Beginners Guide to Access Control Lists (ACLs) in Linux
Access Control List Concepts
Standard Linux file permissions are satisfactory when files are used by only a single owner, and a single designated group of people. However, some use cases require that files are accessed with different file permission sets by multiple named users and groups. Access Control Lists (ACLs) provide this function.
With ACLs, you can grant permissions to multiple users and groups, identified by user name, group name, UID, or GID, using the same permission flags used with regular file permissions: read, write, and execute. These additional users and groups, beyond the file owner and the file’s group affiliation, are called named users and named groups respectively, because they are named not in a long listing, but rather within an ACL.
Users can set ACLs on files and directories that they own. Privileged users, assigned the CAP_FOWNER Linux capability, can set ACLs on any file or directory. New files and subdirectories automatically inherit ACL settings from the parent directory’s default ACL, if they are set. Similar to normal file access rules, the parent directory hierarchy needs at least the other search (execute) permission set to enable named users and named groups to have access.
File-system ACL Support
File systems need to be mounted with ACL support enabled. XFS file systems have built-in ACL support. Other file systems, such as ext3 or ext4 created on CentOS/RHEL 8, have the acl option enabled by default, although on earlier versions you should confirm that ACL support is enabled. To enable file-system ACL support, use the ACL option with the mount command or in the file system’s entry in /etc/fstab configuration file.
Viewing and Interpreting ACL permissions
The ls -l command only outputs minimal ACL setting details:
[user@host content]$ ls -l reports.txt -rwxrw----+ 1 user operators 130 Mar 19 23:56 reports.txt
The plus sign (+) at the end of the 10-character permission string indicates that an extended ACL structure with entries exists on this file.
- user: Shows the user ACL settings, which are the same as the standard user file settings; rwx.
- group: Shows the current ACL mask settings, not the group owner settings; rw.
- other: Shows the other ACL settings, which are the same as the standard other file settings; no access.
View File ACLs
To display ACL settings on a file, use getfacl file:
[user@host content]$ getfacl reports.txt # file: reports.txt # owner: user # group: operators user::rwx user:consultant3:-- user:1005:rwx #effective:rw group::rwx #effective:rw group:consultant1:r- group:2210:rwx #effective:rw mask::rw other::--
Review each section of the previous example:
# file: reports.txt # owner: user # group: operators
The first three lines are comments that identify the file name, owner (user), and group owner (operators). If there are any additional file flags, such as setuid or setgid, then a fourth comment line will appear showing which flags are set.
user::rwx user:consultant3:--- user:1005:rwx #effective:rw-
1. File owner permissions. user has rwx. 2. Named user permissions. One entry for each named user associated with this file. consultant3 has no permissions. 3. Named user permissions. UID 1005 has rwx, but the mask limits the effective permissions to rw only.
group::rwx #effective:rw- group:consultant1:r-- group:2210:rwx #effective:rw-
1. Group owner permissions. operators has rwx, but the mask limits the effective permissions to rw only. 2. Named group permissions. One entry for each named group associated with this file. consultant1 has r only. 3. Named group permissions. GID 2210 has rwx, but the mask limits the effective permissions to rw only.
Mask settings show the maximum permissions possible for all named users, the group owner, and named groups. UID 1005, operators, and GID 2210 cannot execute this file, even though each entry has the execute permission set.
Other or “world” permissions. All other UIDs and GIDs have NO permissions.
Viewing Directory ACLs
To display ACL settings on a directory, use the getfacl directory command:
[user@host content]$ getfacl . # file: . # owner: user # group: operators # flags: -s user::rwx user:consultant3:-- user:1005:rwx group::rwx group:consultant1:r-x group:2210:rwx mask::rwx other::-- default:user::rwx default:user:consultant3:-- default:group::rwx default:group:consultant1:r-x default:mask::rwx default:other::--
Review each section of the previous example:
Opening comment entries:
# file: . # owner: user # group: operators # flags: -s-
The first three lines are comments that identify the directory name, owner (user), and group owner (operators). If there are any additional directory flags **(setuid, setgid, sticky), **then a fourth comment line shows which flags are set; in this case, **setgid **.
Standard ACL entries:
user::rwx user:consultant3:-- user:1005:rwx group::rwx group:consultant1:r-x group:2210:rwx mask::rwx other::--
The ACL permissions on this directory are the same as the file example shown earlier, but apply to the directory. The key difference is the inclusion of the execute permission on these entries (when appropriate) to allow directory search permission.
Default user entries:
1. Default file owner ACL permissions. The file owner will get rwx, read/write on new files and execute on new subdirectories. 2. Default named user ACL permissions. One entry for each named user who will automatically get the default ACL applied to new files or subdirectories. consultant3 always defaults to no permissions.
