On Tue, Feb 13, 2018 at 8:12 PM, Brendan Higgins
<[email protected]> wrote:
> On Tue, Feb 13, 2018 at 1:29 AM, Avi Fishman <[email protected]> wrote:
>> On Tue, Feb 13, 2018 at 10:30 AM, Arnd Bergmann <[email protected]> wrote:
>>> On Tue, Feb 6, 2018 at 12:57 AM, Brendan Higgins
>>> <[email protected]> wrote:
>>>> Add a common device tree for all Nuvoton NPCM750 BMCs and a board
>>>> specific device tree for the NPCM750 (Poleg) evaluation board.
>>>>
>>>> Signed-off-by: Brendan Higgins <[email protected]>
>>>> Reviewed-by: Tomer Maimon <[email protected]>
>>>> Reviewed-by: Avi Fishman <[email protected]>
>>>> Reviewed-by: Joel Stanley <[email protected]>
>>>> Reviewed-by: Rob Herring <[email protected]>
>>>> Tested-by: Tomer Maimon <[email protected]>
>>>> Tested-by: Avi Fishman <[email protected]>
>>> ...
>>>> + enable-method = "nuvoton,npcm7xx-smp";
>>>
>>> I see this has already been reviewed quite a bit, but I'm curious
>>> about the 'npcm7xx'
>>> part here. Shouldn't that be a real chip name rather than a wildcard?
>>
>> There is a family of npcm7xx, some with SMP and some without.
>> For those who has it, it is common for all to use the same
>> "nuvoton,npcm7xx-smp".
>
> I think Arnd is right. In this case it should be "nuvoton,npcm750-smp". We can
> use CPU_METHOD_OF_DECLARE to alias to as many compatible properties as needed.
>
> Although everything in the 7xx family currently is either single core or has
> the
> same SMP behavior, you could paint yourself into a corner if you were to make
> one SMP device in the 7xx family which behaves differently from all other
> devices and then need to come up with new compatible string names just for
> that.
Right, this is the reason for the very simple 'no wildcards in DT
identifiers' rule.
There simply isn't a reason to make an exception here.
Arnd
