On Tue, Sep 21, 2004 at 02:58:37PM -0600, Tim Gardner wrote:
>
> > My personal (simplistic) favourite is still a simple threshold (absolute
> > value / percentage) for incomplete neighbour entries. This way we make
> > sure that we cannot starve 'real' (fully resolved) entries at the cost
> > of incomplete ones.
> >
> > > -Andi
>
> It's not like NUD doesn't already have an overflow policy.
> neigh_forced_gc() performs a cleanup on NUD_INCOMPLETE entries when
> gc_thresh3 is exceeded.
No, that's not what it does. neigh_forced_gc explicitly only elects
INCOMPLETE entries that exist for at least n->parms_retrans_time in
order to avoid flooding the network with too many arp/neighbour
requests (since we could delete an incomplete one before the reply
arrives, and could do this for quite some time over and over again).
That's been my point all over this discussion...
> rtg
> timg@xxxxxxx http://www.tpi.com
> 406-443-5357(MT) 503-601-0234(OR)
--
- Harald Welte <laforge@xxxxxxxxxxxx> http://www.gnumonks.org/
============================================================================
Programming is like sex: One mistake and you have to support it your lifetime
signature.asc
Description: Digital signature
|