1 files changed, 82 insertions, 38 deletions

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 4522579aaca2..be6d22b8190f 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: GPL-2.0-only
 /*
  * INET		An implementation of the TCP/IP protocol suite for the LINUX
  *		operating system.  INET is implemented using the  BSD Socket
@@ -52,12 +53,8 @@ void tcp_mstamp_refresh(struct tcp_sock *tp)
 {
 	u64 val = tcp_clock_ns();
 
-	if (val > tp->tcp_clock_cache)
-		tp->tcp_clock_cache = val;
-
-	val = div_u64(val, NSEC_PER_USEC);
-	if (val > tp->tcp_mstamp)
-		tp->tcp_mstamp = val;
+	tp->tcp_clock_cache = val;
+	tp->tcp_mstamp = div_u64(val, NSEC_PER_USEC);
 }
 
 static bool tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle,
@@ -70,7 +67,7 @@ static void tcp_event_new_data_sent(struct sock *sk, struct sk_buff *skb)
 	struct tcp_sock *tp = tcp_sk(sk);
 	unsigned int prior_packets = tp->packets_out;
 
-	tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
+	WRITE_ONCE(tp->snd_nxt, TCP_SKB_CB(skb)->end_seq);
 
 	__skb_unlink(skb, &sk->sk_write_queue);
 	tcp_rbtree_insert(&sk->tcp_rtx_queue, skb);
@@ -1053,11 +1050,22 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 	tcb = TCP_SKB_CB(skb);
 	memset(&opts, 0, sizeof(opts));
 
-	if (unlikely(tcb->tcp_flags & TCPHDR_SYN))
+	if (unlikely(tcb->tcp_flags & TCPHDR_SYN)) {
 		tcp_options_size = tcp_syn_options(sk, skb, &opts, &md5);
-	else
+	} else {
 		tcp_options_size = tcp_established_options(sk, skb, &opts,
 							   &md5);
+		/* Force a PSH flag on all (GSO) packets to expedite GRO flush
+		 * at receiver : This slightly improve GRO performance.
+		 * Note that we do not force the PSH flag for non GSO packets,
+		 * because they might be sent under high congestion events,
+		 * and in this case it is better to delay the delivery of 1-MSS
+		 * packets and thus the corresponding ACK packet that would
+		 * release the following packet.
+		 */
+		if (tcp_skb_pcount(skb) > 1)
+			tcb->tcp_flags |= TCPHDR_PSH;
+	}
 	tcp_header_size = tcp_options_size + sizeof(struct tcphdr);
 
 	/* if no packet is in qdisc/device queue, then allow XPS to select
@@ -1156,6 +1164,8 @@ 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)) {
@@ -1186,10 +1196,10 @@ static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb)
 	struct tcp_sock *tp = tcp_sk(sk);
 
 	/* Advance write_seq and place onto the write_queue. */
-	tp->write_seq = TCP_SKB_CB(skb)->end_seq;
+	WRITE_ONCE(tp->write_seq, TCP_SKB_CB(skb)->end_seq);
 	__skb_header_release(skb);
 	tcp_add_write_queue_tail(sk, skb);
-	sk->sk_wmem_queued += skb->truesize;
+	sk_wmem_queued_add(sk, skb->truesize);
 	sk_mem_charge(sk, skb->truesize);
 }
 
@@ -1289,6 +1299,7 @@ int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue,
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *buff;
 	int nsize, old_factor;
+	long limit;
 	int nlen;
 	u8 flags;
 
@@ -1299,6 +1310,20 @@ int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue,
 	if (nsize < 0)
 		nsize = 0;
 
+	/* tcp_sendmsg() can overshoot sk_wmem_queued by one full size skb.
+	 * We need some allowance to not penalize applications setting small
+	 * SO_SNDBUF values.
+	 * Also allow first and last skb in retransmit queue to be split.
+	 */
+	limit = sk->sk_sndbuf + 2 * SKB_TRUESIZE(GSO_MAX_SIZE);
+	if (unlikely((sk->sk_wmem_queued >> 1) > limit &&
+		     tcp_queue != TCP_FRAG_IN_WRITE_QUEUE &&
+		     skb != tcp_rtx_queue_head(sk) &&
+		     skb != tcp_rtx_queue_tail(sk))) {
+		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPWQUEUETOOBIG);
+		return -ENOMEM;
+	}
+
 	if (skb_unclone(skb, gfp))
 		return -ENOMEM;
 
@@ -1306,8 +1331,9 @@ int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue,
 	buff = sk_stream_alloc_skb(sk, nsize, gfp, true);
 	if (!buff)
 		return -ENOMEM; /* We'll just try again later. */
+	skb_copy_decrypted(buff, skb);
 
