On Mon, Dec 07, 2020 at 04:52:57PM -0700, Jens Axboe wrote:

On 12/4/20 12:13 PM, Sebastian Andrzej Siewior wrote:

quoted

Controllers with multiple queues have their IRQ-handelers pinned to a
CPU. The core shouldn't need to complete the request on a remote CPU.

Remove this case and always raise the softirq to complete the request.

I don't like this one at all, it'll add a softirq jump for the fast path
for eg nvme devices.

For the real fast path, that is either a polled queue or irq driven
queues that only map to a single CPU we are never reaching this code,
so I'm not too worried.  Not that I'd complain about numbers, preferably
in the commit log.