wrong source ip addr on aggregation

Frank argus-mailinglist-1524134246 at f-block.org
Thu Jun 28 08:41:30 EDT 2018


hi,

i encountered the following issue, resulting in a wrong source ipv6 address:

there are several connections to the same ipv6 address, coming from two
different source addresses, which however share the same /48 bits:

   11:57:23.766559  e           tcp 1111:22:3333:4444::15.21978     ->
caffe:affe::1.https        48      16800   FIN
   12:03:26.631506  e           tcp 1111:22:3333:4444::15.52620     ->
caffe:affe::1.smtp         47      22007   FIN
   12:03:26.631451  e           tcp 1111:22:3333:4444::15.52620     ->
caffe:affe::1.smtp         47      22007   FIN
   12:05:08.037777  e           tcp 1111:22:3333:4444::15.51150     ->
caffe:affe::1.http         11       1844   FIN
   12:05:08.041924  e ipv6-icmp 1111:22:3333:4444::15.0x0080   <->
caffe:affe::1.0x0000        2        178   ECO
   12:05:08.037777  e           tcp 1111:22:3333:5555::15.51150     ->
caffe:affe::1.http         11       1844   FIN

now, with racluster, i get this:

racluster -r /tmp/ipv6traffic -m daddr
   08:01:18.051607  e            ip  1111:22:1111:22::          <->
caffe:affe::1           12119    5900614   CON

with a wrong source net.


there is only one icmp connection in /tmp/ipv6traffic (shown above), and
when i strip it, suddenly the output changes to this:

racluster -r /tmp/ipv6traffic_stripped -m daddr
   08:01:18.051607  e *         tcp          1111:22::           ->
caffe:affe::1           12117    5900436   FIN

with a (more) correct stripped source address.


btw. is there a way to influence the "ip-stripping" process on
aggregation? e.g. in the last case, i would have expected this source
address: 1111:22:3333::

and/or is it possible to somehow indicate on output, if and how many
bits of the source (or destination) address have been stripped?

thanks
frank




More information about the argus mailing list