Under Unix (for example BSD Unix, MacOSX) the login program has the setuid bit set. In Linux login does not have the setuid bit set.
The questions is about /bin/login .
What are the differences between the two architectural decisions and why Linux doesn't have to do it? How does it solve the permissions' problem?
I've seen this as a CS questions somewhere and I can't figure out the answer. So please feel free to help.
Thanks in Advance!
In Linux you normally have a bunch of getty (text console) processes and/or a display manager (graphical) login process running with root privileges. When creating a user session, these are calling /bin/login, still with root privileges. login then drops the privileges by changing the UID/GID and execv-ing the login shell, or launching the graphical session.