Re: [PATCH] gpio: better lookup method for platform GPIOs

[Alex Courbot]

On 11/29/2013 12:54 AM, Andy Shevchenko wrote:
> On Thu, Nov 28, 2013 at 10:46 AM, Alexandre Courbot <acourbot@xxxxxxxxxx> wrote:
> > Change the format of the platform GPIO lookup tables to make them less
> > confusing and improve lookup efficiency.
> >
> > The previous format was a single linked-list that required to compare
> > the device name and function ID of every single GPIO defined for each
> > lookup. Switch that to a list of per-device tables, so that the lookup
> > can be done in two steps, omitting the GPIOs that are not relevant for a
> > particular device.
> >
> > The matching rules are now defined as follows:
> > - The device name must match *exactly*, and can be NULL for GPIOs not
> >    assigned to a particular device,
> > - If the function ID in the lookup table is NULL, the con_id argument of
> >    gpiod_get() will not be used for lookup. However, if it is defined, it
> >    must match exactly.
> > - The index must always match.
>
> Thanks for that, since I'm also was a bit confused of those dev_id/con_id stuff.
> Few comments below (mostly about style).
>
>
> > --- a/Documentation/gpio/board.txt
> > +++ b/Documentation/gpio/board.txt
>
> > @@ -88,16 +89,20 @@ Note that GPIO_LOOKUP() is just a shortcut to GPIO_LOOKUP_IDX() where idx = 0.
> >
> >   A lookup table can then be defined as follows:
> >
> > -       struct gpiod_lookup gpios_table[] = {
> > -       GPIO_LOOKUP_IDX("gpio.0", 15, "foo.0", "led", 0, GPIO_ACTIVE_HIGH),
> > -       GPIO_LOOKUP_IDX("gpio.0", 16, "foo.0", "led", 1, GPIO_ACTIVE_HIGH),
> > -       GPIO_LOOKUP_IDX("gpio.0", 17, "foo.0", "led", 2, GPIO_ACTIVE_HIGH),
> > -       GPIO_LOOKUP("gpio.0", 1, "foo.0", "power", GPIO_ACTIVE_LOW),
> > -       };
> > +struct gpiod_lookup_table gpios_table = {
> > +       .dev_id = "foo.0",
> > +       .size = 4,
> > +       .table = {
> > +       GPIO_LOOKUP_IDX("gpio.0", 15, "led", 0, GPIO_ACTIVE_HIGH),
> > +       GPIO_LOOKUP_IDX("gpio.0", 16, "led", 1, GPIO_ACTIVE_HIGH),
> > +       GPIO_LOOKUP_IDX("gpio.0", 17, "led", 2, GPIO_ACTIVE_HIGH),
> > +       GPIO_LOOKUP("gpio.0", 1, "power", GPIO_ACTIVE_LOW),
>
> Can you use deeper indentation for GPIO_* lines here?

Fixed.

> > --- a/drivers/gpio/gpiolib.c
> > +++ b/drivers/gpio/gpiolib.c
>
> > @@ -2326,72 +2322,77 @@ static struct gpio_desc *acpi_find_gpio(struct device *dev, const char *con_id,
> >          return desc;
> >   }
> >
> > -static struct gpio_desc *gpiod_find(struct device *dev, const char *con_id,
> > -                                   unsigned int idx,
> > -                                   enum gpio_lookup_flags *flags)
> > +static struct gpiod_lookup_table *gpiod_find_lookup_table(struct device *dev)
> >   {
> >          const char *dev_id = dev ? dev_name(dev) : NULL;
> > -       struct gpio_desc *desc = ERR_PTR(-ENODEV);
> > -       unsigned int match, best = 0;
> > -       struct gpiod_lookup *p;
> > +       struct gpiod_lookup_table *table;
> >
> >          mutex_lock(&gpio_lookup_lock);
> >
> > -       list_for_each_entry(p, &gpio_lookup_list, list) {
> > -               match = 0;
> > +       list_for_each_entry(table, &gpio_lookup_list, list) {
> > +               if (table->dev_id && dev_id && strcmp(table->dev_id, dev_id))
>
> Maybe check !dev_id outside of loop?

And create two loops, one for each case? Might complicate the code for 
little benefit IMHO, but please elaborate if I missed your point.

> > +                       continue;
> >
> > -               if (p->dev_id) {
> > -                       if (!dev_id || strcmp(p->dev_id, dev_id))
> > -                               continue;
> > +               if (dev_id != table->dev_id)
> > +                       continue;
> >
> > -                       match += 2;
> > -               }
> > +               return table;
>
> What  about
>
> if (dev_id == table->dev_id)
>   return table;
>
> ?

Actually my algorithm is broken to start with - and dangerous, as the 
missed mutex_unlock() you spotted later testifies. I will rewrite it in 
a (hopefully) sounder way.

> > +static struct gpio_desc *gpiod_find(struct device *dev, const char *con_id,
> > +                                   unsigned int idx,
> > +                                   enum gpio_lookup_flags *flags)
> > +{
> > +       struct gpio_desc *desc = ERR_PTR(-ENODEV);
> > +       struct gpiod_lookup_table *table;
> > +       int i;
> >
> > -               if (match > best) {
> > -                       struct gpio_chip *chip;
>
> Looks like redundant empty line.

Fixed.

> > -                       chip = find_chip_by_name(p->chip_label);
> > +       table = gpiod_find_lookup_table(dev);
> > +       if (!table)
> > +               return desc;
> >
> > -                       if (!chip) {
> > -                               dev_warn(dev, "cannot find GPIO chip %s\n",
> > -                                        p->chip_label);
> > -                               continue;
> > -                       }
> > +       for (i = 0; i < table->size; i++) {
> > +               struct gpio_chip *chip;
> > +               struct gpiod_lookup *p = &table->table[i];
> >
> > -                       if (chip->ngpio <= p->chip_hwnum) {
> > -                               dev_warn(dev, "GPIO chip %s has %d GPIOs\n",
> > -                                        chip->label, chip->ngpio);
> > +               if (p->idx != idx)
> > +                       continue;
> > +
> > +               if (p->con_id) {
> > +                       if (!con_id || strcmp(p->con_id, con_id))
>
> Could be one 'if' and moreover !con_id check might be outside a loop.

Again, wouldn't that require two separate loops?


--
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/