RFC 2292 struct and constant definitions

[Skyb0rg007]

Description

This PR adds the structs and constants as defined by RFC 2292 - Advanced Sockets API for IPv6.
These constants are available for use in get/setsockopt on supported Unix platforms.
The operations defined on the icmp6_filter struct are defined as C macros, so the implementations are not included here.

Sources

https://github.com/bminor/glibc/blob/04e750e75b73957cf1c791535a3f4319534a52fc/inet/netinet/icmp6.h
https://github.com/kraj/musl/blob/kraj/master/include/netinet/icmp6.h
https://github.com/freebsd/freebsd-src/blob/9ae367d11de8abbdf53884836c9ba30908c5c8db/sys/netinet/icmp6.h

Also in the RFC: https://datatracker.ietf.org/doc/html/rfc2292#section-3.2

Checklist

Unfortunately I could not get the libc-tests working locally, so this PR is marked as draft.

• Relevant tests in libc-test/semver have been updated
• No placeholder or unstable values like *LAST or *MAX are
  included (see #3131)
• Tested locally (cd libc-test && cargo test --target mytarget);
  especially relevant for platforms that may not be checked in CI

@rustbot label +stable-nominated

[Skyb0rg007]

Alright, I need help. On *BSD platforms (FreeBSD, macOS, etc), struct in6_addr is defined as the union of uint32_t[4], uint16_t[8], and uint8_t[16], so Rust encodes it as [u8; 16] with repr(align(4)).

But the nd_(router|neighbor)_(solicit|advert) structures are marked as __packed (so their C alignment is 1), even though they include a in6_addr. Rust won't allow using repr(packed) on a struct containing a repr(align(...)) attribute, so I'm stuck.

Linux doesn't mark these structs as __packed, so there's no issue there.

[Skyb0rg007]

I am most interested in the addition of the ICMP6_FILTER constant to libc. If it would help with review and acceptance, I can create a separate PR with just that single change, and leave the other RFC constants and structures for later PRs.

[Skyb0rg007]

I see that some C macros such as the CMSG_ macros were added this package as const functions. It may make sense to add the ICMP6_FILTER_ macros too, especially since the icmp6_filter struct’s field name is different on Solaris.

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[tgross35]

Sorry for the delay, thank you for the changes here. The contents themselves look good, but I think they should be moved out of the root unix module.

View changes since this review

[tgross35]

For any reserved/padding fields, please make them private and wrap them in Padding

[tgross35]

This struct should be opaque, can you make the fields private?

[tgross35]

I think enabling these in the unix module is too high level; do we know anything about AIX, fuchsia, redox, etc?

Instead, could you put these into the linux_like and bsd modules? That will mean some duplication, but that's probably fine because it will simplify some of the repetitive padding config.

[tgross35]

What would be ideal here is to instead put these in src/new/common/bsd/{icmp6.rs,ip6.rs} and src/new/common/linux_like/{icmp6.rs,ip6.rs}, to match the header structure. The reexport them in src/new/{bionic_libc,musl,glibc,openbsd,freebsd,...}/{icmp6.rs,ip6.rs}, similar to what is done with unistd, and update src/new/mod.rs.

I won't require that because this has already been waiting a while, but eventually everything will move to that new structure so anything to help that is appreciated.

[tgross35]

Could you add links to the PR description to the headers for these other platforms that have the specific alignment requirements? (Looks like only glibc/musl/freebsd are there)

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.