Gentoo Sudo(ers) Guide
1.
About Sudo
Granting Permissions
The app-admin/sudo package allows the system administrator to grant
permission to other users to execute one or more applications they would
normally have no right to. Unlike using the setuid bit on these
applications sudo gives a more fine-grained control on who can
execute a certain command and when.
With sudo you can make a clear list who can execute a certain
application. If you would set the setuid bit, any user would be able to run this
application (or any user of a certain group, depending on the permissions used).
You can (and probably even should) require the user to provide a password when
he wants to execute the application and you can even fine-tune the permissions
based on the user's location: logged on from the system itself or through SSH
from a remote site.
Logging Activity
One additional advantage of sudo is that it can log any attempt
(successful or not) to run an application. This is very useful if you want to
track who made that one fatal mistake that took you 10 hours to fix :)
Configuring Sudo
The sudo configuration is managed by the /etc/sudoers file.
This file should never be edited through nano /etc/sudoers or
vim /etc/sudoers or any other editor you might like. When you want
to alter this file, you should use visudo.
This tool makes sure that no two system administrators are editing this file at
the same time, preserves the permissions on the file and performs some syntax
checking to make sure you make no fatal mistakes in the file.
About this Guide
This guide is meant as a quick introduction. The sudo package is a lot
more powerful than what is described in this guide. It has special features for
editing files as a different user (sudoedit), running from within a
script (so it can background, read the password from standard in instead of the
keyboard, ...), etc.
Please read the sudo and sudoers manual pages for more
information.
2.
Sudoers Syntax
Basic Syntax
The most difficult part of sudo is the /etc/sudoers syntax.
The basic syntax is like so:
Code Listing 2.1: Basic /etc/sudoers syntax |
user host = commands
|
This syntax tells sudo that the user, identified by user and
logged on through the system host can execute any of the commands listed
in commands as the root user. A more real-life example might make this
more clear: allow the user swift to execute emerge if he is logged
on from the system (not through SSH):
Code Listing 2.2: Live /etc/sudoers examples |
swift localhost = /usr/bin/emerge
|
A big warning is in place though: do not allow a user to run an
application that can allow people to elevate privileges. For instance, allowing
users to execute emerge as root can indeed grant them full root access
to the system because emerge can be manipulated to change the live file
system to the user's advantage. If you do not trust your sudo users,
don't grant them any rights.
The user name can also be substituted with a group name - in this case you
should start the group name with a % sign. For instance, to allow any
one in the wheel group to execute emerge:
Code Listing 2.3: Allowing the wheel group members to execute emerge |
%wheel localhost = /usr/bin/emerge
|
You can extend the line to allow for several commands (instead of making a
single entry for each command). For instance, to allow the same user to not only
run emerge but also ebuild and emerge-webrsync as root:
Code Listing 2.4: Multiple commands |
swift localhost = /usr/bin/emerge, /usr/bin/ebuild, /usr/sbin/emerge-webrsync
|
You can also specify a precise command and not only the tool itself. This is
useful to restrict the use of a certain tool to a specified set of command
options. The sudo tool allows shell-style wildcards (AKA meta or glob
characters) to be used in pathnames as well as command line arguments in the
sudoers file. Note that these are not regular expressions.
Let us put this to the test:
Code Listing 2.5: Attempt to update the system using sudo |
$ sudo emerge -uDN world
We trust you have received the usual lecture from the local System
Administrator. It usually boils down to these three things:
#1) Respect the privacy of others.
#2) Think before you type.
#3) With great power comes great responsibility.
Password:
|
The password that sudo requires is the user's own password. This is to
make sure that no terminal that you accidentally left open to others is abused
for malicious purposes.
You should know that sudo does not alter the ${PATH} variable: any
command you place after sudo is treated from your environment. If
you want the user to run a tool in for instance /sbin he should
provide the full path to sudo, like so:
Code Listing 2.6: Using the full path to a tool |
$ sudo /usr/sbin/emerge-webrsync
|
Using Aliases
In larger environments having to enter all users over and over again (or hosts,
or commands) can be a daunting task. To ease the administration of
/etc/sudoers you can define aliases. The format to declare
aliases is quite simple:
Code Listing 2.7: Declaring aliases in /etc/sudoers |
Host_Alias hostalias = hostname1, hostname2, ...
User_Alias useralias = user1, user2, ...
Cmnd_Alias cmndalias = command1, command2, ...
|
One alias that always works, for any position, is the ALL alias (to make
a good distinction between aliases and non-aliases it is recommended to use
capital letters for aliases). As you might undoubtedly have guessed, the
ALL alias is an alias to all possible settings.
