On Mon, Dec 22, 2003 at 01:27:49PM +0530, Arvind Narayanan wrote: The ONLY time I've ever seen anything like this it has been by design: the second user logs in before the first user logs out (on a different terminal that is still considered to be the physical console) and there is no "user stack" in the interest of enhancing security through simplicity and avoiding confusing shifts in permission partway through a login session. But in this case it reverts to root, not the previous user. Are you sure that the user doesn't have another login session at the same time? They are refcounted. michaelkjohnson "He that composes himself is wiser than he that composes a book." Linux Application Development -- Ben Franklin http://people.redhat.com/johnsonm/lad/

Have to agree with that, it only got worse with RH 9 and happened very regularly in public clusters. I think that whenever it happened some applications from previous user were hanging and the only way for a non privileged user to fix it is to reboot the machine in question (very annoying). My solution was to patch pam_console code to recognize a new option that forces takeover of ownership regardless of pam counting. That way I can force it for gdm logins. There was already some talk on this list about making sure the logout shuts down more gnome apps (like gconfd-2 and others), hopefully that will also help with this problem. Josko Alan Cox wrote:

On Tue, 23 Dec 2003, Alan Cox wrote: Why dont just approach the problem from a different angle? Rather than assigning ownership of the device to the local user, which just doesnt work for multi-head/multi-Xserver setups (first logged in user 'hogs' local devices), instead have pam_console assign console users a 'local' or 'console' group which then gives them access to devices which should only be avaiable from console (eg sound/webcam), all the other stuff (eg floppies) can have a generic site wide 'users' group ownership (and rw permission). This would work for multi-head setups and wouldnt require constant Ie: current: we futz ownership of local devices as and when users log in and out. problematic. instead: ownership of local devices statically assigned to a 'local' group with rw perms. pam_console then just adds group 'local' to the supplementary groups if the user is local. No potential for dangling ownerships or 'race to be first to login', everyone who is local gets access. ? regards, -- Paul Jakma paul(a)clubi.ie paul(a)jakma.org Key ID: 64A2FF6A warning: do not ever send email to spam(a)dishone.st Fortune: Nothing is impossible for the man who doesn't have to do it himself. -- A.H. Weiler