Hi,

On jeu, sep 29, 2005, ??????? ??????? wrote:
> I have two vlan interfaces and one satellite interface.
> sputnik is satellite iface.
> net.ipv4.conf.sputnik.rp_filter shold be set to 0, which is done by
> this line in /etc/network/interfaces:
>  post-up /sbin/sysctl -w net.ipv4.conf.sputnik.rp_filter=0

 This is one ugly way of doing it, and should work.

> system was converted to using vlans a month ago, and today
> i was disappointed by a fact what rp_filter does not set to 0
> on 'ifup sputnik'.
> it turns out what /etc/network/if-up.d/ip  set this sysctl on it's own,
> and i need additional parameter in /etc/network/interfaces
> specifically to tell vlan's ip script to set rp_filter.

 The /etc/network/if-up.d/ip script will only change settings *IF* you
 set them.  "if [ -n "$IF_IP_PROXY_ARP" ]" means that the code will only
 run if something is written in IF_IP_PROXY_ARP, that is if you
 configured ip_proxy_arp in your interfaces file.

 However, please note that rp_filter only makes sense for IP interfaces.
 If you don't have any IP address on your interface, I suppose
 rp_filter won't work.

> I suggest what /etc/network/if-up.d/ip should act only on vlan interfaces,
> checking them in a manner if-pre-up.d/vlan script does.

 I think that /etc/network/if-up.d/ip applies in general to all people
 wanting to use proxy_arp and/or rp_filter.  This applies to people with
 DMZ or multiple internet connections for example, and these people
 don't necessarily use vlans.  Hence, I think this script should be in
 ifupdown, as the comment points out:
    # This should probably go into ifupdown
    # But usually only those with lots of interfaces (vlans) need these

 ... but this is a separate problem.

 Could you explain your original problem?  I don't think
 /etc/network/if-up.d/ip is messing with your "post-up" stanza, could
 you please the complete definition of your interface and run:
    cat /proc/sys/net/ipv4/conf/sputnik/rp_filter
 after ifuping your interface?

   Thanks,

-- 
Loïc Minier <[EMAIL PROTECTED]>

Reply via email to