Re: [PATCH][2.1.71] BSD revoke() syscall

Matthew Kirkwood (
Thu, 11 Dec 1997 17:53:46 +0000 (GMT)

On Thu, 11 Dec 1997, Pavel Machek wrote:

> > User A dials up to a linux machine, runs a program which opens /dev/tty
> > and hangs around in the background. User A logs out, and User A's
> > program catches signals so it doesn't exit. User B logs in, and starts
> > to do work. User A's program can still execute a few limited tty
> > ioctl's, including ones which can reconfigure termios settings, such
> > as
> Why? Why allow program playing with someone else-s configuration? Is
> it needed for something?

If I'm not much mistaken, any file descriptors held by user A are marked
as "hungup" in the kernel, but not actually closed. Aside from the
hung-up bit (actually a security _improvement_) this is all standard

Which is why a good revoke() call would be useful.

Which is why someone other than me should write it :-)