From mboxrd@z Thu Jan  1 00:00:00 1970
From: =?windows-1252?Q?Roger_Pau_Monn=E9?= <roger.pau@citrix.com>
Subject: Re: [PATCH v5 1/4] xen: introduce a helper to allocate
 non-contiguous memory
Date: Mon, 11 May 2015 14:11:36 +0200
Message-ID: <55509C78.8060807@citrix.com>
References: <1431095651-25935-1-git-send-email-roger.pau@citrix.com>
	<1431095651-25935-2-git-send-email-roger.pau@citrix.com>
	<555079810200007800078B9D@mail.emea.novell.com>
	<55507BEA.7030003@citrix.com>
	<5550A3790200007800078DC1@mail.emea.novell.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xen.org>
Received: from mail6.bemta14.messagelabs.com ([193.109.254.103])
	by lists.xen.org with esmtp (Exim 4.72)
	(envelope-from <roger.pau@citrix.com>) id 1YrmYu-00022T-3K
	for xen-devel@lists.xenproject.org; Mon, 11 May 2015 12:11:56 +0000
In-Reply-To: <5550A3790200007800078DC1@mail.emea.novell.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: Jan Beulich <JBeulich@suse.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>, Tim Deegan <tim@xen.org>, Stefano Stabellini <stefano.stabellini@citrix.com>, Ian Campbell <ian.campbell@citrix.com>, xen-devel@lists.xenproject.org
List-Id: xen-devel@lists.xenproject.org

El 11/05/15 a les 12.41, Jan Beulich ha escrit:
>>>> On 11.05.15 at 11:52, <roger.pau@citrix.com> wrote:
>> El 11/05/15 a les 9.42, Jan Beulich ha escrit:
>>>>>> On 08.05.15 at 16:34, <roger.pau@citrix.com> wrote:
>>> There actually is another problem here: You must not free the
>>> pages before having torn down the mapping. An option (to
>>> avoid the need for a memory allocation) might be to zap the
>>> present bits (or whatever their equivalent on ARM) in a first
>>> pass (making sure to flush TLBs), free, and then vunmap().
>>
>> The aim of this is to catch reads/writes to those mappings while we are
>> tearing them down?
> 
> No, that wouldn't be worth it. The aim is to avoid cache attribute
> conflicts between the to be torn down mapping and possible
> immediate re-uses of the pages elsewhere.

After looking into this I think that this approach is complex to
implement on ARM. OTOH I've realized we could fetch all pages before
vunmap and add them to a temporary linked list (using pg->list) unless
there's a chance this page might be already member of a list. Then
vunmap and finally free all the pages. Does this sound sensible?

Roger.