[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y+I61dp5qPkWK4s2@lothringen>
Date: Tue, 7 Feb 2023 12:49:41 +0100
From: Frederic Weisbecker <frederic@...nel.org>
To: Michal Koutný <mkoutny@...e.com>
Cc: LKML <linux-kernel@...r.kernel.org>,
Roman Gushchin <roman.gushchin@...ux.dev>,
Michal Hocko <mhocko@...e.com>,
Marcelo Tosatti <mtosatti@...hat.com>,
Leonardo <leobras@...hat.com>,
Johannes Weiner <hannes@...xchg.org>,
Shakeel Butt <shakeelb@...gle.com>,
Muchun Song <muchun.song@...ux.dev>,
Andrew Morton <akpm@...ux-foundation.org>,
Peter Zijlstra <peterz@...radead.org>,
Thomas Gleixner <tglx@...utronix.de>, cgroups@...r.kernel.org,
linux-mm@...ck.org
Subject: Re: [PATCH 1/2] sched/isolation: Merge individual nohz_full features
into a common housekeeping flag
On Mon, Feb 06, 2023 at 04:51:09PM +0100, Michal Koutný wrote:
> Hello Frederic.
>
> On Sat, Feb 04, 2023 at 12:24:08AM +0100, Frederic Weisbecker <frederic@...nel.org> wrote:
> > The individual isolation features turned on by nohz_full were initially
> > split in order for each of them to be tunable through cpusets. However
> > plans have changed in favour of an interface (be it cpusets or sysctl)
> > grouping all these features to be turned on/off altogether.
> > Then should the need ever arise, the interface can still be expanded
> > to handle the individual isolation features.
> >
> > Therefore the current isolation split between tick/timer/workqueue/rcu/
> > kthreads/misc doesn't make sense anymore.
>
> Why it doesn't make sense? I think it's a useful annotation of
> respective operations wrt CPU isolation.
But what do we need these annotations for? The only outcome I've ever
seen with these is that it confuses everyone.
>
> The grouping you did into HK_TYPE_KERNEL_NOISE (or even coarser) should
> IMO be done at the place where it'll be exposed into the favored
> interface (like it's with nohz_full=).
That being said I should reserve the grouping to HK_TYPE_KERNEL_NOISE when
I'll introduce the cpuset interface. This way I can add the support for
each part smoothly. For example first patch moves HK_TYPE_TIMER to
HK_TYPE_KERNEL_NOISE and unbound timers are supported by cpuset.kernel_noise,
second patch moves HK_TYPE_WQ to HK_TYPE_KERNEL_NOISE and unbound workqueues
are supported by cpuset.kernel_noise, etc until all of them turned by nohz_full=
are supported... This is what I'm doing in fact but I'm so slow to write this patchset...
Thanks.
Powered by blists - more mailing lists