Take a look on TCP source code (fast and slow path). All that you need
is very well comented there.
I implemented a protocol called TCP-XM sometime ago (2.6.15.4 to be more
precise), and all the doubts and problems that I had (related to the
network subsystem) were answered there.
Mail me in private if you need a more detailed help.
DHAJOGLO escreveu:
I've hit a major wall while developing a small kernel module (2.6.18.8). Basically, my modules loads and handles ip protocol 253 (experimental). I have successfully received the socket buffer, read the contents and am now trying to turn it around and send back a reply. I'm not sure the proper method for this and am rather stuck. Any help would do. Here is a brief rundown:
get skb (OK)
read payload (OK)
send back out-----
skb->pkt_type = PACKET_OUTGOING;
iph = skb->nh.iph;
saddr = iph->saddr;
daddr = iph->daddr;
iph->saddr = daddr;
iph->daddr = saddr;
rt = (struct rtable *) skb->dst;
eth = (struct ethhdr *)skb->mac.raw;
memcpy(swap_mac,eth->h_dest,6);
memcpy(eth->h_dest,eth->h_source,6);
memcpy(eth->h_source,swap_mac,6);
NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL,rt->u.dst.dev,dst_output);
-----
This generates error:
ip_rt_bug: a.b.c.d -> e.f.g.h, eth0
if, instead of sending to NF_HOOK I send directly to dev:
dev_queue_xmit(skb_clone(skb, GFP_ATOMIC));
I get a buggy protocol error:
protocol 0800 is buggy, dev eth0
A tcpdump shows that the packet is being generated but its being dropped before being sent out.
I don't know how to figure out what to fix in my skb.
Thanks,
-dave
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
--
--
Best Regards
Alan Menegotto
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
