On 31/05/2017 19:34, Bjorn Helgaas wrote:
> On Wed, May 31, 2017 at 04:18:44PM +0200, Mason wrote:
>
>> For the record, below is the patch I had in mind:
>>
>> diff --git a/kernel/irq/msi.c b/kernel/irq/msi.c
>> index 8a3e872798f3..edfc95575a37 100644
>> --- a/kernel/irq/msi.c
>> +++ b/kernel/irq/msi.c
>> @@ -91,9 +91,11 @@ int msi_domain_set_affinity(struct irq_data *irq_data,
>> {
>> struct irq_data *parent = irq_data->parent_data;
>> struct msi_msg msg;
>> - int ret;
>> + int ret = -EINVAL;
>> +
>> + if (parent->chip->irq_set_affinity)
>> + ret = parent->chip->irq_set_affinity(parent, mask, force);
>>
>> - ret = parent->chip->irq_set_affinity(parent, mask, force);
>> if (ret >= 0 && ret != IRQ_SET_MASK_OK_DONE) {
>> BUG_ON(irq_chip_compose_msi_msg(irq_data, &msg));
>> irq_chip_write_msi_msg(irq_data, &msg);
>>
>>
>>
>> Then, it would be safe to remove "do-nothing" .irq_set_affinity
>> callbacks in drivers/pci/host
>
> I assume it's obvious that nobody else is going to take this and run
> with it, so nothing will happen with this unless you finish it up by
> doing the cleanup (removing the "do-nothing" callbacks) and adding a
> changelog and signed-off-by.
Smirk.
> This would be more an IRQ patch than a PCI patch, but if I were
> reviewing it, I would look for assurance that *all* the no-op
> .irq_set_affinity callbacks were cleaned up, not just those in
> drivers/pci/host.
Are you saying the patch is *wrong* if not all "do-nothing"
callbacks are cleaned up?
Regards.