Default group entries:
1. Default group owner ACL permissions. The file group owner will get rwx, read/write on new files and execute on new subdirectories. 2. Default named group ACL permissions. One entry for each named group which will automatically get the default ACL. consultant1 will get rx, read-only on new files, and execute on new subdirectories.
Default ACL mask entry:
Default mask settings show the initial maximum permissions possible for all new files or directories created that have named user ACLs, the group owner ACL, or named group ACLs: read and write for new files and execute permission on new subdirectories. New files never get execute permission.
Default other entry:
Default other or “world” permissions. All other UIDs and GIDs have no permissions to new files or new subdirectories.
The default entries in the previous example do not include the named user (UID 1005) and named group (GID 2210); consequently, they will not automatically get initial ACL entries added for them to any new files or new subdirectories. This effectively limits them to files and subdirectories that they already have ACLs on, or if the relevant file owner adds the ACL later using **setfacl **. They can still create their own files and subdirectories.
The ACL Mask
The ACL mask defines the maximum permissions that you can grant to named users, the group owner, and named groups. It does not restrict the permissions of the file owner or other users. All files and directories that implement ACLs will have an ACL mask.
The mask can be viewed with getfacl and explicitly set with setfacl. It will be calculated and added automatically if it is not explicitly set, but it could also be inherited from a parent directory default mask setting. By default, the mask is recalculated whenever any of the affected ACLs are added, modified, or deleted.
ACL Permission Precedence
When determining whether a process (a running program) can access a file, file permissions and ACLs are applied as follows:
- If the process is running as the user that owns the file, then the file’s user ACL permissions apply.
- If the process is running as a user that is listed in a named user ACL entry, then the named user ACL permissions apply (as long as it is permitted by the mask).
- If the process is running as a group that matches the group owner of the file, or as a group with an explicitly named group ACL entry, then the matching ACL permissions apply (as long as it is permitted by the mask).
- Otherwise, the file’s other ACL permissions apply.
Examples of ACL use by the Operating System
CentOS/RHEL has examples that demonstrate typical ACL use for extended permission requirements.
ACLs on Systemd Journal Files
systemd-journald uses ACL entries to allow read access to the /run/log/journal/cb44…8ae2/system.journal file to the adm and wheel groups. This ACL allows the members of the adm and wheel groups to have read access to the logs managed by journalctl without having the need to give special permissions to the privileged content inside /var/log/, like messages, secure or audit.
Due to the systemd-journald configuration, the parent folder of the system.journal file can change, but the systemd-journald applies ACLs to the new folder and file automatically.
[user@host ]$ getfacl /run/log/journal/cb44...8ae2/system.journal getfacl: Removing leading '/' from absolute path names # file: run/log/journal/cb44...8ae2/system.journal # owner: root # group: systemd-journal user::rw group::r- group:adm:r- group:wheel:r- mask::r- other::--
ACL on Systemd Managed Devices
systemd-udev uses a set of udev rules that enable the uaccess tag to some devices, such as CD/DVD players or writers, USB storage devices, sound cards, and many others. The previous mentioned udev rules sets ACLs on those devices to allow users logged in to a graphical user interface (for example gdm) to have full control of those devices.
The ACLs will remain active until the user logs out of the GUI. The next user who logs in to the GUI will have a new ACL applied for the required devices. In the following example, you can see the user has an ACL entry with rw permissions applied to the /dev/sr0 device that is a CD/DVD drive.
[user@host ]$ getfacl /dev/sr0 getfacl: Removing leading '/' from absolute path names # file: dev/sr0 # owner: root # group: cdrom user::rw user:group:rw group::rw mask::rw other::--
Changing ACL File Permissions
Use setfacl to add, modify, or remove standard ACLs on files and directories.
ACLs use the normal file system representation of permissions, “r” for read permission, “w” for write permission, and “x” for execute permission. A “-” (dash) indicates that the relevant permission is absent. When (recursively) setting ACLs, an uppercase “X” can be used to indicate that execute permission should only be set on directories and not regular files, unless the file already has the relevant execute permission. This is the same behavior as chmod.
Adding or Modifying ACLs
ACLs can be set via the command-line using the -m option, or passed in via a file using the M option (use “-” (dash) instead of a file name for stdin). These two options are the “modify” options; they add new ACL entries or replace specific existing ACL entries on a file or directory. Any other existing ACL entries on the file or directory remain untouched.
