netinet6: Pass IPv4-mapped ASM multicast joins/leaves to netinet.

[bms]

Summary

Add comment about future IPv4 source-address selection support.

This has been compile tested with LINT and LINT-NOINET,
but has not yet been runtime tested.

EAFNOSUPPORT change moved to D55233.
Track other dependent refactoring requested by Gleb:
D55344
D55345
D55382

Inspired by: xnu (Clean-roomed from APSL change)
PR: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=193246

Test Plan

Kyua tests will follow in a separate commit. I note that not all of the legacy IP multicast regression tests were ported to ATF over the years.

[bms]

The IPv4 mapped behaviour is limited to ASM memberships only, and is not directly mandated by an RFC. The comments around in6u_lookup_v4mapped_mc() reflect this-- it's something that has seen wider use than one would anticipate, despite it being kind of dog stupid to do IPv4 operations on an IPv6 socket.

This revision is also missing a man page update. Historically Pavlin Radoslavov has handled the multicast routing end of things, here it needs to go in ip6(4).

@bz: The Apple xnu code has cleaned up scope ID handling, so that may be
worth abridging from also. They added scope checks to im6o_mc_filter(), conditionalised on their own in6_embedded_scope global.

They renamed in6_joingroup_locked() to in6_mc_join() which is a little confusing, and it looks like they demoted the per-socket SSM filters from an RB-tree to a growable array. That I can understand; stuff gets merged into the stack-wide RB-tree for each group anyway.

However I didn't check what changes they may have made to imo_multi_filter(), yet.

OT: Apple do not provide meaningful change history in their xnu drops on GitHub, which is extremely inconvenient.

OT: I am in touch separately with core-secretary@ to open a dialogue about regaining full commit access after a 7.5 year hiatus for health & bereavement reasons.

[glebius]

Bruce, very glad to see you back!!!

sys/netinet6/in6_mcast.c

1462

I'd suggest to commit all of the changes from EINVALs to EAFNOSUPPORT as a separate commit.

1894

Let's spell as AF_INET6 for consistency.

1899

What does letter u in in6u_ stands for?

1918

IMHO, this macro should go away. We should search for first AF_INET address on the list of addresses of the given interface instead of searching on the global list. btw, this macro is used only in multicast related code.

You may want to add the suggested code as a function, e.g. in_firstaddr() and place it in in.c next to in_ifhasaddr().

1997–2003

This all can be a static initializer.

2005

Let's add brackets per style(9).

2337–2343

Can be static initializer.

2340

Is the void cast needed at all?

2345

Let's add brackets per style(9).

2354

Why do we need these extra braces?

[bms]

sys/netinet6/in6_mcast.c

1462

Re-raised as D55233.

[bms]

Rebase

[bms]

Correctly claim authorship.

[rgrimes]

Please check that this compiles properly with possible combinations INET, INET6, INET & INET6. It looks to be as if your using INET to protect INET6 only code.

Um, probably wrong on that, as these files well not be pulled in unless compiling INET6

[bms]

In D55031#1266797, @rgrimes wrote:

Please check that this compiles properly with possible combinations INET, INET6, INET & INET6. It looks to be as if your using INET to protect INET6 only code.

Um, probably wrong on that, as these files well not be pulled in unless compiling INET6

The #ifdef INET is not being used in its normal sense; it is there to denote that the IPv4-mapped address support requires INET, despite being referenced from a file that is covered by the INET6 option. So your initial confusion is understandable.

Also, the only reason this trick works at all is that inpcb->inp_moptions exists and is populated even for PF_INET6 sockets.

[pouria]

sys/netinet6/in6_mcast.c

1893

IMHO, there is no need to mention OpenJDK here.

1896

1902–1905

2001–2002

2341–2342

[glebius]

Do you plan to rebase that over removal of IFP_TO_IA() in preference to in_ifprimaryaddr()?

[bms]

Mend and make do.