Re: [PATCH net-next 6/8] net: gre: Implement LCO for GRE over IPv4

Alexander Duyck [off-list ref] wrote:

There isn't any need to add such an indication, nor do we need to
start bitflipping the return value from csum_fold in all cases.  I
think there was just some confusion about UDP checksums vs GRE or TCP
checksums.

Yeah. I think I finally got there. The naive software methods will never  
generate a true 0 unless everything was zero. Real one's complement  
machines did addition in terms of subtraction so that 1 + -1 would never  
produce a -0, only a normal 0. Of course a simple adder would produce a -0,  
making it impossible to get back to a normal 0.

I'd say we are better off keeping this simple.  The original patch
just needs to drop the check for the resultant checksum being 0 since
that is not needed for GRE.

I'm all in favor of simple. I had just started to worry about a possible  
change in behavior that might have interoperability problems with some  
implementations. I wonder if any implementation ever did the addition by  
subtraction, but also failed to make 0 compare equal to -0? I guess if they  
knew enough to do the former, they should not have blown the latter.