When first defining an ACL on a file, if the add operation does not include settings for the file owner, group owner, or other permissions, then these will be set based on the current standard file permissions (these are also known as the base ACL entries and cannot be deleted), and a new mask value will be calculated and added as well. To add or modify a user or named user ACL:
[user@host ~]$ setfacl -m u:name:rX file
If name is left blank, then it applies to the file owner, otherwise name can be a username or UID value. In this example, the permissions granted would be read-only, and if already set, execute (unless file was a directory, in which case the directory would get the execute permission set to allow directory search).
ACL file owner and standard file owner permissions are equivalent; consequently, using chmod on the file owner permissions is equivalent to using setfacl on the file owner permissions. chmod has no effect on named users. To add or modify a group or named group ACL:
[user@host ~]$ setfacl -m g:name:rw file
This follows the same pattern for adding or modifying a user ACL entry. If name is left blank, then it applies to the group owner. Otherwise, specify a group name or GID value for a named group. The permissions would be read and write in this example. chmod has no effect on any group permissions for files with ACL settings, but it updates the ACL mask. To add or modify the other ACL:
[user@host ~]$ setfacl -m o::- file
other only accepts permission settings. Typical permission settings for others are: no permissions at all, set with a dash (-); and read-only permissions set as usual with r. Of course, you can set any of the standard permissions. ACL other and standard other permissions are equivalent, so using chmod on the other permissions is equivalent to using setfacl on the other permissions.
You can add multiple entries with the same command; use a comma-separated list of entries:
[user@host ~]$ setfacl -m u::rwx,g:consultants:rX,o::- file
This sets the file owner to read, write, and execute, sets the named group consultants to readonly and conditional execute, and restricts all other users to no permissions. The group owner maintains existing file or ACL permissions and other “named” entries remain unchanged.
Using getfacl as Input
You can use the output from getfacl as input to setfacl:
[user@host ~]$ getfacl file-A | setfacl --set-file=- file-B
The –set-file option accepts input from a file or from stdin. The dash character (-) specifies the use of stdin. In this case, file-B will have the same ACL settings as file-A.
Setting an Explicit ACL Mask
You can set an ACL mask explicitly on a file or directory to limit the maximum effective permissions for named users, the group owner, and named groups. This restricts any existing permissions that exceed the mask, but does not affect permissions that are less permissive than the mask.
[user@host ~]$ setfacl -m m::r file
This adds a mask value that restricts any named users, the group owner, and any named groups to read-only permission, regardless of their existing settings. The file owner and other users are not impacted by the mask setting. getfacl shows an effective comment beside entries that are restricted by a mask setting.
Recursive ACL Modifications
When setting an ACL on a directory, use the -R option to apply the ACL recursively. Remember that you are likely to want to use the “X” (capital X) permission with recursion so that files with the execute permission set retain the setting and directories get the execute permission set to allow directory search. It is considered good practice to also use the uppercase “X” when non-recursively setting ACLs because it prevents administrators from accidentally adding execute permissions to a regular file.
[user@host ~]$ setfacl -R -m u:name:rX directory
This adds the user name to the [directory] directory and all existing files and subdirectories, setting read-only and conditional execute permissions.
Deleting specific ACL entries follows the same basic format as the modify operation, except that “:perms” is not specified.
[user@host ~]$ setfacl -x u:name,g:name file
This removes only the named user and the named group from the file or directory ACL. Any other existing ACL entries remain active. You can include both the delete (-x) and modify (-m) operations in the same setfacl operation.
The mask can only be deleted if there are no other ACLs set (excluding the base ACL which cannot be deleted), so it must be deleted last. The file will no longer have any ACLs and ls -l will not show the plus sign (+) next to the permissions string. Alternatively, to delete all ACL entries on a file or directory (including default ACL on directories), use the following command:
[user@host ~]$ setfacl -b file
Controlling Default ACL File Permissions
To ensure that files and directories created within a directory inherit certain ACLs, use the default ACL on a directory. You can set a default ACL and any of the standard ACL settings, including a default mask.
The directory itself still requires standard ACLs for access control because the default ACLs do not implement access control for the directory; they only provide ACL permission inheritance support. For example:
[user@host ~]$ setfacl -m d:u:name:rx directory
This adds a default named user (d:u:name) with read-only permission and execute permission on subdirectories. The setfacl command for adding a default ACL for each of the ACL types is exactly the same as for standard ACLs, but prefaced with d:. Alternatively, use the -d option on the command line.
Deleting Default ACL Entries
Delete a default ACL the same way that you delete a standard ACL, prefacing with d:, or use the d option.
[user@host ~]$ setfacl -x d:u:name directory
This removes the default ACL entry that was added in the previous example. To delete all default ACL entries on a directory, use setfacl -k directory.