Re: PROBLEM: csum_partial() / i386 does not handle unaligned address with empty region properly

[David S. Miller]

Andi Kleen writes:
 > AFAIK the standard kernel never passes zero length to csum_partial so there
 > is no bug in it. If IPVS does that it probably needs to be fixed.
 > I would report it to the IPVS maintainers.

Yes, better to put the rare check into the weird callers instead
of punishing everyone who does not allow this to occur by other
means already.

Later,
David S. Miller
[e-mail]