linux-kernel - Re: [PATCH 07/11] PM / devfreq: check get_dev

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] [thread-next>] [day] [month] [year] [list]
Message-ID: <d92b7660-a622-06f3-6e1f-03c99fdf0a55@gmail.com>
Date:   Sat, 13 Mar 2021 01:09:51 +0900
From:   Chanwoo Choi <cwchoi00@...il.com>
To:     Dong Aisheng <dongas86@...il.com>,
        Chanwoo Choi <cw00.choi@...sung.com>
Cc:     Dong Aisheng <aisheng.dong@....com>,
        Linux PM <linux-pm@...r.kernel.org>,
        "moderated list:ARM/FREESCALE IMX / MXC ARM ARCHITECTURE" 
        <linux-arm-kernel@...ts.infradead.org>,
        Sascha Hauer <kernel@...gutronix.de>,
        Shawn Guo <shawnguo@...nel.org>,
        dl-linux-imx <linux-imx@....com>,
        open list <linux-kernel@...r.kernel.org>,
        myungjoo.ham@...sung.com, kyungmin.park@...sung.com,
        Abel Vesa <abel.vesa@....com>
Subject: Re: [PATCH 07/11] PM / devfreq: check get_dev_status before start
 monitor

On 21. 3. 12. 오후 7:57, Dong Aisheng wrote:
> On Thu, Mar 11, 2021 at 2:54 PM Chanwoo Choi <cw00.choi@...sung.com> wrote:
>>
>> On 3/10/21 1:56 PM, Dong Aisheng wrote:
>>> On Wed, Mar 10, 2021 at 11:08 AM Chanwoo Choi <cw00.choi@...sung.com> wrote:
>>>>
>>>> On 3/10/21 11:56 AM, Dong Aisheng wrote:
>>>>> On Wed, Mar 10, 2021 at 12:12 AM Chanwoo Choi <cwchoi00@...il.com> wrote:
>>>>>>
>>>>>> On 21. 3. 10. 오전 12:58, Chanwoo Choi wrote:
>>>>>>> On 21. 3. 9. 오후 9:58, Dong Aisheng wrote:
>>>>>>>> The devfreq monitor depends on the device to provide load information
>>>>>>>> by .get_dev_status() to calculate the next target freq.
>>>>>>>>
>>>>>>>> And this will cause changing governor to simple ondemand fail
>>>>>>>> if device can't support.
>>>>>>>>
>>>>>>>> Signed-off-by: Dong Aisheng <aisheng.dong@....com>
>>>>>>>> ---
>>>>>>>>    drivers/devfreq/devfreq.c                 | 10 +++++++---
>>>>>>>>    drivers/devfreq/governor.h                |  2 +-
>>>>>>>>    drivers/devfreq/governor_simpleondemand.c |  3 +--
>>>>>>>>    3 files changed, 9 insertions(+), 6 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
>>>>>>>> index 7231fe6862a2..d1787b6c7d7c 100644
>>>>>>>> --- a/drivers/devfreq/devfreq.c
>>>>>>>> +++ b/drivers/devfreq/devfreq.c
>>>>>>>> @@ -482,10 +482,13 @@ static void devfreq_monitor(struct work_struct
>>>>>>>> *work)
>>>>>>>>     * to be called from governor in response to DEVFREQ_GOV_START
>>>>>>>>     * event when device is added to devfreq framework.
>>>>>>>>     */
>>>>>>>> -void devfreq_monitor_start(struct devfreq *devfreq)
>>>>>>>> +int devfreq_monitor_start(struct devfreq *devfreq)
>>>>>>>>    {
>>>>>>>>        if (IS_SUPPORTED_FLAG(devfreq->governor->flags, IRQ_DRIVEN))
>>>>>>>> -        return;
>>>>>>>> +        return 0;
>>>>>>>> +
>>>>>>>> +    if (!devfreq->profile->get_dev_status)
>>>>>>>> +        return -EINVAL;
>>>>>>
>>>>>> Again, I think that get_dev_status is not used for all governors.
>>>>>> So that it cause the governor start fail. Don't check whether
>>>>>> .get_dev_status is NULL or not.
>>>>>>
>>>>>
>>>>> I'm not quite understand your point.
>>>>> it is used by governor_simpleondemand.c and tegra_devfreq_governor.
>>>>> get_target_freq -> devfreq_update_stats -> get_dev_status
>>>>
>>>> The devfreq can add the new governor by anyone.
>>>> So these functions like devfreq_monitor_* have to support
>>>> the governors and also must support the governor to be added
>>>> in the future.
>>>
>>> Yes, but devfreq_monitor_* is only used by polling mode, right?
>>> The governor using it has to implement get_dev_status unless
>>> there's an exception in the future.
>>>
>>> Currently this patch wants to address the issue that user can switch
>>> to ondemand governor (polling mode) by sysfs even devices does
>>> not support it (no get_dev_status implemented).
>>
>> As I commented, I'll fix this issue. If devfreq driver doesn't implement
>> the .get_dev_status, don't show it via available_governors. I think that
>> it is fundamental solution to fix this issue.
> 
> Sounds good
> 
>> So on this version,
>> don't add the this conditional statement on this function
>>
> 
> Almost all this patch did is adding a checking for get_dev_status.
> So do you mean drop this patch?
> I wonder it's still a necessary checking to explicitly tell devfreq monitor
> users that get_dev_status is needed during governor startup.

