From: Eric Dumazet <edumazet@google.com>
Date: Fri, 24 Oct 2025 12:07:07 +0000 (+0000)
Subject: tcp: remove one ktime_get() from recvmsg() fast path
X-Git-Tag: v6.19-rc1~170^2~302
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=0ae1ac7335ca2328c42fbbe2b71bc9c3fc8e65d9;p=thirdparty%2Flinux.git

tcp: remove one ktime_get() from recvmsg() fast path

Each time some payload is consumed by user space (recvmsg() and friends),
TCP calls tcp_rcv_space_adjust() to run DRS algorithm to check
if an increase of sk->sk_rcvbuf is needed.

This function is based on time sampling, and currently calls
tcp_mstamp_refresh(tp), which is a wrapper around ktime_get_ns().

ktime_get_ns() has a high cost on some platforms.
100+ cycles for rdtscp on AMD EPYC Turin for instance.

We do not have to refresh tp->tcp_mpstamp, using the last cached value
is enough. We only need to refresh it from __tcp_cleanup_rbuf()
if an ACK must be sent (this is a rare event).

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20251024120707.3516550-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index b79da6d393927..a9345aa5a2e5f 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1556,8 +1556,10 @@ void __tcp_cleanup_rbuf(struct sock *sk, int copied)
 				time_to_ack = true;
 		}
 	}
-	if (time_to_ack)
+	if (time_to_ack) {
+		tcp_mstamp_refresh(tp);
 		tcp_send_ack(sk);
+	}
 }
 
 void tcp_cleanup_rbuf(struct sock *sk, int copied)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 8fc97f4d8a6b2..ff19f6e54d55c 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -928,9 +928,15 @@ void tcp_rcv_space_adjust(struct sock *sk)
 
 	trace_tcp_rcv_space_adjust(sk);
 
-	tcp_mstamp_refresh(tp);
+	if (unlikely(!tp->rcv_rtt_est.rtt_us))
+		return;
+
+	/* We do not refresh tp->tcp_mstamp here.
+	 * Some platforms have expensive ktime_get() implementations.
+	 * Using the last cached value is enough for DRS.
+	 */
 	time = tcp_stamp_us_delta(tp->tcp_mstamp, tp->rcvq_space.time);
-	if (time < (tp->rcv_rtt_est.rtt_us >> 3) || tp->rcv_rtt_est.rtt_us == 0)
+	if (time < (tp->rcv_rtt_est.rtt_us >> 3))
 		return;
 
 	/* Number of bytes copied to user in last RTT */