[PATCH] rtc: stm32: Handle single EXTI IRQ as wake up source

[Alexandre TORGUE]

Hi Marek

On 5/30/23 21:54, Marek Vasut wrote:
> On 5/30/23 17:18, Amelie Delaunay wrote:
>> On 5/30/23 16:13, Marek Vasut wrote:
>>> On 5/30/23 16:02, Amelie Delaunay wrote:
>>>> Hi Marek,
>>>
>>> Hello Amelie,
>>>
>>>> On 5/18/23 02:33, Marek Vasut wrote:
>>>>> The arch/arm/boot/dts/stm32mp151.dtsi specifies one interrupt for the
>>>>> RTC node, while the expectation of the RTC driver is two interrupts on
>>>>> STM32MP15xx SoC, one connected to GIC interrupt controller and another
>>>>> one to EXTI. Per STM32MP15xx reference manual, the two interrupts 
>>>>> serve
>>>>> the same purpose, except the EXTI one can also wake the system up. The
>>>>> EXTI driver already internally handles this GIC and EXTI duality and
>>>>> maps the EXTI interrupt onto GIC during runtime, and only uses the 
>>>>> EXTI
>>>>> for wake up functionality.
>>>>>
>>>>> Therefore, fix the driver such that if on STM32MP15xx there is only 
>>>>> one
>>>>> interrupt specified in the DT, use that interrupt as EXTI interrupt 
>>>>> and
>>>>> set it as wake up source.
>>>>>
>>>>> This fixes the following warning in the kernel log on STM32MP15xx:
>>>>> "
>>>>> stm32_rtc 5c004000.rtc: error -ENXIO: IRQ index 1 not found
>>>>> stm32_rtc 5c004000.rtc: alarm can't wake up the system: -6
>>>>> "
>>>>>
>>>>> This also fixes the system wake up via built-in RTC using e.g.:
>>>>> $ rtcwake -s 5 -m mem
>>>>>
>>>>> Fixes: b72252b6580c ("rtc: stm32: add stm32mp1 rtc support")
>>>>> Signed-off-by: Marek Vasut <marex at denx.de>
>>>>> ---
>>>>> Cc: Alessandro Zummo <a.zummo at towertech.it>
>>>>> Cc: Alexandre Belloni <alexandre.belloni at bootlin.com>
>>>>> Cc: Alexandre Torgue <alexandre.torgue at foss.st.com>
>>>>> Cc: Amelie DELAUNAY <amelie.delaunay at foss.st.com>
>>>>> Cc: Maxime Coquelin <mcoquelin.stm32 at gmail.com>
>>>>> Cc: linux-arm-kernel at lists.infradead.org
>>>>> Cc: linux-rtc at vger.kernel.org
>>>>> Cc: linux-stm32 at st-md-mailman.stormreply.com
>>>>> ---
>>>>>   drivers/rtc/rtc-stm32.c | 9 +++++----
>>>>>   1 file changed, 5 insertions(+), 4 deletions(-)
>>>>>
>>>>> diff --git a/drivers/rtc/rtc-stm32.c b/drivers/rtc/rtc-stm32.c
>>>>> index 229cb2847cc48..72994b9f95319 100644
>>>>> --- a/drivers/rtc/rtc-stm32.c
>>>>> +++ b/drivers/rtc/rtc-stm32.c
>>>>> @@ -780,14 +780,15 @@ static int stm32_rtc_probe(struct 
>>>>> platform_device *pdev)
>>>>>       ret = device_init_wakeup(&pdev->dev, true);
>>>>>       if (rtc->data->has_wakeirq) {
>>>>> -        rtc->wakeirq_alarm = platform_get_irq(pdev, 1);
>>>>> +        rtc->wakeirq_alarm = platform_get_irq_optional(pdev, 1);
>>>>>           if (rtc->wakeirq_alarm > 0) {
>>>>>               ret = dev_pm_set_dedicated_wake_irq(&pdev->dev,
>>>>>                                   rtc->wakeirq_alarm);
>>>>> -        } else {
>>>>> +        } else if (rtc->wakeirq_alarm == -EPROBE_DEFER) {
>>>>>               ret = rtc->wakeirq_alarm;
>>>>> -            if (rtc->wakeirq_alarm == -EPROBE_DEFER)
>>>>> -                goto err;
>>>>> +            goto err;
>>>>> +        } else {
>>>>> +            ret = dev_pm_set_wake_irq(&pdev->dev, rtc->irq_alarm);
>>>>>           }
>>>>>       }
>>>>>       if (ret)
>>>>
>>>> In our downstream [1], dedicated wakeup irq management is dropped: 
>>>> it is neither described in st,stm32-rtc bindings nor used in 
>>>> STM32Fx, STM32Hx, STM32MP1x device trees.
>>>> The pointed patch is going to be upstreamed.
>>>>
>>>> [1] 
>>>> https://github.com/STMicroelectronics/linux/commit/5a45e1f100d59c33b6b50fe98c0f62862bd6f3d2
>>>
>>> Won't that break compatibility with DTs which do use two interrupts 
>>> entries ?
>>
>> I didn't find any DTs using STM32 RTC with two interrupts. Did I miss 
>> something?
> 
> I am not aware of any either (I also did check a couple of repositories 
> to be sure) . However, the DT is an ABI , so there might be users we do 
> not know about, who might be unable to update their DTs , and who would 
> be broken by dropping the support for two interrupts.

Currently if people use 2 interrupts in their DT with an up to date 
kernel I don't think it works fine. At the beginning of the MP1 story, 2 
interrupts were needed to wakeup system from LPSTOP: one for GIC and the 
other one for EXTI. But since maybe 2 years, EXTI and GIC interrupts are 
directly linked inside the EXTI driver. So now, devices only need to 
take one interrupt. With this implementation if one device uses 2 
interrupts in their DT then the GIC interrupt will be mapped two times. 
So I think that current implementation in RTC driver is broken and it 
should be aligned with "new" EXTI implementation. Note also that the use 
of 2 interrupts has never been documented in dt-bindings documentation.

Above words are for STM32 MPU products For STM32 MCU products RTC is 
only mapped to the EXTI (not to the NVIC) so no needs to handle two 
interrupts.

Alex


> On the other hand, I only use one interrupt anyway, so I am not affected 
> and I don't have a strong preference regarding which patch gets in.