* RFC: promote netfilter MARK value from IPv6 packets to sit packets
@ 2003-02-17 14:57 Erik Hensema
2003-02-23 19:33 ` Harald Welte
0 siblings, 1 reply; 6+ messages in thread
From: Erik Hensema @ 2003-02-17 14:57 UTC (permalink / raw)
To: netdev
Hi,
I'm new to the list, so be gentle with your flames ;-)
On my outgoing connection to the internet I provide QoS using ratelimiting
and prioritizing packets. However, it's a cablemodem and my provider only
speaks IPv4. Over the connections I run multiple IPv6 tunnels, and I want
to provide QoS on these tunnels too.
QoS however can only be provided on the physical outgoing interface. On
that level, I can only see IPv4 packets and therefore it's impossible to
provide QoS on IPv6 for me.
In order to be able to provide QoS on tunneled IPv6 connections, I've
created a simple patch (definately not ready for inclusion in the kernel,
since it surely needs a configuration option) which promotes the netfilter
MARK value from the IPv6 packets to the sit packets.
Now I can mark packets using ip6tables, and on the ipv4 level I can still
differentiate between the priorities. Problem solved, I'm happy ;-)
Below is the patch, created on Linux 2.4.19:
--- net/ipv6/sit.c.orig Mon Feb 17 15:30:41 2003
+++ net/ipv6/sit.c Mon Feb 17 15:29:40 2003
@@ -571,6 +571,9 @@
}
if (skb->sk)
skb_set_owner_w(new_skb, skb->sk);
+#ifdef CONFIG_NETFILTER
+ new_skb->nfmark = skb->nfmark;
+#endif
dev_kfree_skb(skb);
skb = new_skb;
}
--
Erik Hensema (erik@hensema.net)
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: RFC: promote netfilter MARK value from IPv6 packets to sit packets
2003-02-17 14:57 RFC: promote netfilter MARK value from IPv6 packets to sit packets Erik Hensema
@ 2003-02-23 19:33 ` Harald Welte
2003-02-23 23:42 ` Erik Hensema
0 siblings, 1 reply; 6+ messages in thread
From: Harald Welte @ 2003-02-23 19:33 UTC (permalink / raw)
To: Erik Hensema; +Cc: netdev, Netfilter Development Mailinglist
[-- Attachment #1: Type: text/plain, Size: 1181 bytes --]
On Mon, Feb 17, 2003 at 03:57:27PM +0100, Erik Hensema wrote:
> In order to be able to provide QoS on tunneled IPv6 connections, I've
> created a simple patch (definately not ready for inclusion in the kernel,
> since it surely needs a configuration option) which promotes the netfilter
> MARK value from the IPv6 packets to the sit packets.
> Now I can mark packets using ip6tables, and on the ipv4 level I can still
> differentiate between the priorities. Problem solved, I'm happy ;-)
I like this patch. I think we should make it a kernel configuration
option, but for all kind of tunnel interfaces. Something like
'propagate NFMARK while tunneling' (or maybe 'preserve' instead of
'propagate' is better language?)
DaveM: Would this be acceptable?
> Erik Hensema (erik@hensema.net)
--
- Harald Welte <laforge@netfilter.org> http://www.netfilter.org/
============================================================================
"Fragmentation is like classful addressing -- an interesting early
architectural error that shows how much experimentation was going
on while IP was being designed." -- Paul Vixie
[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: RFC: promote netfilter MARK value from IPv6 packets to sit packets
2003-02-23 19:33 ` Harald Welte
@ 2003-02-23 23:42 ` Erik Hensema
2003-02-24 13:41 ` jamal
0 siblings, 1 reply; 6+ messages in thread
From: Erik Hensema @ 2003-02-23 23:42 UTC (permalink / raw)
To: Harald Welte, netdev, Netfilter Development Mailinglist
On Sun, Feb 23, 2003 at 08:33:39PM +0100, Harald Welte wrote:
> On Mon, Feb 17, 2003 at 03:57:27PM +0100, Erik Hensema wrote:
>
> > In order to be able to provide QoS on tunneled IPv6 connections, I've
> > created a simple patch (definately not ready for inclusion in the kernel,
> > since it surely needs a configuration option) which promotes the netfilter
> > MARK value from the IPv6 packets to the sit packets.
