This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


RE: [Xen-devel] Page cache shared across domain

To: "Jacob Gorm Hansen" <jacobg@xxxxxxx>, "Mark Williamson" <mark.williamson@xxxxxxxxxxxx>
Subject: RE: [Xen-devel] Page cache shared across domain
From: "Ian Pratt" <m+Ian.Pratt@xxxxxxxxxxxx>
Date: Tue, 31 Jan 2006 10:34:11 -0000
Cc: xen-devel@xxxxxxxxxxxxxxxxxxx, Amit Patel <amitpate@xxxxxxxxx>
Delivery-date: Tue, 31 Jan 2006 10:43:51 +0000
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-id: Xen developer discussion <xen-devel.lists.xensource.com>
List-post: <mailto:xen-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
Thread-index: AcYmT6Si6NRr/BZ+SGeR14jq1BkHEAAAT19A
Thread-topic: [Xen-devel] Page cache shared across domain

> > Another approach is to use a shared interdomain ramdisk and 
> do Execute 
> > In Place (XIP) on the contents.  It's not a full shared page cache, 
> > but it gives you sharing for program text, etc.  I'm implementing 
> > infrastructure to support the dcssblk driver used on IBM zSeries 
> > mainframes, which will give us this functionality.
> Would it make sense to have a Read-Only page type in Xen (if 
> it's not already there), so that pages coming from the 
> blockdev-backend could be made RO by the backend, and the 
> domain would have to CoW them if it needed to modify the contents?

Xen is certainly able to enforce pages as RO, and there have been
several prototypes knocked up over the last couple years that exhibit
this kind of functionality. 

For PV guests its necessary to introduce a new kind of page fault that
requires them to copy-to-write (i.e. they must replace the imutable
'physical' page with one of their own and update all references). This
is exactly the same kind of OS extension that's required by embedded
folk to execute from flash memory (copy into SRAM on writes), so should
hopefully make it into mainstream linux at some point.

> Seems this would allow a shared buffer cache, and also open 
> the door to a voluntary page sharing mechanism.

Xen-devel mailing list

<Prev in Thread] Current Thread [Next in Thread>