Drop CONFIG_PM and CONFIG_PM_SLEEP ifdeffery while converting dw8250_pm_ops
to use new PM macros.
Signed-off-by: Andy Shevchenko <[email protected]>
---
drivers/tty/serial/8250/8250_dw.c | 8 ++------
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
index f71428c85562..adcc869352b1 100644
--- a/drivers/tty/serial/8250/8250_dw.c
+++ b/drivers/tty/serial/8250/8250_dw.c
@@ -691,7 +691,6 @@ static int dw8250_remove(struct platform_device *pdev)
return 0;
}
-#ifdef CONFIG_PM_SLEEP
static int dw8250_suspend(struct device *dev)
{
struct dw8250_data *data = dev_get_drvdata(dev);
@@ -709,9 +708,7 @@ static int dw8250_resume(struct device *dev)
return 0;
}
-#endif /* CONFIG_PM_SLEEP */
-#ifdef CONFIG_PM
static int dw8250_runtime_suspend(struct device *dev)
{
struct dw8250_data *data = dev_get_drvdata(dev);
@@ -733,11 +730,10 @@ static int dw8250_runtime_resume(struct device *dev)
return 0;
}
-#endif
static const struct dev_pm_ops dw8250_pm_ops = {
- SET_SYSTEM_SLEEP_PM_OPS(dw8250_suspend, dw8250_resume)
- SET_RUNTIME_PM_OPS(dw8250_runtime_suspend, dw8250_runtime_resume, NULL)
+ SYSTEM_SLEEP_PM_OPS(dw8250_suspend, dw8250_resume)
+ RUNTIME_PM_OPS(dw8250_runtime_suspend, dw8250_runtime_resume, NULL)
};
static const struct dw8250_platform_data dw8250_dw_apb = {
--
2.35.1
On Thu, Jun 30, 2022 at 9:42 AM Ilpo Järvinen
<[email protected]> wrote:
>
> On Wed, 29 Jun 2022, Andy Shevchenko wrote:
>
> > Drop CONFIG_PM and CONFIG_PM_SLEEP ifdeffery while converting dw8250_pm_ops
> > to use new PM macros.
> >
> > Signed-off-by: Andy Shevchenko <[email protected]>
>
> Reviewed-by: Ilpo Järvinen <[email protected]>
>
> Not directily related to the patch itself but do you have any idea why
> 1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate old ones")
> didn't wrap RUNTIME_PM_OPS() pointers with pm_ptr()? I'm asking this
> because in SET_RUNTIME_PM_OPS() the callbacks are only created with
> #ifdef CONFIG_PM so I'd have expected RUNTIME_PM_OPS() to maintain that
> behavior but it didn't? Was it just an oversight that should be fixed?
I have had the same question, but I think it might be related to how
PM runtime functions when there is no respective configuration option
set.
+Cc: Rafael.
--
With Best Regards,
Andy Shevchenko
On Wed, 29 Jun 2022, Andy Shevchenko wrote:
> Drop CONFIG_PM and CONFIG_PM_SLEEP ifdeffery while converting dw8250_pm_ops
> to use new PM macros.
>
> Signed-off-by: Andy Shevchenko <[email protected]>
Reviewed-by: Ilpo J?rvinen <[email protected]>
Not directily related to the patch itself but do you have any idea why
1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate old ones")
didn't wrap RUNTIME_PM_OPS() pointers with pm_ptr()? I'm asking this
because in SET_RUNTIME_PM_OPS() the callbacks are only created with
#ifdef CONFIG_PM so I'd have expected RUNTIME_PM_OPS() to maintain that
behavior but it didn't? Was it just an oversight that should be fixed?
--
i.
