Re: [Qemu-devel] [PATCH v6 05/13] register: QOMify

[Alistair Francis]

On Fri, Jun 10, 2016 at 3:55 AM, Peter Maydell <address@hidden> wrote:
> On 12 May 2016 at 23:46, Alistair Francis <address@hidden> wrote:
>> From: Peter Crosthwaite <address@hidden>
>>
>> QOMify registers as a child of TYPE_DEVICE. This allows registers to
>> define GPIOs.
>>
>> Define an init helper that will do QOM initialisation.
>
> You should just squash this down into patch 2.

It relies on some work from the previous two patches so it becomes a
pretty big patch if they all get squashed together. I'd rather leave
it separate like this.

>
>> Signed-off-by: Peter Crosthwaite <address@hidden>
>> Signed-off-by: Alistair Francis <address@hidden>
>> Reviewed-by: KONRAD Frederic <address@hidden>
>> ---
>
>>  /**
>> + * Initialize a register. GPIO's are setup as IOs to the specified device.
>> + * Fast paths for eligible registers are enabled.
>> + * @reg: Register to initialize
>> + */
>
> I can't work out what this documentation comment is trying to say.
> How can a register have a GPIO? What does a fast path do, what
> registers are elegible, why do I care whether they're enabled or not?

I have updated it to this, which removes the fast path (that was left
over) and explains more about GPIOs.

/**
 * Initialize a register. This will also setup any GPIO links which are used
 * to connect register updates in one device to other devices. Generally this
 * is useful for interrupt propagation.
 * @reg: Register to initialize
 */

Thanks,

Alistair

>
>> +
>> +void register_init(RegisterInfo *reg);
>> +
>> +/**
>>   * Memory API MMIO write handler that will write to a Register API register.
>>   *  _be for big endian variant and _le for little endian.
>>   * @opaque: RegisterInfo to write to
>> --
>> 2.7.4
>
> thanks
> -- PMM
>