I think that the it is enough to check .get_dev_status in
devfreq_update_stats. We have to check it on where it is used.

> 
>> And on next version, please use the capital letter for first character
>> on patch title as following:
>>
>> - PM / devfreq: Check get_dev_status before start monitor
>>
> 
> Okay to me.
> Thanks for the suggestion.
> 
> Regards
> Aisheng
> 
>>>
>>> Regards
>>> Aisheng
>>>
>>>>
>>>>>
>>>>> Without checking, device can switch to ondemand governor if it does not support.
>>>>>
>>>>> Am i missed something?
>>>>>
>>>>> Regards
>>>>> Aisheng
>>>>>
>>>>>>>>        switch (devfreq->profile->timer) {
>>>>>>>>        case DEVFREQ_TIMER_DEFERRABLE:
>>>>>>>> @@ -495,12 +498,13 @@ void devfreq_monitor_start(struct devfreq *devfreq)
>>>>>>>>            INIT_DELAYED_WORK(&devfreq->work, devfreq_monitor);
>>>>>>>>            break;
>>>>>>>>        default:
>>>>>>>> -        return;
>>>>>>>> +        return -EINVAL;
>>>>>>>>        }
>>>>>>>>        if (devfreq->profile->polling_ms)
>>>>>>>>            queue_delayed_work(devfreq_wq, &devfreq->work,
>>>>>>>>                msecs_to_jiffies(devfreq->profile->polling_ms));
>>>>>>>> +    return 0;
>>>>>>>>    }
>>>>>>>>    EXPORT_SYMBOL(devfreq_monitor_start);
>>>>>>>> diff --git a/drivers/devfreq/governor.h b/drivers/devfreq/governor.h
>>>>>>>> index 5cee3f64fe2b..31af6d072a10 100644
>>>>>>>> --- a/drivers/devfreq/governor.h
>>>>>>>> +++ b/drivers/devfreq/governor.h
>>>>>>>> @@ -75,7 +75,7 @@ struct devfreq_governor {
>>>>>>>>                    unsigned int event, void *data);
>>>>>>>>    };
>>>>>>>> -void devfreq_monitor_start(struct devfreq *devfreq);
>>>>>>>> +int devfreq_monitor_start(struct devfreq *devfreq);
>>>>>>>>    void devfreq_monitor_stop(struct devfreq *devfreq);
>>>>>>>>    void devfreq_monitor_suspend(struct devfreq *devfreq);
>>>>>>>>    void devfreq_monitor_resume(struct devfreq *devfreq);
>>>>>>>> diff --git a/drivers/devfreq/governor_simpleondemand.c
>>>>>>>> b/drivers/devfreq/governor_simpleondemand.c
>>>>>>>> index d57b82a2b570..ea287b57cbf3 100644
>>>>>>>> --- a/drivers/devfreq/governor_simpleondemand.c
>>>>>>>> +++ b/drivers/devfreq/governor_simpleondemand.c
>>>>>>>> @@ -89,8 +89,7 @@ static int devfreq_simple_ondemand_handler(struct
>>>>>>>> devfreq *devfreq,
>>>>>>>>    {
>>>>>>>>        switch (event) {
>>>>>>>>        case DEVFREQ_GOV_START:
>>>>>>>> -        devfreq_monitor_start(devfreq);
>>>>>>>> -        break;
>>>>>>>> +        return devfreq_monitor_start(devfreq);
>>>>>>>>        case DEVFREQ_GOV_STOP:
>>>>>>>>            devfreq_monitor_stop(devfreq);
>>>>>>>>
>>>>>>>
>>>>>>> Need to handle the all points of devfreq_monitor_start() usage.
>>>>>>> please check the tegra30-devfreq.c for this update.
>>>>>>>
>>>>>>> $ grep -rn "devfreq_monitor_start" drivers/
>>>>>>> drivers/devfreq/governor_simpleondemand.c:92:
>>>>>>> devfreq_monitor_start(devfreq);
>>>>>>> drivers/devfreq/tegra30-devfreq.c:744:
>>>>>>> devfreq_monitor_start(devfreq);
>>>>>>> ......
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Best Regards,
>>>>>> Samsung Electronics
>>>>>> Chanwoo Choi
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Best Regards,
>>>> Chanwoo Choi
>>>> Samsung Electronics
>>>
>>>
>>
>>
>> --
>> Best Regards,
>> Chanwoo Choi
>> Samsung Electronics


-- 
Best Regards,
Samsung Electronics
Chanwoo Choi