Dmitry Antipov

I've relaxed `liburing` requirements to >= 2.0, it should work as well.

NOTE: in the last version of the patch, `io_uring` is turned off by default even if supported, and it's required to call `event_base_new_with_config(cfg)` with enabled `event_config_set_flag(cfg, EVENT_BASE_FLAG_IO_URING)` flag.

NOTE: IIUC this requires the following kernel change as well: ``` diff --git a/io_uring/sqpoll.c b/io_uring/sqpoll.c index 559652380672..7fc1b388ed90 100644 --- a/io_uring/sqpoll.c +++ b/io_uring/sqpoll.c @@ -327,7 +327,7 @@ int io_sqpoll_wait_sq(struct io_ring_ctx *ctx)...

> it'd be nice if we have a io_uring_wait_cqes_timeout(ring, nr, timeout) What about making `io_uring_wait_cqes()` return `-ETIME` if no events has been completed or non-negative number of events has been...

> It already does that As of `liburing` 2.2, the manual page explicitly states that ``` On success io_uring_wait_cqes(3) returns 0 and the cqe_ptr parm is filled in. ```