A sample use of the ALL alias to allow any user to execute the
shutdown command if he is logged on locally is:
Code Listing 2.8: Allowing any user to execute shutdown |
ALL localhost = /sbin/shutdown
|
Another example is to allow the user swift to execute the emerge
command as root, regardless of where he is logged in from:
Code Listing 2.9: Allowing a user to run an application regardless of his location |
swift ALL = /usr/bin/emerge
|
More interesting is to define a set of users who can run software administrative
applications (such as emerge and ebuild) on the system and a group
of administrators who can change the password of any user, except root!
Code Listing 2.10: Using aliases for users and commands |
User_Alias SOFTWAREMAINTAINERS = swift, john, danny
User_Alias PASSWORDMAINTAINERS = swift, sysop
Cmnd_Alias SOFTWARECOMMANDS = /usr/bin/emerge, /usr/bin/ebuild
Cmnd_Alias PASSWORDCOMMANDS = /usr/bin/passwd [a-zA-Z0-9_-]*, !/usr/bin/passwd root
SOFTWAREMAINTAINERS localhost = SOFTWARECOMMANDS
PASSWORDMAINTAINERS localhost = PASSWORDCOMMANDS
|
Non-Root Execution
It is also possible to have a user run an application as a different, non-root
user. This can be very interesting if you run applications as a different user
(for instance apache for the web server) and want to allow certain users
to perform administrative steps as that user (like killing zombie processes).
Inside /etc/sudoers you list the user(s) in between
( and ) before the command listing:
Code Listing 2.11: Non-root execution syntax |
users hosts = (run-as) commands
|
For instance, to allow swift to run the kill tool as the
apache or gorg user:
Code Listing 2.12: Non-root execution example |
Cmnd_Alias KILL = /bin/kill, /usr/bin/pkill
swift ALL = (apache, gorg) KILL
|
With this set, the user can run sudo -u to select the user he wants
to run the application as:
Code Listing 2.13: Running pkill as the apache user |
$ sudo -u apache pkill apache
|
You can set an alias for the user to run an application as using the
Runas_Alias directive. Its use is identical to the other _Alias
directives we have seen before.
Passwords and Default Settings
By default, sudo asks the user to identify himself using his own
password. Once a password is entered, sudo remembers it for 5 minutes,
allowing the user to focus on his tasks and not repeatedly re-entering his
password.
Of course, this behavior can be changed: you can set the Defaults:
directive in /etc/sudoers to change the default behavior for a
user.
For instance, to change the default 5 minutes to 0 (never remember):
Code Listing 2.14: Changing the timeout value |
Defaults:swift timestamp_timeout=0
|
A setting of -1 would remember the password indefinitely (until the
system reboots).
A different setting would be to require the password of the user that the
command should be run as and not the users' personal password. This is
accomplished using runaspw. In the following example we
also set the number of retries (how many times the user can re-enter a password
before sudo fails) to 2 instead of the default 3:
Code Listing 2.15: Requiring the root password instead of the user's password |
Defaults:john runaspw, passwd_tries=2
|
Another interesting feature is to keep the DISPLAY variable set so that
you can execute graphical tools:
Code Listing 2.16: Keeping the DISPLAY variable alive |
Defaults:john env_keep=DISPLAY
|
You can change dozens of default settings using the Defaults: directive.
Fire up the sudo manual page and search for Defaults.
If you however want to allow a user to run a certain set of commands without
providing any password whatsoever, you need to start the commands with
NOPASSWD:, like so:
Code Listing 2.17: Allowing emerge to be ran as root without asking for a password |
swift localhost = NOPASSWD: /usr/bin/emerge
|
3.
Using Sudo
Listing Privileges
To inform yourself what your capabilities are, run sudo -l:
Code Listing 3.1: Listing capabilities |
$ sudo -l
User swift may run the following commands on this host:
(root) /usr/libexec/xfsm-shutdown-helper
(root) /usr/bin/emerge
(root) /usr/bin/passwd [a-zA-Z0-9_-]*
(root) !/usr/bin/passwd root
(apache) /usr/bin/pkill
(apache) /bin/kill
|
If you have any command in /etc/sudoers that does not require you
to enter a password, it will not require a password to list the entries either.
Otherwise you might be asked for your password if it isn't remembered.
Prolonging the Password Timeout
By default, if a user has entered his password to authenticate himself to
sudo, it is remembered for 5 minutes. If the user wants to prolong this
period, he can run sudo -v to reset the time stamp so that
it will take another 5 minutes before sudo asks for the password again.
Code Listing 3.2: Prolonging the password timeout |
$ sudo -v
|
The inverse is to kill the time stamp using sudo -k.
The contents of this document are licensed under the Creative Commons -
Attribution / Share Alike license.
|