Re: [PATCH v2 0/2] Optimise io_uring completion waiting

[Jens Axboe]

On 9/24/19 3:49 AM, Peter Zijlstra wrote:
> On Tue, Sep 24, 2019 at 10:36:28AM +0200, Jens Axboe wrote:
> 
>> +struct io_wait_queue {
>> +	struct wait_queue_entry wq;
>> +	struct io_ring_ctx *ctx;
>> +	struct task_struct *task;
> 
> wq.private is where the normal waitqueue stores the task pointer.
> 
> (I'm going to rename that)

If you do that, then we can just base the io_uring parts on that. 

>> +	unsigned to_wait;
>> +	unsigned nr_timeouts;
>> +};
>> +
>> +static inline bool io_should_wake(struct io_wait_queue *iowq)
>> +{
>> +	struct io_ring_ctx *ctx = iowq->ctx;
>> +
>> +	/*
>> +	 * Wake up if we have enough events, or if a timeout occured since we
>> +	 * started waiting. For timeouts, we always want to return to userspace,
>> +	 * regardless of event count.
>> +	 */
>> +	return io_cqring_events(ctx->rings) >= iowq->to_wait ||
>> +			atomic_read(&ctx->cq_timeouts) != iowq->nr_timeouts;
>> +}
>> +
>> +static int io_wake_function(struct wait_queue_entry *curr, unsigned int mode,
>> +			    int wake_flags, void *key)
>> +{
>> +	struct io_wait_queue *iowq = container_of(curr, struct io_wait_queue,
>> +							wq);
>> +
>> +	if (io_should_wake(iowq)) {
>> +		list_del_init(&curr->entry);
>> +		wake_up_process(iowq->task);
> 
> Then you can use autoremove_wake_function() here.
> 
>> +		return 1;
>> +	}
>> +
>> +	return -1;
>> +}
> 
> Ideally we'd get wait_event()'s @cond in a custom wake function. Then we
> can _always_ do this.
> 
> This is one I'd love to have lambda functions for. It would actually
> work with GCC nested functions, because the wake function will always be
> in scope, but we can't use those in the kernel for other reasons :/

I'll be happy enough if I can just call autoremove_wake_function(), I
think that will simplify the case enough for io_uring to not really make
me care too much about going further. I'll leave that to you, if you
have the desire :-)

-- 
Jens Axboe