Re: [PATCH 02/11] gpio: vf610: propagate return value to callers

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] gpio: vf610: vf610_gpio_irq_set_wake() breaks GPIO wakeup on Vybrid SoCs
--

commit 4cb6078cbfdf301b9f65c4307b8d2ea1ba2ec0b8
Author: Peng Fan <peng.fan@nxp.com>

gpio: vf610: propagate return value to callers

This commit updates vf610_gpio_irq_set_wake() to propagate the return value
from enable_irq_wake() and disable_irq_wake() instead of unconditionally
returning 0.

> diff --git a/drivers/gpio/gpio-vf610.c b/drivers/gpio/gpio-vf610.c
> index adfa2c2d5873d..44b846ff1b936 100644
> --- a/drivers/gpio/gpio-vf610.c
> +++ b/drivers/gpio/gpio-vf610.c
> @@ -188,13 +188,14 @@ static int vf610_gpio_irq_set_wake(struct irq_data *d, u32 enable)
>  {
>  	struct vf610_gpio_port *port =
>  		gpiochip_get_data(irq_data_get_irq_chip_data(d));
> +	int ret;
>  
>  	if (enable)
> -		enable_irq_wake(port->irq);
> +		ret = enable_irq_wake(port->irq);
>  	else
> -		disable_irq_wake(port->irq);
> +		ret = disable_irq_wake(port->irq);
>  
> -	return 0;
> +	return ret;

[Severity: Medium]
Does propagating ret here break GPIO wakeup on Vybrid (VF500/VF610) SoCs?

On Vybrid platforms in arch/arm/boot/dts/nxp/vf/vfxxx.dtsi, all
fsl,vf610-gpio controllers (gpio0 through gpio4) inherit mscm_ir as their
interrupt parent:

arch/arm/boot/dts/nxp/vf/vfxxx.dtsi:
	soc {
		#address-cells = <1>;
		#size-cells = <1>;
		compatible = "simple-bus";
		interrupt-parent = <&mscm_ir>;
		ranges;

In drivers/irqchip/irq-vf610-mscm-ir.c, vf610_mscm_ir_irq_chip sets neither
.irq_set_wake nor IRQCHIP_SKIP_SET_WAKE:

drivers/irqchip/irq-vf610-mscm-ir.c:
static struct irq_chip vf610_mscm_ir_irq_chip = {
	.name			= "mscm-ir",
	.irq_mask		= irq_chip_mask_parent,
	.irq_unmask		= irq_chip_unmask_parent,
	.irq_eoi		= irq_chip_eoi_parent,
	.irq_enable		= vf610_mscm_ir_enable,
	.irq_disable		= vf610_mscm_ir_disable,
	.irq_retrigger		= irq_chip_retrigger_hierarchy,
	.irq_set_affinity	= irq_chip_set_affinity_parent,
};

When a consumer calls enable_irq_wake() on a Vybrid GPIO interrupt, the call
chain irq_set_irq_wake() -> set_irq_wake_real() -> vf610_gpio_irq_set_wake()
calls enable_irq_wake(port->irq), which invokes set_irq_wake_real() on
port->irq in kernel/irq/manage.c:

static int set_irq_wake_real(unsigned int irq, unsigned int on)
{
	struct irq_desc *desc = irq_to_desc(irq);
	int ret = -ENXIO;

	if (irq_desc_get_chip(desc)->flags &  IRQCHIP_SKIP_SET_WAKE)
		return 0;

	if (desc->irq_data.chip->irq_set_wake)
		ret = desc->irq_data.chip->irq_set_wake(&desc->irq_data, on);

	return ret;
}

Because vf610_mscm_ir_irq_chip lacks both IRQCHIP_SKIP_SET_WAKE and
.irq_set_wake, set_irq_wake_real(port->irq, 1) returns -ENXIO. Previously,
vf610_gpio_irq_set_wake() ignored this and returned 0, allowing
irq_set_irq_wake() on the child GPIO IRQ to set IRQD_WAKEUP_STATE so that
suspend_device_irq() left the GPIO pin unmasked while port->irq remained
active as a chained interrupt.

With this change (and after the follow-up commit 614715b30c6f ("gpio: vf610:
simplify with irq_set_irq_wake")), irq_set_irq_wake() on any Vybrid GPIO IRQ
fails with -ENXIO, preventing IRQD_WAKEUP_STATE from being set and causing
suspend_device_irq() to mask the wakeup GPIO interrupt in PORT_PCR(n) during
system suspend.

Should vf610_mscm_ir_irq_chip also be updated to set .irq_set_wake =
irq_chip_set_wake_parent (or IRQCHIP_SKIP_SET_WAKE) before propagating this
return value?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261029-gpio-vf610-fixes-v1-0-45b484b4c434@nxp.com?part=2