Date: Tue, 3 Jun 2014 22:32:05 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: "Alexander V. Chernikov" <melifaro@FreeBSD.org> Cc: hackers@freebsd.org Subject: Re: Permit init(8) use its own cpuset group. Message-ID: <20140603193205.GX3991@kib.kiev.ua> In-Reply-To: <538E104D.3000904@FreeBSD.org> References: <538C8F9A.4020301@FreeBSD.org> <20140602164850.GS3991@kib.kiev.ua> <538E104D.3000904@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--hNtiuZBcYkRjyNT4
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Tue, Jun 03, 2014 at 10:13:33PM +0400, Alexander V. Chernikov wrote:
> On 02.06.2014 20:48, Konstantin Belousov wrote:
> > On Mon, Jun 02, 2014 at 06:52:10PM +0400, Alexander V. Chernikov wrote:
> >> Hello list!
> >>
> >> Currently init(8) uses group 1 which is root group.
> >> Modifications of this group affects both kernel and userland threads.
> >> Additionally, such modifications are impossible, for example, in prese=
nce
> >> of multi-queue NIC drivers (like igb or ixgbe) which binds their threa=
ds to
> >> particular cpus.
> >>
> >> Proposed change ("init_cpuset" loader tunable) permits changing cpu
> >> masks for
> >> userland more easily. Restricting user processes to migrate to/from CPU
> >> cores
> >> used for network traffic processing is one of the cases.
> >>
> >> Phabricator: https://phabric.freebsd.org/D141 (the same version attach=
ed
> >> inline)
> >>
> >> If there are no objections, I'll commit this next week.
> > Why is the tunable needed ? IMO it is more reasonable to create a new
> > cpuset always, at least I cannot explain why the existing behaviour
> Initially I've planned to make this behavior to be default. However I=20
> though someone
> will complain on changing defaults. That's why I've introduced the=20
> tunable for that.
>=20
> > is useful. If creating new cpuset unconditionally, you could consider
> > doing it in kernel in start_init(). You would also need to update the
> > cpuset(2) man page, which states that cpuset 1 is set for all processes.
> Please take a look on the new version doing this (also updated in=20
> phabricator,
> https://phabric.freebsd.org/D141 ).
Code changes look fine.
For the documentation changes, I find the formulation like
'userland processes are started in set 2.' incorrect. It would be
cleaner to state that init(8) process is created by kernel with
set 2 assigned.
I also suggest to wait some more before committing the change, for
possible complains from people who do know the reason to do it in the
way of your first patch.
--hNtiuZBcYkRjyNT4
Content-Type: application/pgp-signature
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (FreeBSD)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=i492
-----END PGP SIGNATURE-----
--hNtiuZBcYkRjyNT4--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20140603193205.GX3991>