c. if your hardware is not capable to do TCP offloading, then the LONG way is to start from beginning of packet to end of packet, and calculate checksum.

I've tested that, my hardware apparently doesn't support TCP offloading.

d. if u keep the previous checksum, ie (th->checksum=0 is not needed at all), and u know exactly which byte u modify, then just do a xor of the modified byte with the previous checksum, and u get the new checksum.....this is the preferred and fastest way.

so in summary:

th->check = 0;
th->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
datalen, iph->protocol,
csum_partial((char *)th, datalen, 0));
skb->ip_summed = CHECKSUM_UNNECESSARY;

your setting of "0" above is correct, is because u want to calculate the checksum OVERALL....ie, packet start to packet end......demarcate correctly and u should get the answer....NOT csum_partial().....partial checksumming does not apply here.

--
Regards,
Peter Teoh

Well, the problem is now solved, it seems that the error was with the fact that some skbs are not linear and I was ignoring this fact.

Fixed it by the following:

tcph = tcp_hdr(skb);

if(skb_is_nonlinear(skb))

skb_linearize(skb)

tcplen = (skb->len - (iph->ihl << 2));/* tcplen is the length of the skb - the ip-header length */

tcph->check = 0;

tcph->check = tcp_v4_check(tcplen,

iph->saddr,

iph->daddr,

csum_partial((char*) tcph, tcplen, 0));

skb->ip_summed = CHECKSUM_NONE;/* in case another hardware has TCP offloading */

Thanks a lot, anyhow :D