On 5/10/05, kalkov <kalkotiv@xxxxxxxxx> wrote: > hi, > did u get what u wanted, please update. > > On 5/7/05, rajat swarup <rajats@xxxxxxxxx> wrote: > > On 5/7/05, kalkov <kalkotiv@xxxxxxxxx> wrote: > > > i think you should, as per my knowledge. GROUP, any inputs !!!! > > > > > > I am operating in the netfilter hook in the NF_IP_FORWARD stage. > > > > > > > > > > > > > > So would I need to fill in the MAC layer entries also into the buffer > > > > that I allocate if I need to send ICMP messages? > > > > I want to use icmp_send() method. > > > > > > I am not sure what "grouping" means but does it mean that I should > > take those entries from the arriving packets? > > Hi Kalkov, There were a couple of things. I was working in a Kernel module that was implementing Netfilter hooks, so the raw sockets scheme would not be helpful as creating (raw) sockets using socket() system call is done in userland. But since we are in kernel we cannot have socket() system calls. I figured that there were certain flaws in trying to construct an ICMP packet in the Netfilter hooks as the mac header needs to be populated for which you may need to send ARP requests. So we tried to make modifications in net/ipv4/ip_forward.c so that ICMP traceback could be done by calling icmp_send() method and populating the fields of the new packets (created by alloc_skb() ) by accessing fields using existing sk_buff structure (include/linux/skbuff.h) which the ip_forward function receives as parameter as it surely has all routing information and MAC header info. Please let me know if I am making any sense. Thanks and regards, Rajat. http://rajatswarup.blogspot.com/ -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
