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] [PATCH] paging_enabled and non-HVM guests

To: Simon Kagstrom <simon.kagstrom@xxxxxx>
Subject: Re: [Xen-devel] [PATCH] paging_enabled and non-HVM guests
From: Hollis Blanchard <hollisb@xxxxxxxxxx>
Date: Wed, 10 May 2006 10:42:34 -0500
Cc: xen-devel list <xen-devel@xxxxxxxxxxxxxxxxxxx>
Delivery-date: Wed, 10 May 2006 08:42:40 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxx
In-reply-to: <1147272714.27510.8.camel@xxxxxxxxxxxxxxxxxxxxx>
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>
Organization: IBM Linux Technology Center
References: <87wtcv5t9e.wl%simon.kagstrom@xxxxxx> <1147204426.19485.67.camel@xxxxxxxxxxxxxxxxxxxxx> <87r7325s83.wl%simon.kagstrom@xxxxxx> <1147272714.27510.8.camel@xxxxxxxxxxxxxxxxxxxxx>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
On Wed, 2006-05-10 at 09:51 -0500, Hollis Blanchard wrote:
> On Wed, 2006-05-10 at 08:06 +0200, Simon Kagstrom wrote:
> > 
> > I won't argue for an incorrect fix, but as the code is right now it
> > segmentation faults because the virtual address passed to
> > 
> >         page = page_array[va >> PAGE_SHIFT] << PAGE_SHIFT;
> > 
> > (with libxc incorrectly believing paging is disabled) accesses outside
> > of page_array. I'll keep the patch privately for now since gdbserver
> > breaks without it. 
> Yes, and the reason is that page_array is supposed to be indexed with
> *physical* addresses. The current code thinks that paging is disabled
> (because CR0 is bad), assumes a virtual address is physical, and tries
> to index into the array with it.
> Pretty much every use of page_array needs to be abstracted so that it
> does the right thing on both HVM and normal guests (normal guests will
> have machine addresses in its page tables; HVM guests will have
> physical). It's very unfortunate that the people who worked on this
> code seem not to have tested or even thought about paravirtualized
> guests.

To elaborate on my previous mail, it's not just CR0/paging at fault. For
example, this use of page_array:
    l3p = l4[l4_table_offset(va)] >> PAGE_SHIFT;
    l3p = page_array[l3p];
in map_domain_va_64() is obviously incorrect for paravirtualized

Also, I noticed there's another place that already tests VGCF_HVM_GUEST
before paging_enabled(), which I guess is where you got the idea for
your patch.

Simon, would you care to submit the more complete patch?

Hollis Blanchard
IBM Linux Technology Center

Xen-devel mailing list