RE: [PATCH 1/1] mfd: intel_quark_i2c_gpio: Add Intel Quark X1000 I2C-GPIO MFD Driver

["Chen, Alvin" <alvin.chen@intel.com>]

> 
> On 03/11/14 07:39, Raymond Tan wrote:
> > +	pdata->properties->irq	= pdev->irq;
> > +	pdata->properties->irq_shared	= true;
> 
> OK I see it.
> 
> Thanks.
> 
> My question is. How extensively have edge triggered interrupts been tested on
> the GPIO block ?
> 
> The BSP reference code is quite explicit about not missing edge interrupts.
> 
> Have you tested GPIO input in edge mode ?
We indeed test edge mode. Now all are moved to gpio-dwapb module which has been accepted by maintainer.
The BSP code doesn't meet the requirement of upstream, so we totally re-implement this feature in gpio-dwapb.
This patch only passes the necessary parameters and registers the platform devices.



> 
> +irqreturn_t intel_qrk_gpio_isr(int irq, void *dev_id) {
> +	irqreturn_t ret = IRQ_NONE;
> +	u32 pending = 0, gpio = 0;
> +	void __iomem *reg_pending = reg_base + PORTA_INT_STATUS;
> +	void __iomem *reg_eoi = reg_base + PORTA_INT_EOI;
> +
> +	/* Which pin (if any) triggered the interrupt */
> +	while ((pending = ioread32(reg_pending))) {
> +		/*
> +		 * Acknowledge all the asserted GPIO interrupt lines before
> +		 * serving them, so that we don't lose an edge.
> +		 * This has only effect on edge-triggered interrupts.
> +		 */
> +		iowrite32(pending, reg_eoi);
> +
> +		/* Serve each asserted interrupt */
> +		do {
> +			gpio = __ffs(pending);
> +			generic_handle_irq(
> +				gpio_to_irq(INTEL_QRK_GIP_GPIO_BASE + gpio));
> +			pending &= ~BIT(gpio);
> +			ret = IRQ_HANDLED;
> +		} while(pending);
> +	}
> +
> +	return ret;
> +}