sysctl.conf: Enable TFO in sysctl

Message ID 20181214120332.5372-1-ummeegge@ipfire.org
State Dropped
Headers
Series sysctl.conf: Enable TFO in sysctl |

Commit Message

ummeegge Dec. 14, 2018, 11:03 p.m. UTC
  Fixes #11945

This do not enables TFO support in general there is still the execution of
echo 3 > /proc/sys/net/ipv4/tcp_fastopen
needed after every reboot (rc.local e.g.).

For further information see:
https://tools.ietf.org/html/rfc7413#section-4.2.2
https://www.kernel.org/doc/Documentation/networking/ip-sysctl.txt

Signed-off-by: erik.kapfer <ummeegge@ipfire.org>
---
 config/etc/sysctl.conf | 1 +
 1 file changed, 1 insertion(+)
  

Comments

Michael Tremer Dec. 15, 2018, 1:59 a.m. UTC | #1
Hi,

> On 14 Dec 2018, at 12:03, erik.kapfer <ummeegge@ipfire.org> wrote:
> 
> Fixes #11945
> 
> This do not enables TFO support in general there is still the execution of
> echo 3 > /proc/sys/net/ipv4/tcp_fastopen
> needed after every reboot (rc.local e.g.).
> 

Why does this not enable it? Setting that value to 3 is what the sysctl command does.

I am confused.

> For further information see:
> https://tools.ietf.org/html/rfc7413#section-4.2.2
> https://www.kernel.org/doc/Documentation/networking/ip-sysctl.txt
> 
> Signed-off-by: erik.kapfer <ummeegge@ipfire.org>
> ---
> config/etc/sysctl.conf | 1 +
> 1 file changed, 1 insertion(+)
> 
> diff --git a/config/etc/sysctl.conf b/config/etc/sysctl.conf
> index 4066af767..52b21efa4 100644
> --- a/config/etc/sysctl.conf
> +++ b/config/etc/sysctl.conf
> @@ -13,6 +13,7 @@ net.ipv4.tcp_syncookies = 1
> net.ipv4.tcp_fin_timeout = 30
> net.ipv4.tcp_syn_retries = 3
> net.ipv4.tcp_synack_retries = 3
> +net.ipv4.tcp_fastopen = 3
> 
> net.ipv4.conf.default.arp_filter = 1
> net.ipv4.conf.default.rp_filter = 0
> -- 
> 2.12.2
>
  
ummeegge Dec. 15, 2018, 3:41 a.m. UTC | #2
Hi Michael,

Am Freitag, den 14.12.2018, 14:59 +0000 schrieb Michael Tremer:
> Hi,
> 
> > On 14 Dec 2018, at 12:03, erik.kapfer <ummeegge@ipfire.org> wrote:
> > 
> > Fixes #11945
> > 
> > This do not enables TFO support in general there is still the
> > execution of
> > echo 3 > /proc/sys/net/ipv4/tcp_fastopen
> > needed after every reboot (rc.local e.g.).
> > 
> 
> Why does this not enable it? Setting that value to 3 is what the
> sysctl command does.
> 
> I am confused.

you are right, mixed there testings up but used also old descriptions.
There is no need to echo '3' to tcp_fastopen to survive reboots.
Should i amend the patch and correct the commit message ?

Did now some tests with OpenSSL-1.1.1a whereby unbound includes the TFO
configure options and DoT  seems *really* much faster then DoT on
another system without TFO support for unbound and OpenSSL-1.1.0i but
am currently not able to find some TFO usage evidence except the TFO
key

$ cat /proc/sys/net/ipv4/tcp_fastopen_key                
750532b8-36e6eb1d-800cb58e-3008f1f1

Monitoring examples like in here -->
https://blog.wasin.io/blog/2016/12/26/how-to-enable-fast-tcp-open-on-ubuntu.html
didnĀ“t deliver any results but they are also old 
(echo 3 > /proc/sys/net/ipv4/tcp_fastopen) is in this description 
also included which is outdated, possibly the monitoring examples are too.

Best,

Erik
  

Patch

diff --git a/config/etc/sysctl.conf b/config/etc/sysctl.conf
index 4066af767..52b21efa4 100644
--- a/config/etc/sysctl.conf
+++ b/config/etc/sysctl.conf
@@ -13,6 +13,7 @@  net.ipv4.tcp_syncookies = 1
 net.ipv4.tcp_fin_timeout = 30
 net.ipv4.tcp_syn_retries = 3
 net.ipv4.tcp_synack_retries = 3
+net.ipv4.tcp_fastopen = 3
 
 net.ipv4.conf.default.arp_filter = 1
 net.ipv4.conf.default.rp_filter = 0