Suggesting OS for testing?

Tue Mar 3 17:56:05 EST 2009

Thanks for the beaconing explanation. All testing scenarios/test cases
appreciated too!

Using laptops with onboard radios or USB adapters, I almost always get
"Destination Host Unreachable" error messages when trying to ping one mesh
point from another * immediately after * configuration. (I'll put some PCI
cards back into a desktop and see if it happens there too.)
However, when I ran the "iwlist scan" I found I could subsequently ping the
other meshpoints successfully. Once I was able to ping meshpoints, then I
was able to get good data from the "mpath dump" and "station dump" commands.
(Prior to a ping doesn't dump anything.)

I've also been able to run Python programs over the mesh - eg. automatically
pinging the nodes in the mesh, and also doing basic client/server messaging
back and forth. I was also able to "block" one of the nodes to force a
re-routing, and see the new "hop path" appear in the "mpath dump."

I'll try to refresh my installed builds and share the summary results in a
new thread.

Brian

On Tue, Mar 3, 2009 at 4:43 PM, Andrey Yurovsky <andrey at cozybit.com> wrote:

> 2009/3/3 Brian DeLacey <bdelacey at gmail.com>:
> > I have been working with the following wireless USB adapter on several
> > machines at a time:
> >        EW-7318Ug  USB adapter chipset is called   RT73
> > That was working well - for the past couple of weeks - until I installed
> > something new.
>
> Ok, that's interesting.  I suppose that rt2x00 mesh support is working
> then?  I'll try it out when I have a chance.
>
> > To make it work, I used a Jaunty build and downloaded the latest
> > "compat-wireless-2.6.tar.bz2 ", then just ran make and make install. I
> > wasn't rebuilding any kernel bits - and hoped to avoid that. I'm using
> the
> > Aspire One for testing which is great for mobility but would be slow for
> big
> > builds.
>
> That's fine, you're then replacing mac80211 and other components
> (drivers, etc) with the latest.  It should give you the same results
> as building the latest kernel from wireless-testing.
>
> > Although I wasn't initially "seeing" any of the other mesh point devices
> > immediately after installing Linux Wireless, the following workaround
> would
> > consistently connect up the mesh points:
> >        sudo iwlist mesh0 scan
>
> To see other mesh points:
>
> sudo ifconfig mesh0 up
> sudo iw dev mesh0 station dump
>
> To see what paths (if any) you have, especially after transferring
> some data (ex: ping)
>
> sudo iw dev mesh0 mpath dump
>
> "iwlist scan" should have nothing to do with the mesh, aside from mesh
> points showing up in scan results (that's because they beacon).
> Please try a "station dump" with iw and let us know if it shows you
> other mesh points.  Also please try never running the "scan" to
> confirm that things work without it (otherwise it's likely a driver
> bug that we should report or look in to).
>
> > That seemed to prime the mesh routes, and I could then ping the other
> mesh
> > points from multiple directions. I concluded this connected the mesh, but
> > maybe it was just luck that things worked.
>
> No, that should have worked.  The routes are established as needed
> (ex: when you ping).  The mesh points are discovered as beacons come
> in and get processed.
>
> > Is beaconing's sole function to let other devices know where they are
> are,
> > and when they appear and leave? Or is periodic beaconing required to keep
> > the mesh alive? For some apps, the lack of beaconing could be a security
> > feature keeping the mesh points largely "hidden" from view.
>
> Yes, and mesh beacons are quite similar to AP or Ad-Hoc beacons.
> Beaconing is part of the 802.11s draft, so it's required.
>
> > On Tue, Mar 3, 2009 at 3:35 PM, Andrey Yurovsky <andrey at cozybit.com>
> wrote:
> >>
> >> Hi Brian.  By Ralink USB, do you mean that you were testing rt2x00
> >> mesh or just using the adapter in general?  I'm not aware of any
> >> improvements in rt2x00 support as beaconing was broken in that driver.
> >>  o11s mesh is not sensitive to the kernel version, so to speak, but as
> >> a part of the mac80211 subsystem it is sensitive to the version of the
> >> wireless components (mac80211 and drivers) that you have.  If I
> >> understand correctly, Ubuntu usually picks a kernel and then pulls
> >> additional patches from wireless-testing or elsewhere to further
> >> update mac80211 and related bits.
> >>
> >> Another option is to test the latest and greatest.  In this case, you
> >> would pull wireless-testing git and build it based on your Ubuntu
> >> kernel's config.  One way to do that is described here:
> >> http://o11s.org/trac/wiki/Ubuntu
> >>
> >> 2009/3/3 Brian DeLacey <bdelacey at gmail.com>:
> >> > Any recommendations on the "best" combination of Linux Kernel and
> Linux
> >> > Wireless versions to test mesh with?
> >> > I was running well on a recent build of Ubuntu/Jaunty after installing
> >> > recent builds of Linux Wireless.  That combination got my Ralink-based
> >> > USB
> >> > adapters working, but recent updates seem to have broken that
> >> > functionality.
> >> > Using the latest Jaunty dailies and Linux Wireless, I now find that
> >> > Ralink
> >> > USB no longer works and even built-in Atheros is now only partially
> >> > working. (I'm trying to track down where the break occurred.)
> >> > It seems like I need more recent builds (Linux 2.6.28+) and updated
> >> > Linux
> >> > Wireless to get the Ralink USB working, which is why I've been trying
> to
> >> > keep reasonably current with the latest builds of Jaunty and Linux
> >> > Wireless.
> >> > Is the core mesh functionality most sensitive to changes in the
> kernel,
> >> > changes in Linux Wireless, or both?
> >> > Thanks,
> >> >    Brian
> >> >
> >> > _______________________________________________
> >> > Devel mailing list
> >> > Devel at lists.open80211s.org
> >> > http://open80211s.com/mailman/listinfo/devel
> >> >
> >> >
> >>
> >>
> >>
> >> --
> >> Andrey Yurovsky
> >> cozybit Inc.
> >> _______________________________________________
> >> Devel mailing list
> >> Devel at lists.open80211s.org
> >> http://open80211s.com/mailman/listinfo/devel
> >
> >
> > _______________________________________________
> > Devel mailing list
> > Devel at lists.open80211s.org
> > http://open80211s.com/mailman/listinfo/devel
> >
> >
>
>
>
> --
> Andrey Yurovsky
> cozybit Inc.
> _______________________________________________
> Devel mailing list
> Devel at lists.open80211s.org
> http://open80211s.com/mailman/listinfo/devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://open80211s.com/pipermail/devel/attachments/20090303/c734cebd/attachment-0001.html 