tcp: record last received ipv6 flowlabel

mainline inclusion
from mainline-v6.7-rc1
category: performance

In order to better estimate whether a data packet has been
retransmitted or is the result of a TLP, we save the last received
ipv6 flowlabel.

To make space for this field we resize the "ato" field in
inet_connection_sock as the current value of TCP_DELACK_MAX can be
fully contained in 8 bits and add a compile_time_assert ensuring this
field is the required size.

v2: addressed kernel bot feedback about dccp_delack_timer()
v3: addressed build error introduced by commit bbf80d7 ("tcp:
derive delack_max from rto_min")

Signed-off-by: David Morley <morleyd@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Tested-by: David Morley <morleyd@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
(cherry picked from commit 95b9a87)
Signed-off-by: Wentao Guan <guanwentao@uniontech.com>

Author: David Morley

include/net/inet_connection_sock.h

@@ -117,7 +117,10 @@ struct inet_connection_sock {
 		__u8		  quick;	 /* Scheduled number of quick acks	   */
 		__u8		  pingpong;	 /* The session is interactive		   */
 		__u8		  retry;	 /* Number of attempts			   */
-		__u32		  ato;		 /* Predicted tick of soft clock	   */
+		#define ATO_BITS 8
+		__u32		  ato:ATO_BITS,	 /* Predicted tick of soft clock	   */
+				  lrcv_flowlabel:20, /* last received ipv6 flowlabel	   */
+				  unused:4;
 		unsigned long	  timeout;	 /* Currently scheduled timeout		   */
 		__u32		  lrcvtime;	 /* timestamp of last received data packet */
 		__u16		  last_seg_size; /* Size of last incoming segment	   */

include/net/tcp.h

@@ -133,6 +133,8 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
 #define TCP_FIN_TIMEOUT_MAX (120 * HZ) /* max TCP_LINGER2 value (two minutes) */
 
 #define TCP_DELACK_MAX	((unsigned)(HZ/5))	/* maximal time to delay before sending an ACK */
+static_assert((1 << ATO_BITS) > TCP_DELACK_MAX);
+
 #if HZ >= 100
 #define TCP_DELACK_MIN	((unsigned)(HZ/25))	/* minimal time to delay before sending an ACK */
 #define TCP_ATO_MIN	((unsigned)(HZ/25))

net/dccp/timer.c

@@ -196,8 +196,8 @@ static void dccp_delack_timer(struct timer_list *t)
 	if (inet_csk_ack_scheduled(sk)) {
 		if (!inet_csk_in_pingpong_mode(sk)) {
 			/* Delayed ACK missed: inflate ATO. */
-			icsk->icsk_ack.ato = min(icsk->icsk_ack.ato << 1,
-						 icsk->icsk_rto);
+			icsk->icsk_ack.ato = min_t(u32, icsk->icsk_ack.ato << 1,
+						   icsk->icsk_rto);
 		} else {
 			/* Delayed ACK missed: leave pingpong mode and
 			 * deflate ATO.

net/ipv4/tcp.c

@@ -3813,8 +3813,8 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 		info->tcpi_options |= TCPI_OPT_SYN_DATA;
 
 	info->tcpi_rto = jiffies_to_usecs(icsk->icsk_rto);
-	info->tcpi_ato = jiffies_to_usecs(min(icsk->icsk_ack.ato,
-					      tcp_delack_max(sk)));
+	info->tcpi_ato = jiffies_to_usecs(min_t(u32, icsk->icsk_ack.ato,
+						tcp_delack_max(sk)));
 	info->tcpi_snd_mss = tp->mss_cache;
 	info->tcpi_rcv_mss = icsk->icsk_ack.rcv_mss;
 

net/ipv4/tcp_input.c

@@ -805,6 +805,16 @@ void tcp_rcv_space_adjust(struct sock *sk)
 	tp->rcvq_space.time = tp->tcp_mstamp;
 }
 
+static void tcp_save_lrcv_flowlabel(struct sock *sk, const struct sk_buff *skb)
+{
+#if IS_ENABLED(CONFIG_IPV6)
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	if (skb->protocol == htons(ETH_P_IPV6))
+		icsk->icsk_ack.lrcv_flowlabel = ntohl(ip6_flowlabel(ipv6_hdr(skb)));
+#endif
+}
+
 /* There is something which you must keep in mind when you analyze the
  * behavior of the tp->ato delayed ack timeout interval.  When a
  * connection starts up, we want to ack as quickly as possible.  The
@@ -853,6 +863,7 @@ static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb)
 		}
 	}
 	icsk->icsk_ack.lrcvtime = now;
+	tcp_save_lrcv_flowlabel(sk, skb);
 
 	tcp_ecn_check_ce(sk, skb);
 
@@ -4608,6 +4619,9 @@ static void tcp_rcv_spurious_retrans(struct sock *sk, const struct sk_buff *skb)
 	if (TCP_SKB_CB(skb)->seq == tcp_sk(sk)->duplicate_sack[0].start_seq &&
 	    sk_rethink_txhash(sk))
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDUPLICATEDATAREHASH);
+
+	/* Save last flowlabel after a spurious retrans. */
+	tcp_save_lrcv_flowlabel(sk, skb);
 }
 
 static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb)
@@ -4924,6 +4938,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 	u32 seq, end_seq;
 	bool fragstolen;
 
+	tcp_save_lrcv_flowlabel(sk, skb);
 	tcp_ecn_check_ce(sk, skb);
 
 	if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) {

net/ipv4/tcp_timer.c

@@ -318,7 +318,7 @@ void tcp_delack_timer_handler(struct sock *sk)
 	if (inet_csk_ack_scheduled(sk)) {
 		if (!inet_csk_in_pingpong_mode(sk)) {
 			/* Delayed ACK missed: inflate ATO. */
-			icsk->icsk_ack.ato = min(icsk->icsk_ack.ato << 1, icsk->icsk_rto);
+			icsk->icsk_ack.ato = min_t(u32, icsk->icsk_ack.ato << 1, icsk->icsk_rto);
 		} else {
 			/* Delayed ACK missed: leave pingpong mode and
 			 * deflate ATO.