Smatch issues reported for Marvell a38x DDR4

[Stefan Roese]

Hi Tony,

On 05.08.25 15:11, Tony Dinh wrote:
> Hi Andrew,
> 
> On Mon, Jul 28, 2025 at 9:25 PM Andrew Goodbody
> <andrew.goodbody at linaro.org> wrote:
>>
>> On 25/07/2025 14:02, Stefan Roese wrote:
>>> Hi Andrew,
>>>
>>> On 24.07.25 18:01, Andrew Goodbody wrote:
>>>> Hi Stefan,
>>>>
>>>> Smatch reports the following issues
>>>>
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:1029
>>>> mv_ddr4_center_of_mass_calc() warn: unsigned 'd_min' is never less
>>>> than zero.
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:2192
>>>> mv_ddr4_dm_tuning() warn: unsigned 'reg_val' is never less than zero.
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:2270
>>>> mv_ddr4_dm_tuning() warn: unsigned 'reg_val' is never less than zero.
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:662
>>>> mv_ddr4_centralization() warn: unsigned 'if_id' is never less than zero.
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:662
>>>> mv_ddr4_centralization() warn: we never enter this loop
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:911
>>>> mv_ddr4_dqs_reposition() warn: always true condition '(delta >= 0) =>
>>>> (0-255 >= 0)'
>>>> drivers/ddr/marvell/a38x/mv_ddr4_training_calibration.c:924
>>>> mv_ddr4_dqs_reposition() warn: unsigned 'delta' is never less than zero.
>>>>
>>>> As I have no way to validate the correctness of the code I am not
>>>> confident enough to attempt to make the appropriate fixes in this case
>>>> as it could affect its operation. Perhaps you would like to take a look?
>>>
>>> Sure, thanks for all your work on these smatch issues. Really
>>> appreciated.
>>>
>>> But frankly, it's not easy to test anything in this area. I still have
>>> access to one board (AFAIK), but minor changes usually don't really
>>> make a notable difference in the DDR controller configuration here.
>>>
>>> My suggestion is to send the patches (or one patch) "as best as you can"
>>> to the list for official review. I'll try to review and perhaps someone
>>> else with access to such a board will take a look.
>>>
>>> Thanks,
>>> Stefan
>>
>> OK, patch coming. It was not as invasive as I feared.
>>
> 
> Please add me to the CC list of your patches. I'll be available (in a
> couple weeks) to run regression tests for this Armada 38x DDR4 area.

The (one) patch is already in master. It would be great though, if you
could run some regression tests once time permits though.

Thanks,
Stefan