Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210513175349.959661-1-surenb@google.com> <YJ5iAvqAmIhzJRot@hirez.programming.kicks-ass.net>
 <CAJuCfpHy+MknCepfjx9XYUA1j42Auauv7MFQbt+zOU-tA4gasA@mail.gmail.com>
 <YJ64xHoogrowXTok@hirez.programming.kicks-ass.net> <CAJuCfpGkj9HxbkXnYN58JXJp1j6kVkvQhqscnEfjyB5unKg1NQ@mail.gmail.com>
In-Reply-To: <CAJuCfpGkj9HxbkXnYN58JXJp1j6kVkvQhqscnEfjyB5unKg1NQ@mail.gmail.com>
From:   Suren Baghdasaryan <surenb@google.com>
Date:   Fri, 14 May 2021 11:50:06 -0700
Message-ID: <CAJuCfpH2X47_3VvfZXs_eWhYDziOh13qdUwcfxPJe=Zg_Nkvqw@mail.gmail.com>
Subject: Re: [PATCH 1/1] cgroup: make per-cgroup pressure stall tracking configurable
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Tejun Heo <tj@kernel.org>, Johannes Weiner <hannes@cmpxchg.org>,
        lizefan.x@bytedance.com, Ingo Molnar <mingo@redhat.com>,
        Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Benjamin Segall <bsegall@google.com>, mgorman@suse.de,
        Minchan Kim <minchan@kernel.org>,
        Jonathan Corbet <corbet@lwn.net>, bristot@redhat.com,
        "Paul E . McKenney" <paulmck@kernel.org>, rdunlap@infradead.org,
        Andrew Morton <akpm@linux-foundation.org>,
        Thomas Gleixner <tglx@linutronix.de>, macro@orcam.me.uk,
        Viresh Kumar <viresh.kumar@linaro.org>,
        mike.kravetz@oracle.com, linux-doc@vger.kernel.org,
        LKML <linux-kernel@vger.kernel.org>,
        cgroups mailinglist <cgroups@vger.kernel.org>,
        kernel-team <kernel-team@android.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, May 14, 2021 at 11:20 AM Suren Baghdasaryan <surenb@google.com> wrote:
>
> On Fri, May 14, 2021 at 10:52 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > On Fri, May 14, 2021 at 08:54:47AM -0700, Suren Baghdasaryan wrote:
> >
> > > Correct, for this function CONFIG_CGROUPS=n and
> > > cgroup_disable=pressure are treated the same. True, from the code it's
> > > not very obvious. Do you have some refactoring in mind that would make
> > > it more explicit?
> >
> > Does this make sense?
> >
> > --- a/kernel/sched/psi.c
> > +++ b/kernel/sched/psi.c
> > @@ -744,24 +744,26 @@ static void psi_group_change(struct psi_
> >
> >  static struct psi_group *iterate_groups(struct task_struct *task, void **iter)
> >  {
> > +       if (cgroup_psi_enabled()) {
> >  #ifdef CONFIG_CGROUPS
> > -       struct cgroup *cgroup = NULL;
> > +               struct cgroup *cgroup = NULL;
> >
> > -       if (!*iter)
> > -               cgroup = task->cgroups->dfl_cgrp;
> > -       else if (*iter == &psi_system)
> > -               return NULL;
> > -       else
> > -               cgroup = cgroup_parent(*iter);
> > +               if (!*iter)
> > +                       cgroup = task->cgroups->dfl_cgrp;
> > +               else if (*iter == &psi_system)
> > +                       return NULL;
> > +               else
> > +                       cgroup = cgroup_parent(*iter);
> >
> > -       if (cgroup && cgroup_parent(cgroup)) {
> > -               *iter = cgroup;
> > -               return cgroup_psi(cgroup);
> > -       }
> > -#else
> > -       if (*iter)
> > -               return NULL;
> > +               if (cgroup && cgroup_parent(cgroup)) {
> > +                       *iter = cgroup;
> > +                       return cgroup_psi(cgroup);
> > +               }
> >  #endif
> > +       } else {
> > +               if (*iter)
> > +                       return NULL;
> > +       }
> >         *iter = &psi_system;
> >         return &psi_system;
> >  }
>
> Hmm. Looks like the case when cgroup_psi_enabled()==true and
> CONFIG_CGROUPS=n would miss the "if (*iter) return NULL;" condition.
> Effectively with CONFIG_CGROUPS=n this becomes:
>
>        if (cgroup_psi_enabled()) {           <== assume this is true
> #ifdef CONFIG_CGROUPS                <== compiled out
> #endif
>        } else {
>                if (*iter)                                  <== this
> statement will never execute
>                        return NULL;
>        }
>        *iter = &psi_system;
>         return &psi_system;
>

Ah, sorry. I forgot that CONFIG_CGROUPS=n would force
cgroup_psi_enabled()==false (the way function is defined in cgroup.h),
so (CONFIG_CGROUPS=n && cgroup_psi_enabled()==true) is an invalid
configuration. I think adding a comment to your suggestion would make
it more clear.
So your suggestion seems to work. I'll test it and include it in the
next revision. Thanks!


> >
> > --
> > To unsubscribe from this group and stop receiving emails from it, send an email to kernel-team+unsubscribe@android.com.
> >