[libvirt] Fwd: Re: [PATCH] nwfilter: add XML attribute to control match target

[Stefan Berger]

  On 06/11/2010 10:53 AM, Daniel P. Berrange wrote:
> On Fri, Jun 11, 2010 at 10:41:50AM -0400, Stefan Berger wrote:
>    
>> This patch adds an optional XML attribute to a nwfilter rule to give the
>> user control over whether the rule is supposed to be using the match
>> target or not. A rule may now look like as follows with the nomatch
>> attribute either having value '1' or 'true' (case-insensitive).
>>
>> [...]
>> <rule action='accept' direction='in' nomatch='true'>
>>      
> Having inverted boolean flags is a little wierd. Can't this
> be written as match=false instead ?
>    
Ok, I'll invert the logic.

>    
>> <tcp srcmacaddr='1:2:3:4:5:6'
>>             srcipaddr='10.1.2.3' srcipmask='32'
>>             dscp='33'
>>             srcportstart='20' srcportend='21'
>>             dstportstart='100' dstportend='1111'/>
>> </rule>
>> [...]
>>
>> I am also extending the nwfilter schema and add this attribute to a test
>> case.
>>      
> I'm not sure I really understand what this is doing. Can you give a
> quick example of what the iptables command looks like, with and
> without the nomatch attribute set
>
>    
Background: I can specify rules so that the ssh port 22 (tcp) is 
reachable on my system and no other traffic is possible. If I don't use 
the --match target on this tcp port 22 rule, I could initiate traffic 
inside the VM from TCP port 22 (client port) towards an allowed target. 
If the --match NEW, ESTABLISHED target is added on incoming traffic, I 
get control over directionality due to the NEW. The outgoing direction 
will then have --match ESTABLISHED. Since the NEW will be missing on 
outgoing traffic, I won't be able to initiate that traffic.

In essence it's

iptables -p tcp --dport 22 -j ACCEPT

versus

iptables -p tcp --dport 22 -m state --state NEW, ESTABLISHED



> Regards,
> Daniel
>