Re: rtc: ds1307: call the platform's logic for handle IRQs.

[Enric Balletbo i Serra <enric.balletbo@collabora.com>]

Hi Heiner,

On 24/08/17 22:53, Heiner Kallweit wrote:
> Am 24.08.2017 um 22:44 schrieb Heiner Kallweit:
>> Am 24.08.2017 um 22:13 schrieb Enric Balletbo i Serra:
>>> Hi Heiner
>>>
>>> On 24/08/17 21:16, Heiner Kallweit wrote:
>>>> Am 24.08.2017 um 20:46 schrieb Heiner Kallweit:
>>>>> Hi Enric,
>>>>>
>>>>> I just saw your submitted patch on patchwork. As I haven't been subscribed
>>>>> to linux-rtc list yet, I can't reply to the original mail.
>>>>>
>>>>> Few remarks:
>>>>> I think the same can be achieved easier (apart from the fact that member
>>>>> irq was just removed from struct ds1307).
>>>>> The curent call to device_set_wakeup_capable has to be replaced with
>>>>> device_init_wakeup, in addition we have to call dev_pm_set_wake_irq to
>>>>> register the interrupt with the Linux wakeup core. Then the core takes
>>>>> care of everything.
>>>>>
>>>
>>> Oh I just saw your patches, thanks to advice me about them.
>>>
>>>> After further checking not even this may be necessary.
>>>> If flag I2C_CLIENT_WAKE is set for the i2c client then i2c_device_probe()
>>>> enables the device as wakeup device and registers the interrupt with the
>>>> wakeup core.
>>>> If the i2c client is defined via DT, then of_i2c_register_device() sets
>>>> this flag based on the "wakeup-source" property.
>>>> Is your device configured via DT? If yes, did you check whether wakeup
>>>> works w/o your patch with just setting property wakeup-source ?
>>>>
>>>
>>> Yes, device is configured via DT and without my patch, just setting property
>>> wakeup-source didn't work, the reason is that code:
>>>
>>> http://elixir.free-electrons.com/linux/v4.13-rc6/source/drivers/rtc/rtc-ds1307.c#L1382
>>>
>> This shouldn't be the reason because registering the interrupt with the wakeup core
>> happens earlier in i2c_device_probe().
>> Can you check whether dev_pm_set_wake_irq() is actually called with the right
>> irq number in i2c_device_probe()?
>>

You have reason again, I did something wrong on my tests and read the code in a
wrong way, to sum up, adding both, the interrupt pin and the wakeup-source
propriety, works as expected. So forget this patch. Sorry for the noise and
thanks for replying and look at this.

> See also description of commit 4990d4fe327b9d9a7a
> "PM / Wakeirq: Add automated device wake IRQ handling" from 2015.
> It explains why the code you want to add shouldn't be needed any longer.
> 
>> Would be interesting to see the DT config of your rtc.


        rtc0: rtc@68 {
                compatible = "dallas,ds1339";
                pinctrl-names = "default";
                pinctrl-0 = <&rtc0_irq_pins>;
                interrupt-parent = <&gpio0>;
                interrupts = <23 IRQ_TYPE_EDGE_FALLING>; /* gpio 23 */
                wakeup-source;
                trickle-resistor-ohms = <2000>;
                reg = <0x68>;
        };

>>
>>> I didn't test on top of your patches so let me rebase/rethink my patch on top of
>>> yours.
>>>
>>>>
>>>>> See also rtc-ds1343, although I think the calls to enable/disable_irq_wake
>>>>> are not needed there because the core takes care of this already
>>>>> (enable_irq_wake is called from dev_pm_arm_wake_irq).
>>>>>
>>>>> Rgds, Heiner
>>>>>
>>>>
>>>
>>> Thanks,
>>>  Enric
>>>
>>
>