Re: io_uring and spurious wake-ups from eventfd

[Jens Axboe <axboe@kernel.dk>]

On 1/8/20 11:09 PM, Daurnimator wrote:
> On Thu, 9 Jan 2020 at 03:25, Jens Axboe <axboe@kernel.dk> wrote:
>> I see what you're saying, so essentially only trigger eventfd
>> notifications if the completions happen async. That does make a lot of
>> sense, and it would be cleaner than having to flag this per request as
>> well. I think we'd still need to make that opt-in as it changes the
>> behavior of it.
>>
>> The best way to do that would be to add IORING_REGISTER_EVENTFD_ASYNC or
>> something like that. Does the exact same thing as
>> IORING_REGISTER_EVENTFD, but only triggers it if completions happen
>> async.
>>
>> What do you think?
> 
> 
> Why would a new opcode be cleaner than using a flag for the existing
> EVENTFD opcode?

A few reasons I can think of:

1) We don't consume an IOSQE flag, which is a pretty sparse resource.
2) This is generally behavior where you either want one or the other,
   not a mix. Hence a general setup/modify flag makes more sense to me.

-- 
Jens Axboe