On 10/13/20 2:43 AM, Pavel Begunkov wrote: > Move io_put_req() in io_poll_task_handler() from under spinlock. That's > a good rule to minimise time within spinlock sections, and > performance-wise it should affect only rare cases/slow-path. It's actually more important not to bounce on the lock, especially when the extra path isn't that long. You'll end up with the same hold time, just split, but the downside is that you'll dirty that lock twice instead of one. So while I think the patch is fine as it avoids REQ_F_COMP_LOCKED, I don't think the commit message really reflects reality. I'll adjust it a bit. -- Jens Axboe
