Re: Question on mmap() in 64-bit OS

On Thu, 29 Nov 2007 10:41:33 +0100 Josef Moellers <josef.moellers@xxxxxxxxxxxxxxxxxxx> wrote:
| LaBird wrote:
|> Hi Phil,
|>
|> <phil-news-nospam@xxxxxxxx> wrote:
|>
|>>Come the day we actually need more than 48-bit addresses, there will also
|>>be a lot of structural changes, primarily affecting the kernel and how it
|>>handles things like virtual memory and simultaneous views of kernel space
|>>and user space (things are a whole lot easier if the kernel can at least
|>>have the whole kernel space and one whole VM in its view at once). Given
|>
|>
|> I agree with you. Perhaps with a large address space, the process space
|> should be a global one for all processes in execution, rather than each
|> process having its own process space.
|
| You mean, that each process should be able to peek into another process'
| address space?
| Think again: this is as non-secure as you can ever get.

The ability to peek across can still be controlled. The ability to manage
the swapping as one single address space requires less logic to handle it.
See Multics. But operating systems have gone beyond that need even before
Linux came into being (e.g. OS/MVS).

--
|---------------------------------------/----------------------------------|
| Phil Howard KA9WGN (ka9wgn.ham.org) / Do not send to the address below |
| first name lower case at ipal.net / spamtrap-2007-11-29-0836@xxxxxxxx |
|------------------------------------/-------------------------------------|
.