Jack, when your problem occurs, does brctl report that the spanning tree
algorithm has blocked ports with pathcost 1 in favor of ports with pathcost
300?
> -----Original Message-----
> From: Jack Culpepper [mailto:jack@newdream.net]
> Sent: Monday, July 21, 2003 5:29 PM
> To: bridge@math.leidenuniv.nl
> Subject: [Bridge] problems with setpathcost feature
>
>
> Hi Folks,
>
> I have been using the bridging code to build out a small
> wireless network
> with five nodes. I am not sure if anyone on this list is
> familiar with
> OpenAP (small linux dist for running on x86-based AP
> stations) or the HostAP
> driver (wireless driver for Prism2 chipset), but that's what
> I'm using. The
> nodes in the network are "loosely" connected -- that is, the
number of
> connections each node has varies from about 1-4. I am using
> STP to remove
> loops. However, sometimes this results in STP "settling" on
> using links that
> have low signal strengths. This makes the performance of the
> network drop,
> and thus I would like to take signal strength into account
> via pathcosts.
> Eventually, I will be automatically computing pathcosts by
> using some formula
> that involves the inverse of the signal strength (high signal
> strength => low
> path cost). However, as a quick test to make sure the
> pathcost stuff was
> going to work like I want it to, I tried to manually impose
> my own beliefs
> about which links should be used by simply setting them to
> have pathcost 1,
> and all other links to pathcost 300. To my disappointment,
> this did not work.
> I know it didn't, because I am using the patch I just
> submitted to discover
> the actual path my packets are following. Has anyone on the list used
> pathcosts successfully? Is there a trick to doing it besides simply
> 'brctl setpathcost br0 wlan0wds0 1'? Does anyone have any
> reason to believe
> the pathcost code does or does not work? Thanks very much,
>
> Jack
>
> _______________________________________________
> Bridge mailing list
> Bridge@math.leidenuniv.nl
> http://www.math.leidenuniv.nl/mailman/listinfo/bridge
>