[Olsr-dev] Short netmask lenghts
Mon Aug 9 18:55:46 CEST 2010
I used also the /1 for a while.
Now I want to add /2. So I'll have 3 def.gw prio's (/0, /1 and /2).
Linux has problems with multiple, independent same routes (0/0 to different next_hops, maintained independently).
Now I use different metrics:
metric=0 # DHCP from eth1
metric=1 # from ppp ip-up / ip-down scripts
metric=2 # from olsrd
We run daemon promoting the def.gw to /1 or /2, with some policies (satcom, adsl or 3G, signal quality).
The dyn_gw is adjusted for handling metrics 0 _and_ 1.
But since this update, we experience problems with the functions in ipcalc.
I guess the limitation was there for a long time.
(I can't decode this piece of C myself, Caspar told me the functions don't work for shorter than /8)
We can adjust the ipcalc functions and try to get things stable.
Op 9 aug 2010, om 17:14 heeft Henning Rogge het volgende geschreven:
> Am Montag 09 August 2010, 17:10:07 schrieb Teco Boot:
>> We use very short netmask lengths (/1, /2), for priority def.gws.
>> Functions ipcalc.c prefix_to_netmask and netmask_to_prefix do not support
>> this. We can adjust, but are there side effects?
>> Rationale for not supporting it?
> Why should they not support this ?
> I know they are using a 0.0.0.0/1 HNA4 in Vienna...
> Henning Rogge
> 1) You can't win.
> 2) You can't break even.
> 3) You can't leave the game.
> — The Laws of Thermodynamics, summarized
More information about the Olsr-dev