Re: [PATCH 2/5] gpio: Cygnus: add GPIO driver

[Ray Jui]

On 12/5/2014 6:34 PM, Joe Perches wrote:
> On Fri, 2014-12-05 at 18:14 -0800, Ray Jui wrote:
> > On 12/5/2014 5:28 PM, Joe Perches wrote:
> > > On Fri, 2014-12-05 at 16:40 -0800, Ray Jui wrote:
> > > > +static void bcm_cygnus_gpio_irq_handler(unsigned int irq,
> > > > +		struct irq_desc *desc)
> > > > +{
> > > > +	struct bcm_cygnus_gpio *cygnus_gpio;
> > > > +	struct irq_chip *chip = irq_desc_get_chip(desc);
> > > > +	int i, bit;
> > > > +
> > > > +	chained_irq_enter(chip, desc);
> > > > +
> > > > +	cygnus_gpio = irq_get_handler_data(irq);
> > > > +
> > > > +	/* go through the entire GPIO banks and handle all interrupts */
> > > > +	for (i = 0; i < cygnus_gpio->num_banks; i++) {
> > > > +		unsigned long val = readl(cygnus_gpio->base +
> > > > +				(i * GPIO_BANK_SIZE) +
> > > > +				CYGNUS_GPIO_INT_MSTAT_OFFSET);
> > > > +		if (val) {
> > >
> > > This if (val) and indentation isn't really necessary
> >
> > Note for_each_set_bit in this case iterates 32 times searching for bits
> > that are set.
>
> No it doesn't.
>
> #define for_each_set_bit(bit, addr, size) \
> 	for ((bit) = find_first_bit((addr), (size));		\
> 	     (bit) < (size);					\
> 	     (bit) = find_next_bit((addr), (size), (bit) + 1))
>
> find_first_bit:
>
>   * Returns the bit number of the first set bit.
>   * If no bits are set, returns @size.

You are right. I reviewed for_each_set_bit but didn't notice 
find_next_bit may simply return 32 in our case without doing any 
iterative processing. I will get rid of the redundant if (val) check below.

> >   By having the if (val) check here, it can potentially save
> > some of such processing in the ISR. I agree with you that it introduces
> > one extra indent here but I think it's required.
> >
> > > > +			for_each_set_bit(bit, &val, 32) {
> > >
> > > for_each_set_bit will effectively do the if above.
> > >
> > > 32 bit only code?
> > > otherwise isn't this endian unsafe?
> >
> > Will change 'unsigned long val' to 'u32 val'.
>
> All the bit operations only work on long *

Actually, by reviewing the code more deeply, I'm not sure why using 
for_each_set_bit here is 'endian unsafe'. Isn't that already taken care 
of by macros in bitops.h? Sorry if I'm still missing something here...
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/