tcp & udp timeout hardcoding

tcp & udp timeout hardcoding

[xavier]

Hello,

I saw in the 2.4.22 kernel source : 

/usr/src/linux/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
static unsigned long tcp_timeouts[]
= { 30 MINS,    /*      TCP_CONNTRACK_NONE,     */
    5 DAYS,     /*      TCP_CONNTRACK_ESTABLISHED,      */
    2 MINS,     /*      TCP_CONNTRACK_SYN_SENT, */
    60 SECS,    /*      TCP_CONNTRACK_SYN_RECV, */
    2 MINS,     /*      TCP_CONNTRACK_FIN_WAIT, */
    2 MINS,     /*      TCP_CONNTRACK_TIME_WAIT,        */
    10 SECS,    /*      TCP_CONNTRACK_CLOSE,    */
    60 SECS,    /*      TCP_CONNTRACK_CLOSE_WAIT,       */
    30 SECS,    /*      TCP_CONNTRACK_LAST_ACK, */
    2 MINS,     /*      TCP_CONNTRACK_LISTEN,   */
};


ip_conntrack_proto_udp.c
#define UDP_TIMEOUT (30*HZ)
#define UDP_STREAM_TIMEOUT (180*HZ)


Is there any plan to be able to set thoses values throught /proc ?

in some cases a 5 days timeout on tcp connexions may be too long...

thanks



-- 
xavier

Re: tcp & udp timeout hardcoding

[Antony Stone]

On Monday 17 November 2003 2:42 pm, xavier wrote:

> Hello,
>
> I saw in the 2.4.22 kernel source :
>
> /usr/src/linux/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
> static unsigned long tcp_timeouts[]
> = { 30 MINS,    /*      TCP_CONNTRACK_NONE,     */
>     5 DAYS,     /*      TCP_CONNTRACK_ESTABLISHED,      */
>     2 MINS,     /*      TCP_CONNTRACK_SYN_SENT, */
>     60 SECS,    /*      TCP_CONNTRACK_SYN_RECV, */
>     2 MINS,     /*      TCP_CONNTRACK_FIN_WAIT, */
>     2 MINS,     /*      TCP_CONNTRACK_TIME_WAIT,        */
>     10 SECS,    /*      TCP_CONNTRACK_CLOSE,    */
>     60 SECS,    /*      TCP_CONNTRACK_CLOSE_WAIT,       */
>     30 SECS,    /*      TCP_CONNTRACK_LAST_ACK, */
>     2 MINS,     /*      TCP_CONNTRACK_LISTEN,   */
> };
>
> Is there any plan to be able to set thoses values throught /proc ?
>
> in some cases a 5 days timeout on tcp connexions may be too long...

These are the standard values according to the TCP RFCs.   You can change 
them on your machine if you like, but it involves a kernel recompile (or a 
netfilter module recompile, if you use modules).

Note that you can adjust these sort of timings for the normal Linux TCP/IP 
stack by writing to various things in /proc/sys/net/ipv4, but this doesn't 
affect the netfilter connection racking table timeouts.

See /usr/src/linux/Documentation/networking/ip-sysctl.txt for more details.

Bear in mind that fiddling with this sort of stuff can break your network 
connectivity and/or compatibility.

Regards,

Antony.

-- 

The first ninety percent of an engineering project takes ninety percent
of the time, and the last ten percent takes the remaining ninety percent.

                                                     Please reply to the list;
                                                           please don't CC me.

Re: tcp & udp timeout hardcoding

[Eric Leblond]

[-- Attachment #1: Type: text/plain, Size: 441 bytes --]

Le lun 17/11/2003 à 15:42, xavier a écrit :
> Hello,
> 
> I saw in the 2.4.22 kernel source : 
> 
> /usr/src/linux/net/ipv4/netfilter/ip_conntrack_proto_tcp.c

> Is there any plan to be able to set thoses values throught /proc ?

