Re: [PATCH] i2c: i2c-core-base: fix devicetree alias handling

Re: [PATCH] i2c: i2c-core-base: fix devicetree alias handling

[Jon Hunter]

Hi Andreas,

On 02/03/2026 17:04, Andreas Kemnade wrote:
> Parent device is not set up there, so use the new of_node pointer to
> handle aliases.
> 
> Fixes: 0ab80451c70f ("i2c: allow setting the parent device and OF node through the adapter struct")
> Reported-by: Kalle Niemi <kaleposti@gmail.com>
> Closes: https://lore.kernel.org/lkml/EDF3FB58-4747-442E-8463-6F1C6E568962@gmail.com/
> Signed-off-by: Andreas Kemnade <andreas@kemnade.info>
> ---
>   drivers/i2c/i2c-core-base.c | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
> index dd8cec9b04c6..e22d784202b8 100644
> --- a/drivers/i2c/i2c-core-base.c
> +++ b/drivers/i2c/i2c-core-base.c
> @@ -1657,10 +1657,9 @@ static int __i2c_add_numbered_adapter(struct i2c_adapter *adap)
>    */
>   int i2c_add_adapter(struct i2c_adapter *adapter)
>   {
> -	struct device *dev = &adapter->dev;
>   	int id;
>   
> -	id = of_alias_get_id(dev->of_node, "i2c");
> +	id = of_alias_get_id(adapter->of_node, "i2c");
>   	if (id >= 0) {
>   		adapter->nr = id;
>   		return __i2c_add_numbered_adapter(adapter);

This change is breaking a test we run to check that I2C devices are 
populated as expected on a few of our Tegra boards. For example, on the 
Tegra234 board, the I2C aliases are defined in 
arch/arm64/boot/dts/nvidia/tegra234.dtsi. Here we have ...

         aliases {
                 i2c0 = &gen1_i2c;
                 i2c1 = &gen2_i2c;
                 i2c2 = &cam_i2c;
                 i2c3 = &dp_aux_ch1_i2c;
                 i2c4 = &bpmp_i2c;
                 i2c5 = &dp_aux_ch0_i2c;
                 i2c6 = &dp_aux_ch2_i2c;
                 i2c7 = &gen8_i2c;
                 i2c8 = &dp_aux_ch3_i2c;
         };

Before this change, on the Tegra234 Jetson AGX Orin board 
(tegra234-p3737-0000+p3701-0000.dts) the I2C devices were mapped as 
above ...

/sys/class/i2c-dev/i2c-0/name --> 3160000.i2c
/sys/class/i2c-dev/i2c-1/name --> c240000.i2c
/sys/class/i2c-dev/i2c-4/name --> Tegra BPMP I2C adapter
/sys/class/i2c-dev/i2c-8/name --> 31e0000.i2c

After this change I now see ...

/sys/class/i2c-dev/i2c-9/name --> Tegra BPMP I2C adapter
/sys/class/i2c-dev/i2c-10/name --> 3160000.i2c
/sys/class/i2c-dev/i2c-11/name --> 31e0000.i2c
/sys/class/i2c-dev/i2c-12/name --> c240000.i2c

So the subject says that this fixes device-tree aliases, but it appears 
to break it for Tegra. In fact, this patch appears to have the same 
problem that you reported with Bartosz's change. Reverting this change 
fixes the problem. So I am a bit confused why we are seeing different 
behaviour.

Jon

-- 
nvpublic

Re: [PATCH] i2c: i2c-core-base: fix devicetree alias handling

[Jon Hunter]

On 06/03/2026 10:18, Jon Hunter wrote:
> Hi Andreas,
> 
> On 02/03/2026 17:04, Andreas Kemnade wrote:
>> Parent device is not set up there, so use the new of_node pointer to
>> handle aliases.
>>
>> Fixes: 0ab80451c70f ("i2c: allow setting the parent device and OF node 
>> through the adapter struct")
>> Reported-by: Kalle Niemi <kaleposti@gmail.com>
>> Closes: https://lore.kernel.org/lkml/ 
>> EDF3FB58-4747-442E-8463-6F1C6E568962@gmail.com/
>> Signed-off-by: Andreas Kemnade <andreas@kemnade.info>
>> ---
>>   drivers/i2c/i2c-core-base.c | 3 +--
>>   1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
>> index dd8cec9b04c6..e22d784202b8 100644
>> --- a/drivers/i2c/i2c-core-base.c
>> +++ b/drivers/i2c/i2c-core-base.c
>> @@ -1657,10 +1657,9 @@ static int __i2c_add_numbered_adapter(struct 
>> i2c_adapter *adap)
>>    */
>>   int i2c_add_adapter(struct i2c_adapter *adapter)
>>   {
>> -    struct device *dev = &adapter->dev;
>>       int id;
>> -    id = of_alias_get_id(dev->of_node, "i2c");
>> +    id = of_alias_get_id(adapter->of_node, "i2c");
>>       if (id >= 0) {
>>           adapter->nr = id;
>>           return __i2c_add_numbered_adapter(adapter);
> 
> This change is breaking a test we run to check that I2C devices are 
> populated as expected on a few of our Tegra boards. For example, on the 
> Tegra234 board, the I2C aliases are defined in arch/arm64/boot/dts/ 
> nvidia/tegra234.dtsi. Here we have ...
> 
>          aliases {
>                  i2c0 = &gen1_i2c;
>                  i2c1 = &gen2_i2c;
>                  i2c2 = &cam_i2c;
>                  i2c3 = &dp_aux_ch1_i2c;
>                  i2c4 = &bpmp_i2c;
>                  i2c5 = &dp_aux_ch0_i2c;
>                  i2c6 = &dp_aux_ch2_i2c;
>                  i2c7 = &gen8_i2c;
>                  i2c8 = &dp_aux_ch3_i2c;
>          };

I guess I should add the mapping (from tegra234.dtsi) ...

