2022-11-05 21:46:33

by Samuel Holland

Subject: [PATCH] mfd: axp20x: Do not sleep in the power off handler

Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
function axp20x_power_off() now runs inside a RCU read-side critical
section, so it is not allowed to call msleep(). Use mdelay() instead.

Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
Signed-off-by: Samuel Holland <[email protected]>
---

drivers/mfd/axp20x.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
index 88a212a8168c..880c41fa7021 100644
--- a/drivers/mfd/axp20x.c
+++ b/drivers/mfd/axp20x.c
@@ -842,7 +842,7 @@ static void axp20x_power_off(void)
AXP20X_OFF);

/* Give capacitors etc. time to drain to avoid kernel panic msg. */
- msleep(500);
+ mdelay(500);
}

int axp20x_match_device(struct axp20x_dev *axp20x)
--
2.37.3

2022-11-06 23:14:51

by Dmitry Osipenko

Subject: Re: [PATCH] mfd: axp20x: Do not sleep in the power off handler

On 11/6/22 00:29, Samuel Holland wrote:
> Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
> function axp20x_power_off() now runs inside a RCU read-side critical
> section, so it is not allowed to call msleep(). Use mdelay() instead.
>
> Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
> Signed-off-by: Samuel Holland <[email protected]>
> ---
>
> drivers/mfd/axp20x.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
> index 88a212a8168c..880c41fa7021 100644
> --- a/drivers/mfd/axp20x.c
> +++ b/drivers/mfd/axp20x.c
> @@ -842,7 +842,7 @@ static void axp20x_power_off(void)
> AXP20X_OFF);
>
> /* Give capacitors etc. time to drain to avoid kernel panic msg. */
> - msleep(500);
> + mdelay(500);
> }
>
> int axp20x_match_device(struct axp20x_dev *axp20x)

ARM was doing local_irq_disable() on power-off before
do_kernel_power_off() was introduced, so this should've been incorrect
for a couple years at least. If I'm not missing anything.

--
Best regards,
Dmitry

2022-11-07 09:54:30

by Lee Jones

Subject: Re: [PATCH] mfd: axp20x: Do not sleep in the power off handler

On Sat, 05 Nov 2022, Samuel Holland wrote:

> Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
> function axp20x_power_off() now runs inside a RCU read-side critical
> section, so it is not allowed to call msleep(). Use mdelay() instead.
>
> Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
> Signed-off-by: Samuel Holland <[email protected]>
> ---
>
> drivers/mfd/axp20x.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)

Applied, thanks.

--
Lee Jones [李琼斯]

2022-11-07 10:19:12

by Lee Jones

Subject: Re: [PATCH] mfd: axp20x: Do not sleep in the power off handler

On Mon, 07 Nov 2022, Dmitry Osipenko wrote:

> On 11/6/22 00:29, Samuel Holland wrote:
> > Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
> > function axp20x_power_off() now runs inside a RCU read-side critical
> > section, so it is not allowed to call msleep(). Use mdelay() instead.
> >
> > Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
> > Signed-off-by: Samuel Holland <[email protected]>
> > ---
> >
> > drivers/mfd/axp20x.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
> > index 88a212a8168c..880c41fa7021 100644
> > --- a/drivers/mfd/axp20x.c
> > +++ b/drivers/mfd/axp20x.c
> > @@ -842,7 +842,7 @@ static void axp20x_power_off(void)
> > AXP20X_OFF);
> >
> > /* Give capacitors etc. time to drain to avoid kernel panic msg. */
> > - msleep(500);
> > + mdelay(500);
> > }
> >
> > int axp20x_match_device(struct axp20x_dev *axp20x)
>
> ARM was doing local_irq_disable() on power-off before
> do_kernel_power_off() was introduced, so this should've been incorrect
> for a couple years at least. If I'm not missing anything.

Is that a review? Care to provide a tag?

--
Lee Jones [李琼斯]

2022-11-07 12:05:08

by Dmitry Osipenko

Subject: Re: [PATCH] mfd: axp20x: Do not sleep in the power off handler

On 11/7/22 12:39, Lee Jones wrote:
> On Mon, 07 Nov 2022, Dmitry Osipenko wrote:
>
>> On 11/6/22 00:29, Samuel Holland wrote:
>>> Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
>>> function axp20x_power_off() now runs inside a RCU read-side critical
>>> section, so it is not allowed to call msleep(). Use mdelay() instead.
>>>
>>> Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
>>> Signed-off-by: Samuel Holland <[email protected]>
>>> ---
>>>
>>> drivers/mfd/axp20x.c | 2 +-
>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
>>> index 88a212a8168c..880c41fa7021 100644
>>> --- a/drivers/mfd/axp20x.c
>>> +++ b/drivers/mfd/axp20x.c
>>> @@ -842,7 +842,7 @@ static void axp20x_power_off(void)
>>> AXP20X_OFF);
>>>
>>> /* Give capacitors etc. time to drain to avoid kernel panic msg. */
>>> - msleep(500);
>>> + mdelay(500);
>>> }
>>>
>>> int axp20x_match_device(struct axp20x_dev *axp20x)
>>
>> ARM was doing local_irq_disable() on power-off before
>> do_kernel_power_off() was introduced, so this should've been incorrect
>> for a couple years at least. If I'm not missing anything.
>
> Is that a review? Care to provide a tag?

