NOTE: CentOS Enterprise Linux 5 is built from the Red Hat Enterprise Linux source code. Other than logo and name changes CentOS Enterprise Linux 5 is compatible with the equivalent Red Hat version. This document applies equally to both Red Hat and CentOS Enterprise Linux 5.
This chapter is an overview of SELinux policy, some of its internals, and how it works. It discusses the policy in general terms, while Section 43.8, “Targeted Policy Overview” focuses on the details of the targeted policy as it ships in Red Hat Enterprise Linux. This chapter starts with a brief overview of what policy is and where it resides.
Following on from this, the role of SELinux during the boot process is discussed. This is followed by discussions on file security contexts, object classes and permissions, attributes, types, access vectors, macros, users and roles, constraints, and a brief discussion summarizing special kernel interfaces.
43.7.1. What is the SELinux Policy?
The SELinux Policy is the set of rules that guide the SELinux security engine. It defines types for file objects and domains for processes. It uses roles to limit the domains that can be entered, and has user identities to specify the roles that can be attained. In essence, types and domains are equivalent, the difference being that types apply to objects while domains apply to processes.
18.104.22.168. SELinux Types
A type is a way of grouping items based on their similarity from a security perspective. This is not necessarily related to the unique purpose of an application or the content of a document. For example, a file can have any type of content and be for any purpose, but if it belongs to a user and exists in that user's home directory, it is considered to be of a specific security type, user_home_t.
These object types are considered alike because they are accessible in the same way by the same set of subjects. Similarly, processes tend to be of the same type if they have the same permissions as other subjects. In the targeted policy, programs that run in the unconfined_t domain have an executable file with a type such as sbin_t. From an SELinux perspective, this means they are all equivalent in terms of what they can and cannot do on the system.
For example, the binary executable file object at /usr/bin/postgres has the type postgresql_exec_t. All of the targeted daemons have their own *_exec_t type for their executable applications. In fact, the entire set of PostgreSQL executables such as createlang, pg_dump, and pg_restore have the same type, postgresql_exec_t, and they transition to the same domain, postgresql_t, upon execution.
22.214.171.124.1. Using Policy Rules to Define Type Access
The SELinux policy defines various rules which determine how each domain may access each type. Only what is specifically allowed by the rules is permitted. By default, every operation is denied and audited, meaning it is logged in the $AUDIT_LOG file. In Red Hat Enterprise Linux, this is set to /var/log/messages. The policy is compiled into binary format for loading into the kernel security server, and each time the security server makes a decision, it is cached in the AVC to optimize performance.
The policy can be defined either by modifying the existing files or by adding local Type Enforcement (TE) and File Context (FC) files to the policy tree. These new policies can be loaded into the kernel in real time. Otherwise, the policy is loaded during the boot process by init, as explained in Section 43.7.3, “The Role of Policy in the Boot Process”. Ultimately, every system operation is determined by the policy and the type-labeling of the files.
After loading a new policy, it is recommended that you restart any services that may have new or changed labeling. Generally speaking, this is only the targeted daemons, as listed in Section 43.8.1, “What is the Targeted Policy?”.
126.96.36.199. SELinux and Mandatory Access Control
SELinux is an implementation of Mandatory Access Control (MAC). Depending on the security policy type, SELinux implements either Type Enforcment (TE), Roles Based Access Control (RBAC) or Bell-La Padula Model Multi-Level Security (MLS).
The policy specifies the rules in the implemented environment. It is written in a language created specifically for writing security policy. Policy writers use m4 macros to capture common sets of low-level rules. A number of m4 macros are defined in the existing policy, which facilitate the writing of new policy. These rules are preprocessed into many additional rules as part of building the policy.conf file, which is compiled into the binary policy.
Access rights are divided differently among domains, and no domain is required to act as a master for all other domains. Moving between domains is controlled by the policy, through login programs, userspace programs such as newrole, or by requiring a new process execution in the new domain. This movement between domains is referred to as a transition.
43.7.2. Where is the Policy?
There are two components to the policy: the binary tree and the source tree. The binary tree is provided by the selinux-policy-<policyname> package and supplies the binary policy file.
Alternatively, the binary policy can be built from source when the selinux-policy-devel package is installed.
Information on how to edit, write and compile policy is currently outside the scope of this document.
188.8.131.52. Binary Tree Files
/etc/selinux/targeted/ — this is the root directory for the targeted policy, and contains the binary tree.
/etc/selinux/targeted/policy/ — this is the location of the the binary policy file policy.<xx>. In this guide, the variable SELINUX_POLICY is used for this directory.
/etc/selinux/targeted/contexts/ — this is the location of the security context information and configuration files, which are used during runtime by various applications.
/etc/selinux/targeted/contexts/files/ — contains the default contexts for the entire file system. This is referenced by restorecon when perfoming relabeling operations.
