From: Joel Fernandes Subject: Re: [PATCH RFC] cpuset: Make cpusets get restored on hotplug Date: Thu, 26 Mar 2020 16:23:40 -0400 Message-ID: <20200326202340.GA146657@google.com> References: <20200326191623.129285-1-joel@joelfernandes.org> <20200326192035.GO162390@mtj.duckdns.org> <20200326194448.GA133524@google.com> <972a5c1b-6721-ac20-cec5-617af67e617d@redhat.com> <20200326201649.GQ162390@mtj.duckdns.org> Mime-Version: 1.0 Return-path: DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=joelfernandes.org; s=google; h=date:from:to:cc:subject:message-id:references:mime-version :content-disposition:in-reply-to:user-agent; bh=hJWp5t89/iwbfn8Ca+Lrz1rZRbdYA2KCpWmXHJOj3y4=; b=dY7wSC0838pzka4AzUdCK19Pqo/xSg0AU8qIfHTom1OXY5mFHlZIYMENXzpLnBMDcC NwYnyykFW1mIgRuaJe2SIZNN+JLl1bR6XMi/axhztQ+3Lol92UZB9R5juK0hc6a1xaWl o5DOw2ARWnmKhSSoBhVBErNkXtykPQzMjcX8c= Content-Disposition: inline In-Reply-To: <20200326201649.GQ162390-qYNAdHglDFBN0TnZuCh8vA@public.gmane.org> Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Tejun Heo Cc: Sonny Rao , Waiman Long , linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Dmitry Shmidt , Amit Pundir , kernel-team-z5hGa2qSFaRBDgjK7y7TUQ@public.gmane.org, Jesse Barnes , vpillai-tWEdnPwa2b6ifxm/DiOjgA@public.gmane.org, Peter Zijlstra , Guenter Roeck , Greg Kerr , cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Johannes Weiner , Li Zefan On Thu, Mar 26, 2020 at 04:18:59PM -0400, Tejun Heo wrote: > On Thu, Mar 26, 2020 at 01:05:04PM -0700, Sonny Rao wrote: > > I am surprised if anyone actually wants this behavior, we (Chrome OS) > > The behavior is silly but consistent in that it doesn't allow empty active > cpusets and it has been like that for many many years now. > > > found out about it accidentally, and then found that Android had been > > carrying a patch to fix it. And if it were a desirable behavior then > > why isn't it an option in v2? > > Nobody is saying it's a good behavior (hence the change in cgroup2) and there > are situations where changing things like this is justifiable, but, here: > > * The proposed change makes the interface inconsistent and does so > unconditionally on what is now a mostly legacy interface. > > * There already is a newer version of the interface which includes the > desired behavior. > > * I forgot but as Waiman pointed out, you can even opt-in to the new > behavior, which is more comprehensive without the inconsitencies, > while staying on cgroup1. Thank you Tejun, Waiman and Sonny. I confirmed the cgroup_v2_mode option fixes cgroup v1's broken behavior. We will use this mount option on our systems to fix it. thanks, - Joel