On Mon, Sep 18 2023 at 21:32, Jason Gunthorpe wrote:
> On Tue, Sep 19, 2023 at 02:25:32AM +0200, Thomas Gleixner wrote:
>
>> > I don't think we need to try and fix that..
>>
>> We might want to detect it and yell about it, right?
>
> It strikes me as a good idea, yes. If it doesn't cost anything.
It should not be expensive in the interrupt allocation/deallocation
path, where hardware needs to be accessed anyway. So one extra read is
not the end of the world.
Thanks,
tglx
