Linux permissions 101

Knowing how to control users' access to files is a fundamental system administration skill.
156 readers like this.
Top 5 Linux pain points in 2017

Internet Archive Book Images. Modified by Opensource.com. CC BY-SA 4.0

Understanding Linux permissions and how to control which users have access to files is a fundamental skill for systems administration.

This article will cover standard Linux file systems permissions, dig further into special permissions, and wrap up with an explanation of default permissions using umask.

Understanding the ls command output

Before we can talk about how to modify permissions, we need to know how to view them. The ls command with the long listing argument (-l) gives us a lot of information about a file.

$ ls -lAh
total 20K
-rwxr-xr--+ 1 root root    0 Mar  4 19:39 file1
-rw-rw-rw-. 1 root root    0 Mar  4 19:39 file10
-rwxrwxr--+ 1 root root    0 Mar  4 19:39 file2
-rw-rw-rw-. 1 root root    0 Mar  4 19:39 file8
-rw-rw-rw-. 1 root root    0 Mar  4 19:39 file9
drwxrwxrwx. 2 root root 4.0K Mar  4 20:04 testdir

To understand what this means, let's break down the output regarding the permissions into individual sections. It will be easier to reference each section individually.

Take a look at each component of the final line in the output above:

drwxrwxrwx. 2 root root 4.0K Mar  4 20:04 testdir
Section 1 Section 2 Section 3 Section 4 Section 5 Section 6 Section 7
d rwx rwx rwx  . root root

Section 1 (on the left) reveals what type of file it is.

d Directory
- Regular file
l A soft link

The info page for ls has a full listing of the different file types.

Each file has three modes of access:

  • the owner
  • the group
  • all others

Sections 2, 3, and 4 refer to the user, group, and "other users" permissions. And each section can include a combination of r (read), w (write), and x (executable) permissions.

Each of the permissions is also assigned a numerical value, which is important when talking about the octal representation of permissions.

Permission Octal Value
Read 4
Write 2
Execute 1

Section 5 details any alternative access methods, such as SELinux or File Access Control List (FACL).

Method Character
No other method  -
SELinux  .
FACLs  +
Any combination of methods  +

Sections 6 and 7 are the names of the owner and the group, respectively.

Using chown and chmod

The chown command

The chown (change ownership) command is used to change a file's user and group ownership.

To change both the user and group ownership of the file foo to root, we can use these commands:

$ chown root:root foo
$ chown root: foo

Running the command with the user followed by a colon (:) sets both the user and group ownership.

To set only the user ownership of the file foo to the root user, enter:

$ chown root foo

To change only the group ownership of the file foo, precede the group with a colon:

$ chown :root foo

The chmod command

The chmod (change mode) command controls file permissions for the owner, group, and all other users who are neither the owner nor part of the group associated with the file.

The chmod command can set permissions in both octal (e.g., 755, 644, etc.) and symbolic (e.g., u+rwx, g-rwx, o=rw) formatting.

Octal notation assigns 4 "points" to read, 2 to write, and 1 to execute. If we want to assign the user read permissions, we assign 4 to the first slot, but if we want to add write permissions, we must add 2. If we want to add execute, then we add 1. We do this for each permission type: owner, group, and others.

For example, if we want to assign read, write, and execute to the owner of the file, but only read and execute to group members and all other users, we would use 755 in octal formatting. That's all permission bits for the owner (4+2+1), but only a 4 and 1 for the group and others (4+1).

The breakdown for that is: 4+2+1=7; 4+1=5; and 4+1=5.

If we wanted to assign read and write to the owner of the file but only read to members of the group and all other users, we could use chmod as follows:

$ chmod 644 foo_file

In the examples below, we use symbolic notation in different groupings. Note the letters u, g, and o represent user, group, and other. We use u, g, and o in conjunction with +, -, or = to add, remove, or set permission bits.

To add the execute bit to the ownership permission set:

$ chmod u+x foo_file

To remove read, write, and execute from members of the group:

$ chmod g-rwx foo_file

To set the ownership for all other users to read and write:

$ chmod o=rw

The special bits: Set UID, set GID, and sticky bits

In addition to the standard permissions, there are a few special permission bits that have some useful benefits.

Set user ID (suid)

When suid is set on a file, an operation executes as the owner of the file, not the user running the file. A good example of this is the passwd command. It needs the suid bit to be set so that changing a password runs with root permissions.

$ ls -l /bin/passwd
-rwsr-xr-x. 1 root root 27832 Jun 10  2014 /bin/passwd

An example of setting the suid bit would be:

$ chmod u+s /bin/foo_file_name

Set group ID (sgid)

The sgid bit is similar to the suid bit in the sense that the operations are done under the group ownership of the directory instead of the user running the command.

An example of using sgid would be if multiple users are working out of the same directory, and every file created in the directory needs to have the same group permissions. The example below creates a directory called collab_dir, sets the sgid bit, and changes the group ownership to webdev.

$ mkdir collab_dir
$ chmod g+s collab_dir
$ chown :webdev collab_dir

Now any file created in the directory will have the group ownership of webdev instead of the user who created the file.

