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;
Date:   Fri, 5 Jun 2020 16:02:44 -0700
From:   Andres Freund <andres@anarazel.de>
To:     Jens Axboe <axboe@kernel.dk>
Cc:     io-uring@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        akpm@linux-foundation.org
Subject: Re: [PATCHSET v5 0/12] Add support for async buffered reads
Message-ID: <20200605230244.u342xw4negic2nmr@alap3.anarazel.de>
References: <20200605202028.d57nklzpeolukni7@alap3.anarazel.de>
 <20200605203613.ogfilu2edcsfpme4@alap3.anarazel.de>
 <75bfe993-008d-71ce-7637-369f130bd984@kernel.dk>
 <3539a454-5321-0bdc-b59c-06f60cc64b56@kernel.dk>
 <34aadc75-5b8a-331e-e149-45e1547b543e@kernel.dk>
 <20200605223044.tnh7qsox7zg5uk53@alap3.anarazel.de>
 <20200605223635.6xesl7u4lxszvico@alap3.anarazel.de>
 <0e5b7a2d-eb0e-16e6-cc9f-c2ca5fe8cb92@kernel.dk>
 <20200605225429.s424lrrqhaseoa2h@alap3.anarazel.de>
 <0179d0bf-0b01-3f05-9221-654a00f0452d@kernel.dk>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <0179d0bf-0b01-3f05-9221-654a00f0452d@kernel.dk>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi,

On 2020-06-05 16:56:44 -0600, Jens Axboe wrote:
> On 6/5/20 4:54 PM, Andres Freund wrote:
> > On 2020-06-05 16:49:24 -0600, Jens Axboe wrote:
> >> Yes that's expected, if we have to fallback to ->readpage(), then it'll
> >> go to a worker. read-ahead is what drives the async nature of it, as we
> >> issue the range (plus more, depending on RA window) as read-ahead for
> >> the normal read, then wait for it.
> > 
> > But I assume async would still work for files with POSIX_FADV_RANDOM
> > set, or not? Assuming the system wide setting isn't zero, of course.
> 
> Yes it'll work if FADV_RANDOM is set.

Cool.


> But just not if read-ahead is totally disabled. I guess we could make
> that work too, though not sure that it's super important.

It's not from my end. I was just trying to check if the reduced
performance I saw was related to interactions between PG prefetching and
kernel level prefetching. And changing the /sys entry seemed easier than
making postgres set POSIX_FADV_RANDOM...

Greetings,

Andres Freund