13 KiB
SELinux
Introduction
SELinux is a mandatory access control (MAC) system on Linux which adds a fine granular permission system for access to all resources on the system such as files, devices, networks and inter-process communication.
The most important questions are answered briefly in the FAQ of the SELinux Project. For more details on SELinux and how to actually use and administrate it on your systems have a look at Red Hat Enterprise Linux 7 - SELinux User's and Administrator's Guide. For an simplified (and funny) introduction download the SELinux Coloring Book.
This documentation will use a similar format like the SELinux User's and Administrator's Guide.
Policy
Icinga 2 is providing its own SELinux Policy. At the moment it is not upstreamed to the reference policy because it is under development. Target of the development is a policy package for Red Hat Enterprise Linux 7 and its derivates running the targeted policy which confines Icinga2 with all features and all checks executed.
Installation
Later the policy will be installed by a seperate package and this section will be removed. Now it describes the installation to support development and testing. It assumes that Icinga 2 is already installed from packages and running on the system.
The policy package will run the daemon in a permissive domain so nothing will be denied also if the system runs in enforcing mode, so please make sure to run the system in this mode.
# sestatus
SELinux status: enabled
SELinuxfs mount: /sys/fs/selinux
SELinux root directory: /etc/selinux
Loaded policy name: targeted
Current mode: enforcing
Mode from config file: enforcing
Policy MLS status: enabled
Policy deny_unknown status: allowed
Max kernel policy version: 28
You can change the configured mode by editing /etc/selinux/config
and the current mode by executing setenforce 0
.
As a prerequisite install the git
, selinux-policy-devel
and audit
package. Enable and start the audit daemon afterwards.
# yum install git selinux-policy-devel audit
# systemctl enable auditd.service
# systemctl start auditd.service
After that clone the icinga2 git repository and checkout the feature branch.
# git clone git://git.icinga.org/icinga2.git
# cd icinga2/
# git checkout feature/rpm-selinux-8332
To create and install the policy package run the installation script which also labels the resources. (The script assumes Icinga 2 was started once after system startup, the labeling of the port will only happen once and fail later on.)
# cd tools/selinux/
# ./icinga.sh
Some changes to the systemd scripts are also required to handle file contexts correctly. This is at the moment only included in the feature branch, so it has to be copied manually.
# cp ../../etc/initsystem/{prepare-dirs,safe-reload} /usr/lib/icinga2/
After that restart Icinga 2 and verify it running in its own domain icinga2_t
.
# systemctl restart icinga2.service
# ps -eZ | grep icinga2
system_u:system_r:icinga2_t:s0 2825 ? 00:00:00 icinga2
General
When the SELinux policy package for Icinga 2 is installed, the Icinga 2 daemon (icinga2) runs in its own domain icinga2_t
and is separated from other confined services.
Files have to be labeled correctly for allowing icinga2 access to it. For example it writes to its own log files labeled icinga2_log_t
. Also the API port is labeled icinga_port_t
and icinga2 is allowed to manage it. Furthermore icinga2 can open high ports and unix sockets to connect to databases and features like graphite. It executes the nagios plugins and transitions to their context if those are labeled for example nagios_services_plugin_exec_t
or nagios_system_plugin_exec_t
.
Additional the Apache webserver is allowed to connect to the Command pipe of Icinga 2 to allow web interfaces sending commands to icinga2. This will perhaps change later on while investigating Icinga Web 2 for SELinux!
Types
The command pipe is labeled icinga2_command_t
and other services can request access to it by using the interface icinga2_send_commands
.
The nagios plugins use their own contexts and icinga2 will transition to it. This means plugins have to be labeled correctly for their required permissions. The plugins installed from package should have set their permissions by the corresponding policy module and you can restore them using restorecon -R -v /usr/lib64/nagios/plugins/
. To label your own plugins use chcon -t type /path/to/plugin
, for the type have a look at table below.
Type | Domain | Use case | Provided by policy package |
---|---|---|---|
nagios_admin_plugin_exec_t | nagios_admin_plugin_t | Plugins which require require read access on all file attributes | nagios |
nagios_checkdisk_plugin_exec_t | nagios_checkdisk_plugin_t | Plugins which require read access to all filesystem attributes | nagios |
nagios_mail_plugin_exec_t | nagios_mail_plugin_t | Plugins which access the local mail service | nagios |
nagios_services_plugin_exec_t | nagios_services_plugin_t | Plugins monitoring network services | nagios |
nagios_system_plugin_exec_t | nagios_system_plugin_t | Plugins checking local system state | nagios |
nagios_unconfined_plugin_exec_t | nagios_unconfined_plugin_t | Plugins running without confinement | nagios |
nagios_eventhandler_plugin_exec_t | nagios_eventhandler_plugin_t | Eventhandler (actually running unconfined) | nagios |
nagios_openshift_plugin_exec_t | nagios_openshift_plugin_t | Plugins monitoring openshift | nagios |
nagios_notification_plugin_exec_t | nagios_notification_plugin_t | Notification commands | icinga (will be moved later) |
If one of those plugin domains causes problems you can set it to permissive by executing semanage permissive -a domain
.
The policy provides a role icinga2adm_r
for confining an user which enables an administrative user managing only Icinga 2 on the system. This user will also execute the plugins in their domain instead of the users one, so you can verify their execution with the same restrictions like they have when executed by icinga2.
Booleans
SELinux is based on the least level of access required for a service to run. Using booleans you can grant more access in a defined way. The Icinga 2 policy package provides the following booleans.
icinga2_can_connect_all
Having this boolean enabled allows icinga2 to connect to all ports. This can be neccesary if you use features which connect to unconfined services.
