tcp: add a sysctl to config the tcp_default_init_rwnd

The default initial rwnd is hardcoded to 10.

Now we allow it to be controlled via
  /proc/sys/net/ipv4/tcp_default_init_rwnd
which limits the values from 3 to 100

This is somewhat needed because ipv6 routes are
autoconfigured by the kernel.

See "An Argument for Increasing TCP's Initial Congestion Window"
in https://developers.google.com/speed/articles/tcp_initcwnd_paper.pdf

Change-Id: I386b2a9d62de0ebe05c1ebe1b4bd91b314af5c54
Signed-off-by: JP Abgrall <jpa@google.com>

Conflicts:
	net/ipv4/sysctl_net_ipv4.c
	net/ipv4/tcp_input.c

3 files changed, 24 insertions, 1 deletions

diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index a0bd7a55193e..dd825be50089 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -152,6 +152,21 @@ static int ipv4_ping_group_range(struct ctl_table *table, int write,
 	return ret;
 }
 
+/* Validate changes from /proc interface. */
+static int proc_tcp_default_init_rwnd(ctl_table *ctl, int write,
+				      void __user *buffer,
+				      size_t *lenp, loff_t *ppos)
+{
+	int old_value = *(int *)ctl->data;
+	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
+	int new_value = *(int *)ctl->data;
+
+	if (write && ret == 0 && (new_value < 3 || new_value > 100))
+		*(int *)ctl->data = old_value;
+
+	return ret;
+}
+
 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 				       void __user *buffer, size_t *lenp, loff_t *ppos)
 {
@@ -761,6 +776,13 @@ static struct ctl_table ipv4_table[] = {
 		.proc_handler	= proc_dointvec_ms_jiffies,
 	},
 	{
+		.procname       = "tcp_default_init_rwnd",
+		.data           = &sysctl_tcp_default_init_rwnd,
+		.maxlen         = sizeof(int),
+		.mode           = 0644,
+		.proc_handler   = proc_tcp_default_init_rwnd
+	},
+	{
 		.procname	= "icmp_msgs_per_sec",
 		.data		= &sysctl_icmp_msgs_per_sec,
 		.maxlen		= sizeof(int),
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index d4c51158470f..da34f830f4bc 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -102,6 +102,7 @@ int sysctl_tcp_thin_dupack __read_mostly;
 int sysctl_tcp_moderate_rcvbuf __read_mostly = 1;
 int sysctl_tcp_early_retrans __read_mostly = 3;
 int sysctl_tcp_invalid_ratelimit __read_mostly = HZ/2;
+int sysctl_tcp_default_init_rwnd __read_mostly = TCP_INIT_CWND * 2;
 
 #define FLAG_DATA		0x01 /* Incoming frame contained data.		*/
 #define FLAG_WIN_UPDATE		0x02 /* Incoming ACK was a window update.	*/
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 9bfc39ff2285..9f069bd9de46 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -191,7 +191,7 @@ u32 tcp_default_init_rwnd(u32 mss)
 	 * (RFC 3517, Section 4, NextSeg() rule (2)). Further place a
 	 * limit when mss is larger than 1460.
 	 */
-	u32 init_rwnd = TCP_INIT_CWND * 2;
+	u32 init_rwnd = sysctl_tcp_default_init_rwnd;
 
 	if (mss > 1460)
 		init_rwnd = max((1460 * init_rwnd) / mss, 2U);