On some boards such as i.MX8MQ-EVK, the PCIe driver may take up to
200ms to restore the PCIe link during the no_irq resume phase. This
causes key press events to be lost because the key may be released
before the timer starts running, as interrupts are disabled during
this 200ms window.

Report key press events directly in interrupt handler to prevent event
loss during system suspend.

Signed-off-by: Joy Zou <joy.zou@nxp.com>
---
 drivers/input/keyboard/snvs_pwrkey.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/drivers/input/keyboard/snvs_pwrkey.c b/drivers/input/keyboard/snvs_pwrkey.c
index bab3ab57fdac77256be75a080773ea99372ec9c7..b557c1618d7369e872c6ce708a7b3017264ee385 100644
--- a/drivers/input/keyboard/snvs_pwrkey.c
+++ b/drivers/input/keyboard/snvs_pwrkey.c
@@ -78,6 +78,16 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
 
 	pm_wakeup_event(input->dev.parent, 0);
 
+	/*
+	 * Report key press events directly in interrupt handler to prevent event
+	 * loss during system suspend.
+	 */
+	if (pdev->dev.power.is_suspended) {
+		pdata->keystate = 1;
+		input_report_key(input, pdata->keycode, 1);
+		input_sync(input);
+	}
+
 	regmap_read(pdata->snvs, SNVS_LPSR_REG, &lp_status);
 	if (lp_status & SNVS_LPSR_SPO) {
 		if (pdata->minor_rev == 0) {

-- 
2.37.1

On Thu, Mar 26, 2026 at 06:39:40PM +0800, Joy Zou wrote:
> On some boards such as i.MX8MQ-EVK, the PCIe driver may take up to
> 200ms to restore the PCIe link during the no_irq resume phase. This
> causes key press events to be lost because the key may be released
> before the timer starts running, as interrupts are disabled during
> this 200ms window.

if irq disable, how imx_snvs_pwrkey_interrupt get run?

Frank
>
> Report key press events directly in interrupt handler to prevent event
> loss during system suspend.
>
> Signed-off-by: Joy Zou <joy.zou@nxp.com>
> ---
>  drivers/input/keyboard/snvs_pwrkey.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/drivers/input/keyboard/snvs_pwrkey.c b/drivers/input/keyboard/snvs_pwrkey.c
> index bab3ab57fdac77256be75a080773ea99372ec9c7..b557c1618d7369e872c6ce708a7b3017264ee385 100644
> --- a/drivers/input/keyboard/snvs_pwrkey.c
> +++ b/drivers/input/keyboard/snvs_pwrkey.c
> @@ -78,6 +78,16 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
>
>  	pm_wakeup_event(input->dev.parent, 0);
>
> +	/*
> +	 * Report key press events directly in interrupt handler to prevent event
> +	 * loss during system suspend.
> +	 */
> +	if (pdev->dev.power.is_suspended) {
> +		pdata->keystate = 1;
> +		input_report_key(input, pdata->keycode, 1);
> +		input_sync(input);
> +	}
> +
>  	regmap_read(pdata->snvs, SNVS_LPSR_REG, &lp_status);
>  	if (lp_status & SNVS_LPSR_SPO) {
>  		if (pdata->minor_rev == 0) {
>
> --
> 2.37.1
>

On Thu, Mar 26, 2026 at 03:57:27PM -0400, Frank Li wrote:
> On Thu, Mar 26, 2026 at 06:39:40PM +0800, Joy Zou wrote:
> > On some boards such as i.MX8MQ-EVK, the PCIe driver may take up to
> > 200ms to restore the PCIe link during the no_irq resume phase. This
> > causes key press events to be lost because the key may be released
> > before the timer starts running, as interrupts are disabled during
> > this 200ms window.
> 
> if irq disable, how imx_snvs_pwrkey_interrupt get run?
> 
Thank you for your comments. I might have missed some details in my commit
message. Could you please review the description below and let me know if
it's clear and comprehensive enough?

The driver implements debounce protection using a timer-based mechanism:
when a key interrupt occurs, a timer is scheduled to verify the key state
after DEBOUNCE_TIME before reporting the event. This works well during
normal operation.

However, key press events can be lost during system resume on platforms
like i.MX8MQ-EVK because:
1. During the no_irq resume phase, PCIe driver restoration can take up to
200ms with IRQs disabled.
2. The power key interrupt remains pending during the no_irq phase.
3. If the key is released before IRQs are re-enabled, the timer eventually
runs but sees the key as released and skips reporting the event.

