Jason Thorpe <thorpej@nas.nasa.gov> writes:

>  > Is somebody working on this?
>  > Would it work to borrow from FreeBSD code (which
>  > doesn't map shm into kernel space)?
> 
> I think it might be worth pursuing something like they have done, yes.
> Personally, I see zero reason why it needs to be mapped into kernel
> space in the first place.

There is also a bug in the shm code in 1.2G, as near as I can tell.
When I remove a segment, my machine crashes.  This makes shutting down
while running Postgres a real pain...

--Michael