From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-fsdevel-owner@vger.kernel.org>
Received: from mail-oi0-f41.google.com ([209.85.218.41]:34515 "EHLO
	mail-oi0-f41.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751991AbcENOWa (ORCPT
	<rfc822;linux-fsdevel@vger.kernel.org>);
	Sat, 14 May 2016 10:22:30 -0400
Received: by mail-oi0-f41.google.com with SMTP id k142so211306589oib.1
        for <linux-fsdevel@vger.kernel.org>; Sat, 14 May 2016 07:22:30 -0700 (PDT)
MIME-Version: 1.0
In-Reply-To: <1463184168-3381-4-git-send-email-jonathan.derrick@intel.com>
References: <1463184168-3381-1-git-send-email-jonathan.derrick@intel.com>
	<1463184168-3381-4-git-send-email-jonathan.derrick@intel.com>
Date: Sat, 14 May 2016 07:22:29 -0700
Message-ID: <CAPcyv4ibBEh7xm8jUX-aq0a2ecqbZugDa_L1K=dbFujbKaDVVw@mail.gmail.com>
Subject: Re: [RFCv2 3/3] block: Introduce S_HIPRI inode flag
From: Dan Williams <dan.j.williams@intel.com>
To: Jon Derrick <jonathan.derrick@intel.com>
Cc: linux-block@vger.kernel.org, Jens Axboe <axboe@fb.com>,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	Jeff Moyer <jmoyer@redhat.com>,
	Stephen Bates <stephen.bates@microsemi.com>,
	Keith Busch <keith.busch@intel.com>,
	linux-nvme@lists.infradead.org,
	Christoph Hellwig <hch@infradead.org>,
	"Martin K. Petersen" <martin.petersen@oracle.com>
Content-Type: text/plain; charset=UTF-8
Sender: linux-fsdevel-owner@vger.kernel.org
List-ID: <linux-fsdevel.vger.kernel.org>

[ adding Martin, given his hint infrastructure that is in the works... ]


On Fri, May 13, 2016 at 5:02 PM, Jon Derrick <jonathan.derrick@intel.com> wrote:
> S_HIPRI is a hint that indicates the file (currently only block devices)
> is a high priority file. This hint allows direct-io to the block device
> to poll for completions if polling is available to the block device.
>
> The motivation for this patch comes from tiered caching solutions. A
> user may wish to have low-latency block devices act as a cache for
> higher-latency storage media.
>
> With the introduction of block polling, polling could be enabled on a
> queue of a block device. The preadv2/pwritev2 sets allowed a user to
> specify per-io polling, but removed the ability to poll per-queue.
>
> Instead of having a user modify their software to use preadv2/pwritev2,
> this patch allows a user to set S_HIPRI on a block device file to request
> all direct-io for this file to be polled.

Setting aside whether this should be a per-task ioprio hint vs a block
device inode hint, this single flag loses the granularity of whether
reads or writes are polled.  Low latency reads with posted writes
seems a valid configuration.