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;
Subject: Re: [PATCH v3 1/2] sched/uclamp: Add a new sysctl to control RT
 default boost value
To:     Qais Yousef <qais.yousef@arm.com>,
        Pavan Kondeti <pkondeti@codeaurora.org>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Jonathan Corbet <corbet@lwn.net>,
        Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
        Luis Chamberlain <mcgrof@kernel.org>,
        Kees Cook <keescook@chromium.org>,
        Iurii Zaikin <yzaikin@google.com>,
        Quentin Perret <qperret@google.com>,
        Valentin Schneider <valentin.schneider@arm.com>,
        Patrick Bellasi <patrick.bellasi@matbug.net>,
        linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-fsdevel@vger.kernel.org
References: <20200428164134.5588-1-qais.yousef@arm.com>
 <20200429113255.GA19464@codeaurora.org>
 <20200429123056.otyedhljlugyf5we@e107158-lin>
From:   Dietmar Eggemann <dietmar.eggemann@arm.com>
Message-ID: <d3916860-6ee8-4d17-55ea-be5cada1302a@arm.com>
Date:   Thu, 30 Apr 2020 20:21:31 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.7.0
MIME-Version: 1.0
In-Reply-To: <20200429123056.otyedhljlugyf5we@e107158-lin>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 29/04/2020 14:30, Qais Yousef wrote:
> Hi Pavan
> 
> On 04/29/20 17:02, Pavan Kondeti wrote:
>> Hi Qais,
>>
>> On Tue, Apr 28, 2020 at 05:41:33PM +0100, Qais Yousef wrote:

[...]

>>> @@ -907,8 +935,15 @@ uclamp_tg_restrict(struct task_struct *p, enum uclamp_id clamp_id)
>>>  static inline struct uclamp_se
>>>  uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
>>>  {
>>> -	struct uclamp_se uc_req = uclamp_tg_restrict(p, clamp_id);
>>> -	struct uclamp_se uc_max = uclamp_default[clamp_id];
>>> +	struct uclamp_se uc_req, uc_max;
>>> +
>>> +	/*
>>> +	 * Sync up any change to sysctl_sched_uclamp_util_min_rt_default value.
>>> +	 */
>>> +	uclamp_sync_util_min_rt_default(p);
>>> +
>>> +	uc_req = uclamp_tg_restrict(p, clamp_id);
>>> +	uc_max = uclamp_default[clamp_id];
>>
>> We are calling uclamp_sync_util_min_rt_default() unnecessarily for
>> clamp_id == UCLAMP_MAX case. Would it be better to have a separate
> 
> It was actually intentional to make sure we update the value ASAP. I didn't
> think it's a lot of overhead. I can further protect with a check to verify
> whether the value has changed if it seems heavy handed.

Users of uclamp_eff_value()->uclamp_eff_get() ((like
rt_task_fits_capacity())) always call both ids.

So calling uclamp_sync_util_min_rt_default() only for UCLAMP_MIN would
make sense. It's overhead in the fast path for rt tasks.

Since changes to sched_util_clamp_min_rt_default will be fairly rare,
you might even want to consider only doing the uclamp_se_set(...,
min_rt_default, ...) in case

  uc_se->value != sysctl_sched_uclamp_util_min_rt_default

[...]