WARNING - OLD ARCHIVES

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/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-api

RE: [Xen-API] XCP antispoofing (idea and working patch)

To: 'George Shuklin' <george.shuklin@xxxxxxxxx>, "xen-api@xxxxxxxxxxxxxxxxxxx" <xen-api@xxxxxxxxxxxxxxxxxxx>
Subject: RE: [Xen-API] XCP antispoofing (idea and working patch)
From: Dave Scott <Dave.Scott@xxxxxxxxxxxxx>
Date: Wed, 29 Dec 2010 10:38:24 +0000
Accept-language: en-US
Acceptlanguage: en-US
Cc: Alex Zeffertt <Alex.Zeffertt@xxxxxxxxxxxxx>, Rob Hoes <Rob.Hoes@xxxxxxxxxx>
Delivery-date: Wed, 29 Dec 2010 02:38:40 -0800
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <1293417582.18543.37.camel@xxxxxxxxxxxxxxxx>
List-help: <mailto:xen-api-request@lists.xensource.com?subject=help>
List-id: Discussion of API issues surrounding Xen <xen-api.lists.xensource.com>
List-post: <mailto:xen-api@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-api>, <mailto:xen-api-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-api>, <mailto:xen-api-request@lists.xensource.com?subject=unsubscribe>
References: <1293417582.18543.37.camel@xxxxxxxxxxxxxxxx>
Sender: xen-api-bounces@xxxxxxxxxxxxxxxxxxx
Thread-index: AculbzvXGpek0flHR2GBYC+kAYJGIwB0r0hw
Thread-topic: [Xen-API] XCP antispoofing (idea and working patch)
Hi George,

I've cc:d Alex and Rob, who both might have useful thoughts on this.

> Good day.
> 
> I like to raise a question which bother me (and not only me) long time.

It bothers me too :-)

> How we can protect one guest form malicious actions from other guest?
> One of the simplest way to break network is spoofing (usurping) some
> IP/mac addresses on network by abusing ARP or (even) simply assign
> wrong IP to guest machine network interface.
> 
> Classic Xen shipped with antispoofing scripts, but XCP is not.
> 
> What problem we shall solve for quality antispoofing?
> 
> Restriction:
> 1) Shall be configurable (some interface must be restricted, some not)
> 2) must be applying on EVERY host in the pool
> 3) Shall be removed when VM is down or mirgrated out (do not disturb
> any other VM)
> 4) Shall survive reboot (when dom ID and vif interface int dom0
> changing)
> 5) Shall survive migration
> 6) shall restrict both IP and mac usage.

I think these requirements are sensible.

> I create some simple patch to /etc/xensource/scripts/vif (it called
> every time new vif initializing or deinitializing).

Looks good...

> It depends on openvswitch (and does not work with classic brtools),

This is fine.

> using /etc/xensource/ip_restriction.conf with following syntax:
> 
> vif-uuid ip
> (one per line), f.e.
> 25ca3ed1-06ea-5f58-9283-f2af3e8b373e 192.168.1.144
> 
> This file must be identical on every host in the pool.
> 
> If vif uuid is not listed it this file, it silently do nothing. If
> openvswitch is not used, it silently do nothing.
> 
> If IP (right now only ipv4) is assigned to vif-uuid, that vif is
> restricted to it mac (autodetected via xenstore) too.
> 
> When vif is down (vm down, migrated or rebooted) it clear all
> restriction on openvswitch port right before port unplugging.

Sounds good.

> The main concern is usage of configuration file instead xapi. I think
> it
> must be placed somewhere to vif's other config (but it far beyond my
> knowledge, and I except some help with this). xapi shall place
> ip_restriction to other config and copy it to xenstore for vif to be
> readable from scripts/vif script.

This part needs a bit more thought. I was wondering if we wanted to have 
first-class fields for this kind of VIF/switch configuration... something like

Network.default_secure: bool -- if true then unconfigured VIFs have no access, 
otherwise they have full access
VIF.allowed_addresses: some encoding of values like
  Some (list of (MAC, IPv4) pairs) -- specific addresses to limit the VIF to. 
Probably need '*, *'
  None -- no configuration

What do you think, Rob/ Alex?

Also adding explicit openflow rules will conflict with any attached openflow 
controller (eg nox) so we should make it an explicit choice: disable these 
rules and enable a controller, or disable a controller and enable these rules.

Cheers,
Dave

> 
> This patch is against XCP 0.5 (not 1.0b)
> 

_______________________________________________
xen-api mailing list
xen-api@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/mailman/listinfo/xen-api
<Prev in Thread] Current Thread [Next in Thread>