linux-kernel - Re: [PATCH 2/3] block: add support for device frequency PM QoS tuning

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <191ca54f-0faa-4615-967a-7b4c86d59e0e@acm.org>
Date: Mon, 15 Sep 2025 14:26:34 -0700
From: Bart Van Assche <bvanassche@....org>
To: Wang Jianzheng <wangjianzheng@...o.com>, Jens Axboe <axboe@...nel.dk>,
 Alim Akhtar <alim.akhtar@...sung.com>, Avri Altman <avri.altman@....com>,
 "James E.J. Bottomley" <James.Bottomley@...senPartnership.com>,
 "Martin K. Petersen" <martin.petersen@...cle.com>,
 "Rafael J. Wysocki" <rafael@...nel.org>, Peter Wang
 <peter.wang@...iatek.com>, Bean Huo <beanhuo@...ron.com>,
 "Bao D. Nguyen" <quic_nguyenb@...cinc.com>,
 Adrian Hunter <adrian.hunter@...el.com>, linux-block@...r.kernel.org,
 linux-kernel@...r.kernel.org, linux-scsi@...r.kernel.org,
 linux-pm@...r.kernel.org
Subject: Re: [PATCH 2/3] block: add support for device frequency PM QoS tuning

On 9/14/25 4:45 AM, Wang Jianzheng wrote:
> +#ifdef CONFIG_PM
> +static void blk_mq_dev_frequency_work(struct work_struct *work)
> +{
> +	struct request_queue *q =
> +			container_of(work, struct request_queue, dev_freq_work.work);
> +	unsigned long timeout;
> +	struct dev_pm_qos_request *qos = q->dev_freq_qos;
> +
> +	timeout = msecs_to_jiffies(q->disk->dev_freq_timeout);
> +	if (!q || IS_ERR_OR_NULL(q->dev) || IS_ERR_OR_NULL(qos))
> +		return;
> +
> +	if (q->pm_qos_status == PM_QOS_ACTIVE) {
> +		q->pm_qos_status = PM_QOS_FREQ_SET;
> +		dev_pm_qos_add_request(q->dev, qos, DEV_PM_QOS_MIN_FREQUENCY,
> +				       FREQ_QOS_MAX_DEFAULT_VALUE);
> +	} else {
> +		if (time_after(jiffies, READ_ONCE(q->last_active) + timeout))
> +			q->pm_qos_status = PM_QOS_FREQ_REMOV;
> +	}
> +
> +	if (q->pm_qos_status == PM_QOS_FREQ_REMOV) {
> +		dev_pm_qos_remove_request(qos);
> +		q->pm_qos_status = PM_QOS_ACTIVE;
> +	} else {
> +		schedule_delayed_work(&q->dev_freq_work,
> +				      q->last_active + timeout - jiffies);
> +	}
> +}

The above code is similar in nature to the activity detection by the
run-time power management (RPM) code. Why a new timer mechanism instead
of adding more code in the UFS driver RPM callbacks?

> @@ -3161,6 +3211,8 @@ void blk_mq_submit_bio(struct bio *bio)
>   		goto queue_exit;
>   	}
>   
> +	blk_pm_qos_dev_freq_update(q, bio);

Good luck with adding power-management code in the block layer hot path
... I'm not sure anyone will be enthusiast seeing code being added in
blk_mq_submit_bio().

Thanks,

Bart.