Hi Patrick, Just wondering if you've had a chance to get your new version of the IMQ device completed as yet to fix this deadlooping problem? thanks and regards, Andrew. ----- Original Message ----- From: "Patrick McHardy" <kaber@xxxxxxxxx> To: "Andrew Hall" <temp02@xxxxxxxxxxxxxxx> Cc: <lartc@xxxxxxxxxxxxxxx> Sent: Wednesday, February 19, 2003 9:43 PM Subject: Re: IMQ device problems with iptables: dead looping? > Andrew Hall wrote: > > >Hello, > > > >I am using the IMQ with iptables (latest versions) and asking all packets to > >be enqueued to IMQ0 from both prerouting and postrouting (using different > >iptables rules to mark different streams). When I do this I get the kernel > >saying: > > > >"Dead loop on netdevice imq0, fix it urgently!" and communications stop > >intermittently. If I remove the jump from either preroute or postroute it > >works fine but won't work together. > > > >Is this what is mean't to happen, because as far as I can see there is no > >loop happening here? > > > > You probably forward packets. The loop happening ist > input-device -> prerouting -> imq (xmit_lock locked) -> reinject -> ... > -> postrouting -> imq (LOCKED!!) > > To avoid this one has to drop xmit_lock in imq_xmit function before > reinjecteing and grab it back afterwards. > unfortunately this allows for endless recursion which is not possible in > the kernel due to limited stack space. > I've completed a version which should hopefully fix this as many people > requested such a feature. > I haven't released it yet because i have to do alot for my university > atm and haven't got the time for testing > although i promised the new version to some people for january ;( > In about two weeks my semester-vacation start, i hope to release the new > version very quickly afterwards. > > Thanks for your patience, > Patrick > >
