On Tue, May 28, 2002 at 08:06:33PM +0900, Jonathan Khoo spoke out thus: > Hi Sandip, > > In one of my modifications that I had tried, I did change the dst. What > I have done looks like this: > > src | dst src | dst > A | B A | B > ip6_output_finish ip6_rcv > C | D C | D > +------------->---------+ > > As you can see, I reverted the addresses back to the original before any > checksum calculation is done on the opposite host. Hence, I didn't see > the need to recalculate the header which, anyway, makes things way too > messy. Damn. IP6 scares me off. ;) So you have hooked into the xxx_output_finish and xxx_rcv routines on the hosts? That must have been painful. ;) However, now if your routine in the recipient host is receiving the packet correctly before changing the IP addresses, how can you trust the packet sniffer? Packet sniffers like tcpdump and ethereal examine the packet lower down the network layers. They take ethernet frames and dissect them to give you the IP packet. Ofcourse, your routine wouldn't have had an opportunity to change the IP addresses by then, and the checksum in the sniffer code should fail! > I even toyed around with the udp and tcp, but it gets murkier as I > trudged deeper into the codes. > > Is the checksum calculation dependent on dst? Where exactly is the dst > information (physically) in the skbuff? > Does the skbuff contains any dst information after it is being sent out > onto the wires? No checksum is only calculated on certain parts of the IP header and not the skbuff. -- Sandip Bhattacharya Mindframe Software Work: sandip @ mindsw.com, http://www.mindsw.com Play: sandipb @ bigfoot.com, http://www.sandipb.net ----------------------------------------------- -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
