I need to run a command with administrative privileges. Someone said I should run a command as root. How do I do this?
4 Answers
The main two commandline possibilities are:
- Use
su
and enter the root password when prompted. - Put
sudo
in front of the command, and enter your password when prompted.
Running a shell command as root
sudo (preferred when not running a graphical display)
This is the preferred method on most systems, including Ubuntu, Linux Mint, (arguably) Debian, and others. If you don't know a separate root password, use this method.
Sudo requires that you type your own password. (The purpose is to limit the damage if you leave your keyboard unattended and unlocked, and also to ensure that you really wish to run that command and it wasn't e.g. a typo.) It is often configured to not ask again for a few minutes so you can run several sudo
commands in succession.
Example:
sudo service apache restart
If you need to run several commands as root, prefix each of them with sudo
. Sometimes, it is more convenient to run an interactive shell as root. You can use sudo -i
for that:
$ sudo -i
# command 1
# command 2
...
# exit
Instead of sudo -i
, you can use sudo -s
. The difference is that -i
reinitializes the environment to sane defaults, whereas -s
uses your configuration files for better or for worse.
For more information, see the sudo website, or type man sudo
on your system. Sudo is very configurable; for example it can be configured to let a certain user only execute certain commands as root. Read the sudoers
man page for more information; use sudo visudo
to edit the sudoers file.
su
The su
command exists on most unix-like systems. It lets you run a command as another user, provided you know that user's password. When run with no user specified, su
will default to the root account.
Example:
su -c 'service apache restart'
The command to run must be passed using the -c
option. Note that you need quotes so that the command is not parsed by your shell, but passed intact to the root shell that su
runs.
To run multiple commands as root, it is more convenient to start an interactive shell.
$ su
# command 1
# command 2
...
# exit
On some systems, you need to be in group number 0 (called wheel
) to use su
. (The point is to limit the damage if the root password is accidentally leaked to someone.)
Logging in as root
If there is a root password set and you are in possession of it, you can simply type root
at the login prompt and enter the root password. Be very careful, and avoid running complex applications as root as they might do something you didn't intend. Logging in directly as root is mainly useful in emergency situations, such as disk failures or when you've locked yourself out of your account.
Single User Mode
Single user mode, or run-level 1, also gives you root privileges. This is intended primarily for emergency maintenance situations where booting into a multi-user run-level is not possible. You can boot into single user mode by passing single
or emergency
on the kernel command line. Note that booting into single-user mode is not the same as booting the system normally and logging in as root. Rather, the system will only start the services defined for run-level 1. Typically, this is the smallest number of services required to have a usable system.
You can also get to single user mode by using the telinit command: telinit 1
; however, this command requires you to already have gotten root privileges via some other method in order to run.
On many systems booting into single user mode will give the user access to a root shell without prompting for a password. Notably, systemd
-based systems will prompt you for the root password when you boot this way.
Other programs
Calife
Calife lets you run commands as another user by typing your own password, if authorized. It is similar to the much more widespread sudo (see above). Calife is more light-weight than sudo but also less configurable.
Op
Op lets you run commands as another user, including root. This not a full-blown tool to run arbitrary commands: you type op
followed by a mnemonic configured by the system administrator to run a specific command.
Super
Super lets you run commands as another user, including root. The command must have been allowed by the system administrator.
Running a graphical command as root
See also Wikipedia.
PolicyKit (preferred when using GNOME)
Simply prefix your desired command with the command pkexec
. Be aware that while this works in most cases, it does not work universally.
See man pkexec
for more information.
KdeSu, KdeSudo (preferred when using KDE)
kdesu
and kdesudo
are graphical front-ends to su
and sudo
respectively. They allow you to run X Window programs as root with no hassle. They are part of KDE. Type
kdesu -c 'command --option argument'
and enter the root password, or type
kdesudo -c 'command --option argument'
and enter your password (if authorized to run sudo
). If you check the “keep password” option in KdeSu, you will only have to type the root password once per login session.
Other programs
KtsussKtsuss (“keep the su simple, stupid”) is a graphical version of su.
BeesuBeesu is a graphical front-end to the su command that has replaced Gksu in Red Hat-based operating systems. It has been developed mainly for RHEL and Fedora.
Obsolete methods
gksu
and gksudo
gksu
and gksudo
are graphical front-ends to su
and sudo
respectively. They allow you to run X Window programs as root with no hassle. They are part of Gnome. Type
gksu command --option argument
and enter the root password, or type
gksudo command --option argument
and enter your password (if authorized to run sudo
).
gksu
and gksudo
are obsolete. They have been replaced by PolicyKit in GNOME, and many distributions (such as Ubuntu) no longer install them by default. You should not depend on them being available or working properly.
Manually via one of the shell-based methods
Use one of the methods in the "running a shell command as root section". You will need to ensure that neither the DISPLAY
environment variable nor the XAUTHORITY
environment get reset during the transition to root. This may require additional configuration of those methods that is outside the scope of this question.
Overall, this is a bad idea, mostly because graphical applications will read and write configuration files as root, and when you try to use those applications again as your normal user, those applications won't have permission to read their own configurations.
Editing a file as root
-
2It should be noted that
sudo
will only work if (1) it's installed and (2) if your use is in the sudoers file and it allowed to perform the operation.– nopcornCommented Aug 17, 2011 at 12:19 -
2this needs to be updated.
pkexec
is replacinggksu
/gksudo
in a lot of distributions nowadays.– strugeeCommented Nov 18, 2013 at 3:30 -
1@strugee - were you going to try and update this? The stuff under Redhat isn't correct either.– slm ♦Commented Dec 4, 2013 at 3:44
-
Use su
:
$ su -c command
or
$ su
# command
# exit
In either case, you will be prompted for the root
password. For more information, see the manual page.
Since the question was not Linux specific, here's how you achieve the same goal in Solaris 9+ (or Trusted Solaris 8):
Solaris, since version 9, has included a suite of tools affectionately referred to as RBAC, or Role Based Access Control.
The gist of RBAC is that through the granting of Authorizations and Rights, to Users and/or Role, or the granting of Roles to Users, you can create incredibly fine grained models for who can run what with which privileges.
Essentially, you identify authorization in /etc/security/auth_attr, then grant them to users or roles in /etc/user_attr.
You define profiles in /etc/security/prof_attr. You then associate commands with those profiles in /etc/security/exec_attr, followed by assigning those profiles to users in the /etc/user_attr file.
Once those things are done, you actually run pfexec <command>
to execute the command with privileged or authorizations that are granted to that user for that command.
The nice thing about RBAC is that there are no additional privileges granted to the command itself, or the user, only to the combination of user + command. So it's safer than making a binary +s, or just using sudo to make a user be able to execute pretty much anything. (I know that you can lock down sudo, but in my experience most people don't)
Another advantage of RBAC is that you can make root a role account, and assign that role to users who are able to become root with the 'su' command and the root password. The root user will also be able to log in in Single User Mode, which is better (in my opinion) than the Linux model where you can disable the root password passwd -d root
, or lock the root account passwd -l root
, both of which make logging in as root quite hard when something goes wrong.
Ben Rockwood has a great blog post on RBAC that can be read at Using RBAC on (Open)Solaris.
Or you can give some commands a superpower. I'm talking about special permissions in which commands are executed with the permission of owner.
Get the path of command by #which <command>
and set SUID or SGID for that path.
P.S. - SUID and SGID bits should be given with care. They can make your system insecure.