> On May 20, 2024, at 3:33 AM, Alexander Kokorin <zukimorimk...@gmail.com> > wrote: > > We have noticed that when we receive TCP packets with the wrong > checksum from the internet, on > the receiving node it goes through, NIC compares the checksum, lets packet > further and increases the kernel counter for RX ERR. It doesn't make sense > as nothing can be done with such packets.
Don't be so sure. Years ago when I worked for another company, another team had a product that handled networking traffic. Their product dropped packets that had bad TCP checksums. There were certain features in some software that simply would not work. They spent months trying to figure out why the features only didn't work when traffic went through their product. Eventually they finally realized that there were always some TCP checksum errors when that software was used. They stopped dropping the packets with bad checksums and then it worked fine. Yes, there is (or at least was?) software that abused the TCP checksum to pass some other data through the connection. The point of the hardware checksum check is to allow software not not have to do it when it is good - to optimize the normal case. It is not to drop the bad packets at a lower level, hiding them from TCP. -- Mark Rustad (he/him), Ethernet Products Group, Intel Corporation _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel Ethernet, visit https://community.intel.com/t5/Ethernet-Products/bd-p/ethernet-products
