Fix PIMD RPF lookup mode and nexthop tracking

[nabahr]

Generalized the zapi ZEBRA_NEXTHOP_LOOKUP_MRIB to just ZEBRA_NEXTHOP_LOOKUP and added the safi into the request message.

Moved the RPF lookup mode configuration from zebra into pimd.
Now the command show ip rpf x.x.x.x command will just do a lookup against the MRIB, not according to the lookup mode.
To do a nexthop lookup according to the lookup mode, use the command show ip pim nexthop-lookup x.x.x.x y.y.y.y where x.x.x.x is the source address, and y.y.y.y is the group address, which may affect the lookup.

Refactored the PIM zlookup (the synchronous nexthop lookup using ZEBRA_NEXTHOP_LOOKUP) to take the RPF lookup mode into account, which means we may lookup via the URIB, MRIB, or both.

Refactored the PIM next hop tracking and isolated the nexthop lookup and tracking code into pim_nht.h/c.
Now we track in both the MRIB and URIB tables so that decisions about the nexthop based on the RPF lookup mode can be taken into account, and changes to the lookup mode can be reflected immediately without performing additional lookups.

Updated show commands, documentation, and tests. Including a new PIM mrib test to exercise the RPF lookup mode command and verify that NHT is working correctly.

[nabahr]

New mrib tests pass when built with changes from PR #17254.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[eqvinox]

Now the command show ip rpf x.x.x.x command will just do a lookup against the MRIB, not according to the lookup mode.

Hm. I think this is a little problematic UX-wise because it's a "silent" change users can trip over without noticing. Can we just add a multicast or safi multicast option to the show ip route commands? That'd also have the benefit of making the other show ip route options available for looking at the MRIB

[nabahr]

Now the command show ip rpf x.x.x.x command will just do a lookup against the MRIB, not according to the lookup mode.

Hm. I think this is a little problematic UX-wise because it's a "silent" change users can trip over without noticing. Can we just add a multicast or safi multicast option to the show ip route commands? That'd also have the benefit of making the other show ip route options available for looking at the MRIB

Since the rpf lookup mode was moved out of zebra, the command cannot work as it did before these changes. So should this command show ip rpf [A.B.C.D] be removed completely? Deprecated? Add a warning in the output?

[eqvinox]

So should this command show ip rpf [A.B.C.D] be removed completely?

I think that's the only reasonable option, and/or replace it with a stub that prints "this command doesn't exist anymore because it's in pimd now"

Now that I think about it… the command can be implemented in pimd and vtysh can dispatch it there instead

[eqvinox]

Actually — "duh" — nothing says the command needs to stay in zebra, pimd can just handle it?

(Edit: the variant with the IP address parameter — showing the table is probably not doable/sensible from pimd)

[nabahr]

Ok, so I believe the best path forward to clean up the MRIB UX:

• Move show ip rpf x.x.x.x from zebra to pim, aliased (and probably deprecated) to the existing show ip pim nexthop-lookup ... command.
• Expand existing show ip route ... to include a mrib flag for showing the multicast safi table show ip route mrib ...
• Keep show ip rpf in zebra, but deprecated/aliased to show ip route mrib.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[donaldsharp]

can you please fix this documentation issue, that is now showing up:

make[1]: *** [Makefile:8699: pimd/pim6d] Error 1
make[1]: *** Waiting for unfinished jobs....
/home/sharpd/frr4/doc/user/pim.rst:789: WARNING: undefined label: multicast-rib-commands
make[1]: Leaving directory '/home/sharpd/frr4'

[donaldsharp]

More issues:

pimd/pim_rp.c: In function ‘pim_embedded_rp_new’:
pimd/pim_rp.c:1499:9: warning: implicit declaration of function ‘pim_find_or_track_nexthop’ [-Wimplicit-function-declaration]
 1499 |         pim_find_or_track_nexthop(pim, rp_info->rp.rpf_addr, NULL, rp_info, NULL);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~
  CC       yang/pimd_pim6d-frr-pim-candidate.yang.o
pimd/pim_rp.c:1500:14: warning: implicit declaration of function ‘pim_ecmp_nexthop_lookup’ [-Wimplicit-function-declaration]
 1500 |         if (!pim_ecmp_nexthop_lookup(pim, &rp_info->rp.source_nexthop, rp_info->rp.rpf_addr,
      |              ^~~~~~~~~~~~~~~~~~~~~~~
  CC       yang/pimd_pim6d-frr-gmp.yang.o
  CC       pbrd/pbr_main.o
pimd/pim_rp.c: In function ‘pim_embedded_rp_free’:
pimd/pim_rp.c:1541:9: warning: implicit declaration of function ‘pim_delete_tracked_nexthop’ [-Wimplicit-function-declaration]
 1541 |         pim_delete_tracked_nexthop(pim, rp_info->rp.rpf_addr, NULL, rp_info);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~

[nabahr]

I rebased and fixed the errors from a bad merge with embedded RP. Docs link issue also fixed.