$ cd collab_dir
$ touch file-sgid
$ ls -lah file-sgid
-rw-r--r--. 1 root webdev 0 Jun 12 06:04 file-sgid

The "sticky" bit

The sticky bit denotes that only the owner of a file can delete the file, even if group permissions would otherwise allow it. This setting usually makes the most sense on a common or collaborative directory such as /tmp. In the example below, the t in the execute column of the all others permission set indicates that the sticky bit has been applied.

$ ls -ld /tmp
drwxrwxrwt. 8 root root 4096 Jun 12 06:07 /tmp/

Keep in mind this does not prevent somebody from editing the file; it just keeps them from deleting the contents of a directory.

We set the sticky bit with:

$ chmod o+t foo_dir

On your own, try setting the sticky bit on a directory and give it full group permissions so that multiple users can read, write and execute on the directory because they are in the same group.

From there, create files as each user and then try to delete them as the other.

If everything is configured correctly, one user should not be able to delete users from the other user.

Note that each of these bits can also be set in octal format with SUID=4, SGID=2, and Sticky=1.

$ chmod 4744
$ chmod 2644
$ chmod 1755

Uppercase or lowercase?

If you are setting the special bits and see an uppercase S or T instead of lowercase (as we've seen until this point), it is because the underlying execute bit is not present. To demonstrate, the following example creates a file with the sticky bit set. We can then add/remove the execute bit to demonstrate the case change.

$ touch file cap-ST-demo
$ chmod 1755 cap-ST-demo
$ ls -l cap-ST-demo
-rwxr-xr-t. 1 root root 0 Jun 12 06:16 cap-ST-demo

$ chmod o-x cap-X-demo
$ ls -l cap-X-demo
-rwxr-xr-T. 1 root root 0 Jun 12 06:16 cap-ST-demo

Setting the execute bit conditionally

To this point, we've set the execute bit using a lowercase x, which sets it without asking any questions. We have another option: using an uppercase X instead of lowercase will set the execute bit only if it is already present somewhere in the permission group. This can be a difficult concept to explain, but the demo below will help illustrate it. Notice here that after trying to add the execute bit to the group privileges, it is not applied.

$ touch cap-X-file
$ ls -l cap-X-file
-rw-r--r--. 1 root root 0 Jun 12 06:31 cap-X-file
$ chmod g+X cap-X-file
$ ls -l cap-X-file
-rw-r--r--. 1 root root 0 Jun 12 06:31 cap-X-file

In this similar example, we add the execute bit first to the group permissions using the lowercase x and then use the uppercase X to add permissions for all other users. This time, the uppercase X sets the permissions.

$ touch cap-X-file
$ ls -l cap-X-file
-rw-r--r--. 1 root root 0 Jun 12 06:31 cap-X-file
$ chmod g+x cap-X-file
$ ls -l cap-X-file
-rw-r-xr--. 1 root root 0 Jun 12 06:31 cap-X-file
$ chmod g+x cap-X-file
$ chmod o+X cap-X-file
ls -l cap-X-file
-rw-r-xr-x. 1 root root 0 Jun 12 06:31 cap-X-file

Understanding umask

The umask masks (or "blocks off") bits from the default permission set in order to define permissions for a file or directory. For example, a 2 in the umask output indicates it is blocking the write bit from a file, at least by default.

Using the umask command without any arguments allows us to see the current umask setting. There are four columns: the first is reserved for the special suid, sgid, or sticky bit, and the remaining three represent the owner, group, and other permissions.

$ umask
0022

To understand what this means, we can execute umask with a -S (as shown below) to get the result of masking the bits. For instance, because of the 2 value in the third column, the write bit is masked off from the group and other sections; only read and execute can be assigned for those.

$ umask -S
u=rwx,g=rx,o=rx

To see what the default permission set is for files and directories, let's set our umask to all zeros. This means that we are not masking off any bits when we create a file.

$ umask 000
$ umask -S
u=rwx,g=rwx,o=rwx

$ touch file-umask-000
$ ls -l file-umask-000
-rw-rw-rw-. 1 root root 0 Jul 17 22:03 file-umask-000

Now when we create a file, we see the default permissions are read (4) and write (2) for all sections, which would equate to 666 in octal representation.

We can do the same for a directory and see its default permissions are 777. We need the execute bit on directories so we can traverse through them.

$ mkdir dir-umask-000
$ ls -ld dir-umask-000
drwxrwxrwx. 2 root root 4096 Jul 17 22:03 dir-umask-000/

Conclusion

There are many other ways an administrator can control access to files on a system. These permissions are basic to Linux, and we can build upon these fundamental aspects. If your work takes you into FACLs or SELinux, you will see that they also build upon these first rules of file access.

User profile image.
Alex Juarez is a Principal Engineer at Rackspace, touting 11 years with the company. Alex enjoys all things open-source as well as training and mentoring others, and is incredibly qualified to do so as an RHCA/RHCI. Alex has spoken at a number regional and nationwide open source conferences sharing knowledge and encouraging people to learn more.

2 Comments

Thanks for this article. It's a good reminder.

If I have a python script that I run frequently, I will usually set its execution bit, but you don't have to have it set to do the alternative, which is
python ./somescript.py
Once you learn something about permissions, you learn to look for a permissions problem when you get some message that only says that some operation like saving a file can't be done or some executable can't be found.

Creative Commons LicenseThis work is licensed under a Creative Commons Attribution-Share Alike 4.0 International License.