[PATCH u-boot 4/4] eth/r8152: support RTL8153B/RTL8154B

[Marek Vasut]

On 6/11/20 5:18 AM, Hayes Wang wrote:
> Marek Vasut [mailto:marex at denx.de]
>> Sent: Wednesday, June 10, 2020 9:21 PM
> [...]
>>>>>>> +	/* ADC Bias Calibration:
>>>>>>> +	 * read efuse offset 0x7d to get a 17-bit data. Remove the
>>>> dummy/fake
>>>>>>> +	 * bit (bit3) to rebuild the real 16-bit data. Write the data to the
>>>>>>> +	 * ADC ioffset.
>>>>>>> +	 */
>>>>>>> +	ocp_data = r8152_efuse_read(tp, 0x7d);
>>>>>>> +	data = (u16)(((ocp_data & 0x1fff0) >> 1) | (ocp_data & 0x7));
>>>>>>
>>>>>> Are these type casts really needed ?
>>>>>
>>>>> Is there not a warning from u32 to u16?
>>>>
>>>> There is, but it seems justified to warn about conversion from u32 to
>>>> u16, so maybe that needs to be fixed ?
>>>
>>> Excuse me. I don't understand what you mean.
>>> The original is 17-bit data, so I use u32 to store it. And,
>>> the final desired data is 16-bit. That is why
>>> I have to convert it from u32 to u16.
>>
>> So what happens to that bit 17 , is it dropped ?
> 
> I have descripted is in the code, such as following.
> 
> 	/* ADC Bias Calibration:
> 	 * read efuse offset 0x7d to get a 17-bit data. Remove the dummy/fake
> 	 * bit (bit3) to rebuild the real 16-bit data. Write the data to the
> 	 * ADC ioffset.
> 	 */
> 
> The real data (16-bit) would be inserted a dummy bit,
> and store the 17-bit to efuse offset 0x7d. Therefore, when
> reading the 17-bit data from efuse, we have to remove the
> dummy to get the real data.

Ah, hmm, then let's use u32 type and be done with it. That solves the
typecasts and is safe. Would that work ?