On 2/18/21 3:32 PM, Pavel Begunkov wrote: > Don't bother to take a ctx->refs for io_req_task_cancel() because it > take uring_lock before putting a request, and the context is promised to > stay alive until unlock happens. I agree this is fine, as we don't destroy the ctx ref before grabbing the lock. I'll defer this one for 5.13 though. Which is another way of saying "please resend it" when the merge window is over ;-) -- Jens Axboe
