Re: [PATCH v3 4/4] Input: Don't send fake button presses to wake system

[Hans de Goede <hansg@kernel.org>]

Hi,

On 26-Jun-25 20:48, Dmitry Torokhov wrote:
> On Thu, Jun 26, 2025 at 01:20:54PM -0500, Mario Limonciello wrote:
>> On 6/26/2025 1:07 PM, Dmitry Torokhov wrote:
>>> On Thu, Jun 26, 2025 at 12:53:02PM -0500, Mario Limonciello wrote:
>>>>
>>>>
>>>> On 6/26/25 12:44 PM, Dmitry Torokhov wrote:
>>>>> Hi Mario,
>>>>>
>>>>> On Thu, Jun 26, 2025 at 06:33:08AM -0500, Mario Limonciello wrote:
>>>>>>
>>>>>>
>>>>>> On 6/26/25 3:35 AM, Hans de Goede wrote:
>>>>>>> Hi Mario,
>>>>>>>
>>>>>>> On 25-Jun-25 23:58, Mario Limonciello wrote:
>>>>>>>> From: Mario Limonciello <mario.limonciello@amd.com>
>>>>>>>>
>>>>>>>> Sending an input event to wake a system does wake it, but userspace picks
>>>>>>>> up the keypress and processes it.  This isn't the intended behavior as it
>>>>>>>> causes a suspended system to wake up and then potentially turn off if
>>>>>>>> userspace is configured to turn off on power button presses.
>>>>>>>>
>>>>>>>> Instead send a PM wakeup event for the PM core to handle waking the system.
>>>>>>>>
>>>>>>>> Cc: Hans de Goede <hansg@kernel.org>
>>>>>>>> Fixes: 0f107573da417 ("Input: gpio_keys - handle the missing key press event in resume phase")
>>>>>>>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
>>>>>>>> ---
>>>>>>>>     drivers/input/keyboard/gpio_keys.c | 7 +------
>>>>>>>>     1 file changed, 1 insertion(+), 6 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/drivers/input/keyboard/gpio_keys.c b/drivers/input/keyboard/gpio_keys.c
>>>>>>>> index 773aa5294d269..4c6876b099c43 100644
>>>>>>>> --- a/drivers/input/keyboard/gpio_keys.c
>>>>>>>> +++ b/drivers/input/keyboard/gpio_keys.c
>>>>>>>> @@ -420,12 +420,7 @@ static irqreturn_t gpio_keys_gpio_isr(int irq, void *dev_id)
>>>>>>>>     		pm_stay_awake(bdata->input->dev.parent);
>>>>>>>>     		if (bdata->suspended  &&
>>>>>>>>     		    (button->type == 0 || button->type == EV_KEY)) {
>>>>>>>> -			/*
>>>>>>>> -			 * Simulate wakeup key press in case the key has
>>>>>>>> -			 * already released by the time we got interrupt
>>>>>>>> -			 * handler to run.
>>>>>>>> -			 */
>>>>>>>> -			input_report_key(bdata->input, button->code, 1);
>>>>>>>> +			pm_wakeup_event(bdata->input->dev.parent, 0);
>>>>>
>>>>> There is already pm_stay_awake() above.
>>>>
>>>> But that doesn't help with the fact that userspace gets KEY_POWER from this
>>>> and reacts to it.
>>>>
>>>>>
>>>>>>>>     		}
>>>>>>>>     	}
>>>>>>>
>>>>>>> Hmm, we have the same problem on many Bay Trail / Cherry Trail
>>>>>>> windows 8 / win10 tablets, so  this has been discussed before and e.g.
>>>>>>> Android userspace actually needs the button-press (evdev) event to not
>>>>>>> immediately go back to sleep, so a similar patch has been nacked in
>>>>>>> the past.
>>>>>>>
>>>>>>> At least for GNOME this has been fixed in userspace by ignoring
>>>>>>> power-button events the first few seconds after a resume from suspend.
>>>>>>>
>>>>>>
>>>>>> The default behavior for logind is:
>>>>>>
>>>>>> HandlePowerKey=poweroff
>>>>>>
>>>>>> Can you share more about what version of GNOME has a workaround?
>>>>>> This was actually GNOME (on Ubuntu 24.04) that I found this issue.
>>>>>>
>>>>>> Nonetheless if this is dependent on an Android userspace problem could we
>>>>>> perhaps conditionalize it on CONFIG_ANDROID_BINDER_DEVICES?
