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@xxxxxxx>
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@xxxxxxxxxx>
Fixes: 0f107573da417 ("Input: gpio_keys - handle the missing key press event in resume phase")
Signed-off-by: Mario Limonciello <mario.limonciello@xxxxxxx>
---
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);
}
}
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?
Most people not using Android would be compiling with that enabled in
their kernel I'd expect.