> > Now I can mark packets using ip6tables, and on the ipv4 level I can still
> > differentiate between the priorities. Problem solved, I'm happy ;-)
>
> I like this patch. I think we should make it a kernel configuration
> option, but for all kind of tunnel interfaces. Something like
> 'propagate NFMARK while tunneling' (or maybe 'preserve' instead of
> 'propagate' is better language?)
It certainly should be configurable. I've already sent it to the list, but
you can also download it from
http://dexter.hensema.net/~erik/patches/sit-promote-mark-2.4.21-pre4.diff
It should be easy to port this patch to gre and maybe ipip (don't know the
code of the latter, but I assume it's similar to gre and sit). I'll work on
that tomorrow, when I've got access to my development machine again.
In my current patch the configuration option is called 'IPv6: Promote
netfilter MARK value to sit packets'.
I don't think we should call it 'preserve', because technically that's not
what is happening. The tunnel interface creates a fresh new packet, with a
fresh new nfmark. Propagate seems to be the right term to me (as a
non-native english speaker).
--
Erik Hensema (erik@hensema.net)
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: RFC: promote netfilter MARK value from IPv6 packets to sit packets
2003-02-23 23:42 ` Erik Hensema
@ 2003-02-24 13:41 ` jamal
2003-02-24 14:41 ` Harald Welte
0 siblings, 1 reply; 6+ messages in thread
From: jamal @ 2003-02-24 13:41 UTC (permalink / raw)
To: Erik Hensema; +Cc: Harald Welte, netdev, Netfilter Development Mailinglist
If this is to be a config option, it should not be restricted to
netfilter specifics but rather skb specifics. Example the tcindex
(maybe even the cb) etc.
cheers,
jamal
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: RFC: promote netfilter MARK value from IPv6 packets to sit packets
2003-02-24 13:41 ` jamal
@ 2003-02-24 14:41 ` Harald Welte
2003-02-25 2:30 ` jamal
0 siblings, 1 reply; 6+ messages in thread
From: Harald Welte @ 2003-02-24 14:41 UTC (permalink / raw)
To: jamal; +Cc: Erik Hensema, Harald Welte, netdev,
Netfilter Development Mailinglist
[-- Attachment #1: Type: text/plain, Size: 782 bytes --]
On Mon, Feb 24, 2003 at 08:41:10AM -0500, jamal wrote:
> If this is to be a config option, it should not be restricted to
> netfilter specifics but rather skb specifics. Example the tcindex
> (maybe even the cb) etc.
No problem with me. I do understand the usefulness of tcindex, but what
would a totally different protcol (or the user) do with the cb of a
different protocol?
> cheers,
> jamal
--
- Harald Welte <laforge@netfilter.org> http://www.netfilter.org/
============================================================================
"Fragmentation is like classful addressing -- an interesting early
architectural error that shows how much experimentation was going
on while IP was being designed." -- Paul Vixie
[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: RFC: promote netfilter MARK value from IPv6 packets to sit packets
2003-02-24 14:41 ` Harald Welte
@ 2003-02-25 2:30 ` jamal
0 siblings, 0 replies; 6+ messages in thread
From: jamal @ 2003-02-25 2:30 UTC (permalink / raw)
To: Harald Welte; +Cc: Erik Hensema, netdev, Netfilter Development Mailinglist
On Mon, 24 Feb 2003, Harald Welte wrote:
> On Mon, Feb 24, 2003 at 08:41:10AM -0500, jamal wrote:
> > If this is to be a config option, it should not be restricted to
> > netfilter specifics but rather skb specifics. Example the tcindex
> > (maybe even the cb) etc.
>
> No problem with me. I do understand the usefulness of tcindex, but what
> would a totally different protcol (or the user) do with the cb of a
> different protocol?
>
cb is a maybe - it could be useful i think since the inner and outer
headers may be closely related and so share the same state.
I gacve tcindex as an example; others are: priority and some of the
other netfilter stuff (is nfcache still used?) etc.
cheers,
jamal
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2003-02-25 2:30 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-02-17 14:57 RFC: promote netfilter MARK value from IPv6 packets to sit packets Erik Hensema
2003-02-23 19:33 ` Harald Welte
2003-02-23 23:42 ` Erik Hensema
2003-02-24 13:41 ` jamal
2003-02-24 14:41 ` Harald Welte
2003-02-25 2:30 ` jamal
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).