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;
Date:   Tue, 27 Jul 2021 11:16:48 +0100
From:   Quentin Perret <qperret@google.com>
To:     Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc:     mingo@redhat.com, peterz@infradead.org, vincent.guittot@linaro.org,
        qais.yousef@arm.com, rickyiu@google.com, wvw@google.com,
        patrick.bellasi@matbug.net, xuewen.yan94@gmail.com,
        linux-kernel@vger.kernel.org, kernel-team@android.com
Subject: Re: [PATCH v4 2/2] sched: Skip priority checks with
 SCHED_FLAG_KEEP_PARAMS
Message-ID: <YP/dEMeULqozIqZd@google.com>
References: <20210719161656.3833943-1-qperret@google.com>
 <20210719161656.3833943-3-qperret@google.com>
 <ad30be79-8fb2-023d-9936-01f7173164e4@arm.com>
 <YP6++lClPCQvTLcK@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YP6++lClPCQvTLcK@google.com>
Precedence: bulk

On Monday 26 Jul 2021 at 14:56:10 (+0100), Quentin Perret wrote:
> On Thursday 22 Jul 2021 at 10:47:33 (+0200), Dietmar Eggemann wrote:
> > (*) This changes the behaviour when setting uclamp values on a DL task.
> > 
> > Before uclamp values could be set but now, because of
> > 
> >   void __getparam_dl(struct task_struct *p, struct sched_attr *attr)
> >     ..
> >     attr->sched_flags = dl_se->flags
> > 
> > SCHED_FLAG_UTIL_CLAMP gets overwritten and  __sched_setscheduler() bails in:
> > 
> >     if (unlikely(policy == p->policy)) {
> >       ...
> >       retval = 0;
> >       goto unlock;
> >     }
> >   change:
> > 
> > I.e. the:
> > 
> >       if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP)
> >         goto change;
> > 
> > can't trigger anymore.
> 
> Bah, as you said it doesn't seem to be a big deal, but clearly that was
> unintentional. Let me try and fix this.

While looking at this I found existing bugs in the area. Fixes are here:

https://lore.kernel.org/lkml/20210727101103.2729607-1-qperret@google.com/

And with the above series applied this patch should behave correctly
now.

Thanks,
Quentin