Re: [PATCH] of: Introduce of_node_get_match_data() helper

[Geert Uytterhoeven]

Hi Vladimir,

On Fri, Jun 9, 2017 at 9:30 AM, Vladimir Zapolskiy
<vladimir_zapolskiy@xxxxxxxxxx> wrote:
> On 04/25/2017 08:40 PM, Geert Uytterhoeven wrote:
>> If CONFIG_OF=n, code using
>>
>>     info = of_match_node(matchs, np)->data;
>>
>> fails to compile:
>>
>>     warning: dereferencing âvoid *â pointer
>>     error: request for member âdataâ in something not a structure or union
>
> IMHO firstly all the cases like one above must be fixed in place, in
> parallel do an overall conversion of subcases:
>
> -       of_id = of_match_node(some_dt_ids, dev->of_node);
> +       of_id = of_match_device(some_dt_ids, dev);
>
>> Follow the example set by of_device_get_match_data(), and introduce a
>> new helper of_node_get_match_data().  This will allow to increase
>> compile-testing coverage later.
>
> Unfortunately the proposed of_node_get_match_data() is not so similar
> to the of_device_get_match_data(), the latter one has only one input
> argument in opposite to of_match_device(), which is truly convenient.

Thanks to your comment, I realized I can just use of_device_get_match_data().
Hence I'd like to withdraw my patch.

Background: I encountered this in a clock driver, which started life as
a CLK_OF_DECLARE() driver, but ended up as a platform driver anyway.
CLK_OF_DECLARE() drivers don't have a struct device to use, so the
driver still used of_match_node() instead of of_match_device().

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds