Re: git: 7b8151b0cd1b - main - sctp: improve handling of SO_SETFIB socket option
- In reply to: Michael Tuexen : "git: 7b8151b0cd1b - main - sctp: improve handling of SO_SETFIB socket option"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Mon, 26 Oct 2026 20:19:10 UTC
On Mon, Oct 26, 2026 at 07:19:04PM +0000, Michael Tuexen wrote:
M> commit 7b8151b0cd1b4f077b382b6be2c02c936d1988fa
M> Author: Albin Hellqvist <albin.hellqvist@ericsson.com>
M> AuthorDate: 2026-10-26 19:07:03 +0000
M> Commit: Michael Tuexen <tuexen@FreeBSD.org>
M> CommitDate: 2026-10-26 19:07:03 +0000
M>
M> sctp: improve handling of SO_SETFIB socket option
M>
M> Since SCTP keeps fibnum in the SCTP protocol control block,
M> the code can't just defer the processing to the IPv4 or IPv6
M> layer. Therefore it needs to be implemented in the the SCTP
M> layer and it needs to call the socket layer function.
M>
M> PR: 300091
M> Reviewed by: tuexen
M> Fixes: caccbaef8e26 ("socket: Move SO_SETFIB handling to protocol layers")
M> MFC after: 3 days
M> MFC to: stable/14
M> MFC to: stable/15
This change reminded me of the mess we have. The inpcb already stores
the fib in inp_inc.inc_fibnum and some protocols (TCP and SCTP) copy
it from so->so_fibnum.
In general FIB number is definitely not a socket feature - a unix(4),
netlink(4), netgraph(4), list to be continued can't have a notion of
FIB number.
This all needs to be cleaned up.
--
Gleb Smirnoff