Re: [PATCH AUTOSEL 5.10 14/36] media: rkisp1: uapi: change hist_bins array type from __u16 to __u32

[Hans Verkuil <hverkuil-cisco@xs4all.nl>]

On 09/02/2021 14:02, Greg Kroah-Hartman wrote:
> On Tue, Feb 09, 2021 at 01:45:35PM +0100, Dafna Hirschfeld wrote:
>>
>>
>> Am 08.02.21 um 21:46 schrieb Hans Verkuil:
>>> On 08/02/2021 18:57, Sasha Levin wrote:
>>>> From: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
>>>>
>>>> [ Upstream commit 31f190e0ccac8b75d33fdc95a797c526cf9b149e ]
>>>>
>>>> Each entry in the array is a 20 bits value composed of 16 bits unsigned
>>>> integer and 4 bits fractional part. So the type should change to __u32.
>>>> In addition add a documentation of how the measurements are done.
>>>
>>> Dafna, Helen, does it make sense at all to backport these three patches to
>>> when rkisp1 was a staging driver?
>>>
>>> I would be inclined not to backport this.
>>
>> I also don't think it makes sense since this changes the uapi and it is not really a bug fix.
> 
> Why was it ok to change the uapi in a newer kernel and not an older one?

In the older kernels this was a staging driver and the driver API was not public.
It's debatable whether there is any benefit from trying to backport patches like
this to a staging driver like that.

Also, these backports are incomplete, there are other patches that would need to
be applied to make this work. Applying just these three patches without the other
three (commits 66d81de7ea9d, fc672d806bd7 and ef357e02b6c4) makes it very messy
indeed.

I'd just leave the staging driver in older kernels as-is. Certainly don't just
apply these three patches without the other three commits, that would make it
even worse.

Regards,

	Hans

> 
> thanks,
> 
> greg k-h
>