Report key press events directly in interrupt handler to prevent event
loss during system suspend. This is safe because:

1. Only one event is reported per suspend cycle.
2. Normal operation retains the existing timer-based debounce mechanism.
BR
Joy Zou
> Frank
> >
> > Report key press events directly in interrupt handler to prevent event
> > loss during system suspend.
> >
> > Signed-off-by: Joy Zou <joy.zou@nxp.com>
> > ---
> >  drivers/input/keyboard/snvs_pwrkey.c | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> >
> > diff --git a/drivers/input/keyboard/snvs_pwrkey.c b/drivers/input/keyboard/snvs_pwrkey.c
> > index bab3ab57fdac77256be75a080773ea99372ec9c7..b557c1618d7369e872c6ce708a7b3017264ee385 100644
> > --- a/drivers/input/keyboard/snvs_pwrkey.c
> > +++ b/drivers/input/keyboard/snvs_pwrkey.c
> > @@ -78,6 +78,16 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
> >
> >  	pm_wakeup_event(input->dev.parent, 0);
> >
> > +	/*
> > +	 * Report key press events directly in interrupt handler to prevent event
> > +	 * loss during system suspend.
> > +	 */
> > +	if (pdev->dev.power.is_suspended) {
> > +		pdata->keystate = 1;
> > +		input_report_key(input, pdata->keycode, 1);
> > +		input_sync(input);
> > +	}
> > +
> >  	regmap_read(pdata->snvs, SNVS_LPSR_REG, &lp_status);
> >  	if (lp_status & SNVS_LPSR_SPO) {
> >  		if (pdata->minor_rev == 0) {
> >
> > --
> > 2.37.1
> >

On Tue, Mar 31, 2026 at 06:46:55PM +0800, Joy Zou wrote:
> On Thu, Mar 26, 2026 at 03:57:27PM -0400, Frank Li wrote:
> > On Thu, Mar 26, 2026 at 06:39:40PM +0800, Joy Zou wrote:
> > > On some boards such as i.MX8MQ-EVK, the PCIe driver may take up to
> > > 200ms to restore the PCIe link during the no_irq resume phase. This
> > > causes key press events to be lost because the key may be released
> > > before the timer starts running, as interrupts are disabled during
> > > this 200ms window.
> >
> > if irq disable, how imx_snvs_pwrkey_interrupt get run?
> >
> Thank you for your comments. I might have missed some details in my commit
> message. Could you please review the description below and let me know if
> it's clear and comprehensive enough?
>
> The driver implements debounce protection using a timer-based mechanism:
> when a key interrupt occurs, a timer is scheduled to verify the key state
> after DEBOUNCE_TIME before reporting the event. This works well during
> normal operation.
>
> However, key press events can be lost during system resume on platforms
> like i.MX8MQ-EVK because:
> 1. During the no_irq resume phase, PCIe driver restoration can take up to
> 200ms with IRQs disabled.
> 2. The power key interrupt remains pending during the no_irq phase.
> 3. If the key is released before IRQs are re-enabled, the timer eventually
> runs but sees the key as released and skips reporting the event.
>
> Report key press events directly in interrupt handler to prevent event
> loss during system suspend. This is safe because:
>
> 1. Only one event is reported per suspend cycle.
> 2. Normal operation retains the existing timer-based debounce mechanism.

much better. Thanks

Frank

> BR
> Joy Zou
> > Frank
> > >
> > > Report key press events directly in interrupt handler to prevent event
> > > loss during system suspend.
> > >
> > > Signed-off-by: Joy Zou <joy.zou@nxp.com>
> > > ---
> > >  drivers/input/keyboard/snvs_pwrkey.c | 10 ++++++++++
> > >  1 file changed, 10 insertions(+)
> > >
> > > diff --git a/drivers/input/keyboard/snvs_pwrkey.c b/drivers/input/keyboard/snvs_pwrkey.c
> > > index bab3ab57fdac77256be75a080773ea99372ec9c7..b557c1618d7369e872c6ce708a7b3017264ee385 100644
> > > --- a/drivers/input/keyboard/snvs_pwrkey.c
> > > +++ b/drivers/input/keyboard/snvs_pwrkey.c
> > > @@ -78,6 +78,16 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
> > >
> > >  	pm_wakeup_event(input->dev.parent, 0);
> > >
> > > +	/*
> > > +	 * Report key press events directly in interrupt handler to prevent event
> > > +	 * loss during system suspend.
> > > +	 */
> > > +	if (pdev->dev.power.is_suspended) {
> > > +		pdata->keystate = 1;
> > > +		input_report_key(input, pdata->keycode, 1);
> > > +		input_sync(input);
> > > +	}
> > > +
> > >  	regmap_read(pdata->snvs, SNVS_LPSR_REG, &lp_status);
> > >  	if (lp_status & SNVS_LPSR_SPO) {
> > >  		if (pdata->minor_rev == 0) {
> > >
> > > --
> > > 2.37.1
> > >