inpcb: declare struct in_conninfo as a single declaration

[glebius]

Summary

This removes just one level of #define mess that is needed to reach into
an inpcbs IPv4 address. And makes the declaration easier to read. No
functional change.

[glebius]

Note: union in_dependaddr can be anonymized after small cleanup of lbgroup code and siftr(4). This removes another level of of defines for both IPv6 and IPv4. This may need universe/exp-run, though.

The idea is that instead of union in_dependaddr ie_dependfaddr, ie_dependladdr; two anonymous unions are used. Identical in their layout, but one has members with f prefix and other with l.

If there is some consensus on this being a good idea, I can run test builds and post the review,

[glebius]

The struct in_endpoints can also be anonymized easily, but that would highly likely affect something outside of FreeBSD. Another level of #define to be removed. Let me know what you think about this and I will dig deeper into this if we all agree.

Alternatively we can leave publicly declared struct in_conninfo as is and just stop using it internally. But this review (D58273) probably should go in. It must be safe.

[markj]

sys/netinet/in_pcb.h

41–45

BTW, there is nothing providing this guarantee, I believe: the natural alignment of this structure is 4, not 8.

51–52

What does this comment mean?

553

This last sentence is stale.

[glebius]

sys/netinet/in_pcb.h

41–45

AFAIU the comment, it says that the existing members at the beginning of the structure align the IPv6 addresses to 8.

(kgdb) set $inc = (struct in_conninfo *)0
(kgdb) p $inc->inc_ie.ie_dependfaddr
Cannot access memory at address 0x8

And you are right, there is no natural alignment, and inside inpcb the addresses are not aligned:

(kgdb) p/x (uintptr_t)&$inp->inp_inc->inc_ie.ie_dependfaddr & (8-1)
$6 = 0x4

Do you suggest to remove the comment?

51–52

The was pad one? I'd be glad to remove it.

553

Thanks!

[markj]

sys/netinet/in_pcb.h

41–45

I would at least remove the claim that it is aligned today, maybe add an XXX comment describing your observation.

51–52

Yes, I'm not sure what it's trying to say.

[glebius]

Fixup comments:

• Remove comment about alignment. inpcb doesn't have IPv6 addresses aligned and the RFC talks about alignment of API structures, specifically sockaddr_in6.
• Remove old mentioning of old pad
• Fixup mentioning of a structure that is just deleted