Re: [PATCH v3] scsi: ufs: core: skip UFS clkscale if host asynchronous scan in progress

[Ziqi Chen <quic_ziqichen@quicinc.com>]

On 5/14/2025 5:05 PM, Peter Wang (王信友) wrote:
> On Wed, 2025-05-14 at 15:25 +0800, Ziqi Chen wrote:
>>
>> Hi Bart,
>>
>> I tried the scan_mutex, from debugging logs, it seems okay for now.
>> I will provide to our internal test team for stability test.
>> And I will try to collect the extra time spent on clock scaling
>> path with applying scan_mutex.
>> If everything is fine, I will update a new version.
>>
>> BRs,
>> Ziqi
>>
>>
>>
> 
> Hi Ziqi,
> 
> Could we enable devfreq when checking if hba->luns_avail equals 1
> in ufshcd_device_configure?
> I think we can use flow to ensure correctness; it doesn't
> necessarily need to be protected by a mutex.
> 
> Thanks
> Peter

Hi Peter,

Thanks for your suggestion, and sorry for the late response—I was out of
office last week.

It looks like checking hba->luns_avail == 1 in ufshcd_device_configure
before enabling devfreq can ensure correctness. But I think the
function name 'ufshcd_device_configure' suggests that enabling devfreq
inside it is not reasonable. That’s another same reason why we moved
ufshcd_devfreq_init() out of ufshcd_add_lus() in this patch.

BRs,
Ziqi

> 
>