Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 9/9] io_uring: pass submission ref to async
To:     Jens Axboe <axboe@kernel.dk>, io-uring@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <cover.1583078091.git.asml.silence@gmail.com>
 <29efa25e63ea86b9b038fff202a5f7423b5482c8.1583078091.git.asml.silence@gmail.com>
 <fb27a289-717c-b911-7981-db72cbc51c26@gmail.com>
 <fab1f954-98f0-3576-9142-966982988bc0@gmail.com>
 <902c439f-81ff-7593-e6b4-eee3f217c292@kernel.dk>
From:   Pavel Begunkov <asml.silence@gmail.com>
Message-ID: <19a08aaa-069b-27d1-1cb7-5033b74c2e78@gmail.com>
Date:   Mon, 2 Mar 2020 18:26:01 +0300
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101
 Thunderbird/68.5.0
MIME-Version: 1.0
In-Reply-To: <902c439f-81ff-7593-e6b4-eee3f217c292@kernel.dk>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 3/2/2020 6:12 PM, Jens Axboe wrote:
> On 3/2/20 8:08 AM, Pavel Begunkov wrote:
>> On 3/2/2020 12:39 AM, Pavel Begunkov wrote:
>>> On 01/03/2020 19:18, Pavel Begunkov wrote:
>>>> Currenlty, every async work handler accepts a submission reference,
>>>> which it should put. Also there is a reference grabbed in io_get_work()
>>>> and dropped in io_put_work(). This patch merge them together.
>>>>
>>>> - So, ownership of the submission reference passed to io-wq, and it'll
>>>> be put in io_put_work().
>>>> - io_get_put() doesn't take a ref now and so deleted.
>>>> - async handlers don't put the submission ref anymore.
>>>> - make cancellation bits of io-wq to call {get,put}_work() handlers
>>>
>>> Hmm, it makes them more like {init,fini}_work() and unbalanced/unpaired. May be
>>> no a desirable thing.
>>
>> Any objections against replacing {get,put}_work() with
>> io_finilise_work()? It will be called once and only once, and a work
>> must not go away until it happened. It will be enough for now, but not
>> sure whether you have some plans for this get/put pinning.
> 
> I have no further plans there, the get/put work only exist to ensure that
> the work item stays valid in case of cancelation lookups.

Great, it'll go into v2 as well.

-- 
Pavel Begunkov