Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <489a3a2ded1daf962b8bfa3c936e20526b975d1a.camel@trillion01.com>
Subject: Re: [PATCH] io_uring: reduce latency by reissueing the operation
From:   Olivier Langlois <olivier@trillion01.com>
To:     Jens Axboe <axboe@kernel.dk>,
        Pavel Begunkov <asml.silence@gmail.com>,
        io-uring@vger.kernel.org, linux-kernel@vger.kernel.org
Date:   Fri, 18 Jun 2021 17:38:11 -0400
In-Reply-To: <ff0aab90-9aef-4ec6-f00f-89d4ffa21ef6@kernel.dk>
References: <60c13bec.1c69fb81.f2c1e.6444SMTPIN_ADDED_MISSING@mx.google.com>
         <ff0aab90-9aef-4ec6-f00f-89d4ffa21ef6@kernel.dk>
Organization: Trillion01 Inc
Content-Type: text/plain; charset="ISO-8859-1"
User-Agent: Evolution 3.40.2 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Wed, 2021-06-16 at 06:48 -0600, Jens Axboe wrote:
> On 6/9/21 4:08 PM, Olivier Langlois wrote:
> > It is quite frequent that when an operation fails and returns
> > EAGAIN,
> > the data becomes available between that failure and the call to
> > vfs_poll() done by io_arm_poll_handler().
> > 
> > Detecting the situation and reissuing the operation is much faster
> > than going ahead and push the operation to the io-wq.
> 
> I think this is obviously the right thing to do, but I'm not too
> crazy
> about the 'ret' pointer passed in. We could either add a proper
> return
> type instead of the bool and use that, or put the poll-or-queue-async
> into a helper that then only needs a bool return, and use that return
> value for whether to re-issue or not.
> 
> Care to send an updated variant?
> 
No problem!

It is going to be pleasure to send an updated version with the
requested change!

I will take that opportunity to try out my new patch sending setup ;-)