>>>>>
>>>>> No it is not only Android, other userspace may want to distinguish
>>>>> between normal and "dark" resume based on keyboard or other user
>>>>> activity.
>>>>>
>>>>> Thanks.
>>>>>
>>>> In this specific case does the key passed up to satisfy this userspace
>>>> requirement and keep it awake need to specifically be a fabricated
>>>> KEY_POWER?
>>>>
>>>> Or could we find a key that doesn't require some userspace to ignore
>>>> KEY_POWER?
>>>>
>>>> Maybe something like KEY_RESERVED, KEY_FN, or KEY_POWER2?
>>>
>>> The code makes no distinction between KEY_POWER and KEY_A or KEY_B, etc.
>>> It simply passes event to userspace for processing.
>>
>> Right.  I don't expect a problem with most keys, but my proposal is to
>> special case KEY_POWER while suspended.  If a key press event must be sent
>> to keep Android and other userspace happy I suggest sending something
>> different just for that situation.
> 
> I do not know if userspace specifically looks for KEY_POWER or if it
> looks for user input in general, and I'd rather be on safe side and not
> mangle user input.
> 
> As Hans mentioned, at least some userspace already prepared to deal with
> this issue. And again, this only works if by the time ISR/debounce
> runs the key is already released. What if it is still pressed? You still
> going to observe KEY_POWER and need to suppress turning off the screen.
> 
>>
>> Like this:
>>
>> diff --git a/drivers/input/keyboard/gpio_keys.c
>> b/drivers/input/keyboard/gpio_keys.c
>> index 773aa5294d269..66e788d381956 100644
>> --- a/drivers/input/keyboard/gpio_keys.c
>> +++ b/drivers/input/keyboard/gpio_keys.c
>> @@ -425,7 +425,10 @@ static irqreturn_t gpio_keys_gpio_isr(int irq, void
>> *dev_id)
>>                          * already released by the time we got interrupt
>>                          * handler to run.
>>                          */
>> -                       input_report_key(bdata->input, button->code, 1);
>> +                       if (button->code == KEY_POWER)
>> +                               input_report_key(bdata->input, KEY_WAKEUP,
>> 1);
> 
> Just FYI: Here your KEY_WAKEUP is stuck forever.
> 
>> +                       else
>> +                               input_report_key(bdata->input, button->code,
>> 1);
>>                 }
>>         }
>>
>>
>>
>>>
>>> You need to fix your userspace. Even with your tweak it is possible for
>>> userspace to get a normal key event "too early" and turn off the screen
>>> again, so you still need to handle this situation.
>>>
>>> Thanks.
>>>
>>
>> I want to note this driver works quite differently than how ACPI power
>> button does.
>>
>> You can see in acpi_button_notify() that the "keypress" is only forwarded
>> when not suspended [1].  Otherwise it's just wakeup event (which is what my
>> patch was modeling).
>>
>> https://github.com/torvalds/linux/blob/v6.16-rc3/drivers/acpi/button.c#L461
>> [1]
> 
> If you check acpi_button_resume() you will see that the events are sent
> from there. Except that for some reason they chose to use KEY_WAKEUP and
> not KEY_POWER, oh well. Unlike acpi button driver gpio_keys is used on
> multiple other platforms.

Interesting, but the ACPI button code presumably only does this on resume
for a normal press while the system is awake it does use KEY_POWER, right ?

Regards,

Hans


> 
> Thanks.
>