On Wed, 7 Sep 2022, Gleb Smirnoff wrote:
On Wed, Sep 07, 2022 at 09:54:49AM +0000, Bjoern A. Zeeb wrote:
B> On Tue, 6 Sep 2022, Gleb Smirnoff wrote:
B>
B> > On Sun, Sep 04, 2022 at 06:39:53AM +0000, Bjoern A. Zeeb wrote:
B> > B> > B> > Author: Gleb Smirnoff <[email protected]>
B> > B> > B> > AuthorDate: 2022-08-30 22:09:21 +0000
B> > B> > B> > Commit: Gleb Smirnoff <[email protected]>
B> > B> > B> > CommitDate: 2022-08-30 22:09:21 +0000
B> > B> > B> >
B> > B> > B> > divert(4): make it compilable and working without INET
B> > B> > B> >
B> > B> > B> > Differential revision: https://reviews.freebsd.org/D36383
B> > B> > B>
B> > B> > B> Well, almost at least....
B> > B> >
B> > B> > I wonder why LINT-NOINET builds without this??
B> > B>
B> > B> I wondered the same last night.. I would be good to find out?
B> >
B> > Ok, I think that assumption that sysctl parts of divert require
B> > INET were just wrong, that's why LINT-NOINET builds.
B>
B> Not entirely. It's because sys/netinet/in_proto.c uncodonitionally
B> currently provides _net_inet for the sysctl.
B> The reason for that is that a lot of AF independent sysctls were initially
B> put under inet but are equally used for inet6 or neither in the network
stack.
B>
B> You will need at least #if defined(INET6) || defined(INET) around the
sysctls
B> in divert.c then.
B>
B> A LINT-NOIP kernel will complain about the missing symbol if we keep
compiling
B> divert without INET or INET6 in the kernel:
My goal was not to make divert(4) supported on NOIP kernel. The goal was to make
it supported on NOINET, an IPv6 only kernel. I don't see any value for divert(4)
on a kernel without either stack.
then we should stop compiling it in that case.
--
Bjoern A. Zeeb r15:7
