From mboxrd@z Thu Jan  1 00:00:00 1970
From: David Vrabel <david.vrabel@citrix.com>
Subject: Re: Migration memory corruption - PV backends need to
	quiesce
Date: Wed, 2 Jul 2014 11:03:54 +0100
Message-ID: <53B3D90A.4050601@citrix.com>
References: <53ADA10D.8030206@citrix.com>		
	<1404117494.3827.8.camel@kazak.uk.xensource.com>		
	<53B1324F.1090304@citrix.com>	
	<1404121988.5000.27.camel@kazak.uk.xensource.com>	
	<53B28D81.5060904@citrix.com>
	<1404295352.4776.1.camel@kazak.uk.xensource.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xen.org>
In-Reply-To: <1404295352.4776.1.camel@kazak.uk.xensource.com>
List-Unsubscribe: <http://lists.xen.org/cgi-bin/mailman/options/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xen.org>
List-Help: <mailto:xen-devel-request@lists.xen.org?subject=help>
List-Subscribe: <http://lists.xen.org/cgi-bin/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=subscribe>
Sender: xen-devel-bounces@lists.xen.org
Errors-To: xen-devel-bounces@lists.xen.org
To: Ian Campbell <Ian.Campbell@citrix.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>, Tim Deegan <tim@xen.org>, Xen-devel List <xen-devel@lists.xen.org>, Paul Durrant <Paul.Durrant@citrix.com>, Jan Beulich <JBeulich@suse.com>, Ian Jackson <Ian.Jackson@eu.citrix.com>
List-Id: xen-devel@lists.xenproject.org

On 02/07/14 11:02, Ian Campbell wrote:
> On Tue, 2014-07-01 at 11:29 +0100, David Vrabel wrote:
>> On 30/06/14 10:53, Ian Campbell wrote:
>>> On Mon, 2014-06-30 at 10:47 +0100, David Vrabel wrote:
>>>
>>>> Using persistent grants means that backend writes may be lost since the
>>>> memcpy in the backend does not update the dirty bitmap and guest writes
>>>> to the shared ring may make the backends response writes visible.  I
>>>> think we may need to disable persistent grant support from blkback until
>>>> a mechanism for updating the dirty log is in place.
>>>
>>> Yes, sadly I think you might be right here.
>>
>> The copy in the frontend from the persistently granted pages to the
>> pages in the bio dirties those pages so this is actually safe.
> 
> What about if the pause+copy happens after the I/O has completed
> (response on the ring) and the copy out to the bio pages? Meaning that
> on resume the copy would pick up the stale data from the persistent
> grant.
> 
> I think the answer is that the frontend won't consider things actually
> complete until after the copy and therefore that read will be reissued.

Yes.

David