GNU Linux/Permissions/POSIX ACLs

From WhyAskWhy.org Wiki
Jump to: navigation, search


Overview

[1] [2] [3]

POSIX Access Control Lists (ACLs) allow different permissions for different users or groups to be assigned to files or directories, independent of the original owner or the owning group. This means, in addition to the file owner, the file group, and others, additional users and groups can be granted or denied access by using POSIX ACLs. Said another way, the permissions defined by ACLs are a superset of the permissions specified by the 9-bit UNIX permission model. Read, write and execute permissions are the only capabilities that the ACL system deals with. Additional features such as setuid and sticky bits are handled exclusively through the traditional mode bits.


Requirements for using ACLs

  • Supported by the filesystem used to store content
  • The filesystem is mounted with the acl option
  • The appropriate package containing the command-line setfacl and getfacl tools is installed
    • the package is named acl on Ubuntu
  • Current version of file utils (ls, cp, mv, etc) with support for working with ACLs


Enabling ACL support

Confirm Kernel has support built-in

[4]

Here we're looking in the @/boot/config-2.6.32-73-server@ kernel config file on an Ubuntu 10.04 LTS server to verify that the kernel was built with ACL support for the filesystem(s) that we're using. In our case we're only using Ext4, but as you can see below this kernel includes support for the other filesystems listed in the conf as well.

$ grep _ACL /boot/config-$(uname -r)
CONFIG_EXT2_FS_POSIX_ACL=y
CONFIG_EXT3_FS_POSIX_ACL=y
CONFIG_EXT4_FS_POSIX_ACL=y
CONFIG_REISERFS_FS_POSIX_ACL=y
CONFIG_JFS_POSIX_ACL=y
CONFIG_FS_POSIX_ACL=y
CONFIG_XFS_POSIX_ACL=y
CONFIG_OCFS2_FS_POSIX_ACL=y
CONFIG_BTRFS_FS_POSIX_ACL=y
CONFIG_GENERIC_ACL=y
CONFIG_TMPFS_POSIX_ACL=y
CONFIG_NFS_V3_ACL=y
CONFIG_NFSD_V2_ACL=y
CONFIG_NFSD_V3_ACL=y
CONFIG_NFS_ACL_SUPPORT=m

Confirm filesystem is mounted with proper support

[5] [6]

First, you can use tune2fs to check the default mount options for the filesystem in question:

sudo tune2fs -l /dev/sdaX | grep acl
Default mount options:    user_xattr acl

If it's not there, it would need to be added as a mount option for any filesystem that supports ACLs. Also, you should double-check /proc/mounts to confirm that ACL support hasn't been overridden.

$ grep noacl /proc/mounts

Enable acl mount option

[5]

If if it is not already enabled, you can set acl as a default mount option by using the tune2fs -o option partition command:

sudo tune2fs -o acl /dev/sdXY

Using the default mount options instead of an entry in /etc/fstab is very useful for external drives as those partitions will be mounted with the acl option enabled on other Linux machines. In that case there is no need to edit /etc/fstab on every machine.


acl is specified as default mount option when creating an ext2/3/4 filesystem. This is configured in /etc/mke2fs.conf.


ACL Types

[2] [1]

Every object can be thought of as having associated with it an ACL that governs the discretionary access to that object; this ACL is referred to as an access ACL. In addition, a directory may have an associated ACL that governs the initial access ACL for objects created within that directory; this ACL is referred to as a default ACL.

  • Use access ACLs to grant permission to a specific file or directory
  • Use default ACLs to set permissions at the directory level for all files in the directory

Access ACLs

  •  ?


Default ACLs

[7] [8]

Directories can have a default ACL, which is a special kind of ACL defining the access permissions that objects in the directory inherit when they are created. A default ACL affects both subdirectories and files.

The following are the ways in which the permissions of a directory's default ACLs are passed to the files and subdirectories in it:

  • A subdirectory inherits the default ACLs of the parent directory both as its default ACL and as an access ACL.
  • If a file inside that directory does not have an ACL (already), it inherits the permissions of the default ACLs of the directory as its access ACL.

