gnrc_nib: Initial import of a general information base for GNRC

[miri64]

Next step on the route to a new NDP.

Depends on #5269.

[OlegHahm]

Can you give a little more information, about the purpose of this module? E.g. how many instances of gnrc_nib_entry_t do you expect to exist on a typical 6LoWPAN node? The struct lukes quiet huge.

[miri64]

It's not much bigger than a neighbor cache entry (or probably smaller due to the extimer_event_ts instead of xtimer_t + msg_t). This module is to be targeted to replace the neighbor cache along with the prefix and default router list (as a list of gnrc_nib_entry_ts). Additional it introduces the destination cache which wasn't present in GNRC up until now.

[OlegHahm]

Wow, haven't realized how huge the neighbor cache entries are. Thanks for the explanation anyway.

[OlegHahm]

Where are the actual IPv6 addresses stored?

[miri64]

In some central array, to avoid duplicate addresses.

[BytesGalore]

How about using universal_address_ts [1] for prefix and next_hop?
Using this container allows for sharing the stored address with the FIB and other modules.
However, it requires to use the provided functions for thread safety, and creating a new container with an address requires to iterate/search the current container entries.

[1] https://github.com/RIOT-OS/RIOT/blob/master/sys/include/universal_address.h#L63

[miri64]

To be honest I would like to merge the NIB and the FIB in the long term, since they are basically doing the same thing (except that the FIB is providing source routing too and the NIB in addition is containing link-layer and prefix list information). Since your FIB however had proven to be more of a GNRC thing (since other stack implementations come with their own FIB in most cases), I'm not sure if the universality universal_address_t is providing isn't too much overkill

[kYc0o]

This is a new feature and is waiting for other PR, so I think it won't make it for this release... postponing...

[miri64]

Considering #5704, I'm not sure I will continue the work on this PR the way I intended to, so I close it for now.