Am 2020-06-05 12:48, schrieb Andy Shevchenko:
On Fri, Jun 5, 2020 at 1:09 PM Michael Walle <michael@xxxxxxxx> wrote:
Am 2020-06-05 10:01, schrieb Andy Shevchenko:
> On Fri, Jun 5, 2020 at 12:16 AM Michael Walle <michael@xxxxxxxx> wrote:
...
>> + bool "Kontron sl28 core driver"
>> + depends on I2C=y
>
> Why not module?
There are users of the interupt lines provided by the interrupt
controller.
For example, the gpio-button driver. If this is compiled into the
kernel
(which it is by default in the arm64 defconfig), probing will fail
because
the interrupt is not found. Is there a better way for that? I guess
the
same
is true for the GPIO driver.
And GPIO nicely handles this via deferred probe mechanism. Why it
can't be used here?
So, we really need to have a strong argument to limit module nowadays
to be only builtin.
Was that a question for me? TBH thats how other interrupt drivers doing
it for now. And it would be the users who need to be fixed, right? Or
even the platform code? Because it will complain with
[ 2.962990] irq: no irq domain found for interrupt-controller@1c !
[ 2.975762] gpio-keys buttons0: Found button without gpio or irq
[ 2.981872] gpio-keys: probe of buttons0 failed with error -22
>> + depends on OF
>
> I didn't find an evidence this is needed.
>> +#include <linux/of_platform.h>
>
> No evidence of user of this.
> I think you meant mod_devicetable.h.
devm_of_platform_populate(), so I need CONFIG_OF, too right?
Ah, this explains header, thanks!
But it doesn't explain depends OF.
So, perhaps,
depends OF || COMPILE_TEST will be more informative, i.e.
tells "okay, this driver can be compiled w/o OF, but won't be
functional".
ok
--
-michael