All system calls that create file system objects use a mode parameter that defines the access permissions for the newly created file system object. If the parent directory does not have a default ACL, the permission bits as defined by the umask are subtracted from the permissions as passed by the mode parameter, with the result being assigned to the new object. If a default ACL exists for the parent directory, the permission bits assigned to the new object correspond to the overlapping portion of the permissions of the mode parameter and those that are defined in the default ACL. The umask is disregarded in this case.


  • An access ACL set for an individual file can override the default ACL permissions
  • You cannot set a default ACL on a regular file (it is intended for directories only to set inheritance)
  • Setting the default ACL on the top-level folder does NOT set the actual permissions for that directory, only those created beneath it. You'll need to explicitly set the ACL entries at the top-most level.


ACL Entries

[7] [2]

Type Man Page Name Text Form Description
Owner ACL_USER_OBJ user::rwx This entry denotes access rights for the file owner.
Named User ACL_USER user:name:rwx These entries denote access rights for users identified by the entry's qualifier.
Owning Group ACL_GROUP_OBJ group::rwx This entry denotes access rights for the file group.
Named Group ACL_GROUP group:name:rwx These entries denote access rights for groups identified by the entry's qualifier.
Mask ACL_MASK mask::rwx This entry denotes the maximum access rights that can be granted by entries of type ACL_USER, ACL_GROUP_OBJ, or ACL_GROUP.
Other ACL_OTHER other::rwx This entry denotes access rights for processes that do not match any other entry in the ACL.

When an access check is performed, the ACL_USER_OBJ (owner) and ACL_USER (named user) entries are tested against the effective user ID. The effective group ID, as well as all supplementary group IDs are tested against the ACL_GROUP_OBJ (owning group) and ACL_GROUP (named group) entries.

Correspondence Between ACL Entries And File Permission Bits

[7] [2]

There is a correspondence between the file owner, group, and other permissions and specific ACL entries:

  • The owner permissions correspond to the permissions of the ACL_USER_OBJ (owner, user::rwx) entry.
  • If the ACL has an ACL_MASK entry, the group permissions correspond to the permissions of the ACL_MASK entry. Otherwise, if the ACL has no ACL_MASK entry, the group permissions correspond to the permissions of the ACL_GROUP_OBJ (owning group, group::rwx) entry.
  • The other permissions correspond to the permissions of the ACL_OTHER_OBJ (other, other::rwx) entry.


  • The file owner, group, and other permissions always match the permissions of the corresponding ACL entry.
  • Modification of the file permission bits results in the modification of the associated ACL entries, and modification of these ACL entries results in the modification of the file permission bits.


Valid ACLs

[7] [2]

  • A valid ACL contains exactly one entry with each of the ACL_USER_OBJ (owner), ACL_GROUP_OBJ (owning group), and ACL_OTHER (other) tag types.
  • Entries with ACL_USER (named user) and ACL_GROUP (named group) tag types may appear zero or more times in an ACL.
  • An ACL that contains entries of ACL_USER (named user) or ACL_GROUP (named group) tag types must contain exactly one entry of the ACL_MASK tag type.
  • If an ACL contains no entries of ACL_USER (named user) or ACL_GROUP (named group) tag types, the ACL_MASK entry is optional.


Mask

[7] [3] [9] [10]


Unfortunately I'm finding conflicting answers for whether the traditional Unix Group (ACL_GROUP_OBJ, owning group) is affected by the mask.

  • All permissions defined in the owner and owning group entries are always effective. [7]
  • The traditional Unix User (owner), Group (owning group), and Other entries are not affected by the mask [10]

Other sources say:

  • The file owner and others permissions are not affected by the effective rights mask; all other entries are. [9]
  • The mask limits the access that the ACL can confer upon all named users, all named groups, and the default group. [3]

According to the Access Check Algorithm given on the man page I am led to believe that the latter two sources are correct (with the first two being false).


