> > The "implementations" for csum_partial_copy_fromuser in 2.1.x are all
> > currently just #define's over to csum_partial_copy. This is bad since
> > userspace can pass bogus pointers and no exception handling is done
> > resulting in an Oops. (I verified this with a small test program.)
>
> Yep. If someone cares to add the exceptions to it I'll add the patches.
> I've also had email from someone who claims to have an even fast
> csum routine.
Alan, I asked months ago if you guys needed somebody to do that. Now
I'm not bored enough to do it :-) But I may be bored in three or four
weeks if nobody has picked it up by then.
Tom.