Re: [PATCH 1/2] pinctrl/nomadik: add STn8815 ASIC support

[Linus Walleij]

On Fri, Jun 8, 2012 at 6:39 PM, Stephen Warren <swarren@xxxxxxxxxxxxx> wrote:
> On 06/08/2012 03:07 AM, Linus Walleij wrote:
>> This adds support for the STN8815 ASIC for the Nomadik pin
>> controller.
>
>> diff --git a/drivers/pinctrl/pinctrl-nomadik.c b/drivers/pinctrl/pinctrl-nomadik.c
>
>> @@ -1717,6 +1717,8 @@ static int __devinit nmk_pinctrl_probe(struct platform_device *pdev)
>>                       of_match_device(nmk_pinctrl_match, &pdev->dev)->data;
>>
>>       /* Poke in other ASIC variants here */
>> +     if (version == PINCTRL_NMK_STN8815)
>> +             nmk_pinctrl_stn8815_init(&npct->soc);
>>       if (version == PINCTRL_NMK_DB8500)
>>               nmk_pinctrl_db8500_init(&npct->soc);
>
> One comment that came up in other reviews is that we shouldn't have a
> single driver that switches on the SoC type it's running on and then
> dispatches to different ${soc}_init() functions, but rather should have
> multiple separate drivers, where each probe calls some utility function
> with the appropriate SoC parameterization structures/tables.

I discussed this with Arnd, the patch doesn't give the context of where this
identifier comes from:

static const struct platform_device_id nmk_pinctrl_id[] = {
        { "pinctrl-stn8815", PINCTRL_NMK_STN8815 },
        { "pinctrl-db8500", PINCTRL_NMK_DB8500 },
};

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,
};

And the probe looks like so:

static int __devinit nmk_pinctrl_probe(struct platform_device *pdev)
{
        const struct platform_device_id *platid = platform_get_device_id(pdev);
(...)
        if (platid)
                version = platid->driver_data;

So the same driver handles several platform device names, then
the name is used to select a variant. IIRC I asked Arnd about this
and he preferred this, and I was told by Mark Brown in the past
to do things this way (c.f. drivers/mfs/ab8500-core.c).

Doing it the other way is also possible but leads to a proliferation
of probe() calls and struct platform_driver blocks, and result in
more lines of code.

Both ways have precedents in the kernel so I actually think both
are OK, there is two ways to skin this cat simply and I'm not that
sensitive to which one is used.

Yours,
Linus Walleij
--
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/