Re: [PATCH] si4713: Fix oops when si4713_platform_data is marked as __initdata

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Em 18-05-2010 10:24, Jarkko Nikula escreveu:
> On Tue, 18 May 2010 15:55:27 +0300
> Eduardo Valentin <eduardo.valentin@nokia.com> wrote:
> 
>> I'm probably fine with this patch, and the driver must check for the pointer
>> before using it, indeed.
>>
>> But, I'm a bit skeptic about marking its platform data as __initdata. Would it make sense?
>> What happens if driver is built as module and loaded / unload / loaded again?
>>
>> Maybe the initdata flag does not apply in this case. Not sure (and not tested the above case).
>>
> Yep, it doesn't work or make sense for modules if platform data is
> marked as __initdata but with built in case it can save some bytes which
> are not needed after kernel is initialized.
> 
> Like with this driver the i2c_bus number and i2_board_info data are not
> needed after probing but only pointer to set_power must be preserved.
> 

Hi Eduardo,

This patch is still on my queue. It is not clear to me what "proably fine" means...
Please ack or nack on it for me to move ahead ;)

Cheers,
Mauro