From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?ISO-8859-1?Q?Roger_Pau_Monn=E9?= Subject: Re: [PATCH] libxl: set permissions for xs frontend entry pointing to xs backend Date: Tue, 10 Sep 2013 17:19:27 +0200 Message-ID: <522F387F.4010200@citrix.com> References: <1378823007.10928.12.camel@kazak.uk.xensource.com> <1378824892-20789-1-git-send-email-roger.pau@citrix.com> <21039.14046.599923.505120@mariner.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from mail6.bemta14.messagelabs.com ([193.109.254.103]) by lists.xen.org with esmtp (Exim 4.72) (envelope-from ) id 1VJPj4-00051Z-21 for xen-devel@lists.xenproject.org; Tue, 10 Sep 2013 15:19:34 +0000 In-Reply-To: <21039.14046.599923.505120@mariner.uk.xensource.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Ian Jackson Cc: xen-devel@lists.xenproject.org, Ian Campbell , security@xen.org List-Id: xen-devel@lists.xenproject.org On 10/09/13 17:12, Ian Jackson wrote: > Roger Pau Monne writes ("[PATCH] libxl: set permissions for xs frontend entry pointing to xs backend"): >> libxl doesn't currently set the permissions of entries like: >> >> /local/domain//device///backend >> >> This allows the guest to change this xenstore entries to point to a >> different backend path, or to malicious xenstore path forged by the >> guest itself. libxl currently relies on this path being valid in order >> to perform the unplug of devices in libxl__devices_destroy, so we >> should prevent the guest from modifying this xenstore entry. > > Is it sufficient to set the permissions on "backend" - does that > prevent the guest deleting the whole subtree ? No, the guest can still delete the whole subtree, but it can not recreate it (because the parent directory /local/domain//device// is not writeable by the guest). > Really it would be better to make the unplug not depend on this path. > > This is a security issue, so CCing security@. It appears to have > been discovered in public on xen-devel, so shouldn't be embargoed. > > Ian. >