From mboxrd@z Thu Jan  1 00:00:00 1970
From: Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH] make blkfront/blktagp2 respect the
	elevator=xyz command line option
Date: Mon, 12 Oct 2009 10:34:33 +0200
Message-ID: <4AD2EA19.7010804@redhat.com>
References: <C6F54C18.17195%keir.fraser@eu.citrix.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xensource.com>
In-Reply-To: <C6F54C18.17195%keir.fraser@eu.citrix.com>
List-Unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xensource.com>
List-Help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-Subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
Sender: xen-devel-bounces@lists.xensource.com
Errors-To: xen-devel-bounces@lists.xensource.com
To: Keir Fraser <keir.fraser@eu.citrix.com>
Cc: "xen-devel@lists.xensource.com" <xen-devel@lists.xensource.com>, Daniel Stodden <Daniel.Stodden@citrix.com>
List-Id: xen-devel@lists.xenproject.org

>> For blkfront, I agree. Is there a use case for defaulting to something
>> different than a system default?
>
> Noone else does it, except maybe some old s390 drivers. We'll kill it then.

The rationale was that guest I/O delays depend on the I/O done by all 
other guests at the same time.  So, in theory there is no need to 
schedule it on the guest---assuming hypercalls are fast enough it should 
be passed down to the host immediately and left to be scheduled together 
with all the other host I/O.

However, actual benchmarks show that this does not always hold 
(sometimes disastrously) so killing the elevator_init call is safe and 
is fine by me.

Paolo