Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 7 Oct 2020 13:33:31 -0500
From:      Kyle Evans <kevans@freebsd.org>
To:        freebsd-net <freebsd-net@freebsd.org>, FreeBSD Hackers <freebsd-hackers@freebsd.org>
Subject:   Re: Domain Patches
Message-ID:  <CACNAnaFOng71XaKJGCo0SpXBDqLfB3dNebktkWE1TAOdaD1gew@mail.gmail.com>
In-Reply-To: <CACNAnaHGWCGrNMs57zM3TWoWJjF=Or8rZ1_oL13TdTjUKWb3UA@mail.gmail.com>
References:  <CACNAnaHGWCGrNMs57zM3TWoWJjF=Or8rZ1_oL13TdTjUKWb3UA@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Oct 7, 2020 at 11:45 AM Kyle Evans <kevans@freebsd.org> wrote:
>
> Hi,
>
> I have a couple of domain patches in review, if anyone would like to
> comment/review the area:
>
> https://reviews.freebsd.org/D25062 adds a dom_probe callback so that
> domains can indicate whether they should be supported at all or not.
> This avoids some dirty stuff in domains that may not be supported
> based on the hardware that's present (e.g. hvsock included in the
> review, which shouldn't be available outside of HyperV). Instead of
> having to account for circumstances like that in individual callbacks,
> this is a single hammer to prevent any further reach into the domain.
>
> D25062 also updates domain(9) to reflect some changes from the past 11
> years (splitting initialization out into domain_init, dom_destroy
> addition).
>
> https://reviews.freebsd.org/D25459 should make it safe to add domains
> after domainfinalize (i.e. they can be loadable at runtime). The big
> issue with doing this previously is that pfslowtimo/pffasttimo have
> already been setup and may be called on domains that aren't completely
> initialized. There are some additional races re: socket creation that
> should be addressed, but this review should be sufficient for at least
> removing the warning.

Add https://reviews.freebsd.org/D26709 to the list -- in D25459, mjg
suggested adding some lists with domains or protocols that actually
have {slow,fast}timo callbacks to avoid additional overhead from
branching+atomics in those callbacks.

Thanks,

Kyle Evans



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CACNAnaFOng71XaKJGCo0SpXBDqLfB3dNebktkWE1TAOdaD1gew>