tcp: add tcp_rto_max_ms sysctl

JIRA: https://issues.redhat.com/browse/RHEL-115191
Upstream Status: linux.git
Conflicts:\
- Minor chunk difference due to missing upstream commit f086ede ("tcp: add
  sysctl_tcp_rto_min_us") in c9s.
- Context differences due to missing upstream commits 54b771e ("doc: net:
  Fix .rst rendering of net_cachelines pages"), f086ede ("tcp: add
  sysctl_tcp_rto_min_us"), 1c106eb ("net: ipv{6,4}: Remove the now
  superfluous sentinel elements from ctl_table array") and 562b1fd ("tcp:
  Set pingpong threshold via sysctl") in c9s.

commit 1280c26
Author: Eric Dumazet <edumazet@google.com>
Date:   Fri Feb 7 15:28:30 2025 +0000

    tcp: add tcp_rto_max_ms sysctl

    Previous patch added a TCP_RTO_MAX_MS socket option
    to tune a TCP socket max RTO value.

    Many setups prefer to change a per netns sysctl.

    This patch adds /proc/sys/net/ipv4/tcp_rto_max_ms

    Its initial value is 120000 (120 seconds).

    Keep in mind that a decrease of tcp_rto_max_ms
    means shorter overall timeouts, unless tcp_retries2
    sysctl is increased.

    Signed-off-by: Eric Dumazet <edumazet@google.com>
    Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
    Reviewed-by: Neal Cardwell <ncardwell@google.com>
    Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
    Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Signed-off-by: Antoine Tenart <atenart@redhat.com>

Author: atenart

Documentation/networking/ip-sysctl.rst

@@ -697,6 +697,8 @@ tcp_retries2 - INTEGER
 	seconds and is a lower bound for the effective timeout.
 	TCP will effectively time out at the first RTO which exceeds the
 	hypothetical timeout.
+	If tcp_rto_max_ms is decreased, it is recommended to also
+	change tcp_retries2.
 
 	RFC 1122 recommends at least 100 seconds for the timeout,
 	which corresponds to a value of at least 8.
@@ -1041,6 +1043,17 @@ tcp_challenge_ack_limit - INTEGER
 	in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
 	Default: 1000
 
+tcp_rto_max_ms - INTEGER
+	Maximal TCP retransmission timeout (in ms).
+	Note that TCP_RTO_MAX_MS socket option has higher precedence.
+
+	When changing tcp_rto_max_ms, it is important to understand
+	that tcp_retries2 might need a change.
+
+	Possible Values: 1000 - 120,000
+
+	Default: 120,000
+
 UDP variables
 =============
 

Documentation/networking/net_cachelines/netns_ipv4_sysctl.rst

@@ -84,6 +84,7 @@ u8                              sysctl_tcp_sack                              -
 u8                              sysctl_tcp_window_scaling                    -                   -                   tcp_syn_options,tcp_parse_options
 u8                              sysctl_tcp_timestamps                                                                
 u8                              sysctl_tcp_early_retrans                     read_mostly         -                   tcp_schedule_loss_probe(tcp_write_xmit)
+u32                             sysctl_tcp_rto_max_ms                        -                   -
 u8                              sysctl_tcp_recovery                          -                   -                   tcp_fastretrans_alert
 u8                              sysctl_tcp_thin_linear_timeouts              -                   -                   tcp_retrans_timer(on_thin_streams)
 u8                              sysctl_tcp_slow_start_after_idle             -                   -                   unlikely(tcp_cwnd_validate-network-not-starved)

include/net/netns/ipv4.h

@@ -171,6 +171,7 @@ struct netns_ipv4 {
 	u8 sysctl_tcp_sack;
 	u8 sysctl_tcp_window_scaling;
 	u8 sysctl_tcp_timestamps;
+	int sysctl_tcp_rto_max_ms;
 	u8 sysctl_tcp_recovery;
 	u8 sysctl_tcp_thin_linear_timeouts;
 	u8 sysctl_tcp_slow_start_after_idle;

net/ipv4/sysctl_net_ipv4.c

@@ -41,6 +41,7 @@ static int tcp_adv_win_scale_max = 31;
 static int tcp_app_win_max = 31;
 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
 static int tcp_min_snd_mss_max = 65535;
+static int tcp_rto_max_max = TCP_RTO_MAX_SEC * MSEC_PER_SEC;
 static int ip_privileged_port_min;
 static int ip_privileged_port_max = 65535;
 static int ip_ttl_min = 1;
@@ -1425,6 +1426,15 @@ static struct ctl_table ipv4_net_table[] = {
 		.extra1		= SYSCTL_ZERO,
 		.extra2		= SYSCTL_ONE,
 	},
+	{
+		.procname	= "tcp_rto_max_ms",
+		.data		= &init_net.ipv4.sysctl_tcp_rto_max_ms,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= SYSCTL_ONE_THOUSAND,
+		.extra2		= &tcp_rto_max_max,
+	},
 	{ }
 };
 

net/ipv4/tcp.c

@@ -417,6 +417,7 @@ void tcp_init_sock(struct sock *sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
+	int rto_max_ms;
 
 	tp->out_of_order_queue = RB_ROOT;
 	sk->tcp_rtx_queue = RB_ROOT;
@@ -426,8 +427,8 @@ void tcp_init_sock(struct sock *sk)
 
 	icsk->icsk_rto = TCP_TIMEOUT_INIT;
 
-	/* Use a sysctl ? */
-	icsk->icsk_rto_max = TCP_RTO_MAX;
+	rto_max_ms = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rto_max_ms);
+	icsk->icsk_rto_max = msecs_to_jiffies(rto_max_ms);
 
 	icsk->icsk_rto_min = TCP_RTO_MIN;
 

net/ipv4/tcp_ipv4.c

@@ -3222,6 +3222,8 @@ static int __net_init tcp_sk_init(struct net *net)
 
 	net->ipv4.sysctl_tcp_shrink_window = 0;
 
+	net->ipv4.sysctl_tcp_rto_max_ms = TCP_RTO_MAX_SEC * MSEC_PER_SEC;
+
 	return 0;
 fail:
 	tcp_sk_exit(net);