-	sk->sk_wmem_queued += buff->truesize;
+	sk_wmem_queued_add(sk, buff->truesize);
 	sk_mem_charge(sk, buff->truesize);
 	nlen = skb->len - len - nsize;
 	buff->truesize += nlen;
@@ -1388,7 +1414,7 @@ static int __pskb_trim_head(struct sk_buff *skb, int len)
 		} else {
 			shinfo->frags[k] = shinfo->frags[i];
 			if (eat) {
-				shinfo->frags[k].page_offset += eat;
+				skb_frag_off_add(&shinfo->frags[k], eat);
 				skb_frag_size_sub(&shinfo->frags[k], eat);
 				eat = 0;
 			}
@@ -1417,7 +1443,7 @@ int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
 
 	if (delta_truesize) {
 		skb->truesize	   -= delta_truesize;
-		sk->sk_wmem_queued -= delta_truesize;
+		sk_wmem_queued_add(sk, -delta_truesize);
 		sk_mem_uncharge(sk, delta_truesize);
 		sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
 	}
@@ -1457,8 +1483,7 @@ static inline int __tcp_mtu_to_mss(struct sock *sk, int pmtu)
 	mss_now -= icsk->icsk_ext_hdr_len;
 
 	/* Then reserve room for full set of TCP options and 8 bytes of data */
-	if (mss_now < 48)
-		mss_now = 48;
+	mss_now = max(mss_now, sock_net(sk)->ipv4.sysctl_tcp_min_snd_mss);
 	return mss_now;
 }
 
@@ -1861,8 +1886,9 @@ static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len,
 	buff = sk_stream_alloc_skb(sk, 0, gfp, true);
 	if (unlikely(!buff))
 		return -ENOMEM;
+	skb_copy_decrypted(buff, skb);
 
-	sk->sk_wmem_queued += buff->truesize;
+	sk_wmem_queued_add(sk, buff->truesize);
 	sk_mem_charge(sk, buff->truesize);
 	buff->truesize += nlen;
 	skb->truesize -= nlen;
@@ -2038,7 +2064,7 @@ static bool tcp_can_coalesce_send_queue_head(struct sock *sk, int len)
 		if (len <= skb->len)
 			break;
 
-		if (unlikely(TCP_SKB_CB(skb)->eor))
+		if (unlikely(TCP_SKB_CB(skb)->eor) || tcp_has_tx_tstamp(skb))
 			return false;
 
 		len -= skb->len;
@@ -2126,10 +2152,11 @@ static int tcp_mtu_probe(struct sock *sk)
 	nskb = sk_stream_alloc_skb(sk, probe_size, GFP_ATOMIC, false);
 	if (!nskb)
 		return -1;
-	sk->sk_wmem_queued += nskb->truesize;
+	sk_wmem_queued_add(sk, nskb->truesize);
 	sk_mem_charge(sk, nskb->truesize);
 
 	skb = tcp_send_head(sk);
+	skb_copy_decrypted(nskb, skb);
 
 	TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(skb)->seq;
 	TCP_SKB_CB(nskb)->end_seq = TCP_SKB_CB(skb)->seq + probe_size;
@@ -2154,6 +2181,7 @@ static int tcp_mtu_probe(struct sock *sk)
 			 * we need to propagate it to the new skb.
 			 */
 			TCP_SKB_CB(nskb)->eor = TCP_SKB_CB(skb)->eor;
+			tcp_skb_collapse_tstamp(nskb, skb);
 			tcp_unlink_write_queue(skb, sk);
 			sk_wmem_free_skb(sk, skb);
 		} else {
@@ -2237,6 +2265,18 @@ 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,
@@ -2442,7 +2482,7 @@ bool tcp_schedule_loss_probe(struct sock *sk, bool advancing_rto)
 	/* Don't do any loss probe on a Fast Open connection before 3WHS
 	 * finishes.
 	 */
-	if (tp->fastopen_rsk)
+	if (rcu_access_pointer(tp->fastopen_rsk))
 		return false;
 
 	early_retrans = sock_net(sk)->ipv4.sysctl_tcp_early_retrans;
@@ -2750,7 +2790,7 @@ static bool tcp_collapse_retrans(struct sock *sk, struct sk_buff *skb)
 		if (next_skb_size <= skb_availroom(skb))
 			skb_copy_bits(next_skb, 0, skb_put(skb, next_skb_size),
 				      next_skb_size);
-		else if (!skb_shift(skb, next_skb, next_skb_size))
+		else if (!tcp_skb_shift(skb, next_skb, 1, next_skb_size))
 			return false;
 	}
 	tcp_highest_sack_replace(sk, next_skb, skb);
@@ -3092,7 +3132,6 @@ void tcp_send_fin(struct sock *sk)
 		tskb = skb_rb_last(&sk->tcp_rtx_queue);
 