/etc/selinux/targeted/contexts/users/ — in the targeted policy, only the root file is in this directory. These files are used for determining context when a user logs in. For example, for the root user, the context is user_u:system_r:unconfined_t.
/etc/selinux/targeted/modules/active/booleans* — this is where the runtime Booleans are configured.
These files should never be manually changed. You should use the getsebool, setsebool and semanage tools to manipulate runtime Booleans.
184.108.40.206. Source Tree Files
For developing policy modules, the selinux-policy-devel package includes all of the interface files used to build policy. It is recommended that people who build policy use these files to build the policy modules.
This package installs the policy interface files under /usr/share/selinux/devel/include and has make files installed in /usr/share/selinux/devel/Makefile.
To help applications that need the various SELinux paths, libselinux provides a number of functions that return the paths to the different configuration files and directories. This negates the need for applications to hard-code the paths, especially since the active policy location is dependent on the SELINUXTYPE setting in /etc/selinux/config.
For example, if SELINUXTYPE is set to strict, the active policy location is under /etc/selinux/strict.
To view the list of available functions, use the following command:
man 3 selinux_binary_policy_path
This man page is available only if you have the libselinux-devel RPM installed.
The use of libselinux and related functions is outside the scope of this document.
43.7.3. The Role of Policy in the Boot Process
SELinux plays an important role during the early stages of system start-up. Because all processes must be labeled with their correct domain, init performs some essential operations early in the boot process to maintain synchronization between labeling and policy enforcement.
After the kernel has been loaded during the boot process, the initial process is assigned the predefined initial SELinux ID (initial SID)kernel. Initial SIDs are used for bootstrapping before the policy is loaded.
/sbin/init mounts /proc/, and then searches for the selinuxfs file system type. If it is present, that means SELinux is enabled in the kernel.
If init does not find SELinux in the kernel, or if it is disabled via the selinux=0 boot parameter, or if /etc/selinux/config specifies that SELINUX=disabled, the boot process proceeds with a non-SELinux system.
At the same time, init sets the enforcing status if it is different from the setting in /etc/selinux/config. This happens when a parameter is passed during the boot process. The default mode is permissive until the policy is loaded, then enforcement is set by the configuration file or by the parameters enforcing=0 or enforcing=1.
If SELinux is present, /selinux/ is mounted.
The kernel checks /selinux/policyvers for the supported policy version. init instpects /etc/selinux/config to determine which policy is active, such as the targeted policy, and loads the associated file at $SELINUX_POLICY/policy.<version>.
If the binary policy is not the version supported by the kernel, init attempts to load the policy file if it is a previous version. This provides backward compatibility with older policy versions.
If the local settings in /etc/selinux/targeted/booleans are different from those compiled in the policy, init modifies the policy in memory based on the local settings prior to loading the policy into the kernel.
By this stage of the process, the policy is fully loaded into the kernel. The initial SIDs are then mapped to security contexts in the policy. In the case of the targeted policy, the new domain is user_u:system_r:unconfined_t. The kernel can now begin to retrieve security contexts dynamically from the in-kernel security server.
init then re-executes itself so that it can transition to a different domain, if the policy defines it. For the targeted policy, there is no transition defined and init remains in the unconfined_t domain.
At this point, init continues with its normal boot process.
The reason that init re-executes itself is to accommodate stricter SELinux policy controls. The objective of re-execution is to transition to a new domain with its own granular rules. The only way that a process can enter a domain is during execution, which means that such processes are the only entry points into the domains.
For example, if the policy has a specific domain for init, such as init_t, a method is required to change from the initial SID, such as kernel, to the correct runtime domain for init. Because this transition may need to occur, init is coded to re-execute itself after loading the policy.
This init transition occurs if the domain_auto_trans(kernel_t, init_exec_t, <target_domain_t>) rule is present in the policy. This rule states that an automatic transition occurs on anything executing in the kernel_t domain that executes a file of type init_exec_t. When this execution occurs, the new process is assigned the domain <target_domain_t>, using an actual target domain such as init_t.
43.7.4. Object Classes and Permissions
SELinux defines a number of classes for objects, making it easier to group certain permissions by specific classes. For example:
File-related classes include filesystem for file systems, file for files, and dir for directories. Each class has its own associated set of permissions.
The filesystem class can mount, unmount, get attributes, set quotas, relabel, and so forth. The file class has common file permissions such as read, write, get and set attributes, lock, relabel, link, rename, append, etc.
Network related classes include tcp_socket for TCP sockets, netif for network interfaces, and node for network nodes.
The netif class, for example, can send and receive on TCP, UDP and raw sockets (tcp_recv, tcp_send, udp_send, udp_recv, rawip_recv, and rawip_send.)
The object classes have matching declarations in the kernel, meaning that it is not trivial to add or change object class details. The same is true for permissions. Development work is ongoing to make it possible to dynamically register and unregister classes and permissions.
Permissions are the actions that a subject can perform on an object, if the policy allows it. These permissions are the access requests that SELinux actively allows or denies.