On 25 Feb 2004, Greg Troxel wrote:

>   0&0xa006402      255.255.0.0       UGS         0    14513      -  fxp0 =>
> 
> That appears to be a route with a non-normal mask, where the
> destination is 0.0.0.0, but the mask is 10.0.100.2.

Ah, I see what it's saying now. Non-contiguous bitmasks are shown as &
followed by it in hex rather than the regular /<number of bits> suffix.

> This was likely created with
> route add default -netmask 10.0.100.2 255.255.0.0

Odd thing is that I'm pretty certain that this hasn't been entered
manually. I know nothing about the 10.0.100.2 machine (it's out of my
administrative control). I'll kick some people and see if they've been
playing about.

> I ran that and ping of 1.0.0.0 (note that this has no bits set which
> are in the match) gave me the same kernel msg:
> arpresolve: can't allocate llinfo on fxp0 for 255.255.0.0
> 
> # route delete default -netmask 10.0.100.2
> 
> removed the bogus route.

Yes, it did here too. Thanks.

> I would recommend kevlar engtex chaps when playing with non-normal
> routes:
>    http://www.stihlusa.com/apparel/protect_chaps.html

I'll bear that in mind if the urge ever takes me to do it.

> But seriously, I wonder if a sysctl to allow non-normal masks is in
> order, defaulting to not allowed.

That would be a good idea.

Thanks,

-- 
Stephen