linux-kernel - Re: [PATCH] cpuidle: menu: Optimize bucket assignment when next_timer_ns equals KTIME

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: <1c09b717-3f67-47a4-a34b-d4eed1492386@quicinc.com>
Date: Thu, 3 Apr 2025 18:30:13 +0800
From: Zhongqiu Han <quic_zhonhan@...cinc.com>
To: Christian Loehle <christian.loehle@....com>, <rafael@...nel.org>,
        <daniel.lezcano@...aro.org>
CC: <linux-pm@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] cpuidle: menu: Optimize bucket assignment when
 next_timer_ns equals KTIME_MAX

On 4/3/2025 6:06 PM, Christian Loehle wrote:
> On 4/3/25 11:01, Zhongqiu Han wrote:
>> On 4/3/2025 5:34 PM, Christian Loehle wrote:
>>> On 4/3/25 10:28, Zhongqiu Han wrote:
>>>> Directly assign the last bucket value instead of calling which_bucket()
>>>> when next_timer_ns equals KTIME_MAX, the largest possible value that
>>>> always falls into the last bucket. This avoids unnecessary calculations
>>>> and enhances performance.
>>>>
>>>> Signed-off-by: Zhongqiu Han <quic_zhonhan@...cinc.com>
>>>
>>> Reviewed-by: Christian Loehle <christian.loehle@....com>
>>>
>>>> ---
>>>>    drivers/cpuidle/governors/menu.c | 7 ++++++-
>>>>    1 file changed, 6 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c
>>>> index 39aa0aea61c6..8fc7fbed0052 100644
>>>> --- a/drivers/cpuidle/governors/menu.c
>>>> +++ b/drivers/cpuidle/governors/menu.c
>>>> @@ -255,7 +255,12 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev,
>>>>             */
>>>>            data->next_timer_ns = KTIME_MAX;
>>>>            delta_tick = TICK_NSEC / 2;
>>>> -        data->bucket = which_bucket(KTIME_MAX);
>>>> +        /*
>>>> +         * Assign the last bucket value directly instead of calling
>>>> +         * which_bucket(), since KTIME_MAX is the largest possible
>>>> +         * value that always falls into the last bucket.
>>>> +         */
>>>
>>> comment almost seems overkill.
>>>
>>>> +        data->bucket = BUCKETS - 1;
>>>>        }
>>>>          if (unlikely(drv->state_count <= 1 || latency_req == 0) ||
>>>
>> Thanks Christian for the review~
>>
>> Actually I just want to add a comment to indicate that which_bucket()
>> was once called here, in case which_bucket() changes in the future,
>> and however, we stayed with the original approach, leading to the
>> inconsistency.
>>
>> Could you please review the comment below and let me know if it's okay
>> or if I should not add any log? Thanks a lot~
>>
>> /* KTIME_MAX falls into the last bucket, skip which_bucket(). */
>>
>>
>>
>> I will collect review comments before arise patch V2.
> 
> Honestly I'd be fine without a comment, it's pretty obvious that
> everything containing "bucket =" needs to be changed if the bucket
> logic ever changes.
Thanks Christian, let me wait more other aspects comments~

-- 
Thx and BRs,
Zhongqiu Han