TCP checksum offload

TCP checksum offload: during the recent network protocol analysis, it was found that the WireShark displays some data packets sent by the local machine, incorrect Checksum, and prompts maybe caused by "TCP checksumoffload" Goole to search, you can change the ChecksumOffload settings in the system. It is a function of the network card. It can replace the system's TCP/IP protocol stack to calculate the TCP packet checksum. The Checksum Offload in the Window system is as follows (if supported by the NIC): My Network Places (right click Properties)-Local Area Connection (right click Properties)-General (click Configure ...) -After Advanced is enabled, you can see the Checksum Offload button next to it and change it to Rx TCP/IP Checksum on the right. There are four types of Checksum Offload settings: whether or not they are valid for Rx or Tx, or both. For Tx, after setting the Checksum Offload to be valid, the transfer layer of Windows will randomly fill in the TCP CheckSum, so the packet captured on the local machine is Bad Checksum. Then, the network adapter automatically calculates the correct verification code and sends it. Therefore, the other party still receives the correct TCP packet. For Rx, when the Checksum Offload is set to be effective, the NIC will fill in an NDIS_TCP_IP_CHECKSUM_PACKET_INFO structure and set the flag. If the flag fails for some reason, no flag is set, data verification is completed by the TCP/IP protocol stack in Windows. CheckSum Offload is actually a part of the work of the transport layer handed over to the hardware to save the system's CPU resources. Microsoft tests show that it can save up to 30% of CPU resources. Ibm aix documentation pointed out: for PCI interface Gigabit Nic is not as good as let the CPU above MHz to calculate the checksum, And the PCI-X Gigabit Nic can achieve line speed after enabling this, this saves CPU resources.