Re: ip_route_me_harder question...

[Jon <jon_webb@binary-one.com>]

This is a good question. I've noticed that sometimes it's needed and 
sometimes it isn't. We implemented a routing protocol where all the 
packets were bumped up to userspace and after fiddling with the routing 
system for a long time, we ended up using some code from iproute2 to 
change the table before reinjecting the packet (unmodified). This 
alleviated our route_me_harder() issues. But another protocol we were 
looking at during our implementation used route_me_harder() and it 
seemed to work fine. Very confusing.

Regards,
Jon Webb
Computer Communications Research Group
University of California at Santa Cruz

Pablo Cingolani wrote:

>Refered to a question posted by "Zeeshan Ali <zeelists@yahoo.com>" a couple of days ago, I've suggested him to
>use ip_route_me_harder after changing a packet's saddr/daddr (at NF_IP_PRE_ROUTING).
>I've noticed that it doesn't work because after the hook changes the packet, ip_rcv_finish() calls
> skb->dst->input(skb) which seems to setted to something like 'dst_discard' (instead of 'ip_forward()' which is
>what I was expecting).
>Can someone give me a hint on what is wrong / what should be done to (correctly) change the routes 
>so that the packet is forwarded? 
>Thank you
>   Pablo
>
>P.S.: Packet's saddr is changed to a local one.
>
>  
>