Re: kcopy(9)

[David Holland <dholland-tech%netbsd.org@localhost>]

On Thu, Apr 06, 2017 at 06:11:20PM +0800, Paul Goyette wrote:
 > Man page for kcopy(9) makes no mention of what happens if the src and dst
 > memory regions overlap.  :)  While we probably don't expect it to happen,
 > it would be nice if the behavior were documented.  :)

The amd64 version supports it, and since it's a descendent of bcopy
it's probably meant to be supported. However, it's anyone's guess if
this is true on every port. Anyone feel like auditing? :-/

The sentence in there about "the fault handler" is also confusing. I'm
pretty sure it should say something like

   The implementation on ports that use a hook for handling fatal page
   faults must save and restore the hook value, since [...]

Comments anyone?

Plus it should explicitly say "The mechanism for generating the abort
is machine-dependent".

-- 
David A. Holland
dholland%netbsd.org@localhost