&gen1_i2c --> i2c@3160000
&gen2_i2c --> i2c@c240000
&bpmp_i2c --> Tegra BPMP I2C adapter
dp_aux_ch3_i2c --> i2c@31e0000

> Before this change, on the Tegra234 Jetson AGX Orin board (tegra234- 
> p3737-0000+p3701-0000.dts) the I2C devices were mapped as above ...
> 
> /sys/class/i2c-dev/i2c-0/name --> 3160000.i2c
> /sys/class/i2c-dev/i2c-1/name --> c240000.i2c
> /sys/class/i2c-dev/i2c-4/name --> Tegra BPMP I2C adapter
> /sys/class/i2c-dev/i2c-8/name --> 31e0000.i2c

Hence this looks correct.

> After this change I now see ...
> 
> /sys/class/i2c-dev/i2c-9/name --> Tegra BPMP I2C adapter
> /sys/class/i2c-dev/i2c-10/name --> 3160000.i2c
> /sys/class/i2c-dev/i2c-11/name --> 31e0000.i2c
> /sys/class/i2c-dev/i2c-12/name --> c240000.i2c

But this does not.

Jon

-- 
nvpublic

Re: [PATCH] i2c: i2c-core-base: fix devicetree alias handling

[Andreas Kemnade]

On Fri, 6 Mar 2026 10:18:09 +0000
Jon Hunter <jonathanh@nvidia.com> wrote:

[...]

> /sys/class/i2c-dev/i2c-0/name --> 3160000.i2c
> /sys/class/i2c-dev/i2c-1/name --> c240000.i2c
> /sys/class/i2c-dev/i2c-4/name --> Tegra BPMP I2C adapter
> /sys/class/i2c-dev/i2c-8/name --> 31e0000.i2c
> 
> After this change I now see ...
> 
> /sys/class/i2c-dev/i2c-9/name --> Tegra BPMP I2C adapter
> /sys/class/i2c-dev/i2c-10/name --> 3160000.i2c
> /sys/class/i2c-dev/i2c-11/name --> 31e0000.i2c
> /sys/class/i2c-dev/i2c-12/name --> c240000.i2c
> 
> So the subject says that this fixes device-tree aliases, but it appears 
> to break it for Tegra. In fact, this patch appears to have the same 
> problem that you reported with Bartosz's change. Reverting this change 
> fixes the problem. So I am a bit confused why we are seeing different 
> behaviour.
> 
Hmm, i2c-tegra.c was forgotten in Bartosz's change because
the code looks a bit different.

There is still
i2c_dev->adapter.dev.of_node = i2c_dev->dev->of_node;
        i2c_dev->adapter.dev.parent = i2c_dev->dev;
there.
compared to 
  adap->parent = &pdev->dev;
        adap->of_node = pdev->dev.of_node;

in omap code. So I think, the easiest to get forward is to
also adapt i2c-tegra.c.

Regards,
Andreas

Re: [PATCH] i2c: i2c-core-base: fix devicetree alias handling

[Bartosz Golaszewski]

On Sat, Mar 7, 2026 at 6:27 PM Andreas Kemnade <andreas@kemnade.info> wrote:
>
> On Fri, 6 Mar 2026 10:18:09 +0000
> Jon Hunter <jonathanh@nvidia.com> wrote:
>
> [...]
>
> > /sys/class/i2c-dev/i2c-0/name --> 3160000.i2c
> > /sys/class/i2c-dev/i2c-1/name --> c240000.i2c
> > /sys/class/i2c-dev/i2c-4/name --> Tegra BPMP I2C adapter
> > /sys/class/i2c-dev/i2c-8/name --> 31e0000.i2c
> >
> > After this change I now see ...
> >
> > /sys/class/i2c-dev/i2c-9/name --> Tegra BPMP I2C adapter
> > /sys/class/i2c-dev/i2c-10/name --> 3160000.i2c
> > /sys/class/i2c-dev/i2c-11/name --> 31e0000.i2c
> > /sys/class/i2c-dev/i2c-12/name --> c240000.i2c
> >
> > So the subject says that this fixes device-tree aliases, but it appears
> > to break it for Tegra. In fact, this patch appears to have the same
> > problem that you reported with Bartosz's change. Reverting this change
> > fixes the problem. So I am a bit confused why we are seeing different
> > behaviour.
> >
> Hmm, i2c-tegra.c was forgotten in Bartosz's change because
> the code looks a bit different.
>

It was not forgotten, rather my series only changed a part of the
users and I was planning to continue the work but it seems like Johan
committed to attempting to rework i2c using a different approach[1].

Wolfram: do you want to drop these patches from your v7.1 queue then?

Bartosz

> There is still
> i2c_dev->adapter.dev.of_node = i2c_dev->dev->of_node;
>         i2c_dev->adapter.dev.parent = i2c_dev->dev;
> there.
> compared to
>   adap->parent = &pdev->dev;
>         adap->of_node = pdev->dev.of_node;
>
> in omap code. So I think, the easiest to get forward is to
> also adapt i2c-tegra.c.
>
> Regards,
> Andreas

[1] https://lore.kernel.org/all/aar3rj7Db6NmTVS_@hovoldconsulting.com/

Re: [PATCH] i2c: i2c-core-base: fix devicetree alias handling

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 180 bytes --]

Hi Bart,

> Wolfram: do you want to drop these patches from your v7.1 queue then?

Yes, I think this makes sense. Let's see how Johan's approach goes?

Happy hacking,

   Wolfram


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]