Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 4 Jan 2019 17:07:41 +0000
From:   Quentin Perret <quentin.perret@arm.com>
To:     Chunyan Zhang <chunyan.zhang@unisoc.com>
Cc:     Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Vincent Wang <vincent.wang@unisoc.com>,
        linux-kernel@vger.kernel.org, Chunyan Zhang <zhang.lyra@gmail.com>
Subject: Re: [PATCH] sched/cpufreq: calculate util / max firstly in
 get_next_freq()
Message-ID: <20190104170741.h7bm6yivjf4z3tbd@queper01-lin>
References: <1545634747-22401-1-git-send-email-chunyan.zhang@unisoc.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1545634747-22401-1-git-send-email-chunyan.zhang@unisoc.com>
User-Agent: NeoMutt/20171215
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi,

On Monday 24 Dec 2018 at 14:59:07 (+0800), Chunyan Zhang wrote:
> From: Vincent Wang <vincent.wang@unisoc.com>
> 
> When a task that is in_iowait state is enqueued, cpufreq_update_util() will
> be invoked with SCHED_CPUFREQ_IOWAIT flag. In this case,the value of util
> and max, which are parameters used in get_next_freq(), will be cpu
> frequency, instead of cpu util or capactiy.
> 
> For some 32bit architectures, the size of unsigned long is 32. When
> calculating freq, there may be an overflow error in this expression:
> 
> freq = (freq + (freq >> 2)) * util / max;

This has moved to a different file recently -- see 938e5e4b0d15
("sched/cpufreq: Prepare schedutil for Energy Aware Scheduling") so you
probably want to rebase this patch :-)

> 
> This patch will fix this overflow risk by calulating util / max firstly,
> whether they be frequency or util.
> 
> Signed-off-by: Vincent Wang <vincent.wang@unisoc.com>
> Signed-off-by: Chunyan Zhang <chunyan.zhang@unisoc.com>
> ---
>  kernel/sched/cpufreq_schedutil.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> index 3fffad3..7c372db1 100644
> --- a/kernel/sched/cpufreq_schedutil.c
> +++ b/kernel/sched/cpufreq_schedutil.c
> @@ -166,8 +166,9 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy,
>  	struct cpufreq_policy *policy = sg_policy->policy;
>  	unsigned int freq = arch_scale_freq_invariant() ?
>  				policy->cpuinfo.max_freq : policy->cur;
> +	unsigned int ratio = util * 100 / max;
>  
> -	freq = (freq + (freq >> 2)) * util / max;
> +	freq = (freq + (freq >> 2)) * ratio / 100;

Maybe shift by SCHED_CAPACITY_SCALE instead of doing divisions ? Or you
could also just keep the original structure and use proper u64 types
instead of adding (useless) extra steps for 64 bits.

Thanks,
Quentin