Commit a842fe14 authored by Eric Dumazet's avatar Eric Dumazet Committed by David S. Miller
Browse files

tcp: add optional per socket transmit delay 



Adding delays to TCP flows is crucial for studying behavior
of TCP stacks, including congestion control modules.

Linux offers netem module, but it has unpractical constraints :
- Need root access to change qdisc
- Hard to setup on egress if combined with non trivial qdisc like FQ
- Single delay for all flows.

EDT (Earliest Departure Time) adoption in TCP stack allows us
to enable a per socket delay at a very small cost.

Networking tools can now establish thousands of flows, each of them
with a different delay, simulating real world conditions.

This requires FQ packet scheduler or a EDT-enabled NIC.

This patchs adds TCP_TX_DELAY socket option, to set a delay in
usec units.

  unsigned int tx_delay = 10000; /* 10 msec */

  setsockopt(fd, SOL_TCP, TCP_TX_DELAY, &tx_delay, sizeof(tx_delay));

Note that FQ packet scheduler limits might need some tweaking :

man tc-fq

PARAMETERS
   limit
       Hard  limit  on  the  real  queue  size. When this limit is
       reached, new packets are dropped. If the value is  lowered,
       packets  are  dropped so that the new limit is met. Default
       is 10000 packets.

   flow_limit
       Hard limit on the maximum  number  of  packets  queued  per
       flow.  Default value is 100.

Use of TCP_TX_DELAY option will increase number of skbs in FQ qdisc,
so packets would be dropped if any of the previous limit is hit.

Use of a jump label makes this support runtime-free, for hosts
never using the option.

Also note that TSQ (TCP Small Queues) limits are slightly changed
with this patch : we need to account that skbs artificially delayed
wont stop us providind more skbs to feed the pipe (netem uses
skb_orphan_partial() for this purpose, but FQ can not use this trick)

Because of that, using big delays might very well trigger
old bugs in TSO auto defer logic and/or sndbuf limited detection.

Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent e0ffbd37

include/linux/tcp.h

+2 −0
Original line number Diff line number Diff line
@@ -245,6 +245,7 @@ struct tcp_sock { 
		syn_smc:1;	/* SYN includes SMC */

	u32	tlp_high_seq;	/* snd_nxt at the time of TLP retransmit. */



	u32	tcp_tx_delay;	/* delay (in usec) added to TX packets */

	u64	tcp_wstamp_ns;	/* departure time for next sent data packet */

	u64	tcp_clock_cache; /* cache last tcp_clock_ns() (see tcp_mstamp_refresh()) */
@@ -436,6 +437,7 @@ struct tcp_timewait_sock { 
	u32			  tw_last_oow_ack_time;



	int			  tw_ts_recent_stamp;

	u32			  tw_tx_delay;

#ifdef CONFIG_TCP_MD5SIG

	struct tcp_md5sig_key	  *tw_md5_key;

#endif

include/net/tcp.h

+19 −0
Original line number Diff line number Diff line
@@ -2232,4 +2232,23 @@ void clean_acked_data_disable(struct inet_connection_sock *icsk); 
void clean_acked_data_flush(void);

#endif



DECLARE_STATIC_KEY_FALSE(tcp_tx_delay_enabled);

static inline void tcp_add_tx_delay(struct sk_buff *skb,

				    const struct tcp_sock *tp)

{

	if (static_branch_unlikely(&tcp_tx_delay_enabled))

		skb->skb_mstamp_ns += (u64)tp->tcp_tx_delay * NSEC_PER_USEC;

}



static inline void tcp_set_tx_time(struct sk_buff *skb,

				   const struct sock *sk)

{

	if (static_branch_unlikely(&tcp_tx_delay_enabled)) {

		u32 delay = (sk->sk_state == TCP_TIME_WAIT) ?

			tcp_twsk(sk)->tw_tx_delay : tcp_sk(sk)->tcp_tx_delay;



		skb->skb_mstamp_ns = tcp_clock_ns() + (u64)delay * NSEC_PER_USEC;

	}

}



#endif	/* _TCP_H */

include/uapi/linux/tcp.h

+3 −0
Original line number Diff line number Diff line
@@ -127,6 +127,9 @@ enum { 


#define TCP_CM_INQ		TCP_INQ



#define TCP_TX_DELAY		37	/* delay outgoing packets by XX usec */





#define TCP_REPAIR_ON		1

#define TCP_REPAIR_OFF		0

#define TCP_REPAIR_OFF_NO_WP	-1	/* Turn off without window probes */

net/ipv4/tcp.c

+24 −0
Original line number Diff line number Diff line
@@ -2736,6 +2736,21 @@ static int tcp_repair_options_est(struct sock *sk, 
	return 0;

}



DEFINE_STATIC_KEY_FALSE(tcp_tx_delay_enabled);

EXPORT_SYMBOL(tcp_tx_delay_enabled);



static void tcp_enable_tx_delay(void)

{

	if (!static_branch_unlikely(&tcp_tx_delay_enabled)) {

		static int __tcp_tx_delay_enabled = 0;



		if (cmpxchg(&__tcp_tx_delay_enabled, 0, 1) == 0) {

			static_branch_enable(&tcp_tx_delay_enabled);

			pr_info("TCP_TX_DELAY enabled\n");

		}

	}

}



/*

 *	Socket option code for TCP.

 */
@@ -3087,6 +3102,11 @@ static int do_tcp_setsockopt(struct sock *sk, int level, 
		else

			tp->recvmsg_inq = val;

		break;

	case TCP_TX_DELAY:

		if (val)

			tcp_enable_tx_delay();

		tp->tcp_tx_delay = val;

		break;

	default:

		err = -ENOPROTOOPT;

		break;
@@ -3546,6 +3566,10 @@ static int do_tcp_getsockopt(struct sock *sk, int level, 
		val = tp->fastopen_no_cookie;

		break;



	case TCP_TX_DELAY:

		val = tp->tcp_tx_delay;

		break;



	case TCP_TIMESTAMP:

		val = tcp_time_stamp_raw() + tp->tsoffset;

		break;

net/ipv4/tcp_ipv4.c

+6 −4
Original line number Diff line number Diff line
@@ -767,9 +767,11 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb) 
	arg.uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL);

	local_bh_disable();

	ctl_sk = this_cpu_read(*net->ipv4.tcp_sk);

	if (sk)

	if (sk) {

		ctl_sk->sk_mark = (sk->sk_state == TCP_TIME_WAIT) ?

				   inet_twsk(sk)->tw_mark : sk->sk_mark;

		tcp_set_tx_time(skb, sk);

	}

	ip_send_unicast_reply(ctl_sk,

			      skb, &TCP_SKB_CB(skb)->header.h4.opt,

			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
@@ -859,9 +861,9 @@ static void tcp_v4_send_ack(const struct sock *sk, 
	arg.uid = sock_net_uid(net, sk_fullsock(sk) ? sk : NULL);

	local_bh_disable();

	ctl_sk = this_cpu_read(*net->ipv4.tcp_sk);

	if (sk)

	ctl_sk->sk_mark = (sk->sk_state == TCP_TIME_WAIT) ?

			   inet_twsk(sk)->tw_mark : sk->sk_mark;

	tcp_set_tx_time(skb, sk);

	ip_send_unicast_reply(ctl_sk,

			      skb, &TCP_SKB_CB(skb)->header.h4.opt,

			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,

CRA Git | Maintained and supported by SUSTech CRA and CCSE