It's a review and my point was that the "Fixes" tag in this patch
doesn't look correct. But this is minor comment, otherwise patch looks
okay. Hope that the relevant I2C controller driver supports atomic
transfer that are required for power-off to work.

Reviewed-by: Dmitry Osipenko <[email protected]>

--
Best regards,
Dmitry

2022-11-07 16:20:23

by Samuel Holland

Subject: Re: [PATCH] mfd: axp20x: Do not sleep in the power off handler

On 11/6/22 16:40, Dmitry Osipenko wrote:
> On 11/6/22 00:29, Samuel Holland wrote:
>> Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
>> function axp20x_power_off() now runs inside a RCU read-side critical
>> section, so it is not allowed to call msleep(). Use mdelay() instead.
>>
>> Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
>> Signed-off-by: Samuel Holland <[email protected]>
>> ---
>>
>> drivers/mfd/axp20x.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
>> index 88a212a8168c..880c41fa7021 100644
>> --- a/drivers/mfd/axp20x.c
>> +++ b/drivers/mfd/axp20x.c
>> @@ -842,7 +842,7 @@ static void axp20x_power_off(void)
>> AXP20X_OFF);
>>
>> /* Give capacitors etc. time to drain to avoid kernel panic msg. */
>> - msleep(500);
>> + mdelay(500);
>> }
>>
>> int axp20x_match_device(struct axp20x_dev *axp20x)
>
> ARM was doing local_irq_disable() on power-off before
> do_kernel_power_off() was introduced, so this should've been incorrect
> for a couple years at least. If I'm not missing anything.

Yes, you're right. Then the correct tag is

Fixes: 179dc63d06c5 ("mfd: axp20x: Add a 500ms delay at the end of
axp20x_power_off")

Regards,
Samuel

2022-11-14 11:18:28

by Lee Jones

Subject: Re: [PATCH] mfd: axp20x: Do not sleep in the power off handler

On Mon, 07 Nov 2022, Dmitry Osipenko wrote:

> On 11/7/22 12:39, Lee Jones wrote:
> > On Mon, 07 Nov 2022, Dmitry Osipenko wrote:
> >
> >> On 11/6/22 00:29, Samuel Holland wrote:
> >>> Since commit 856c288b0039 ("ARM: Use do_kernel_power_off()"), the
> >>> function axp20x_power_off() now runs inside a RCU read-side critical
> >>> section, so it is not allowed to call msleep(). Use mdelay() instead.
> >>>
> >>> Fixes: 856c288b0039 ("ARM: Use do_kernel_power_off()")
> >>> Signed-off-by: Samuel Holland <[email protected]>
> >>> ---
> >>>
> >>> drivers/mfd/axp20x.c | 2 +-
> >>> 1 file changed, 1 insertion(+), 1 deletion(-)
> >>>
> >>> diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c
> >>> index 88a212a8168c..880c41fa7021 100644
> >>> --- a/drivers/mfd/axp20x.c
> >>> +++ b/drivers/mfd/axp20x.c
> >>> @@ -842,7 +842,7 @@ static void axp20x_power_off(void)
> >>> AXP20X_OFF);
> >>>
> >>> /* Give capacitors etc. time to drain to avoid kernel panic msg. */
> >>> - msleep(500);
> >>> + mdelay(500);
> >>> }
> >>>
> >>> int axp20x_match_device(struct axp20x_dev *axp20x)
> >>
> >> ARM was doing local_irq_disable() on power-off before
> >> do_kernel_power_off() was introduced, so this should've been incorrect
> >> for a couple years at least. If I'm not missing anything.
> >
> > Is that a review? Care to provide a tag?
>
> It's a review and my point was that the "Fixes" tag in this patch
> doesn't look correct. But this is minor comment, otherwise patch looks
> okay. Hope that the relevant I2C controller driver supports atomic
> transfer that are required for power-off to work.
>
> Reviewed-by: Dmitry Osipenko <[email protected]>

Thanks.

--
Lee Jones [李琼斯]