Page MenuHomeFreeBSD

udp: honor IPV6_TCLASS cmsg for UDP/IPv4 packets
ClosedPublic

Authored by tuexen on Oct 24 2025, 8:46 PM.
Tags
None
Referenced Files
F137867327: D53347.id165083.diff
Wed, Nov 26, 3:27 PM
Unknown Object (File)
Mon, Nov 17, 7:31 AM
Unknown Object (File)
Sat, Nov 15, 6:07 PM
Unknown Object (File)
Sun, Nov 2, 11:42 PM
Unknown Object (File)
Sat, Nov 1, 5:00 PM
Unknown Object (File)
Sat, Nov 1, 3:47 PM
Unknown Object (File)
Sat, Nov 1, 3:24 AM
Unknown Object (File)
Thu, Oct 30, 7:35 PM

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

Why don't embed this entire block into udp_v4mapped_pktinfo()?

Why don't embed this entire block into udp_v4mapped_pktinfo()?

Because we are neither handling the cmsg of type IPV6_2292PKTINFO or IPV6_PKTINFO. There is more work needed to clean things up. But I want bz@ in that loop and he is short of time now. We can have a function which does all the mappings, but I would suggest to do this refactoring separately and also deal with the TTL/hop limit stuff.
I would like to get this in FreeBSD 15. It is used when doing ECN for QUIC.

Why don't embed this entire block into udp_v4mapped_pktinfo()?

Because we are neither handling the cmsg of type IPV6_2292PKTINFO or IPV6_PKTINFO. There is more work needed to clean things up. But I want bz@ in that loop and he is short of time now. We can have a function which does all the mappings, but I would suggest to do this refactoring separately and also deal with the TTL/hop limit stuff.
I would like to get this in FreeBSD 15. It is used when doing ECN for QUIC.

If you want @bz in the loop remove v4mapped sockets, remove IPv4, be done with this. /SCNR

There's also bits in TCP which would need proper dealing with some of this still and just have placeholders. I think it'll all need a rather proper makeover and unification between all bits and it can properly be done nicely enough setting letting a function know which bits we are interested and return them like we do in the IPv6 side.

I believe even if not neat and I don't like any of this, this is right.

This revision is now accepted and ready to land.Oct 26 2025, 5:55 PM