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: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 52 insertions, 8 deletions

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bd0856ac680a..5542e3d778e6 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -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;
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index f059fbd81a84..1b7e9e1fbd3b 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -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;
+	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,
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index 11011e8386dc..8bcaf2586b68 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -274,7 +274,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo)
 		tcptw->tw_ts_recent_stamp = tp->rx_opt.ts_recent_stamp;
 		tcptw->tw_ts_offset	= tp->tsoffset;
 		tcptw->tw_last_oow_ack_time = 0;
-
+		tcptw->tw_tx_delay	= tp->tcp_tx_delay;
 #if IS_ENABLED(CONFIG_IPV6)
 		if (tw->tw_family == PF_INET6) {
 			struct ipv6_pinfo *np = inet6_sk(sk);
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index f429e856e263..d954ff9069e8 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -1153,6 +1153,8 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 	memset(skb->cb, 0, max(sizeof(struct inet_skb_parm),
 			       sizeof(struct inet6_skb_parm)));
 
+	tcp_add_tx_delay(skb, tp);
+
 	err = icsk->icsk_af_ops->queue_xmit(sk, skb, &inet->cork.fl);
 
 	if (unlikely(err > 0)) {
@@ -2234,6 +2236,18 @@ static bool tcp_small_queue_check(struct sock *sk, const struct sk_buff *skb,
 			      sock_net(sk)->ipv4.sysctl_tcp_limit_output_bytes);
 	limit <<= factor;
 
+	if (static_branch_unlikely(&tcp_tx_delay_enabled) &&
+	    tcp_sk(sk)->tcp_tx_delay) {
+		u64 extra_bytes = (u64)sk->sk_pacing_rate * tcp_sk(sk)->tcp_tx_delay;
+
+		/* TSQ is based on skb truesize sum (sk_wmem_alloc), so we
+		 * approximate our needs assuming an ~100% skb->truesize overhead.
+		 * USEC_PER_SEC is approximated by 2^20.
+		 * do_div(extra_bytes, USEC_PER_SEC/2) is replaced by a right shift.
+		 */
+		extra_bytes >>= (20 - 1);
+		limit += extra_bytes;
+	}
 	if (refcount_read(&sk->sk_wmem_alloc) > limit) {
 		/* Always send skb if rtx queue is empty.
 		 * No need to wait for TX completion to call us back,
@@ -3212,6 +3226,7 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,
 	int tcp_header_size;
 	struct tcphdr *th;
 	int mss;
+	u64 now;
 
 	skb = alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC);
 	if (unlikely(!skb)) {
@@ -3243,13 +3258,14 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,
 	mss = tcp_mss_clamp(tp, dst_metric_advmss(dst));
 
 	memset(&opts, 0, sizeof(opts));
+	now = tcp_clock_ns();
 #ifdef CONFIG_SYN_COOKIES
 	if (unlikely(req->cookie_ts))
 		skb->skb_mstamp_ns = cookie_init_timestamp(req);
 	else
 #endif
 	{
-		skb->skb_mstamp_ns = tcp_clock_ns();
+		skb->skb_mstamp_ns = now;
 		if (!tcp_rsk(req)->snt_synack) /* Timestamp first SYNACK */
 			tcp_rsk(req)->snt_synack = tcp_skb_timestamp_us(skb);
 	}
@@ -3292,8 +3308,9 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,
 	rcu_read_unlock();
 #endif
 
-	/* Do not fool tcpdump (if any), clean our debris */
-	skb->tstamp = 0;
+	skb->skb_mstamp_ns = now;
+	tcp_add_tx_delay(skb, tp);
+
 	return skb;
 }
 EXPORT_SYMBOL(tcp_make_synack);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index ad7039137a20..5606b2131b65 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -892,6 +892,7 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32
 		} else {
 			mark = sk->sk_mark;
 		}
+		tcp_set_tx_time(buff, sk);
 	}
 	fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark) ?: mark;
 	fl6.fl6_dport = t1->dest;