb, copy), copy, nskb->csum);
+
+		if (skb->len <= copy) {
+			/* We've eaten all the data from this skb.
+			 * Throw it away. */
+			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags;
+			__skb_unlink(skb, &sk->sk_write_queue);
+			sk_stream_free_skb(sk, skb);
+		} else {
+			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags &
+			                           ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
+			if (!skb_shinfo(skb)->nr_frags) {
+				skb_pull(skb, copy);
+				if (skb->ip_summed != CHECKSUM_HW)
+					skb->csum = csum_partial(skb->data, skb->len, 0);
+			} else {
+				__pskb_trim_head(skb, copy);
+				tcp_set_skb_tso_segs(sk, skb, mss_now);
+			}
+			TCP_SKB_CB(skb)->seq += copy;
+		}
+
+		len += copy;
+		skb = next;
+	}
+	tcp_init_tso_segs(sk, nskb, nskb->len);
+
+	/* We're ready to send.  If this fails, the probe will
+	 * be resegmented into mss-sized pieces by tcp_write_xmit(). */
+	TCP_SKB_CB(nskb)->when = tcp_time_stamp;
+	if (!tcp_transmit_skb(sk, nskb, 1, GFP_ATOMIC)) {
+		/* Decrement cwnd here because we are sending
+		* effectively two packets. */
+		tp->snd_cwnd--;
+		update_send_head(sk, tp, nskb);
+
+		icsk->icsk_mtup.probe_size = tcp_mss_to_mtu(sk, nskb->len);
+		icsk->icsk_mtup.probe_seq_start = TCP_SKB_CB(nskb)->seq;
+		icsk->icsk_mtup.probe_seq_end = TCP_SKB_CB(nskb)->end_seq;
+
+		return 1;
+	}
+
+	return -1;
+}
+
+
 /* This routine writes packets to the network.  It advances the
  * send_head.  This happens as incoming acks open up the remote
  * window for us.
@@ -1076,6 +1261,7 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
 	struct sk_buff *skb;
 	unsigned int tso_segs, sent_pkts;
 	int cwnd_quota;
+	int result;
 
 	/* If we are closed, the bytes will have to remain here.
 	 * In time closedown will finish, we empty the write queue and all
@@ -1085,6 +1271,14 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
 		return 0;
 
 	sent_pkts = 0;
+
+	/* Do MTU probing. */
+	if ((result = tcp_mtu_probe(sk)) == 0) {
+		return 0;
+	} else if (result > 0) {
+		sent_pkts = 1;
+	}
+
 	while ((skb = sk->sk_send_head)) {
 		unsigned int limit;
 
@@ -1455,9 +1649,15 @@ void tcp_simple_retransmit(struct sock *sk)
 int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
  	unsigned int cur_mss = tcp_current_mss(sk, 0);
 	int err;
 
+	/* Inconslusive MTU probe */
+	if (icsk->icsk_mtup.probe_size) {
+		icsk->icsk_mtup.probe_size = 0;
+	}
+
 	/* Do not sent more than we queued. 1/4 is reserved for possible
 	 * copying overhead: fragmentation, tunneling, mangling etc.
 	 */
@@ -1883,6 +2083,7 @@ static void tcp_connect_init(struct sock *sk)
 	if (tp->rx_opt.user_mss)
 		tp->rx_opt.mss_clamp = tp->rx_opt.user_mss;
 	tp->max_window = 0;
+	tcp_mtup_init(sk);
 	tcp_sync_mss(sk, dst_mtu(dst));
 
 	if (!tp->window_clamp)
@@ -2180,3 +2381,4 @@ EXPORT_SYMBOL(tcp_make_synack);
 EXPORT_SYMBOL(tcp_simple_retransmit);
 EXPORT_SYMBOL(tcp_sync_mss);
 EXPORT_SYMBOL(sysctl_tcp_tso_win_divisor);
+EXPORT_SYMBOL(tcp_mtup_init);
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index e1880959614a..7c1bde3cd6cb 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -119,8 +119,10 @@ static int tcp_orphan_retries(struct sock *sk, int alive)
 /* A write timeout has occurred. Process the after effects. */
 static int tcp_write_timeout(struct sock *sk)
 {
-	const struct inet_connection_sock *icsk = inet_csk(sk);
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	struct tcp_sock *tp = tcp_sk(sk);
 	int retry_until;
+	int mss;
 
 	if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
 		if (icsk->icsk_retransmits)
@@ -128,25 +130,19 @@ static int tcp_write_timeout(struct sock *sk)
 		retry_until = icsk->icsk_syn_retries ? : sysctl_tcp_syn_retries;
 	} else {
 		if (icsk->icsk_retransmits >= sysctl_tcp_retries1) {
-			/* NOTE. draft-ietf-tcpimpl-pmtud-01.txt requires pmtu black
-			   hole detection. :-(
-
-			   It is place to make it. It is not made. I do not want
-			   to make it. It is disgusting. It does not work in any
-			   case. Let me to cite the same draft, which requires for
-			   us to implement this:
-
-   "The one security concern raised by this memo is that ICMP black holes
-   are often caused by over-zealous security administrators who block
-   all ICMP messages.  It is vitally important that those who design and
-   deploy security systems understand the impact of strict filtering on
-   upper-layer protocols.  The safest web site in the world is worthless
-   if most TCP implementations cannot transfer data from it.  It would
-   be far nicer to have all of the black holes fixed rather than fixing
-   all of the TCP implementations."
-
-                           Golden words :-).
-		   */
+			/* Black hole detection */
+			if (sysctl_tcp_mtu_probing) {
+				if (!icsk->icsk_mtup.enabled) {
+					icsk->icsk_mtup.enabled = 1;
+					tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
+				} else {
+					mss = min(sysctl_tcp_base_mss,
+					          tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low)/2);
+					mss = max(mss, 68 - tp->tcp_header_len);
+					icsk->icsk_mtup.search_low = tcp_mss_to_mtu(sk, mss);
+					tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
+				}
+			}
 
 			dst_negative_advice(&sk->sk_dst_cache);
 		}