Re: [PATCH 4/7] gpio/mxs: convert to use linear irqdomain

[Rob Herring <robherring2@gmail.com>]

On 10/18/2012 01:22 AM, Shawn Guo wrote:
> On Tue, Oct 16, 2012 at 09:22:56PM +0200, Linus Walleij wrote:
>> The MXS driver tries to do the work of irq_domain_add_linear()
>> by reserving a bunch of descriptors somewhere and keeping track
>> of the base offset, then calling irq_domain_add_legacy(). Let's
>> stop doing that and simply use the linear IRQ domain.
>>
>> Cc: Rob Herring <rob.herring@calxeda.com>
>> Cc: Grant Likely <grant.likely@secretlab.ca>
>> Cc: Shawn Guo <shawn.guo@linaro.org>
>> Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
>> ---
>>  drivers/gpio/gpio-mxs.c | 19 ++++++-------------
>>  1 file changed, 6 insertions(+), 13 deletions(-)
>>
>> diff --git a/drivers/gpio/gpio-mxs.c b/drivers/gpio/gpio-mxs.c
>> index 796fb13..71fd5b1 100644
>> --- a/drivers/gpio/gpio-mxs.c
>> +++ b/drivers/gpio/gpio-mxs.c
>> @@ -223,7 +223,6 @@ static int __devinit mxs_gpio_probe(struct platform_device *pdev)
>>  	static void __iomem *base;
>>  	struct mxs_gpio_port *port;
>>  	struct resource *iores = NULL;
>> -	int irq_base;
> 
> ...
> 
>>  	int err;
>>  
>>  	port = devm_kzalloc(&pdev->dev, sizeof(*port), GFP_KERNEL);
>> @@ -272,16 +271,10 @@ static int __devinit mxs_gpio_probe(struct platform_device *pdev)
>>  	/* clear address has to be used to clear IRQSTAT bits */
>>  	writel(~0U, port->base + PINCTRL_IRQSTAT(port) + MXS_CLR);
>>  
>> -	irq_base = irq_alloc_descs(-1, 0, 32, numa_node_id());
>> -	if (irq_base < 0)
>> -		return irq_base;
>> -
>> -	port->domain = irq_domain_add_legacy(np, 32, irq_base, 0,
>> +	port->domain = irq_domain_add_linear(np, 32,
>>  					     &irq_domain_simple_ops, NULL);
>> -	if (!port->domain) {
>> -		err = -ENODEV;
>> -		goto out_irqdesc_free;
>> -	}
>> +	if (!port->domain)
>> +		return -ENODEV;
>>  
>>  	/* gpio-mxs can be a generic irq chip */
>>  	mxs_gpio_init_gc(port, irq_base);
>                                ^^^^^^^^
> 
> So I know this one is not compile-tested.
> 
> This is exactly the reason why I have to use irq_domain_add_legacy
> other than irq_domain_add_linear when I add irqdomain support for the
> driver.  The driver uses generic-irq infrastructural which needs
> irq_base for setup.  So sadly, before generic-irq gets improved, any
> irq chip that uses generic-irq will have to use irq_domain_add_legacy.

We should get someone to work on that. ;)

I think I need to go back and just fix this aspect rather than try and
integrate irqdomains into generic-irqchip.

Rob