Re: [PATCH v7 3/4] i2c: tegra: Add logic to support different register offsets

[Kartik Rajput <kkartik@nvidia.com>]

Hi Jon,

Thanks for reviewing the patch!

On 20/01/26 02:31, Jon Hunter wrote:
> 
> 
> On 13/01/2026 16:59, Kartik Rajput wrote:
>> Tegra410 use different offsets for existing I2C registers, update
>> the logic to use appropriate offsets per SoC.
>>
>> As the registers offsets are now also defined for dvc and vi, following
>> function are not required and they are removed:
>>   - tegra_i2c_reg_addr(): No translation required.
>>   - dvc_writel(): Replaced with i2c_writel() with DVC check.
> 
> This says dvc_writel is replace, but ...
> 

Yes, the commit message needs to be updated.

>>   - dvc_readl(): Replaced with i2c_readl().
>>
>> Signed-off-by: Kartik Rajput <kkartik@nvidia.com>
>> ---
>> Changes in v7:
>>     * Fix Tegra256 reg offsets, change it to tegra264_i2c_regs as it
>>       supports SW mutex.
>> Changes in v6:
>>     * Do not remove dvc_writel().
> 
> This says it isn't and ...
> 
>>   /**
>> @@ -348,45 +466,26 @@ static void dvc_writel(struct tegra_i2c_dev *i2c_dev, u32 val,
>>       writel_relaxed(val, i2c_dev->base + reg);
>>   }
>> -static u32 dvc_readl(struct tegra_i2c_dev *i2c_dev, unsigned int reg)
>> -{
>> -    return readl_relaxed(i2c_dev->base + reg);
>> -}
> 
> If we are removing dvc_readl() I am not sure why we have not removed dvc_writel().
> 
>>   static void tegra_i2c_dma_complete(void *args)
>> @@ -621,12 +720,12 @@ static void tegra_dvc_init(struct tegra_i2c_dev *i2c_dev)
>>   {
>>       u32 val;
>> -    val = dvc_readl(i2c_dev, DVC_CTRL_REG3);
>> +    val = i2c_readl(i2c_dev, DVC_CTRL_REG3);
>>       val |= DVC_CTRL_REG3_SW_PROG;
>>       val |= DVC_CTRL_REG3_I2C_DONE_INTR_EN;
>>       dvc_writel(i2c_dev, val, DVC_CTRL_REG3);
>> -    val = dvc_readl(i2c_dev, DVC_CTRL_REG1);
>> +    val = i2c_readl(i2c_dev, DVC_CTRL_REG1);
>>       val |= DVC_CTRL_REG1_INTR_EN;
>>       dvc_writel(i2c_dev, val, DVC_CTRL_REG1);
>>   }
> 
> We could just call writel_relaxed() directly if we can't use i2c_writel().
> 
> Jon
> 

This makes sense, we are not doing anything special in the dvc_writel(), it can be
replaced with writel_relaxed() directly.

Thanks,
Kartik