Linux, Red Hat, Uncategorized

Every time you disable SELinux God kills a kitten

Anything that states you must disable selinux to get it to work should never be run in production or on a system outside a lab.

This post is to enable you to run selinux when others don’t know how. This post will make you better with linux if you have had trouble with selinux.  This post may even improve your life and all sort of other stuff.

Why care about selinux?

Too many articles out there that default to “just disable selinux” and it works.  From my windows days, this is very much the same as the “just disable User Account Control “UAC” and the windows firewall.”  The obvious reason not to do this is security and although security is hard, it is important. SELinux provides additional security and mandatory access controls in linux.  It confines applications so they don’t compromise or harm other processes.  Another way to think about this, is that SELinux is written to secure apps that aren’t inherently secure or as secure as they could be.  In short, you want to be secure and you want your stuff to work.

This guide will show you how to uncover anything selinux is blocking and add policies to allow the behavior with some pretty cool tools available.

How do I make this all work?

It’s important to note selinux has 3 modes of operations.  One is disable, this turns off everything and really should never be used unless you are in a locked down environment and require the sliver of performance lost by adding selinux.  enforcing mode turns selinux on and enforces the security that is set.  permissive turns selinux on but does not enforce the rules.  Use permissive instead of disabled.  When anyone says disabled selinux, they should be saying, make selinux permissive because I don’t know how it works.

Security Hardening

This guide is designed to harden the security on centos, rhel, fedora and any system that can install the tools and use selinux. The changes suggested here may impact your system and cause services or features to become unavailable without proper testing. By following this guide and understanding what you are doing, you should be able to prevent this. This guide will show you how to apply SELinux to your system and be alerted for anything that is blocked.

Minimum requirements

  • Linux kernel supporting selinux (should be anything after 2.6) – ideally a VM or testing installation
    • RHEL, Centos or fedora (ideally RHEL 7 or Centos 7 or higher)
    • other distros can work but I’m using yum below (use dnf for fedora)
  • SSH or terminal access to your system
  • Familiarity with vim or your favorite editor
  • Ability to use 2 terminals at once or screen

Installation

We first want to set selinux to permissive. Before we do this, it’s best to have your disks created and the structure setup. Permissive allows us to log all SELinux violations without preventing their actions. This is good because we can see what will be blocked. In addition, we will install some tools that almost tell us exactly how to unblock anything that does get blocked.

Let’s go ahead and install the tools, set permissive and reboot for it to take effect (you can do this without reboot but you’ll miss any issues that present themselves on reboot).

Install the extra tools

# yum -y install setroubleshoot-server setools mcstrans

Now let’s enable selinux in permissive mode. We need to reboot after to catch the logs. You can use vi on /etc/selinux/config or copy/paste the line below (check those file paths are the same on your system)

# sed -i 's/^SELINUX=.*/SELINUX=permissive/g' /etc/selinux/config && cat /etc/selinux/config
# systemctl reboot

Now once the system reboots, you will get some log entries that we can find out if SELinux is blocking anything. We want to find entries that indicate something it being blocked. Let’s search the messages log.

# cat /var/log/messages | grep "SELinux is preventing"

If no errors, let’s search for just “SELinux”

# cat /var/log/messages | grep "SELinux"

We should get something back like this if everything is ok.

Aug 10 15:43:09 Rocky3 kernel: SELinux: Initializing.
Aug 10 15:43:13 Rocky3 kernel: SELinux: Disabled at runtime.
Aug 10 15:51:15 rocky3 kernel: SELinux: Initializing.
Aug 10 15:51:17 rocky3 kernel: SELinux: Disabled at runtime.
Aug 10 15:52:39 rocky3 kernel: SELinux: Initializing.
Aug 10 15:52:41 rocky3 kernel: SELinux: Permission validate_trans in class security not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_iscsi_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_fib_lookup_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_connector_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_netfilter_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_generic_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_scsitransport_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_rdma_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class netlink_crypto_socket not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Permission audit_read in class capability2 not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: Class binder not defined in policy.
Aug 10 15:52:41 rocky3 kernel: SELinux: the above unknown classes and permissions will be allowed
Aug 10 15:52:41 rocky3 systemd[1]: Successfully loaded SELinux policy in 69.596ms.