> ---
> drivers/tty/serial/8250/8250_dw.c | 8 ++------
> 1 file changed, 2 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/tty/serial/8250/8250_dw.c b/drivers/tty/serial/8250/8250_dw.c
> index f71428c85562..adcc869352b1 100644
> --- a/drivers/tty/serial/8250/8250_dw.c
> +++ b/drivers/tty/serial/8250/8250_dw.c
> @@ -691,7 +691,6 @@ static int dw8250_remove(struct platform_device *pdev)
> return 0;
> }
>
> -#ifdef CONFIG_PM_SLEEP
> static int dw8250_suspend(struct device *dev)
> {
> struct dw8250_data *data = dev_get_drvdata(dev);
> @@ -709,9 +708,7 @@ static int dw8250_resume(struct device *dev)
>
> return 0;
> }
> -#endif /* CONFIG_PM_SLEEP */
>
> -#ifdef CONFIG_PM
> static int dw8250_runtime_suspend(struct device *dev)
> {
> struct dw8250_data *data = dev_get_drvdata(dev);
> @@ -733,11 +730,10 @@ static int dw8250_runtime_resume(struct device *dev)
>
> return 0;
> }
> -#endif
>
> static const struct dev_pm_ops dw8250_pm_ops = {
> - SET_SYSTEM_SLEEP_PM_OPS(dw8250_suspend, dw8250_resume)
> - SET_RUNTIME_PM_OPS(dw8250_runtime_suspend, dw8250_runtime_resume, NULL)
> + SYSTEM_SLEEP_PM_OPS(dw8250_suspend, dw8250_resume)
> + RUNTIME_PM_OPS(dw8250_runtime_suspend, dw8250_runtime_resume, NULL)
> };
>
> static const struct dw8250_platform_data dw8250_dw_apb = {
>
Hi Ilpo,
Le jeu., juin 30 2022 at 10:41:40 +0300, Ilpo J?rvinen
<[email protected]> a ?crit :
> On Wed, 29 Jun 2022, Andy Shevchenko wrote:
>
>> Drop CONFIG_PM and CONFIG_PM_SLEEP ifdeffery while converting
>> dw8250_pm_ops
>> to use new PM macros.
>>
>> Signed-off-by: Andy Shevchenko <[email protected]>
>
> Reviewed-by: Ilpo J?rvinen <[email protected]>
>
> Not directily related to the patch itself but do you have any idea why
> 1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate old ones")
> didn't wrap RUNTIME_PM_OPS() pointers with pm_ptr()? I'm asking this
> because in SET_RUNTIME_PM_OPS() the callbacks are only created with
> #ifdef CONFIG_PM so I'd have expected RUNTIME_PM_OPS() to maintain
> that
> behavior but it didn't? Was it just an oversight that should be fixed?
The RUNTIME_PM_OPS() does not wrap pointers with pm_ptr(), because the
pointer to the dev_pm_ops should only ever be used wrapped with
pm_ptr() or pm_sleep_ptr().
Which is not done here.
Andy:
The deference of dw8250_pm_ops should be pm_ptr(&dw8250_pm_ops). If you
only had system suspend/resume functions, you'd use pm_sleep_ptr()
there.
Cheers,
-Paul
> --
> i.
