Revert "tcp: get rid of sysctl_tcp_adv_win_scale"

This reverts commit dfa2f04. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
geliangtang · May 2, 2024 · 8014fce · 8014fce
1 parent 09137a5
commit 8014fce
Show file tree

Hide file tree

Showing 5 changed files with 18 additions and 20 deletions.
diff --git a/Documentation/networking/ip-sysctl.rst b/Documentation/networking/ip-sysctl.rst
@@ -321,7 +321,6 @@ tcp_abort_on_overflow - BOOLEAN
 	option can harm clients of your server.
 
 tcp_adv_win_scale - INTEGER
-	Obsolete since linux-6.6
 	Count buffering overhead as bytes/2^tcp_adv_win_scale
 	(if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
 	if it is <= 0.

diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
@@ -179,7 +179,7 @@ struct netns_ipv4 {
 	u8 sysctl_tcp_abort_on_overflow;
 	u8 sysctl_tcp_fack; /* obsolete */
 	int sysctl_tcp_max_reordering;
-	int sysctl_tcp_adv_win_scale; /* obsolete */
+	int sysctl_tcp_adv_win_scale;
 	u8 sysctl_tcp_dsack;
 	u8 sysctl_tcp_app_win;
 	u8 sysctl_tcp_frto;

diff --git a/include/net/tcp.h b/include/net/tcp.h
@@ -1519,7 +1519,11 @@ static inline int __tcp_win_from_space(u8 scaling_ratio, int space)
 
 static inline int tcp_win_from_space(const struct sock *sk, int space)
 {
-	return __tcp_win_from_space(tcp_sk(sk)->scaling_ratio, space);
+	int tcp_adv_win_scale = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_adv_win_scale);
+
+	return tcp_adv_win_scale <= 0 ?
+		(space>>(-tcp_adv_win_scale)) :
+		space - (space>>tcp_adv_win_scale);
 }
 
 /* inverse of __tcp_win_from_space() */

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
@@ -1707,7 +1707,7 @@ EXPORT_SYMBOL(tcp_peek_len);
 /* Make sure sk_rcvbuf is big enough to satisfy SO_RCVLOWAT hint */
 int tcp_set_rcvlowat(struct sock *sk, int val)
 {
-	int space, cap;
+	int cap;
 
 	if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
 		cap = sk->sk_rcvbuf >> 1;
@@ -1722,10 +1722,10 @@ int tcp_set_rcvlowat(struct sock *sk, int val)
 	if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
 		return 0;
 
-	space = tcp_space_from_win(sk, val);
-	if (space > sk->sk_rcvbuf) {
-		WRITE_ONCE(sk->sk_rcvbuf, space);
-		WRITE_ONCE(tcp_sk(sk)->window_clamp, val);
+	val <<= 1;
+	if (val > sk->sk_rcvbuf) {
+		WRITE_ONCE(sk->sk_rcvbuf, val);
+		WRITE_ONCE(tcp_sk(sk)->window_clamp, tcp_win_from_space(sk, val));
 	}
 	return 0;
 }

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
@@ -232,16 +232,6 @@ static void tcp_measure_rcv_mss(struct sock *sk, const struct sk_buff *skb)
 	 */
 	len = skb_shinfo(skb)->gso_size ? : skb->len;
 	if (len >= icsk->icsk_ack.rcv_mss) {
-		/* Note: divides are still a bit expensive.
-		 * For the moment, only adjust scaling_ratio
-		 * when we update icsk_ack.rcv_mss.
-		 */
-		if (unlikely(len != icsk->icsk_ack.rcv_mss)) {
-			u64 val = (u64)skb->len << TCP_RMEM_TO_WIN_SCALE;
-
-			do_div(val, skb->truesize);
-			tcp_sk(sk)->scaling_ratio = val ? val : 1;
-		}
 		icsk->icsk_ack.rcv_mss = min_t(unsigned int, len,
 					       tcp_sk(sk)->advmss);
 		/* Account for possibly-removed options */
@@ -756,8 +746,8 @@ void tcp_rcv_space_adjust(struct sock *sk)
 
 	if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf) &&
 	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
+		int rcvmem, rcvbuf;
 		u64 rcvwin, grow;
-		int rcvbuf;
 
 		/* minimal window to cope with packet losses, assuming
 		 * steady state. Add some cushion because of small variations.
@@ -769,7 +759,12 @@ void tcp_rcv_space_adjust(struct sock *sk)
 		do_div(grow, tp->rcvq_space.space);
 		rcvwin += (grow << 1);
 
-		rcvbuf = min_t(u64, tcp_space_from_win(sk, rcvwin),
+		rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER);
+		while (tcp_win_from_space(sk, rcvmem) < tp->advmss)
+			rcvmem += 128;
+
+		do_div(rcvwin, tp->advmss);
+		rcvbuf = min_t(u64, rcvwin * rcvmem,
 			       READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[2]));
 		if (rcvbuf > sk->sk_rcvbuf) {
 			WRITE_ONCE(sk->sk_rcvbuf, rcvbuf);