It is already available in P-O-M :
http://www.netfilter.org/documentation/pomlist/pom-combined.html#tcp-window-tracking

BR,
-- 
Eric Leblond
NuFW, Now User Filtering Works (http://www.nufw.org)

[-- Attachment #2: Ceci est une partie de message numériquement signée. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: tcp & udp timeout hardcoding

[Antony Stone]

On Monday 17 November 2003 6:09 pm, Ramin Dousti wrote:

> On Mon, Nov 17, 2003 at 04:30:05PM +0000, Antony Stone wrote:
> > On Monday 17 November 2003 2:42 pm, xavier wrote:
> > > Hello,
> > >
> > > I saw in the 2.4.22 kernel source :
> > >
> > > /usr/src/linux/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
> > > static unsigned long tcp_timeouts[]
> > > = { 30 MINS,    /*      TCP_CONNTRACK_NONE,     */
> > >     5 DAYS,     /*      TCP_CONNTRACK_ESTABLISHED,      */
> > >     2 MINS,     /*      TCP_CONNTRACK_SYN_SENT, */
> > >     60 SECS,    /*      TCP_CONNTRACK_SYN_RECV, */
> > >     2 MINS,     /*      TCP_CONNTRACK_FIN_WAIT, */
> > >     2 MINS,     /*      TCP_CONNTRACK_TIME_WAIT,        */
> > >     10 SECS,    /*      TCP_CONNTRACK_CLOSE,    */
> > >     60 SECS,    /*      TCP_CONNTRACK_CLOSE_WAIT,       */
> > >     30 SECS,    /*      TCP_CONNTRACK_LAST_ACK, */
> > >     2 MINS,     /*      TCP_CONNTRACK_LISTEN,   */
> > > };
> > >
> > > Is there any plan to be able to set thoses values throught /proc ?
> > >
> > > in some cases a 5 days timeout on tcp connexions may be too long...
> >
> > These are the standard values according to the TCP RFCs.
>
> Which RFC talks about TCP_CONNTRACK_ESTABLISHED, again? I don't think
> it's written yet ;-)

Agreed :)   The above table is obviously an extended version of the timeouts 
used in a normal TCP/IP stack, with the addition of things only needed for a 
connection tracking system.

The timeouts are based, I believe, on RFC 793 (esp. the wonderful diagram on 
page 23).

Antony.

-- 

Wanted: telepath.   You know where to apply.

                                                     Please reply to the list;
                                                           please don't CC me.

Re: tcp & udp timeout hardcoding

[Ramin Dousti]

On Mon, Nov 17, 2003 at 04:30:05PM +0000, Antony Stone wrote:

> On Monday 17 November 2003 2:42 pm, xavier wrote:
> 
> > Hello,
> >
> > I saw in the 2.4.22 kernel source :
> >
> > /usr/src/linux/net/ipv4/netfilter/ip_conntrack_proto_tcp.c
> > static unsigned long tcp_timeouts[]
> > = { 30 MINS,    /*      TCP_CONNTRACK_NONE,     */
> >     5 DAYS,     /*      TCP_CONNTRACK_ESTABLISHED,      */
> >     2 MINS,     /*      TCP_CONNTRACK_SYN_SENT, */
> >     60 SECS,    /*      TCP_CONNTRACK_SYN_RECV, */
> >     2 MINS,     /*      TCP_CONNTRACK_FIN_WAIT, */
> >     2 MINS,     /*      TCP_CONNTRACK_TIME_WAIT,        */
> >     10 SECS,    /*      TCP_CONNTRACK_CLOSE,    */
> >     60 SECS,    /*      TCP_CONNTRACK_CLOSE_WAIT,       */
> >     30 SECS,    /*      TCP_CONNTRACK_LAST_ACK, */
> >     2 MINS,     /*      TCP_CONNTRACK_LISTEN,   */
> > };
> >
> > Is there any plan to be able to set thoses values throught /proc ?
> >
> > in some cases a 5 days timeout on tcp connexions may be too long...
> 
> These are the standard values according to the TCP RFCs.

Which RFC talks about TCP_CONNTRACK_ESTABLISHED, again? I don't think
it's written yet ;-)

Ramin