mpls-linux-general

[mpls-linux-general] MPLS and 802.11e

[<luc...@li...>]

Hello, I'm still working with MPLS over WiFi.
I'm trying to enable wireless QoS 
extensions (802.11e) together with MPLS, but it seems not to succeed in 
classifying the flows once they are encapsulated in MPLS.
I try to explain: 
802.11e does a matching between the DSCP field of IP header and a class (Access 
Category) .
When I send a flow with MPLS, if I sniff the traffic with wireshark 
I can see the DSCP field correctly written, but the wireless QoS information do 
not match with the right class.
Otherwise, if I send the same flow without 
MPLS, everything works fine.
I'd like to understand why, can anybody help me?

Thanks a lot,
Luca

Re: [mpls-linux-general] MPLS and 802.11e

[Adrian P. <adr...@gm...>]

Hello

How do you map the DSCP to the wireless class? Which tool do you use for
this? It might be that the tool/module you're using expects IP traffic, not
MPLS...

On Wed, Jun 4, 2008 at 12:35 PM, luc...@li... <luc...@li...>
wrote:

> Hello, I'm still working with MPLS over WiFi.
> I'm trying to enable wireless QoS
> extensions (802.11e) together with MPLS, but it seems not to succeed in
> classifying the flows once they are encapsulated in MPLS.
> I try to explain:
> 802.11e does a matching between the DSCP field of IP header and a class
> (Access
> Category) .
> When I send a flow with MPLS, if I sniff the traffic with wireshark
> I can see the DSCP field correctly written, but the wireless QoS
> information do
> not match with the right class.
> Otherwise, if I send the same flow without
> MPLS, everything works fine.
> I'd like to understand why, can anybody help me?
>
> Thanks a lot,
> Luca
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general
>

Re: [mpls-linux-general] MPLS and 802.11e

[Jeremy J. <je...@co...>]

I think MadWifi driver has it's own mapping policy, which might make it
more difficult.  I think it should be configurable, like vlan skb_prio
-> vlan priority mapping, but maybe it's not yet.

You should look at the wireless driver code.

On Wed, 2008-06-04 at 11:35 +0200, luc...@li... wrote:
> Hello, I'm still working with MPLS over WiFi.
> I'm trying to enable wireless QoS 
> extensions (802.11e) together with MPLS, but it seems not to succeed in 
> classifying the flows once they are encapsulated in MPLS.
> I try to explain: 
> 802.11e does a matching between the DSCP field of IP header and a class (Access 
> Category) .
> When I send a flow with MPLS, if I sniff the traffic with wireshark 
> I can see the DSCP field correctly written, but the wireless QoS information do 
> not match with the right class.
> Otherwise, if I send the same flow without 
> MPLS, everything works fine.

Re: [mpls-linux-general] MPLS and 802.11e

[Luca P. <luc...@li...>]

Do you mean it could be that MadWifi driver doesn't "understand" the 
meaning of the packet's fields but only extracts them with some kind of 
fixed-size masks?


Jeremy Jackson ha scritto:
> I think MadWifi driver has it's own mapping policy, which might make it
> more difficult.  I think it should be configurable, like vlan skb_prio
> -> vlan priority mapping, but maybe it's not yet.
>
> You should look at the wireless driver code.
>
> On Wed, 2008-06-04 at 11:35 +0200, luc...@li... wrote:
>   
>> Hello, I'm still working with MPLS over WiFi.
>> I'm trying to enable wireless QoS 
>> extensions (802.11e) together with MPLS, but it seems not to succeed in 
>> classifying the flows once they are encapsulated in MPLS.
>> I try to explain: 
>> 802.11e does a matching between the DSCP field of IP header and a class (Access 
>> Category) .
>> When I send a flow with MPLS, if I sniff the traffic with wireshark 
>> I can see the DSCP field correctly written, but the wireless QoS information do 
>> not match with the right class.
>> Otherwise, if I send the same flow without 
>> MPLS, everything works fine.
>>     
>
>
> .
>
>   

Re: [mpls-linux-general] MPLS and 802.11e

[Adrian P. <adr...@gm...>]

Yes, it could be offset-based, but for this you'll have to inspect the
code...
>From what I've read the tool you are using allows you to work with the wifi
module. I don't see why it would need to understand what MPLS/IP is.

On Thu, Jun 5, 2008 at 11:22 AM, Luca Pilosu <luc...@li...> wrote:

> Do you mean it could be that MadWifi driver doesn't "understand" the
> meaning of the packet's fields but only extracts them with some kind of
> fixed-size masks?
>
>
> Jeremy Jackson ha scritto:
> > I think MadWifi driver has it's own mapping policy, which might make it
> > more difficult.  I think it should be configurable, like vlan skb_prio
> > -> vlan priority mapping, but maybe it's not yet.
> >
> > You should look at the wireless driver code.
> >
> > On Wed, 2008-06-04 at 11:35 +0200, luc...@li... wrote:
> >
> >> Hello, I'm still working with MPLS over WiFi.
> >> I'm trying to enable wireless QoS
> >> extensions (802.11e) together with MPLS, but it seems not to succeed in
> >> classifying the flows once they are encapsulated in MPLS.
> >> I try to explain:
> >> 802.11e does a matching between the DSCP field of IP header and a class
> (Access
> >> Category) .
> >> When I send a flow with MPLS, if I sniff the traffic with wireshark
> >> I can see the DSCP field correctly written, but the wireless QoS
> information do
> >> not match with the right class.
> >> Otherwise, if I send the same flow without
> >> MPLS, everything works fine.
> >>
> >
> >
> > .
> >
> >
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general
>

Re: [mpls-linux-general] MPLS and 802.11e

[Jorge B. [DTI2] <jo...@dt...>]

Luca Pilosu escribió:
> Do you mean it could be that MadWifi driver doesn't "understand" the 
> meaning of the packet's fields but only extracts them with some kind of 
> fixed-size masks?

