On 7/4/07, Rajendra Stalekar <rajendra.stalekar@xxxxxxxxx> wrote:
Hi, I find that in the __netpoll_rx function in the netpoll.c , the iph->protocol is only checking for the UDP protocol in the transport layer for getting the true length of the frame. Why is this check not done for IPPROTO_TCP?
More important is why do you need even the UDP length at ethernet layer at all during receiving???? AFAICS, this is something for better performance. If you have a bad packet there is no point is pushing it to upper layers, because eventually they 'll also reject it. Doing it prematurely means less headache because of a supposedly bad packet. Thats why there is even a fast checksum calculation for the packet few lines below. OTOH, tcp layer is a connection oriented protocol. If you discard the packet this way here only, how are the pending socket queues coming to know that packet was a bad seed??? They need to be carried up to the TCP layer for checking because of the nature of the TCP protocol. UDP is a conncetion less and UDP layer don't care if there is no notification of packet dropping. I may be wrong , so feel free to correct me. Thanks ~psr -- play the game -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
