On 24/02/2020 18:27, Jens Axboe wrote:
> On 2/24/20 2:15 AM, Pavel Begunkov wrote:
>> It seems no opcode may return -EAGAIN for non-blocking case and expect
>> to be reissued. Remove retry code from io_wq_submit_work().
> 
> There's actually a comment right there on how that's possible :-)

Yeah, I saw it and understand the motive, and how it may happen, but can't
find a line, which can actually return -EAGAIN. Could you please point to an
example?

E.g. I suppose for io_read() it may happen in call_read_iter(), but its result
(i.e. res2) will be written to cqe, but not returned.
> Normally, for block IO, we can wait for request slots if we run out.
> For polled IO, that isn't possible since the task itself is the one
> that will find completions and hence free request slots as well. If
> the submitting task is allowed to sleep waiting for requests that it
> itself are supposed to find and complete, then we'd hang. Hence we
> return -EAGAIN for that case, and have no other choice for polled
> IO than to retry.


-- 
Pavel Begunkov