Index | Thread | Search

From:
"Theo de Raadt" <deraadt@openbsd.org>
Subject:
Re: bgpd: add infrastructure to support rtr sessions with tcp md5 and ipsec
To:
tech@openbsd.org
Date:
Wed, 02 Oct 2024 09:28:29 -0600

Download raw body.

Thread
I've got a weird vibe about this imsg request being IMSG_SOCKET_CLOSE,
which looks too close to close().

The way things work is with IMSG_SOCKET_CONN the speaker asks the master
(who sets up ipsec-or-md5-network configuration, then provides a socket fd,
but closes this fd in the master itself), then the speaker moves data, and
eventually the speaker calls [the only] close() on the fd, and tells the
the master to deconfigure the ipsec-or-md5-network configuration.  This
operation is being called IMSG_SOCKET_CLOSE.

Well, isn't actually a close (the imsg receiver does not have a copy of
the fd).

I just find the names IMSG_SOCKET_CONN and IMSG_SOCKET_CLOSE a bit weird.