Bluetooth: Mesh: Filter duplicates in brg subnets list

[ludvigsj]

Adds functionality to filter out duplicate entries in the
Bridged Subnets List message.

This is done by iterating through the part of the table we have already
processed, to check if this is the first time we see a given key pair or
not.

Also, adds a babblesim test that checks

• That duplicate net key index pairs are filtered out in the bridged
  subnets list.
• That indexing into the filtered list works as expected (indexing
  happens into the already filtered list, not the full bridging table)

[ludvigsj]

Quick explanation here

I considered several solutions:

1. The current solution, just iterating through the (start of the) table again to identify duplicates.
   • $O(n^2)$ time. Creates an overhead of < 10ms on nRF52840 in the worst case, with a maxed out table (255 entries) where all subnet pairs are unique. Goes above 100ms when close to 1000 lines in the table (which is currently not allowed by Kconfig)
2. Store the already seen pairs in a separate buffer/list
   • Will require 3 bytes to store each unique pair, meaning potentially 768 bytes of buffer if the table size is maxed out.
   • Does not give any advantage in the worst case where every pair is unique - in this case it will just become a copy of the bridging table and the runtime will be the same as solution 1.
   • Might give a small runtime advantage for long tables with a small/medium number of unique pairs.
     • Note: Also identical to solution 1 if there is only one unique pair (meaning every idx1 in the table is identical and every idx2 in the table is identical). They are both $O(n)$ in this best case.
3. Sort the bridging table on insertion.
   • This allows $O(n)$ time to construct the filtered list, compared to $O(n^2)$ for the two other solutions.
   • If we want to support >1000 entries, I believe this is the solution, it is performant enough that the table size is limited by available RAM before we see slowdown.
   • I have not implemented this since we currently only support 255 entries, to save complexity in the code change.

[alxelax]

In general, it looks Ok to me apart test scenario. Could you elaborate, how do you check start index? I would like to see that filtering exactly works.

[PavelVPV]

LGTM, but I'd add few more entries into subnet_pairs to check filtering.

[PavelVPV]

Approving. Up to you if you want to fix the minor comments.

[ludvigsj]

Updated with Pavels requested style changes

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud