Commit c41a38ef authored by Paolo Abeni's avatar Paolo Abeni
Browse files

Merge branch 'tcp-save-flowlabel-and-use-for-receiver-repathing' 

David Morley says:

====================
tcp: save flowlabel and use for receiver repathing

This patch series stores the last received ipv6 flowlabel. This last
received flowlabel is then used to help decide whether a packet is
likely an RTO retransmit or the result of a TLP. This new information
is used to better inform the flowlabel change decision for data
receivers.
====================

Link: https://lore.kernel.org/r/20231006011841.3558307-1-morleyd.kernel@gmail.com


Signed-off-by: default avatarPaolo Abeni <pabeni@redhat.com>
parents 8cea95b0 93946301

include/net/inet_connection_sock.h

+4 −1
Original line number Diff line number Diff line
@@ -114,7 +114,10 @@ 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

+2 −0
Original line number Diff line number Diff line
@@ -131,6 +131,8 @@ 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

+2 −2
Original line number Diff line number Diff line
@@ -196,7 +196,7 @@ 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_ack.ato = min_t(u32, icsk->icsk_ack.ato << 1,

						   icsk->icsk_rto);

		} else {

			/* Delayed ACK missed: leave pingpong mode and

net/ipv4/tcp.c

+2 −2
Original line number Diff line number Diff line
@@ -3756,7 +3756,7 @@ 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,

	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

+26 −3
Original line number Diff line number Diff line
@@ -778,6 +778,16 @@ 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
@@ -826,6 +836,7 @@ 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);
@@ -4513,12 +4524,23 @@ static void tcp_rcv_spurious_retrans(struct sock *sk, const struct sk_buff *skb) 
{

	/* When the ACK path fails or drops most ACKs, the sender would

	 * timeout and spuriously retransmit the same segment repeatedly.

	 * The receiver remembers and reflects via DSACKs. Leverage the

	 * DSACK state and change the txhash to re-route speculatively.

	 * If it seems our ACKs are not reaching the other side,

	 * based on receiving a duplicate data segment with new flowlabel

	 * (suggesting the sender suffered an RTO), and we are not already

	 * repathing due to our own RTO, then rehash the socket to repath our

	 * packets.

	 */

	if (TCP_SKB_CB(skb)->seq == tcp_sk(sk)->duplicate_sack[0].start_seq &&

#if IS_ENABLED(CONFIG_IPV6)

	if (inet_csk(sk)->icsk_ca_state != TCP_CA_Loss &&

	    skb->protocol == htons(ETH_P_IPV6) &&

	    (tcp_sk(sk)->inet_conn.icsk_ack.lrcv_flowlabel !=

	     ntohl(ip6_flowlabel(ipv6_hdr(skb)))) &&

	    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);

#endif

}



static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb)
@@ -4835,6 +4857,7 @@ 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))) {