From mboxrd@z Thu Jan  1 00:00:00 1970
From: tglx@linutronix.de (Thomas Gleixner)
Date: Fri, 2 Sep 2016 21:10:43 +0200 (CEST)
Subject: [PATCH v3 6/9] pinctrl: Add IRQ support to STM32 gpios
In-Reply-To: <1472830398-13275-7-git-send-email-alexandre.torgue@st.com>
References: <1472830398-13275-1-git-send-email-alexandre.torgue@st.com>
 <1472830398-13275-7-git-send-email-alexandre.torgue@st.com>
Message-ID: <alpine.DEB.2.20.1609022057490.5647@nanos>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Fri, 2 Sep 2016, Alexandre TORGUE wrote:
> +static int stm32_gpio_domain_translate(struct irq_domain *d,
> +				       struct irq_fwspec *fwspec,
> +				       unsigned long *hwirq,
> +				       unsigned int *type)
> +{
> +	if ((fwspec->param_count != 2) ||
> +		(fwspec->param[0] >= STM32_GPIO_IRQ_LINE))
> +		return -EINVAL;

Just a nitpick. This is unnecessarily hard to parse because you indented
the line break like a conditional statement

> +	if ((fwspec->param_count != 2) ||
> +	    (fwspec->param[0] >= STM32_GPIO_IRQ_LINE))
> +		return -EINVAL;

Makes it immediately obvious that the second line belongs to the if.

> +static void stm32_gpio_domain_activate(struct irq_domain *d,
> +				       struct irq_data *irq_data)
> +{
> +	struct stm32_gpio_bank *bank = d->host_data;
> +	struct stm32_pinctrl *pctl = dev_get_drvdata(bank->gpio_chip.parent);
> +
> +	if (gpiochip_lock_as_irq(&bank->gpio_chip, irq_data->hwirq)) {
> +		dev_err(pctl->dev,
> +			"Unable to configure STM32 %s%ld as IRQ\n",
> +			bank->gpio_chip.label, irq_data->hwirq);
> +		return;

Hmm, that's nasty. When an interrupt is mapped then we don't expect the
activate function to fail. You really should lock that interrupt when it's
mapped.

> +	}
> +	regmap_field_write(pctl->irqmux[irq_data->hwirq], bank->range.id);
> +}

> +static int stm32_gpio_domain_alloc(struct irq_domain *domain,
> +				   unsigned int virq,
> +				   unsigned int nr_irqs, void *data)
> +{
> +	struct irq_fwspec *fwspec = data;
> +	struct irq_fwspec parent_fwspec;
> +	struct stm32_pinctrl *pctl = domain->host_data;
> +	irq_hw_number_t hwirq;
> +	unsigned int i;
> +
> +	hwirq = fwspec->param[0];
> +	for (i = 0; i < nr_irqs; i++)
> +		irq_domain_set_hwirq_and_chip(domain, virq + i, hwirq + i,
> +					      &stm32_gpio_irq_chip, pctl);
> +
> +	parent_fwspec.fwnode = domain->parent->fwnode;
> +	parent_fwspec.param_count = 2;
> +	parent_fwspec.param[0] = fwspec->param[0];
> +	parent_fwspec.param[1] = fwspec->param[1];
> +
> +	return irq_domain_alloc_irqs_parent(domain, virq, nr_irqs,
> +			&parent_fwspec);

So doing it here would be probably the right thing to do:


	ret = gpiochip_lock_as_irq();
	if (ret)
		return ret;

   	ret = irq_domain_alloc_irqs_parent(domain, virq, nr_irqs,
	      				   &parent_fwspec);
	if (ret)
		gpiochip_unlock_as_irq();

	return ret;

So of course you need your own free() function which undoes that lock
thingy.

Thanks,

	tglx