>
>> ---
>> drivers/tty/serial/8250/8250_dw.c | 8 ++------
>> 1 file changed, 2 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/tty/serial/8250/8250_dw.c
>> b/drivers/tty/serial/8250/8250_dw.c
>> index f71428c85562..adcc869352b1 100644
>> --- a/drivers/tty/serial/8250/8250_dw.c
>> +++ b/drivers/tty/serial/8250/8250_dw.c
>> @@ -691,7 +691,6 @@ static int dw8250_remove(struct platform_device
>> *pdev)
>> return 0;
>> }
>>
>> -#ifdef CONFIG_PM_SLEEP
>> static int dw8250_suspend(struct device *dev)
>> {
>> struct dw8250_data *data = dev_get_drvdata(dev);
>> @@ -709,9 +708,7 @@ static int dw8250_resume(struct device *dev)
>>
>> return 0;
>> }
>> -#endif /* CONFIG_PM_SLEEP */
>>
>> -#ifdef CONFIG_PM
>> static int dw8250_runtime_suspend(struct device *dev)
>> {
>> struct dw8250_data *data = dev_get_drvdata(dev);
>> @@ -733,11 +730,10 @@ static int dw8250_runtime_resume(struct
>> device *dev)
>>
>> return 0;
>> }
>> -#endif
>>
>> static const struct dev_pm_ops dw8250_pm_ops = {
>> - SET_SYSTEM_SLEEP_PM_OPS(dw8250_suspend, dw8250_resume)
>> - SET_RUNTIME_PM_OPS(dw8250_runtime_suspend, dw8250_runtime_resume,
>> NULL)
>> + SYSTEM_SLEEP_PM_OPS(dw8250_suspend, dw8250_resume)
>> + RUNTIME_PM_OPS(dw8250_runtime_suspend, dw8250_runtime_resume,
>> NULL)
>> };
>>
>> static const struct dw8250_platform_data dw8250_dw_apb = {
>>
On Thu, 30 Jun 2022, Paul Cercueil wrote:
> Hi Ilpo,
>
> Le jeu., juin 30 2022 at 10:41:40 +0300, Ilpo J?rvinen
> <[email protected]> a ?crit :
> > On Wed, 29 Jun 2022, Andy Shevchenko wrote:
> >
> > > Drop CONFIG_PM and CONFIG_PM_SLEEP ifdeffery while converting
> > > dw8250_pm_ops
> > > to use new PM macros.
> > >
> > > Signed-off-by: Andy Shevchenko <[email protected]>
> >
> > Reviewed-by: Ilpo J?rvinen <[email protected]>
> >
> > Not directily related to the patch itself but do you have any idea why
> > 1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate old ones")
> > didn't wrap RUNTIME_PM_OPS() pointers with pm_ptr()? I'm asking this
> > because in SET_RUNTIME_PM_OPS() the callbacks are only created with
> > #ifdef CONFIG_PM so I'd have expected RUNTIME_PM_OPS() to maintain that
> > behavior but it didn't? Was it just an oversight that should be fixed?
>
> The RUNTIME_PM_OPS() does not wrap pointers with pm_ptr(), because the pointer
> to the dev_pm_ops should only ever be used wrapped with pm_ptr() or
> pm_sleep_ptr().
>
> Which is not done here.
Ok, thanks a lot for the explanation. It's really appreciated.
--
i.
> Andy:
> The deference of dw8250_pm_ops should be pm_ptr(&dw8250_pm_ops). If you only
> had system suspend/resume functions, you'd use pm_sleep_ptr() there.
On Thu, Jun 30, 2022 at 09:44:07AM +0100, Paul Cercueil wrote:
> Le jeu., juin 30 2022 at 10:41:40 +0300, Ilpo J?rvinen
> <[email protected]> a ?crit :
> > On Wed, 29 Jun 2022, Andy Shevchenko wrote:
> >
> > > Drop CONFIG_PM and CONFIG_PM_SLEEP ifdeffery while converting
> > > dw8250_pm_ops
> > > to use new PM macros.
> > >
> > > Signed-off-by: Andy Shevchenko <[email protected]>
> >
> > Reviewed-by: Ilpo J?rvinen <[email protected]>
I will drop this for v2.
> > Not directily related to the patch itself but do you have any idea why
> > 1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate old ones")
> > didn't wrap RUNTIME_PM_OPS() pointers with pm_ptr()? I'm asking this
> > because in SET_RUNTIME_PM_OPS() the callbacks are only created with
> > #ifdef CONFIG_PM so I'd have expected RUNTIME_PM_OPS() to maintain that
> > behavior but it didn't? Was it just an oversight that should be fixed?
>
> The RUNTIME_PM_OPS() does not wrap pointers with pm_ptr(), because the
> pointer to the dev_pm_ops should only ever be used wrapped with pm_ptr() or
> pm_sleep_ptr().
>
> Which is not done here.
>
> Andy:
> The deference of dw8250_pm_ops should be pm_ptr(&dw8250_pm_ops). If you only
> had system suspend/resume functions, you'd use pm_sleep_ptr() there.
Right, it's a shame how I forgot that while telling everybody to use them.
Thanks, Paul!
--
With Best Regards,
Andy Shevchenko