Cumulus FRR assign LLAs to point-to-point OSPF interfaces without addresses

[arifogel]

• enables forwarding across unnumbered links

[batfish-bot]

This change is 

[codecov]

Codecov Report

Merging #7038 (48f99ce) into master (9053587) will increase coverage by 0.01%.
The diff coverage is 96.42%.

@@             Coverage Diff              @@
##             master    #7038      +/-   ##
============================================
+ Coverage     71.89%   71.91%   +0.01%     
- Complexity    37699    37711      +12     
============================================
  Files          3036     3036              
  Lines        154740   154761      +21     
  Branches      18564    18570       +6     
============================================
+ Hits         111258   111290      +32     
+ Misses        34592    34585       -7     
+ Partials       8890     8886       -4     

Impacted Files	Coverage Δ
...tion/cumulus/CumulusConcatenatedConfiguration.java	73.63% <96.42%> (+1.08%)	⬆️
...ar/cumulus_frr/CumulusFrrConfigurationBuilder.java	78.23% <0.00%> (+0.12%)	⬆️
...ain/java/org/batfish/storage/FileBasedStorage.java	84.76% <0.00%> (+0.26%)	⬆️
...src/main/java/org/batfish/coordinator/WorkMgr.java	75.81% <0.00%> (+0.38%)	⬆️
...rg/batfish/datamodel/ConcreteInterfaceAddress.java	96.87% <0.00%> (+6.25%)	⬆️
...n/java/org/batfish/datamodel/LinkLocalAddress.java	89.28% <0.00%> (+7.14%)	⬆️

[dhalperi]

Reviewed 3 of 3 files at r1.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @arifogel and @corinaminer)

---

projects/batfish/src/main/java/org/batfish/representation/cumulus/CumulusConcatenatedConfiguration.java, line 226 at r1 (raw file):

        .forEach(
            (iname, iface) -> {
              if (iface.getAllAddresses().size() == 0

It feels like there is a missing && <some vs property> that indicates this interface was configured for unnumbered.

Above, the function was labeled something like isUsedForBgpUnnumbered(ifacename, bgp VS model). I imagine something generic there.

---

projects/batfish/src/main/java/org/batfish/representation/cumulus/CumulusConcatenatedConfiguration.java, line 226 at r1 (raw file):

        .forEach(
            (iname, iface) -> {
              if (iface.getAllAddresses().size() == 0

From discussion: seems like we should add an LLA unconditionally? Otherwise there might not be an L3 edge.

Also from discussion: it seems like there should not even be able to be a concrete address assigned. Unnumbered should be mutually exclusive with concrete.

[arifogel]

Reviewable status: 0 of 4 files reviewed, 3 unresolved discussions (waiting on @arifogel, @corinaminer, and @dhalperi)

a discussion (no related file):
After #7039

---

[arifogel]

Reviewable status: 0 of 4 files reviewed, 3 unresolved discussions (waiting on @arifogel, @corinaminer, and @dhalperi)

---

projects/batfish/src/main/java/org/batfish/representation/cumulus/CumulusConcatenatedConfiguration.java, line 226 at r1 (raw file):

Previously, dhalperi (Dan Halperin) wrote…

From discussion: seems like we should add an LLA unconditionally? Otherwise there might not be an L3 edge.

Also from discussion: it seems like there should not even be able to be a concrete address assigned. Unnumbered should be mutually exclusive with concrete.

I don't believe we said they should be mutually exclusive. An interface could win one unnumbered and address and lose another. In that case it would have a concrete address, but still may need the LLA for L3 edge.

In any case, now adding the LLA even for winning interface.

---

projects/batfish/src/main/java/org/batfish/representation/cumulus/CumulusConcatenatedConfiguration.java, line 226 at r1 (raw file):

Previously, dhalperi (Dan Halperin) wrote…

It feels like there is a missing && <some vs property> that indicates this interface was configured for unnumbered.

Above, the function was labeled something like isUsedForBgpUnnumbered(ifacename, bgp VS model). I imagine something generic there.

Now relying on ospfAddresses being non-empty. This happens when at least 1 address was assigned in VS interfaces or frr file.

[dhalperi]

Reviewed 4 of 4 files at r2.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @corinaminer)