LOOP interface is set to node 0 regardless of yaml file or routing argument

[PietroAngeli]

Whatever setting in a yaml file or routing argument, the loop interface is always set to 0.
If a remote node with node address 0 sends packets, they get therefore dropped (see /lib/csp/CHANGELOG line 101).
The loop interface should be on the local node, not 0.

[johandc]

The issue is really with the node, which have address zero. This address has been reserved for a long time, and it was time to put it into use, when adding the loopback interface. I acknowledge that we still see some older nodes starting up with node 0 as a legacy thing. Therefore we manually compile csh without the loopback interface sometimes, to get a CSH capable of talking to an external address on node 0. But this should become increasingly rare, as products stop using address zero as default.

[PietroAngeli]

The point is that setting the LOOP to node 0 is wrong, regardless of legacy compatibility:
Run an instance of CSH with node X and do:
ident X --> no reply, as there is not a specific rule for routing packets destined to node X on the LOOP interface
ident 0 --> the machine running the instance idents itself
the node of the loop interface should be the same set for the local instance of CSH, for example for node 2, KISS interface and CSP V1 the routing table will look like:
2/5 LOOP
0/0 KISS

This means that if we do "ident 2" our own machine replies, while all other nodes interface correctly over KISS.

[PietroAngeli]

By the way, I tested this with satctl hash 04090ec (building with meson 0.62.0 for compatibility) and it works perfectly, the routing table is as i showed and there are no conflicts. I also worked with other implementations of libcsp and they route this way, without the need to reserve node 0.