Re: [PATCHSET 0/2] io_uring support for linked timeouts

[Jens Axboe <axboe@kernel.dk>]

On 11/15/19 2:22 PM, Pavel Begunkov wrote:
> On 15/11/2019 22:34, Jens Axboe wrote:
>> How about something like this? Should work (and be valid) to have any
>> sequence of timeout links, as long as there's something in front of it.
>> Commit message has more details.
> 
> If you don't mind, I'll give a try rewriting this. A bit tight
> on time, so hopefully by this Sunday.
> 
> In any case, there are enough of edge cases, I need to spend some
> time to review and check it.

Of course, appreciate more eyes on this for sure. We'll see what happens
with 5.4 release, I suspect it won't happen until 11/24. In any case,
this is not really staged yet, just sitting in for-5.5/io_uring-post
as part of a series that'll likely go in after the initial merge.

> REQ1 -> LINKED_TIMEOUT -> REQ2 -> REQ3
> Is this a valid case? Just to check that I got this "can't have both" right.
> If no, why so? I think there are a lot of use cases for this.

Yes, it's valid. With the recently posted stuff, the only invalid case
is having a linked timeout as the first entry since that's nonsensical.
It has to be linked from a previous request. We no longer need to
restrict where the linked timeout appears otherwise.

-- 
Jens Axboe