Hi Qasim,

On Friday, June 1, 2012, Qasim Javed wrote:

> Queuing discipline functionality is only invoked from within
> dev_queue_xmit.
>
> This function can be called in two possible cases:
>
> 1. When a source mesh station sends an IP packet it ends up calling
> dev_queue_xmit so the queuing discipline is exercised on the source.
>
> 2. As a result of ieee80211_deliver_skb which in turn is invoked from
> ieee80211_rx_h_data.
>
> So, basically when a mesh station is forwarding a frame, the qdisc
> (even if installed on a mesh interface using tc) will not be
> exercised. I had never thought about that. This means I cannot use
> netem for introducing delay on the intermediate hops.
>
> Since wmediumd does not introduce delays, I guess my test case would
> be probabilistic rather than deterministic. This is because, the bug
> that I am trying to reproduce would only trigger if the PREQs arrive
> at the target in a certain order as described previously.
>
> I will look some more into it and get back to you later.
>
> By the way, is it OK if I can show you a packet capture which clearly
> demonstrates that the wrong originator sequence number is being used
> in the PREP?


A test script would help us ensure there are no regressions on this, but if
it's going to be such a time sink, a capture will do.
Adding delays to wmediumd is on the todo list.

Thanks,

Javier


> Thanks,
> -Qasim
>
> On Thu, May 31, 2012 at 3:24 PM, Qasim Javed <[email protected]> wrote:
> > Hi Javier,
> >
> > I am trying to introduce delay on one of the links using netem qdisc.
> > When I use netem on the source node, I can see the delay is added
> > (verified using ping) but when I add it to one of the intermediate
> > hops, the delay does not show up in pcap file.
> >
> > Logically the queuing disciplines should be invoked irrespective of
> > whether a mesh node is forwarding traffic or not.
> >
> > I have wmediumd working fine for me, but I need a deterministic delay
> > (in addition to the loss which makes the metric value for one of the
> > paths higher, thus making it undesirable) on one of the paths to
> > emulate the scenario in which the wrong usage of the originator
> > sequence number (in a PREP) will be triggered.
> >
> > Any ideas?
> >
> > Thanks,
> > -Qasim
> >
> > On Wed, May 30, 2012 at 7:08 PM, Qasim Javed <[email protected]> wrote:
> >> Thanks. That will do.
> >>
> >> -Qasim
> >>
> >> On Wed, May 30, 2012 at 6:09 PM, Javier Cardona <[email protected]>
> wrote:
> >>> Hi Qasim,
> >>>
> >>> On Wed, May 30, 2012 at 2:55 PM, Qasim Javed <[email protected]> wrote:
> >>>> Hi Javier,
> >>>>
> >>>> I have set up disto11s, but where do I access the existing hwsim test
> >>>> cases that you have?
> >>>
> >>> Our test suite is not public but here is a template that should give
> >>> you a jump start:
> >>> https://github.com/cozybit/open80211s/wiki/HwsimTestTemplate
> >>> As you will see, it is trivial to modify to test your scenario.
> >>>
> >>> Cheers,
> >>>
> >>> Javier
> >>>
> >>>> On Tue, May 29, 2012 at 12:59 PM, Javier Cardona <[email protected]>
> wrote:
> >>>>> Hi Qasim,
> >>>>>
> >>>>> Thanks for looking into this.  If your explanation is accurate, then
> >>>>> it looks like the originator and target sequence numbers are swapped
> >>>>> for PREPs.  That said, the scenario you describe should be easily
> >>>>> testable using hwsim.  In fact we have a very similar test on our
> test
> >>>>> suite that is currently passing.  Would you consider writing a hwsim
> >>>>> test case to test your patch?
> >>>>>
> >>>>> Cheers,
> >>>>>
> >>>>> Javier
> >>>>>
> >>>>> PS.  I've narrowed down the audience of this thread to avoid boring
> >>>>> everyone in linux, netdev and linux-wireless on obscure mesh
> >>>>> discussions... :)
> >>>>>
> >>>>> On Thu, May 24, 2012 at 10:02 PM, Qasim Javed <[email protected]>
> wrote:
> >>>>>> Hi,
> >>>>>>
> >>>>>> I have been doing some experiments using the 802.11s functionality
> in the mac80211 stack. Today I stumbled across something which I believe is
> a critical bug in the usage of originator sequence number for a Path Reply
> message upon the reception of a Path Request message.
> >>>>>>
> >>>>>> Consider the following topology:
> >>>>>>
> >>>>>>                    +---+
> >>>>>>                    | S |
> >>>>>>                    +---+
> >>>>>>                  /      \
> >>>>>>                 /        \
> >>>>>>            +---+          +---+
> >>>>>>            | A |



-- 
Javier Cardona
cozybit Inc.
http://www.cozybit.com
_______________________________________________
Devel mailing list
[email protected]
http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel

Reply via email to