Re: [PATCH v5 4/9] drivers: base cacheinfo: Add support for ACPI based firmware tables

From: Rafael J. Wysocki
Date: Tue Dec 12 2017 - 18:03:09 EST


On Tue, Dec 12, 2017 at 11:55 PM, Jeremy Linton <jeremy.linton@xxxxxxx> wrote:
> Hi,
>
>
> On 12/12/2017 11:25 AM, Rafael J. Wysocki wrote:
>>

[cut]

>>>>
>>>>
>>>>
>>>> What about converting this to using struct fwnode instead of adding
>>>> fields to it?
>>>
>>>
>>>
>>> I didn't really want to add another field here, but I've also pointed out
>>> how I thought converting it to a fwnode wasn't a good choice.
>>>
>>> https://lkml.org/lkml/2017/11/20/502
>>>
>>> Mostly because IMHO its even more misleading (lacking any
>>> fwnode_operations)
>>> than misusing the of_node as a void *.
>>
>>
>> I'm not sure what you mean.
>
>
> Converting the DT drivers/cacheinfo.c code to use a fwnode_handle is
> straightforward. But IMHO it doesn't solve the readability problem of either
> casting the ACPI/PPTT token directly to the resulting fwnode_handle *, or
> alternatively an actual fwnode_handle with bogus fwnode_operations to wrap
> that token.

I'm not talking about that at all.

>>
>> Anyway, the idea is to have one pointer in there instead of two that
>> cannot be used at the same time and there's no reason why of_node
>> should be special.
>
>
> Avoid two pointers for size, or readability? Because the last
> version had a union with of_node, which isn't strictly necessary as I can
> just cast the pptt token to of_node. There is exactly one line of code after
> that which uses the token and it doesn't care about type.

So call this field "token" or similar. Don't call it "of_node" and
don't introduce another "firmware_node" thing in addition to that.
That just is a mess, sorry.

Thanks,
Rafael