Re: [PATCH 03/10] watchdog: rzg2l_wdt: Check return status of pm_runtime_put()

[Guenter Roeck]

On 1/22/24 23:02, claudiu beznea wrote:
> On 22.01.2024 19:31, Guenter Roeck wrote:
> > On 1/22/24 03:11, Claudiu wrote:
> > > From: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
> > >
> > > pm_runtime_put() may return an error code. Check its return status.
> > >
> > > Fixes: 2cbc5cd0b55f ("watchdog: Add Watchdog Timer driver for RZ/G2L")
> > > Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@xxxxxxxxxxxxxx>
> > > ---
> > >    drivers/watchdog/rzg2l_wdt.c | 6 +++++-
> > >    1 file changed, 5 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/watchdog/rzg2l_wdt.c b/drivers/watchdog/rzg2l_wdt.c
> > > index 4ab9e7c5e771..0554965027cd 100644
> > > --- a/drivers/watchdog/rzg2l_wdt.c
> > > +++ b/drivers/watchdog/rzg2l_wdt.c
> > > @@ -144,9 +144,13 @@ static int rzg2l_wdt_start(struct watchdog_device
> > > *wdev)
> > >    static int rzg2l_wdt_stop(struct watchdog_device *wdev)
> > >    {
> > >        struct rzg2l_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > > +    int ret;
> > >          rzg2l_wdt_reset(priv);
> > > -    pm_runtime_put(wdev->parent);
> > > +
> > > +    ret = pm_runtime_put(wdev->parent);
> > > +    if (ret < 0)
> > > +        return ret;
> > >          return 0;
> > >    }
> >
> > A simple
> >      return pm_runtime_put();
> > might do.
>
> pm_runtime_put() may return 1 if the device is already suspended though
> this call trace:
>
> pm_runtime_put() ->
>     __pm_runtime_idle() ->
>         rpm_idle() ->
>             rpm_suspend() ->
>                 rpm_check_suspend_allowed() [1]
>
> That return value is not considered error thus I wanted to consider it
> here, too.
Good point.

> [1]
> https://elixir.bootlin.com/linux/latest/source/drivers/base/power/runtime.c#L278
>
> > However, one question: Given that pm_runtime_put() returns -ENOSYS if
> > CONFIG_PM is disabled, that means the driver will depend on CONFIG_PM=y.
>
> Indeed, the driver depends on CONFIG_PM=y for proper working. It is for
> devices selecting ARCH_RZG2L and RZ/V2M (ARM64 based uarch) which select
> CONFIG_PM=y:
> https://elixir.bootlin.com/linux/latest/source/drivers/soc/renesas/Kconfig#L45
>
> The driver is written with CONFIG_PM=y dependency in mind (e.g. the clocks
> are enabled though runtime PM APIs).
>
> > Assuming this is intentional, would it make sense to explicitly declare
> > that dependency in Kconfig ? It doesn't seem to make any sense to build
> > the driver if it won't work anyway.
>
> The dependency exists there for ARCH_RZG2L and RZ/V2M devices but not
> directly and it is not strict (in the sense that we allow to build the
> driver w/o CONFIG_PM (I think this is good to check build on different
> configurations, the COMPILE_TEST is there anyway in [1]) ). E.g.:
>
> RENESAS_RZG2LWDT depends on ARCH_RENESAS [1]
> ARCH_RENESAS is the ARMv8 uarch flag [2]
> SOC_RENESAS is set if ARCH_RENESAS [3]
> ARCH_RZG2L is visible only if SOC_RENESAS [4]
> ARCH_RZG2L selects PM [5]
> RZ/V2M selects PM [6]
>
> Please let me know what do you think about it?
If the driver indeed depends on CONFIG_PM, that should be made explicit.

Guenter

> Thank you,
> Claudiu Beznea
>
>
> [1]
> https://elixir.bootlin.com/linux/latest/source/drivers/watchdog/Kconfig#L913
> [2]
> https://elixir.bootlin.com/linux/latest/source/arch/arm64/Kconfig.platforms#L273
> [3]
> https://elixir.bootlin.com/linux/latest/source/drivers/soc/renesas/Kconfig#L2
> [4]
> https://elixir.bootlin.com/linux/latest/source/drivers/soc/renesas/Kconfig#L9
> [5]
> https://elixir.bootlin.com/linux/latest/source/drivers/soc/renesas/Kconfig#L45
> [6]
> https://elixir.bootlin.com/linux/latest/source/drivers/soc/renesas/Kconfig#L328
>
>
> > Thanks,
> > Guenter