Re: [PATCH v5 02/23] of: device: Export of_device_{get_modalias, uvent_modalias} to modules

[Javier Martinez Canillas]

Hello ChenYu

On Fri, Nov 11, 2016 at 12:01 AM, Chen-Yu Tsai <wens@xxxxxxxx> wrote:
> On Fri, Nov 11, 2016 at 5:42 AM, Rob Herring <robh@xxxxxxxxxx> wrote:

[snip]

>>
>> Do you have patches using this already. If not, it is starting to get
>> a bit late for v4.10.
>>
>> I can apply this, but then you'll just be pulling in other DT patches.
>
> Not sure what you mean by "using" this...
>
> I have patches which use this to add DT-based modalias entries for
> module auto-loading to i2c and sunxi-rsb that I haven't sent.
>

Unfortunately the I2C core can't be changed without breaking a lot of
I2C drivers that are relying on the current behavior. I've already
posted a RFC patch [0] for I2C that does this about a year ago and
enumerated the issues that have to be addressed before the change can
be made (and fixed some of the issues mentioned) on this series [1].

Another issue is that an I2C device ID table is a requirement anyways
since I2C drivers expect an i2c_device_id as an argument of their
probe function. Kieran already have patches [2] to change that which
should land soon.

I plan to fix the remaining I2C drivers once his patches are merged
and re-post the RFC patch as a proper one.

> As far as DT usage goes, we already need this for the axp20x mfd driver.
> There are 2 variants, i2c and sunxi-rsb. For the I2C variant a fix was
> sent to fix module auto-loading by using the I2C client ID table:
>
>     mfd: axp20x-i2c: Add i2c-ids to fix module auto-loading
>     https://git.kernel.org/cgit/linux/kernel/git/lee/mfd.git/commit/?h=for-mfd-next&id=b7142a19321484bd7681aa547c1d50148c8e2825
>

Yes, this is the workaround used by most DT-only I2C drivers. The only
reason that these drivers have an I2C device ID is due the
restrictions imposed by the I2C core.

Best regards,
Javier