Commit 3e59cb0d authored by Yuchung Cheng's avatar Yuchung Cheng Committed by David S. Miller
Browse files

tcp: remove bad timeout logic in fast recovery 



tcp_timeout_skb() was intended to trigger fast recovery on timeout,
unfortunately in reality it often causes spurious retransmission
storms during fast recovery. The particular sign is a fast retransmit
over the highest sacked sequence (SND.FACK).

Currently the RTO timer re-arming (as in RFC6298) offers a nice cushion
to avoid spurious timeout: when SND.UNA advances the sender re-arms
RTO and extends the timeout by icsk_rto. The sender does not offset
the time elapsed since the packet at SND.UNA was sent.

But if the next (DUP)ACK arrives later than ~RTTVAR and triggers
tcp_fastretrans_alert(), then tcp_timeout_skb() will mark any packet
sent before the icsk_rto interval lost, including one that's above the
highest sacked sequence. Most likely a large part of scorebard will be
marked.

If most packets are not lost then the subsequent DUPACKs with new SACK
blocks will cause the sender to continue to retransmit packets beyond
SND.FACK spuriously. Even if only one packet is lost the sender may
falsely retransmit almost the entire window.

The situation becomes common in the world of bufferbloat: the RTT
continues to grow as the queue builds up but RTTVAR remains small and
close to the minimum 200ms. If a data packet is lost and the DUPACK
triggered by the next data packet is slightly delayed, then a spurious
retransmission storm forms.

As the original comment on tcp_timeout_skb() suggests: the usefulness
of this feature is questionable. It also wastes cycles walking the
sack scoreboard and is actually harmful because of false recovery.

It's time to remove this.

Signed-off-by: default avatarYuchung Cheng <ycheng@google.com>
Acked-by: default avatarEric Dumazet <edumazet@google.com>
Acked-by: default avatarNeal Cardwell <ncardwell@google.com>
Acked-by: default avatarNandita Dukkipati <nanditad@google.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 3cc7587b

include/linux/tcp.h

+0 −1
Original line number Diff line number Diff line
@@ -246,7 +246,6 @@ struct tcp_sock { 


	/* from STCP, retrans queue hinting */

	struct sk_buff* lost_skb_hint;

	struct sk_buff *scoreboard_skb_hint;

	struct sk_buff *retransmit_skb_hint;



	struct sk_buff_head	out_of_order_queue; /* Out of order segments go here */

include/net/tcp.h

+0 −1
Original line number Diff line number Diff line
@@ -1193,7 +1193,6 @@ static inline void tcp_mib_init(struct net *net) 
static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)

{

	tp->lost_skb_hint = NULL;

	tp->scoreboard_skb_hint = NULL;

}



static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)

net/ipv4/tcp_input.c

+1 −64
Original line number Diff line number Diff line
@@ -1255,8 +1255,6 @@ static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *skb, 


	if (skb == tp->retransmit_skb_hint)

		tp->retransmit_skb_hint = prev;

	if (skb == tp->scoreboard_skb_hint)

		tp->scoreboard_skb_hint = prev;

	if (skb == tp->lost_skb_hint) {

		tp->lost_skb_hint = prev;

		tp->lost_cnt_hint -= tcp_skb_pcount(prev);
@@ -1964,20 +1962,6 @@ static bool tcp_pause_early_retransmit(struct sock *sk, int flag) 
	return true;

}



static inline int tcp_skb_timedout(const struct sock *sk,

				   const struct sk_buff *skb)

{

	return tcp_time_stamp - TCP_SKB_CB(skb)->when > inet_csk(sk)->icsk_rto;

}



static inline int tcp_head_timedout(const struct sock *sk)

{

	const struct tcp_sock *tp = tcp_sk(sk);



	return tp->packets_out &&

	       tcp_skb_timedout(sk, tcp_write_queue_head(sk));

}



/* Linux NewReno/SACK/FACK/ECN state machine.

 * --------------------------------------

 *
@@ -2084,12 +2068,6 @@ static bool tcp_time_to_recover(struct sock *sk, int flag) 
	if (tcp_dupack_heuristics(tp) > tp->reordering)

		return true;



	/* Trick#3 : when we use RFC2988 timer restart, fast

	 * retransmit can be triggered by timeout of queue head.

	 */

	if (tcp_is_fack(tp) && tcp_head_timedout(sk))

		return true;



	/* Trick#4: It is still not OK... But will it be useful to delay

	 * recovery more?

	 */
@@ -2126,44 +2104,6 @@ static bool tcp_time_to_recover(struct sock *sk, int flag) 
	return false;

}



/* New heuristics: it is possible only after we switched to restart timer

 * each time when something is ACKed. Hence, we can detect timed out packets

 * during fast retransmit without falling to slow start.

 *

 * Usefulness of this as is very questionable, since we should know which of

 * the segments is the next to timeout which is relatively expensive to find

 * in general case unless we add some data structure just for that. The

 * current approach certainly won't find the right one too often and when it

 * finally does find _something_ it usually marks large part of the window

 * right away (because a retransmission with a larger timestamp blocks the

 * loop from advancing). -ij

 */

static void tcp_timeout_skbs(struct sock *sk)

{

	struct tcp_sock *tp = tcp_sk(sk);

	struct sk_buff *skb;



	if (!tcp_is_fack(tp) || !tcp_head_timedout(sk))

		return;



	skb = tp->scoreboard_skb_hint;

	if (tp->scoreboard_skb_hint == NULL)

		skb = tcp_write_queue_head(sk);



	tcp_for_write_queue_from(skb, sk) {

		if (skb == tcp_send_head(sk))

			break;

		if (!tcp_skb_timedout(sk, skb))

			break;



		tcp_skb_mark_lost(tp, skb);

	}



	tp->scoreboard_skb_hint = skb;



	tcp_verify_left_out(tp);

}



/* Detect loss in event "A" above by marking head of queue up as lost.

 * For FACK or non-SACK(Reno) senders, the first "packets" number of segments

 * are considered lost. For RFC3517 SACK, a segment is considered lost if it
@@ -2249,8 +2189,6 @@ static void tcp_update_scoreboard(struct sock *sk, int fast_rexmit) 
		else if (fast_rexmit)

			tcp_mark_head_lost(sk, 1, 1);

	}



	tcp_timeout_skbs(sk);

}



/* CWND moderation, preventing bursts due to too big ACKs
@@ -2842,7 +2780,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked, 
		fast_rexmit = 1;

	}



	if (do_lost || (tcp_is_fack(tp) && tcp_head_timedout(sk)))

	if (do_lost)

		tcp_update_scoreboard(sk, fast_rexmit);

	tcp_cwnd_reduction(sk, newly_acked_sacked, fast_rexmit);

	tcp_xmit_retransmit_queue(sk);
@@ -3075,7 +3013,6 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets, 


		tcp_unlink_write_queue(skb, sk);

		sk_wmem_free_skb(sk, skb);

		tp->scoreboard_skb_hint = NULL;

		if (skb == tp->retransmit_skb_hint)

			tp->retransmit_skb_hint = NULL;

		if (skb == tp->lost_skb_hint)

CRA Git | Maintained and supported by SUSTech CRA and CCSE