In message <45D3A7FC-6E3F-11D7-9F28-000A957650EC@wasabisystems.com>,
Jason Thorpe  writes:
>
>On Saturday, April 12, 2003, at 06:56  AM, David Laight wrote:

>The kernel has always had "ovbcopy()" for overlapping bcopy.  So, I'd 
>think bcopy() should behave like memcpy(), i.e. not bother checking for 
>overlap.

This sounds like a bad idea: yes, the kernel has had ovbcopy(), but on
at least two of our ports, bcopy() has historically checked for
overlap and given userspace-bcopy() semantics.

Why change it (potentially introducing breakage) if it ain't broken,
and if we're meant to be using memcpy()/memmove() anyway?