On Tue, Jan 22, 2013 at 10:10 PM, Gregory CLEMENT
<gregory.clem...@free-electrons.com> wrote:
> Now that pca953x driver can handle GPIO expanders with more than 32
> bits this patch adds the support for the pca9505 which cam with 40
> GPIOs.
>
> Signed-off-by: Gregory CLEMENT <gregory.clem...@free-electrons.com>
Patch applied, thanks.
But guys, this driver contains some horrific stuff, look at this:
static int pxa_gpio_nums(void)
{
int count = 0;
#ifdef CONFIG_ARCH_PXA
if (cpu_is_pxa25x()) {
#ifdef CONFIG_CPU_PXA26x
count = 89;
gpio_type = PXA26X_GPIO;
#elif defined(CONFIG_PXA25x)
count = 84;
gpio_type = PXA26X_GPIO;
#endif /* CONFIG_CPU_PXA26x */
} else if (cpu_is_pxa27x()) {
count = 120;
gpio_type = PXA27X_GPIO;
} else if (cpu_is_pxa93x()) {
count = 191;
gpio_type = PXA93X_GPIO;
} else if (cpu_is_pxa3xx()) {
count = 127;
gpio_type = PXA3XX_GPIO;
}
#endif /* CONFIG_ARCH_PXA */
#ifdef CONFIG_ARCH_MMP
if (cpu_is_pxa168() || cpu_is_pxa910()) {
count = 127;
gpio_type = MMP_GPIO;
} else if (cpu_is_mmp2()) {
count = 191;
gpio_type = MMP_GPIO;
}
#endif /* CONFIG_ARCH_MMP */
return count;
}
This is totally killing all attempts at a single kernel for multiple
machines in this family. The above should not be #ifdef's but instead
use either platform data or the compatible property to figure out which
one to use.
It's fine to introduce new compatible= strings or device names to
distinguish between these.
As an example, in pinctrl-nomadik.c we have this:
static const struct of_device_id nmk_pinctrl_match[] = {
{
.compatible = "stericsson,nmk_pinctrl",
.data = (void *)PINCTRL_NMK_DB8500,
},
{},
};
static const struct platform_device_id nmk_pinctrl_id[] = {
{ "pinctrl-stn8815", PINCTRL_NMK_STN8815 },
{ "pinctrl-db8500", PINCTRL_NMK_DB8500 },
{ "pinctrl-db8540", PINCTRL_NMK_DB8540 },
{ }
};
static struct platform_driver nmk_pinctrl_driver = {
.driver = {
.owner = THIS_MODULE,
.name = "pinctrl-nomadik",
.of_match_table = nmk_pinctrl_match,
},
.probe = nmk_pinctrl_probe,
.id_table = nmk_pinctrl_id,
};
The first match is for DT boot the latter for using the platform
device name directly.
Then in the probefunction we do:
static int nmk_pinctrl_probe(struct platform_device *pdev)
{
const struct platform_device_id *platid = platform_get_device_id(pdev);
(...)
if (platid)
version = platid->driver_data;
else if (np) {
const struct of_device_id *match;
match = of_match_device(nmk_pinctrl_match, &pdev->dev);
if (!match)
return -ENODEV;
version = (unsigned int) match->data;
}
(...)
if (version == PINCTRL_NMK_STN8815)
nmk_pinctrl_stn8815_init(&npct->soc);
if (version == PINCTRL_NMK_DB8500)
nmk_pinctrl_db8500_init(&npct->soc);
if (version == PINCTRL_NMK_DB8540)
nmk_pinctrl_db8540_init(&npct->soc);
}
Surely a scheme like this must be possible to use to distinguish between
the different versions at runtime rather than using these #ifdefs?
Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
