Weirdness of ofnode_get_addr_size()

[Jan Kiszka]

On 04.05.21 22:01, Dario Binacchi wrote:
> Hi Jan,
>
>> Il 04/05/2021 17:26 Simon Glass <sjg at chromium.org> ha scritto:
>>
>>
>> Hi Jan,
>>
>> On Sun, 2 May 2021 at 01:53, Jan Kiszka <jan.kiszka at web.de> wrote:
>>>
>>> Hi,
>>>
>>> I'm trying to make some sense of ofnode_get_addr_size() in order to fix
>>> [1] properly.
>>>
>>> First, the documentation if this functions says "This does no address
>>> translation". But the node-pointer path happily calls
>>> of_translate_address(), as the result of a6a45cd32539. For not
>>> offset-bases path, it calls fdtdec_get_addr_size() which does no
>>> translation.
>>>
>>> Related to [1]: The node-pointer path cleanly calls
>>> of_n_addr/size_cells() in order to retrieve the configured number of
>>> cells. But the offset-based path simply calls fdtdec_get_addr_size()
>>> which assumes that the number of cells is derived from the physical
>>> address width of that platform.
>>>
>>> So, what is that functions actually supposed to do?
>>
>> +Dario Binacchi +Mario Six
>>
>> I suspect the code has got ahead of the docs.
>>
>> d64b9cdcd47 fdt: translate address if #size-cells = <0>
>
> I submitted a patch to revert the commit:
> https://patchwork.ozlabs.org/project/uboot/patch/20210501150527.10273-6-dariobin@libero.it/
>

But of_translate_address is still part of ofnode_get_addr_size, just
called under less conditions.

And that is only one of the inconsistencies, at least as it looks to me.

Jan