Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH RFC] modpost: Support I2C Aliases from OF tables
To:     Wolfram Sang <wsa@the-dreams.de>
Cc:     Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>,
        Masahiro Yamada <yamada.masahiro@socionext.com>,
        Michal Marek <michal.lkml@markovi.net>,
        linux-kbuild@vger.kernel.org,
        open list <linux-kernel@vger.kernel.org>,
        linux-renesas-soc@vger.kernel.org,
        Lee Jones <lee.jones@linaro.org>,
        Alexandre Belloni <alexandre.belloni@bootlin.com>,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Mark Brown <broonie@kernel.org>
References: <20190710193918.31135-1-kieran.bingham+renesas@ideasonboard.com>
 <0e1b6e0b-1c94-4b00-7fda-c2a303ee3816@redhat.com>
 <20190731194419.GB4084@kunai>
From:   Javier Martinez Canillas <javierm@redhat.com>
Message-ID: <f4a78e93-6aaa-bc72-cf94-06fc2574451c@redhat.com>
Date:   Tue, 6 Aug 2019 00:25:34 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <20190731194419.GB4084@kunai>
Content-Type: text/plain; charset=windows-1252
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hello Wolfram,

On 7/31/19 9:44 PM, Wolfram Sang wrote:
> Hi Javier,
> 
> thank you for providing the extra information.
> 
> (And Kieran, thanks for the patch!)
> 
>> The other option is to remove i2c_of_match_device() and don't make OF match
>> to fallback to i2c_of_match_device_sysfs(). This is what happens in the ACPI
>> case, since i2c_device_match() just calls acpi_driver_match_device() directly
>> and doesn't have a wrapper function that fallbacks to sysfs matching.
>>
>> In this case an I2C device ID table would be required if the devices have to
>> be instantiated through sysfs. That way the I2C table would be used both for
>> auto-loading and also to match the device when it doesn't have an of_node.
> 
> That would probably mean that only a minority of drivers will not add an I2C
> device ID table because it is easy to add an you get the sysfs feature?
>

I believe so yes.
 
> Then we are back again with the situation that most drivers will have
> multiple tables. With the minor change that the I2C device id table is
> not required anymore by the core, but it will be just very useful to
> have? Or?
>

Yes, it won't be needed anymore if you are only instantiating all your devices
from your firmware interface (e.g: OF, ACPI).

>> If the former is the correct way to solve this then the patch looks good to me.
>>
>> Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
> 
> For this actual patch from Kieran, I'd like to hear an opinion from the
> people maintaining modpost. The aproach looks okay to me, yet I can't
> tell how "easy" we are with adding new types like 'i2c_of'.
>

As Masahiro-san mentioned, this approach will still require to add a new macro
MODULE_DEVICE_TABLE(i2c_of, bar_of_match) so the OF device table is used twice.

One to expose the "of:N*T*Cfoo,bar" and another one to expose it as "i2c:bar".

I expect that many developers would miss adding this macro for new drivers that
are DT-only and so sysfs instantiation would not work there. So whatever is the
approach taken we should clearly document all this so drivers authors are aware.

> Thanks everyone,
> 
>    Wolfram
> 

Best regards,
-- 
Javier Martinez Canillas
Software Engineer - Desktop Hardware Enablement
Red Hat