Re: [PATCH RESEND 1/2] soc: qcom: cmd-db: Mark device as having no PM support

[Pavan Kondeti]

Hi Stephen,
On Fri, Oct 14, 2022 at 05:49:33PM -0700, Stephen Boyd wrote:
> This driver purely exposes information from memory to the kernel. Let's
> mark it as not having any device PM functionality, so that during
> suspend we skip even trying to call a suspend function on this device.
> This clears up suspend logs more than anything else, but it also shaves
> a few cycles off suspend.
> 
> Cc: Konrad Dybcio <konrad.dybcio@xxxxxxxxxxxxxx>
> Signed-off-by: Stephen Boyd <swboyd@xxxxxxxxxxxx>
> ---
> 
> Resend to fix Bjorn's email.
> 
>  drivers/soc/qcom/cmd-db.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/soc/qcom/cmd-db.c b/drivers/soc/qcom/cmd-db.c
> index 629a7188b576..33856abd560c 100644
> --- a/drivers/soc/qcom/cmd-db.c
> +++ b/drivers/soc/qcom/cmd-db.c
> @@ -338,6 +338,8 @@ static int cmd_db_dev_probe(struct platform_device *pdev)
>  
>  	debugfs_create_file("cmd-db", 0400, NULL, NULL, &cmd_db_debugfs_ops);
>  
> +	device_set_pm_not_required(&pdev->dev);
> +
>  	return 0;
>  }
>  
> 

Sorry for asking this question here, thought it would be easy for you to
get the context.

I was recently reading this driver and learned about
device_set_pm_not_required() API. However, it is not clear how this is
working in practice? 

The driver is calling device_set_pm_not_required() in probe(), however
all the checks for device_pm_not_required() happen before probe() is
called. For ex: dpm_sysfs_add() and device_pm_add() both happens before
probe(). I was expecting not to see
/sys/bus/platform/devices/XXXXXX.aop-cmd-db/power directory but it is
still present.

Also, I believe once driver set this flag, the device clean up might not
remove sysfs entries as we have device_pm_not_required() checks active
by that time.

Would it make sense to add code after probe() and take some action if
device_pm_not_required() returns true?

Thanks,
Pavan