On Tue, 13 Jun 2000, Neil Moore wrote:
> James Sutherland writes:
> > In a normal workstation, that is exactly equivalent to losing the machine!
> > Admittedly, if the machine is a network server, it will continue to
> > perform that function until a scheduled reboot.
> >
> > One objection I have to the microkernel approach is the need to give
> > userspace programs hardware access. That opens up a whole new can of
> > worms; just saying "userspace things cannot touch hardware" is a lot
> > simpler.
>
> Except that's not true in Linux. There exist /dev/kmem, /proc/kcore,
> iopl(), and ioperm().
OK, userspace things CAN go monkeying with hardware things if they really
want. However, not having them do that normally makes life much easier :-)
There are a couple of fairly straightforward ways for a root process to
get more or less direct hardware access if it really needs it, but if
every driver were implemented in this way, things would be a truly
horrible mess. And performance would have disappeared. And Linus wouldn't
be happy :-)
Also, on a secure system, you would tend to disable these interfaces
anyway, IIRC; see the thread about disabling kernel module loading etc.
James.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:29 EST