[Xorp-hackers] XORP enhancement for wireless mesh network routing

[Ben Greear]

On 06/01/2012 12:07 PM, Jiangxin Hu wrote:
>>> First, I believe it would be possible to want this feature enabled on
> non-wireless interfaces, so maybe instead of having>>  a 'wireless'
> attribute, we could call it something like 'allow-disconnected-routes' or
> something like that.
>>> And maybe we should just always allow those routes to be added and not
> even bother with all the framework to set the flag?
>
> Agree if we allow such routes for wired network. I don't know it is
> meaningful for wired network or not.
>
>>> At least some of the changes do not appear directly related to the
> 'wireless'
>>> flag.  Maybe there was some cleanup included?  If so, it would be nice if
> that were a separate patch.
>
> There are two things in the code changes:
> 1. the parameter 'wireless' (fea, ifmgr, mirror, etc.)
> 2. the execution part (fea, rib)
>     in order to insert such route into kernel, the add route function call
> must declare such route as scope-link type route
>     also, theoretically, the route -- destination net: 192.168.0.0/24 next
> hop: 192.168.0.1 interface: eth0 is a valid route
>     for node 1 configure such as eth0:10.0.0.1/24. however, I don't think any
> wireless routing protocol generated such rotue now.

I think any interface can be used in a mesh, even if it is mostly
wireless devices in the real world.  If the routing protocol
thinks a route should be added, and the kernel doesn't complain,
I see no reason for FEA to complain...

> * This work is done on Fedora core 16, so other system may not work.
> * 'wireless' parameter is for interface only, any vif under the interface
> will be treated as wireless vif

Would you care to re-do the patch so that you just change FEA to
always allow these types of routes, and skip adding the new
interface flag (and all the related xrl stuff)?

Thanks,
Ben


-- 
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com