In more complex ACLs, the traditional group permission bits correspond to a special ACL entry called the mask rather than the owning group (group::rwx) ACL entry. Also known as the effective rights mask, the mask entry further limits the permissions granted by all named users, named groups, and the default (aka, "primary" or "owning") group entry by defining which of the permissions in those entries are effective (allowed) and which are masked (disallowed).

In other words, the mask specifies an upper bound on the access that the ACL can assign to individual groups and users. If the mask is more restrictive than the ACL permissions that you grant, then the mask takes precedence.

ACL entries for named users, named groups, and the default group can include permission bits that are not present in the mask, but the kernel simply ignores them. As a result, the traditional mode bits can never understate the access allowed by the ACL as a whole. Furthermore, clearing a bit from the group portion of the traditional mode clears the corresponding bit in the ACL mask and thereby forbids this permission to everyone but the file's owner and those who fall in the category of "other".


  • While conceptually similar to the umask, the ACL mask is always in effect and specifies the allowed permissions rather than the permissions to be denied
  • If permissions exist in one of the mentioned entries as well as the mask, they are effective
  • Permissions contained only in the mask or only in the actual entry are not effective--meaning the permissions are not granted
  • Modifying group permissions on an ACL-bearing file (i.e., via chmod) affects the mask


Order of evaluation / Access Check Algorithm

[2]

A process may request read, write, or execute/search access to a file object protected by an ACL. The access check algorithm determines whether access to the object will be granted.

if the effective user ID of the process matches the user ID of the file object owner
    if the ACL_USER_OBJ (owner) entry contains the requested permissions
        access is granted
    else 
        access is denied
else 
    if the effective user ID of the process matches the qualifier of any entry of type ACL_USER (named user)
        if the matching ACL_USER (named user) entry and the ACL_MASK entry contain the requested permissions
            access is granted
        else 
            access is denied
else 
    if the effective group ID or any of the supplementary group IDs of the process match the file group (owning group) or the qualifier of any entry of type ACL_GROUP (named group)
        if the ACL contains an ACL_MASK entry, then
            if the ACL_MASK entry and any of the matching ACL_GROUP_OBJ (owning group) or ACL_GROUP (named group) entries contain the requested permissions
                access is granted
            else
                access is denied
        else (note that there can be no ACL_GROUP (named group) entries without an ACL_MASK entry) 
            if the ACL_GROUP_OBJ (owning group) entry contains the requested permissions
                access is granted
            else access is denied
else 
    if the ACL_OTHER entry contains the requested permissions
        access is granted
else 
    access is denied


Viewing ACLs

ACLs are viewed using the getfacl command-line tool.

TODO

add examples here


Setting ACLs

[11]

The setfacl utility sets Access Control Lists (ACLs) of files and directories. On the command line, a sequence of commands is followed by a sequence of files (which in turn can be followed by another sequence of commands, ...).

The file owner and processes capable of CAP_FOWNER are granted the right to modify ACLs of a file. This is analogous to the permissions required for accessing the file mode. (On current Linux systems, root is the only user with the CAP_FOWNER capability.)

