On Fri, May 13 2022 at 16:45, Ricardo Neri wrote:
> On Fri, May 13, 2022 at 10:50:09PM +0200, Thomas Gleixner wrote:
>> > Also, if lapic_nmi_controller.irq_set_affinity() is NULL, then irq_chips
>> > INTEL-IR, AMD-IR, those using msi_domain_set_affinity() need to check for
>> > NULL.
>> > They currently unconditionally call the parent irq_chip's
>> > irq_set_affinity().
>> > I see that there is a irq_chip_set_affinity_parent() function. Perhaps it
>> > can
>> > be used for this check?
>>
>> Yes, this lacks obviously a NMI specific set_affinity callback and this
>> can be very trivial and does not have any of the complexity of interrupt
>> affinity assignment. First online CPU in the mask with a fallback to any
>> online CPU.
>
> Why would we need a fallback to any online CPU? Shouldn't it fail if it cannot
> find an online CPU in the mask?
Might as well fail. Let me think about it.
>> I did not claim that this is complete. This was for illustration.
>
> In the reworked patch, may I add a Co-developed-by with your name and your
> SOB?
Suggested-by is good enough.
Thanks,
tglx