 	if (tskb) {
-coalesce:
 		TCP_SKB_CB(tskb)->tcp_flags |= TCPHDR_FIN;
 		TCP_SKB_CB(tskb)->end_seq++;
 		tp->write_seq++;
@@ -3103,16 +3142,14 @@ coalesce:
 			 * if FIN had been sent. This is because retransmit path
 			 * does not change tp->snd_nxt.
 			 */
-			tp->snd_nxt++;
+			WRITE_ONCE(tp->snd_nxt, tp->snd_nxt + 1);
 			return;
 		}
 	} else {
 		skb = alloc_skb_fclone(MAX_TCP_HEADER, sk->sk_allocation);
-		if (unlikely(!skb)) {
-			if (tskb)
-				goto coalesce;
+		if (unlikely(!skb))
 			return;
-		}
+
 		INIT_LIST_HEAD(&skb->tcp_tsorted_anchor);
 		skb_reserve(skb, MAX_TCP_HEADER);
 		sk_forced_mem_schedule(sk, skb->truesize);
@@ -3185,7 +3222,7 @@ int tcp_send_synack(struct sock *sk)
 			tcp_rtx_queue_unlink_and_free(skb, sk);
 			__skb_header_release(nskb);
 			tcp_rbtree_insert(&sk->tcp_rtx_queue, nskb);
-			sk->sk_wmem_queued += nskb->truesize;
+			sk_wmem_queued_add(sk, nskb->truesize);
 			sk_mem_charge(sk, nskb->truesize);
 			skb = nskb;
 		}
@@ -3218,6 +3255,7 @@ 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)) {
@@ -3249,12 +3287,17 @@ 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);
+		skb->skb_mstamp_ns = cookie_init_timestamp(req, now);
 	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);
+	}
 
 #ifdef CONFIG_TCP_MD5SIG
 	rcu_read_lock();
@@ -3294,8 +3337,9 @@ 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);
@@ -3382,14 +3426,14 @@ static void tcp_connect_init(struct sock *sk)
 	tp->snd_una = tp->write_seq;
 	tp->snd_sml = tp->write_seq;
 	tp->snd_up = tp->write_seq;
-	tp->snd_nxt = tp->write_seq;
+	WRITE_ONCE(tp->snd_nxt, tp->write_seq);
 
 	if (likely(!tp->repair))
 		tp->rcv_nxt = 0;
 	else
 		tp->rcv_tstamp = tcp_jiffies32;
 	tp->rcv_wup = tp->rcv_nxt;
-	tp->copied_seq = tp->rcv_nxt;
+	WRITE_ONCE(tp->copied_seq, tp->rcv_nxt);
 
 	inet_csk(sk)->icsk_rto = tcp_timeout_init(sk);
 	inet_csk(sk)->icsk_retransmits = 0;
@@ -3403,9 +3447,9 @@ static void tcp_connect_queue_skb(struct sock *sk, struct sk_buff *skb)
 
 	tcb->end_seq += skb->len;
 	__skb_header_release(skb);
-	sk->sk_wmem_queued += skb->truesize;
+	sk_wmem_queued_add(sk, skb->truesize);
 	sk_mem_charge(sk, skb->truesize);
-	tp->write_seq = tcb->end_seq;
+	WRITE_ONCE(tp->write_seq, tcb->end_seq);
 	tp->packets_out += tcp_skb_pcount(skb);
 }
 
@@ -3542,11 +3586,11 @@ int tcp_connect(struct sock *sk)
 	/* We change tp->snd_nxt after the tcp_transmit_skb() call
 	 * in order to make this packet get counted in tcpOutSegs.
 	 */
-	tp->snd_nxt = tp->write_seq;
+	WRITE_ONCE(tp->snd_nxt, tp->write_seq);
 	tp->pushed_seq = tp->write_seq;
 	buff = tcp_send_head(sk);
 	if (unlikely(buff)) {
-		tp->snd_nxt	= TCP_SKB_CB(buff)->seq;
+		WRITE_ONCE(tp->snd_nxt, TCP_SKB_CB(buff)->seq);
 		tp->pushed_seq	= TCP_SKB_CB(buff)->seq;
 	}
 	TCP_INC_STATS(sock_net(sk), TCP_MIB_ACTIVEOPENS);