Configuration Examples
Confining a plugin
Download and install a plugin, for example check_mysql_health.
# wget http://labs.consol.de/download/shinken-nagios-plugins/check_mysql_health-2.1.9.2.tar.gz
# tar xvzf check_mysql_health-2.1.9.2.tar.gz
# cd check_mysql_health-2.1.9.2/
# ./configure --libexecdir /usr/lib64/nagios/plugins
# make
# make install
It is labeled nagios_unconfined_plugins_exec_t
by default, so it runs without restrictions.
# ls -lZ /usr/lib64/nagios/plugins/check_mysql_health
-rwxr-xr-x. root root system_u:object_r:nagios_unconfined_plugin_exec_t:s0 /usr/lib64/nagios/plugins/check_mysql_health
In this case the plugin is monitoring a service, so it should be labeled nagios_services_plugin_exec_t
to restrict its permissions.
# chcon -t nagios_services_plugin_exec_t /usr/lib64/nagios/plugins/check_mysql_health
# ls -lZ /usr/lib64/nagios/plugins/check_mysql_health
-rwxr-xr-x. root root system_u:object_r:nagios_services_plugin_exec_t:s0 /usr/lib64/nagios/plugins/check_mysql_health
The plugin still runs fine but if someone changes the script to do weird stuff it will fail to do so.
Allow icinga to connect to all ports.
You are running graphite on a different port than 2003
and want icinga2
to connect to it.
Change the port value for the graphite feature according to your graphite installation before enabling it.
# cat /etc/icinga2/features-enabled/graphite.conf
/**
* The GraphiteWriter type writes check result metrics and
* performance data to a graphite tcp socket.
*/
library "perfdata"
object GraphiteWriter "graphite" {
//host = "127.0.0.1"
//port = 2003
port = 2004
}
# icinga2 feature enable graphite
Before you restart the icinga2 service allow it to connect to all ports by enabling the boolean ´icinga2_can_connect_all` (now and permanent).
# setsebool icinga2_can_connect_all true
# setsebool -P icinga2_can_connect_all true
If you restart the daemon now it will successfully connect to graphite.
Confining a user
Start by adding the Icinga 2 administrator role icinga2adm_r
to the administrative SELinux user staff_u
.
# semanage user -m -R "staff_r sysadm_r system_r unconfined_r icinga2adm_r" staff_u
Confine your user login and create a sudo rule.
# semanage login -a dirk -s staff_u
# echo "dirk ALL=(ALL) NOPASSWD: ALL" > /etc/sudoers.d/dirk
Login to the system using ssh and verify your id.
$ id -Z
staff_u:staff_r:staff_t:s0-s0:c0.c1023
Try to execute some commands as root using sudo.
$ sudo id -Z
staff_u:staff_r:staff_t:s0-s0:c0.c1023
$ sudo vi /etc/icinga2/icinga2.conf
"/etc/icinga2/icinga2.conf" [Permission Denied]
$ sudo cat /var/log/icinga2/icinga2.log
cat: /var/log/icinga2/icinga2.log: Keine Berechtigung
$ sudo systemctl reload icinga2.service
Failed to get D-Bus connection: No connection to service manager.
Those commands fail because you only switch to root but do not change your SELinux role. Try again but tell sudo also to switch the SELinux role and type.
$ sudo -r icinga2adm_r -t icinga2adm_t id -Z
staff_u:icinga2adm_r:icinga2adm_t:s0-s0:c0.c1023
$ sudo -r icinga2adm_r -t icinga2adm_t vi /etc/icinga2/icinga2.conf
"/etc/icinga2/icinga2.conf"
$ sudo -r icinga2adm_r -t icinga2adm_t cat /var/log/icinga2/icinga2.log
[2015-03-26 20:48:14 +0000] information/DynamicObject: Dumping program state to file '/var/lib/icinga2/icinga2.state'
$ sudo -r icinga2adm_r -t icinga2adm_t systemctl reload icinga2.service
Now the commands will work, but you have always to remember to add the arguments, so change the sudo rule to set it by default.
# echo "dirk ALL=(ALL) ROLE=icinga2adm_r TYPE=icinga2adm_t NOPASSWD: ALL" > /etc/sudoers.d/dirk
Now try the commands again without providing the role and type and they will work, but if you try to read apache logs or restart apache for example it will still fail.
$ sudo cat /var/log/httpd/error_log
/bin/cat: /var/log/httpd/error_log: Keine Berechtigung
$ sudo systemctl reload httpd.service
Failed to issue method call: Access denied
Bugreports
If you experience any problems while running in enforcing mode try to reproduce it in permissive mode. If the problem persists it is not related to SELinux because in permissive mode SELinux will not deny anything.
For now Icinga 2 is running in a permissive domain and adds also some rules for other necessary services so no problems should occure at all. But you can help to enhance the policy by testing Icinga 2 running confined by SELinux.
When filing a bug report please add the following information additionally to the normal ones:
- Output of
semodule -l | grep -e icinga2 -e nagios -e apache
- Output of
ps -eZ | grep icinga2
- Output of
semanage port -l | grep icinga2
- Output of
audit2allow -li /var/log/audit/audit.log
If access to a file is blocked and you can tell which one please provided the output of ls -lZ /path/to/file
(and perhaps the directory above).
If asked for full audit.log add -w /etc/shadow -p w
to /etc/audit/rules.d/audit.rules
, restart the audit daemon, reproduce the problem and add /var/log/audit/audit.log
to the bug report. With the added audit rule it will include the path of files access was denied to.
If asked to provide full audit log with dontaudit rules disabled executed semodule -DB
before reproducing the problem. After that enable the rules again to prevent auditd spamming your logfile by executing semodule -B
.