Options

$ setfacl --help
setfacl 2.2.49 -- set file access control lists
Usage: setfacl [-bkndRLP] { -m|-M|-x|-X ... } file ...
  -m, --modify=acl        modify the current ACL(s) of file(s)
  -M, --modify-file=file  read ACL entries to modify from file
  -x, --remove=acl        remove entries from the ACL(s) of file(s)
  -X, --remove-file=file  read ACL entries to remove from file
  -b, --remove-all        remove all extended ACL entries
  -k, --remove-default    remove the default ACL
      --set=acl           set the ACL of file(s), replacing the current ACL
      --set-file=file     read ACL entries to set from file
      --mask              do recalculate the effective rights mask
  -n, --no-mask           don't recalculate the effective rights mask
  -d, --default           operations apply to the default ACL
  -R, --recursive         recurse into subdirectories
  -L, --logical           logical walk, follow symbolic links
  -P, --physical          physical walk, do not follow symbolic links
      --restore=file      restore ACLs (inverse of `getfacl -R')
      --test              test mode (ACLs are not modified)
  -v, --version           print version and exit
  -h, --help              this help text
  • When using setfacl, you are allowed to use multiple -m (modify) or -x (delete) options on the same line.
  • The options -m, and -x expect an ACL on the command line. Multiple ACL entries are separated by comma characters (','). The options -M, and -X read an ACL from a file or from standard input. The ACL entry format is described in Section ACL ENTRIES.
  • The --set and --set-file options set the ACL of a file or a directory. The previous ACL is replaced. ACL entries for this operation must include permissions.
  • The -m (--modify) and -M (--modify-file) options modify the ACL of a file or directory. ACL entries for this operation must include permissions.
  • The -x (--remove) and -X (--remove-file) options remove ACL entries. It is not an error to remove an entry which does not exist. Only ACL entries without the perms field are accepted as parameters, unless POSIXLY_CORRECT is defined.
  • When reading from files using the -M, and -X options, setfacl accepts the output getfacl produces. There is at most one ACL entry per line. After a Pound sign ('#'), everything up to the end of the line is treated as a comment.
  • If setfacl is used on a file system which does not support ACLs, setfacl operates on the file mode permission bits. If the ACL does not fit completely in the permission bits, setfacl modifies the file mode permission bits to reflect the ACL as closely as possible, writes an error message to standard error, and returns with an exit status greater than 0.

Examples

Explain this:

 setfacl -d -m group:rwx /path/to/your/dir

It appears to be setting the Default ACL for the owning group to rwx (octal 777) for a specific directory. Presumably this means that inheritance would push those settings down to any newly created files/directories.

  • Q: What about existing files?
  • Q: What about existing directories?

Removing POSIX ACLs

To remove all the permissions for a user, groups, or others, use the following command [12]:

setfacl -x ACL entry type file

For example, to remove all permissions from the user antony:

setfacl -x u:antony /mnt/gluster/data/test-file


Changes to the file utilities

[11]

On a system that supports ACLs, the file utilities ls, cp, and mv change their behavior in the following way:

  • For files that have a default ACL or an access ACL that contains more than the three required ACL entries, the ls utility in the longform produced by ls -l displays a plus sign (+) after the permission string.
  • If the -p flag is specified, the cp utility also preserves ACLs. If this is not possible, a warning is produced.
  • The mv utility always preserves ACLs. If this is not possible, a warning is produced.

The effect of the chmod utility, and of the chmod system call, on the access ACL is described in Correspondence Between ACL Entries And File Permission Bits.


Misc

IBM's Tivoli Storage Manager [13]

File System ACL Support
XFS YES
EXT2 YES
EXT3 YES
EXT4 YES
GPFS YES
JFS NO
VxFS NO
EXT2 YES
NSS YES


References

Directly used

  1. 1.0 1.1 Support > Product Documentation > Red Hat Storage > 3.0 > Administration Guide > 7.6. POSIX Access Control Lists
  2. 2.0 2.1 2.2 2.3 2.4 2.5 acl(5) - Linux man page
  3. 3.0 3.1 3.2 UNIX and Linux System Administration Handbook, 4th Edition
  4. ACL(Access Control List) Configuration in Debian
  5. 5.0 5.1 Arch Linux Wiki - ACLs
  6. Ubuntu wiki - ACLs
  7. 7.0 7.1 7.2 7.3 7.4 7.5 Documentation > Security Guide > Local Security > Chapter 9. Access Control Lists in Linux
  8. The University of Tennessee Knoxville > Resources > IT Support > IT Knowledge Base > Linux/POSIX ACLs
  9. 9.0 9.1 getfacl(1) - Linux man page
  10. 10.0 10.1 Using ACLs with Fedora Core 2 (Linux Kernel 2.6.5)
  11. 11.0 11.1 setfacl(1) - Linux man page
  12. Support > Product Documentation > Red Hat Storage > 2.0 > Administration Guide > 9.5.3. Removing POSIX ACLs
  13. IBM > Tivoli Software > File system and ACL support

Queued up