Archived from groups: comp.dcom.lans.ethernet (
More info?)
"Affan Syed" <asyed@usc.edu> wrote in message
news:csv7l8$9oa$1@gist.usc.edu...
> > :> and if the s/w stack is correctly implemented, the
> > :> incorrect destination should then drop it.
> >
> > If the s/w stack is correctly implimented, the "incorrect" destination
> > will offer a way of capturing all packets, such as for tcpdump /
> > ethereal. When such a mode is activated, the exact details of what
> > happens for mismatched IPs with the correct MAC varies. For example for
> > some OS's, an icmp unreachable will be generated for protocols -other-
> > than TCP or UDP, but the TCP and UDP stacks filter at a different point
> > and know enough not to act on the packets. These differences in
> > operation are exploited by "anti-sniffer" software designed to
> > locate equipment that is snooping on the net.
>
> Could you explain a little bit more on this? My question is based more on
> curiosity of how deterimental the effect of having duplicate MAC's on the
> same LAN would be. As far as I can see, if they are on the same LAN (again
> with switches we would have the port ringin => delays), the effect shouldt
> be that significant, esp for say and TCP/UDP traffic.
i will Walter answer your direct Q - but there is an invalid assumption
here.
my previous answer was about what happens if there are duplicate MACs on a
hub - which doesnt seem to be a problem in practice.
if you use switches and have duplicate active MAC addresses at 2 different
points, then you may have major problems with a duplicate MAC causing side
effects.
The problem is that noticing that the MAC "appears" at 2 different points
and altering the switch tables and setup within the network is a processor
driven activity.
So - high traffic levels on 2 or more ports from duplicate MACs is likely to
generate lots of processor load - and severely overloading a switch
processor tends to cause lots of problems
in the worst case i have seen loss of management, then loss of control
protocols such as spanning tree, or layer 3 control - and in turn that can
cause instability across a campus
>
> Thanks everyone for satisfying the curisosity of a student.
>
> Regards
>
> Affan
--
Regards
Stephen Hope - return address needs fewer xxs