io-uring.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Pavel Begunkov <asml.silence@gmail.com>
To: Bob Liu <bob.liu@oracle.com>, Jens Axboe <axboe@kernel.dk>,
	io-uring@vger.kernel.org, linux-block@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 2/2] io_uring: io_queue_link*() right after submit
Date: Wed, 6 Nov 2019 12:31:32 +0300	[thread overview]
Message-ID: <2600cd84-a953-734c-1972-2c6ae0125ce5@gmail.com> (raw)
In-Reply-To: <81c2db6f-e262-328a-5917-71b30d9390a5@gmail.com>

On 11/6/2019 12:06 PM, Pavel Begunkov wrote:
> On 11/6/2019 11:36 AM, Bob Liu wrote:
>> On 11/6/19 5:22 AM, Pavel Begunkov wrote:
>>> After a call to io_submit_sqe(), it's already known whether it needs
>>> to queue a link or not. Do it there, as it's simplier and doesn't keep
>>> an extra variable across the loop.
>>>
>>> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
>>> ---
>>>  fs/io_uring.c | 22 ++++++++++------------
>>>  1 file changed, 10 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>>> index ebe2a4edd644..82c2da99cb5c 100644
>>> --- a/fs/io_uring.c
>>> +++ b/fs/io_uring.c
>>> @@ -2687,7 +2687,6 @@ static int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr,
>>>  	struct io_submit_state state, *statep = NULL;
>>>  	struct io_kiocb *link = NULL;
>>>  	struct io_kiocb *shadow_req = NULL;
>>> -	bool prev_was_link = false;
>>>  	int i, submitted = 0;
>>>  	bool mm_fault = false;
>>>  
>>> @@ -2710,17 +2709,6 @@ static int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr,
>>>  			}
>>>  		}
>>>  
>>> -		/*
>>> -		 * If previous wasn't linked and we have a linked command,
>>> -		 * that's the end of the chain. Submit the previous link.
>>> -		 */
>>> -		if (!prev_was_link && link) {
>>> -			io_queue_link_head(ctx, link, &link->submit, shadow_req);
>>> -			link = NULL;
>>> -			shadow_req = NULL;
>>> -		}
>>> -		prev_was_link = (s.sqe->flags & IOSQE_IO_LINK) != 0;
>>> -
>>>  		if (link && (s.sqe->flags & IOSQE_IO_DRAIN)) {
>>>  			if (!shadow_req) {
>>>  				shadow_req = io_get_req(ctx, NULL);
>>> @@ -2741,6 +2729,16 @@ static int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr,
>>>  		trace_io_uring_submit_sqe(ctx, s.sqe->user_data, true, async);
>>>  		io_submit_sqe(ctx, &s, statep, &link);
>>>  		submitted++;
>>> +
>>> +		/*
>>> +		 * If previous wasn't linked and we have a linked command,
>>> +		 * that's the end of the chain. Submit the previous link.
>>> +		 */
>>> +		if (!(s.sqe->flags & IOSQE_IO_LINK) && link) 
>> The behavior changed to 'current seq' instead of previous after dropping prev_was_link?
>>
> The old behaviour was to remember @prev_was_link for current sqe, and
> use at the beginning of the next iteration, where it becomes
> "previous/last sqe". See, prev_was_link was set after io_queue_link_head.
> 
> If @i is iteration idx, then timeline was:
> i:   sqe[i-1].is_link -> io_queue_link_head() # if (prev_was_link)
> i:   sqe[i].is_link = prev_was_link = (sqe[i].flags & LINK)
> i+1: sqe[i].is_link -> io_queue_link_head() # if (prev_was_link)
> i+1: sqe[i+1].is_link = ...
> 
> 
> After the change, it's done at the same loop iteration by swapping order
> of checking @prev_was_link and io_queue_link_head().
> 
> i:   sqe[i].is_link = ...
> i:   sqe[i].is_link -> io_queue_link_head()
> i+1: sqe[i+1].is_link = ...
> i+1: sqe[i+1].is_link -> io_queue_link_head()
> 
> Shouldn't change the behavior, if I'm not missing something.
> 
And the same goes for ordering with io_submit_sqe(), which assembles a link.

i:   prev_was_link = ... # for sqe[i]
i:   io_submit_sqe() # for sqe[i]
i+1: prev_was_link -> io_queue_link_head # for sqe[i]

after:
i:   io_submit_sqe() # for sqe[i]
i:   is_link = ... # for sqe[i]
i:   is_link -> io_queue_link_head # for sqe[i]

> 
>>> +			io_queue_link_head(ctx, link, &link->submit, shadow_req);
>>> +			link = NULL;
>>> +			shadow_req = NULL;
>>> +		}
>>>  	}
>>>  
>>>  	if (link)
>>>
>>

  reply	other threads:[~2019-11-06  9:31 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-11-05 21:22 [PATCH v2 0/2] cleanup of submission path Pavel Begunkov
2019-11-05 21:22 ` [PATCH v2 1/2] io_uring: Merge io_submit_sqes and io_ring_submit Pavel Begunkov
2019-11-06  8:57   ` Bob Liu
2019-11-06  9:07     ` Pavel Begunkov
2019-11-05 21:22 ` [PATCH v2 2/2] io_uring: io_queue_link*() right after submit Pavel Begunkov
2019-11-06  8:36   ` Bob Liu
2019-11-06  9:06     ` Pavel Begunkov
2019-11-06  9:31       ` Pavel Begunkov [this message]
2019-11-06 11:20         ` Bob Liu
2019-11-06 14:10 ` [PATCH v2 0/2] cleanup of submission path Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=2600cd84-a953-734c-1972-2c6ae0125ce5@gmail.com \
    --to=asml.silence@gmail.com \
    --cc=axboe@kernel.dk \
    --cc=bob.liu@oracle.com \
    --cc=io-uring@vger.kernel.org \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).