LARTC - Mar 2003 - IMQ - bug when doing ingress and egress shaping on the same interface.

http://www.mail-archive.com/lartc@mailman.ds9a.nl/msg03827.html

you can''t add ingress and egress packets into 
the same IMQ device from the same physical device.

iptables -t mangle -A POSTROUTING -o eth0 -j IMQ
iptables -t mangle -A PRETROUTING -i eth0 -j IMQ

Is this problem solved?


****************************************************

Is it?


Kjell



_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

Kjell Chris Flor wrote:
>http://www.mail-archive.com/lartc@mailman.ds9a.nl/msg03827.html
>
>you can''t add ingress and egress packets into 
>the same IMQ device from the same physical device.
>
>iptables -t mangle -A POSTROUTING -o eth0 -j IMQ
>iptables -t mangle -A PRETROUTING -i eth0 -j IMQ
>
>Is this problem solved?
>
>
>****************************************************
>
>Is it?
>
>
>Kjell
>  
>
i''m sorry to disapoint you, it is not solved and might possibly never
be.
I''m thinking about dropping imq maintenance, i don''t need it
myself
anymore and due to lack of time i cannot provide even the most-requested
features. i think the imq users would be better of with a maintainer who
actually uses imq himself.

Volunteers are welcome!

If anyone wants to "fix" the issue described in this mail,
it''s
basically a fact
of dropping dev->xmit_lock before reinjecting the packet in imq_xmit and
grabbing it back afterwards. look in net/sched/sch_generic.c::sch_restart()
in order to understand what to do exactly with the locks. make sure to
catch too deep recursion, one or two levels should be enough.


Bye,
Patrick

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/