The current situation:
# su --pty - kzak
$ ll $(tty)
crw--w---- 1 root tty 136, 9 Feb 23 11:53 /dev/pts/9
$ mesg
mesg: cannot open /dev/pts/9: Permission denied
the pseudo-terminal is still owned by the original user.
New version:
# su --pty - kzak
# ll $(tty)
crw--w---- 1 kzak tty 136, 9 Feb 23 11:56 /dev/pts/9
# mesg
is y
The patch follows login(1) to change the pty owner and group. It
follows "TTYPERM" and "TTYGROUP" from login.defs (or econf lib).
Signed-off-by: Karel Zak <kzak@redhat.com>
pam_set_item() man page:
PAM_TTY
The terminal name: prefixed by /dev/ if it is a device file;
for graphical, X-based, applications the value for this item
should be the $DISPLAY variable.
It seems for example pam_timestamp module is not robust enough to
differentiate between /dev/ and pty/0 and it assumes that '/' in the
path always means '/dev/' prefix ...
Fixes: https://github.com/karelzak/util-linux/issues/1242
Signed-off-by: Karel Zak <kzak@redhat.com>
When the requested shell matches the restricted shell, there is no reason
to issue a warning, since we will be doing precisely as requested.
Signed-off-by: Jouke Witteveen <j.witteveen@gmail.com>
Now the code is duplicate on many places, but all we usually need is to
remember child status. It seems good enough to have very simple
callback child_die() to inform application about a change.
The patch also add PID to all signal related callbacks.
Signed-off-by: Karel Zak <kzak@redhat.com>
With users coming from LDAP, it is often the case that the entry in LDAP
does not contain one or more attributes required by su or, because of
misconfigured access control rights, the attribute might not be readable
by the LDAP client. In that case, su just tells the user that the user
does not exist.
It might be more user-friendly to tell the user to check the user entry
for all required fields.
We use PAM and if pam_strerror() returns nothing we have no clue why
authentication failed. It's mistake to blame incorrect password if
there are many possible reason...
Addresses: https://github.com/karelzak/util-linux/issues/778
Signed-off-by: Karel Zak <kzak@redhat.com>
* use proper winsize rather than uninitialized variable (Oops...)
* set the current terminal to the raw mode
* disable ECHO for non-terminal execution to be compatible with
non-pty output
Addresses: https://github.com/karelzak/util-linux/issues/767
Signed-off-by: Karel Zak <kzak@redhat.com>
ENV_SUPATH and ENV_ROOTPATH are equivalent and ENV_ROOTPATH takes
precedence in both login and su. It makes no sense. More logical would be
precedence of ENV_SUPATH in su and ENV_ROOTPATH in login.
Signed-off-by: Stanislav Brabec <sbrabec@suse.cz>
su(1) documentation says:
/etc/default/su command specific logindef config file
/etc/login.defs global logindef config file
It indirectly indicates that /etc/default/su should take precedence
over /etc/login.defs.
But the reverse is true. It is not possible to define ENV_PATH in
/etc/login.defs and then make su specific customization in
/etc/default/su. We need to change read order to match the documented
behavior.
Signed-off-by: Stanislav Brabec <sbrabec@suse.cz>
man utmp:
String fields are terminated by a null byte ('\0') if they are shorter
than the size of the field.
Signed-off-by: Karel Zak <kzak@redhat.com>
* usable with --login to whitelist specified environment variables
* the list is ignored for the core variables like HOME, SHELL, USER,
LOGNAME and PATH (su --login always resets these variables)
Note that su(1) requires password and after successful authentication
user has full control over the session, so he can set arbitrary
environment variables. The whitelist makes things more user friendly
only.
The patch removes unnecessary optimization when allocate environ[]. It
seems better to keep all in glibc hands and just reset the environment
array only.
Addresses: https://github.com/karelzak/util-linux/issues/221
Signed-off-by: Karel Zak <kzak@redhat.com>
Let's make it possible to use debug.h without environment variables.
Suggested-by: J William Piggott <elseifthen@gmx.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
It seems better to deallocate logindefs.conf stuff in long time
running (=waiting) processes like login(1) and su(1).
Signed-off-by: Karel Zak <kzak@redhat.com>
We want to use waitpid() only when child is terminated or stopped to
pick up child status, otherwise PTY proxy has to be active. This is
difference between "su" and "su --pty". For "su" we keep parent all
time in waitpid().
It would be possible to use separate code based on signalfd_siginfo,
but it seems better to keep all this stuff on one place -- it means
wait_for_child().
Signed-off-by: Karel Zak <kzak@redhat.com>
Not sure why I have problem with this years ago for script(1), but it
seems .fd=-1 is really enough to the ignore the FD.
Reported-by: Vaclav Dolezal <vdolezal@redhat.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
The signal mask is used by pty_init_slave(), but it has never been
uninitialized before fork(), so child gets 0 as a mask :-(
Note that script(1) has no this issue because it opens signal-fd
before fork().
Reported-by: Vaclav Dolezal <vdolezal@redhat.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
This patch a little bit reorders signals initialization. The original
code unblocks SIGINT SIGQUIT before signal handler is set for the
signals. It means there is a small possible race.
It seems better to compose wanted mask, setup handlers and then
unblock all the wanted signals.
Signed-off-by: Karel Zak <kzak@redhat.com>
The patch from master branch, somehow lost during su refactoring
rebase.
Reported-by: Tobias Stöckmann <tobias@stoeckmann.org>
Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak
Rosen Penev
Jouke Witteveen
Thorsten Kukuk
Jakub Hrozek
Stanislav Brabec
Ruediger Meier
Samuel Thibault
Sami Kerola