	Driver is not responsible of classifying packets, protocol stack 
(net80211) is. You have to look at function ieee80211_classify() on the 
file ieee80211_output.c. There you'll see that it only classifies VLAN 
or IP packets. You have to modify and teach that function how to 
classify MPLS/IP packets.

	Regards,

	-Jorge

==============================================================
Jorge Boncompte - Ingenieria y Gestion de RED
DTI2 - Desarrollo de la Tecnologia de las Comunicaciones
--------------------------------------------------------------
C/ Abogado Enriquez Barrios, 5   14004 CORDOBA (SPAIN)
Tlf: +34 957 761395 / FAX: +34 957 450380
==============================================================
- Sin pistachos no hay Rock & Roll...
- Without wicker a basket cannot be made.
==============================================================

Re: [mpls-linux-general] MPLS and 802.11e

[James R. L. <jl...@mi...>]

On Thu, Jun 05, 2008 at 12:20:42PM +0200, Jorge Boncompte [DTI2] wrote:
> Luca Pilosu escribió:
> > Do you mean it could be that MadWifi driver doesn't "understand" the 
> > meaning of the packet's fields but only extracts them with some kind of 
> > fixed-size masks?
> 
> 	Driver is not responsible of classifying packets, protocol stack 
> (net80211) is. You have to look at function ieee80211_classify() on the 
> file ieee80211_output.c. There you'll see that it only classifies VLAN 
> or IP packets. You have to modify and teach that function how to 
> classify MPLS/IP packets.

... or more generically classify based off of tcindex/nfmark so that
other protocols besides IP can take advantage of this ...

> 
> 	Regards,
> 
> 	-Jorge
> 
> ==============================================================
> Jorge Boncompte - Ingenieria y Gestion de RED
> DTI2 - Desarrollo de la Tecnologia de las Comunicaciones
> --------------------------------------------------------------
> C/ Abogado Enriquez Barrios, 5   14004 CORDOBA (SPAIN)
> Tlf: +34 957 761395 / FAX: +34 957 450380
> ==============================================================
> - Sin pistachos no hay Rock & Roll...
> - Without wicker a basket cannot be made.
> ==============================================================
> 
> 
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu
jl...@mi...

Re: [mpls-linux-general] MPLS and 802.11e

[Jeremy J. <je...@co...>]

There was a time when madwifi had it's own classifying code, which I
disagree with.  IMO, it should respect the skb_priority field, which can
be set by iptables or generic networking stack code.

in net/ipv4/ip_input.c::ip_forward()

skb->priority = rt_tos2priority(iph->tos);

I believe the correct implementation might be for MPLS to set
skb->priority. Perhaps in addition to outgoing interface/label, there
should be a priority attribute for label switching, what would get
copied here for each packet as it is label-switched.

In madwifi 0.9.4 net80211/ieee80211_output.c::ieee80211_classify() you
have the details, if WME is enabled, then VLAN priority tag and IP TOS
are mapped hardcoded to skb-> priority, overwriting the linux network
core's setting.  If the packet is not VLAN or IP, it's prio is
overwritten always to WME_AC_BE.  This is the problem you're having with
MPLS.  It would be interesting to see how the ath5k driver does this,
since madwifi-ng is obsolete.

I think wifi code should just use the skb-> priority, instead of
implementing it's own policy in 802.11 stack.  It's not feasible to
duplicate the decoding of priority from all possible protocols, and
having it hardcoded in wifi driver.

On Thu, 2008-06-05 at 12:20 +0200, Jorge Boncompte [DTI2] wrote:
> Luca Pilosu escribió:
> > Do you mean it could be that MadWifi driver doesn't "understand" the 
> > meaning of the packet's fields but only extracts them with some kind of 
> > fixed-size masks?
> 
> 	Driver is not responsible of classifying packets, protocol stack 
> (net80211) is. You have to look at function ieee80211_classify() on the 
> file ieee80211_output.c. There you'll see that it only classifies VLAN 
> or IP packets. You have to modify and teach that function how to 
> classify MPLS/IP packets.
> 
> 	Regards,
> 
> 	-Jorge
> 
> ==============================================================
> Jorge Boncompte - Ingenieria y Gestion de RED
> DTI2 - Desarrollo de la Tecnologia de las Comunicaciones
> --------------------------------------------------------------
> C/ Abogado Enriquez Barrios, 5   14004 CORDOBA (SPAIN)
> Tlf: +34 957 761395 / FAX: +34 957 450380
> ==============================================================
> - Sin pistachos no hay Rock & Roll...
> - Without wicker a basket cannot be made.
> ==============================================================
> 
> 
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general
-- 
Jeremy Jackson
Coplanar Networks
(519)489-4903
http://www.coplanar.net
je...@co...