On 08/02/16 14:48, Antoine Tenart wrote:
> On Mon, Feb 08, 2016 at 02:29:12PM +0000, Marc Zyngier wrote:
>> On 08/02/16 14:17, Antoine Tenart wrote:
>>> Thomas,
>>>
>>> On Mon, Feb 08, 2016 at 11:31:47AM +0100, Thomas Gleixner wrote:
>>>> On Mon, 8 Feb 2016, Antoine Tenart wrote:
>>>>> +static int alpine_msix_set_affinity(struct irq_data *irq_data,
>>>>> + const struct cpumask *mask, bool force)
>>>>> +{
>>>>> + int ret;
>>>>> +
>>>>> + ret = irq_chip_set_affinity_parent(irq_data, mask, force);
>>>>> + return ret == IRQ_SET_MASK_OK ? IRQ_SET_MASK_OK_DONE : ret;
>>>>
>>>> What's the point of this exercise? Why can't you just set the affinity
>>>> callback to irq_chip_set_affinity_parent() ?
>>>
>>> That's what done in irq-gic-v2m.c. Besides that, I see no point. I'll
>>> update for v2.
>>
>> That's because there is no need to do another compose_msi_msg/write_msg
>> in msi_domain_set_affinity() once the affinity has been updated at the
>> GIC level. Alternatively, updating the GIC driver to always return
>> IRQ_SET_MASK_OK_DONE would be perfectly acceptable.
>
> I'm using drivers/irqchip/irq-gic-v3.c which is indeed always returning
> IRQ_SET_MASK_OK. I'll make a new patch in the v2 of this series to
> return IRQ_SET_MASK_OK_DONE instead in the GIC driver (and then patch
> irq-gic-v2m.c).
/me puzzled. GICv3, but no ITS??? WTF???
M.
--
Jazz is not dead. It just smells funny...