If something is being blocked we would see something like:

Aug 10 16:10:18 rocky3 setroubleshoot: SELinux is preventing /usr/sbin/nmbd from getattr access on the file /etc/samba/smb.conf

If you have any errors, please see the troubleshooting part of this document before enabling the enforcing mode. For clean installs without any additional services, you should not see any errors.

Once again, if you see errors, go to the troubleshooting section. If you installed a program, you’ll likely see selinux blocking something.

Let’s enable enforcing now that we have no errors.

# sed -i 's/^SELINUX=.*/SELINUX=enforcing/g' /etc/selinux/config && cat /etc/selinux/config
# setenforce 1

It is a good idea to now check for errors again but you should not see any if this was done right. We can run the sestatus command to check selinux status.  If you see errors, resolve them before you reboot!!!

# sestatus

Should show the following:

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: 30

now let’s reboot and check once again to ensure selinux is enforcing and we don’t have additional errors

systemctl reboot
getenforce
sealert -a /var/log/audit/audit.log

Troubleshooting

Resolving log messages

Some other places to look for errors include utilizing the sealert tool, this tool not only shows the error but shows how to fix the error with confidence scores indicating how sure it thinks it is.

When we search in /var/log/messages, we may have a message similar to the one below. Note the last sentence, which says what command to run.

Aug 10 16:11:05 rocky3 setroubleshoot: SELinux is preventing /usr/sbin/nmbd from read access on the file smb.conf. For complete SELinux messages. run sealert -l 109d0656-c57b-4fdd-9ef3-09c6cee6351a

When you run that command, you’ll get an output similar to the one below. Note the confidence scores and resolutions. Look up what they do (don’t blindly trust everything please). The following output indicates that SELinux is preventing the smbd service from modifying the smb.conf file.

SELinux is preventing /usr/sbin/smbd from open access on the file /etc/samba/smb.conf.

                                            *****  Plugin restorecon (60.0 confidence) suggests   ************************

                                            If you want to fix the label. 
                                            /etc/samba/smb.conf default label should be samba_etc_t.
                                            Then you can run restorecon.
                                            Do
                                            # /sbin/restorecon -v /etc/samba/smb.conf

                                            *****  Plugin samba_share (30.8 confidence) suggests   ***********************

                                            If you want to allow smbd to have open access on the smb.conf file
                                            Then you need to change the label on '/etc/samba/smb.conf'
                                            Do
                                            # semanage fcontext -a -t samba_share_t '/etc/samba/smb.conf'
                                            # restorecon  -v '/etc/samba/smb.conf'

                                            *****  Plugin catchall_boolean (5.26 confidence) suggests   ******************

                                            If you want to allow samba to export all ro
                                            Then you must tell SELinux about this by enabling the 'samba_export_all_ro' boolean.

                                            Do
                                            setsebool -P samba_export_all_ro 1

                                            *****  Plugin catchall_boolean (5.26 confidence) suggests   ******************

                                            If you want to allow samba to export all rw
                                            Then you must tell SELinux about this by enabling the 'samba_export_all_rw' boolean.

                                            Do
                                            setsebool -P samba_export_all_rw 1

                                            *****  Plugin catchall (1.09 confidence) suggests   **************************

                                            If you believe that smbd should be allowed open access on the smb.conf file by default.
                                            Then you should report this as a bug.
                                            You can generate a local policy module to allow this access.
                                            Do
                                            allow this access for now by executing:
                                            # grep smbd /var/log/audit/audit.log | audit2allow -M mypol
                                            # semodule -i mypol.pp

To resolve this, we want to have smbd access to the smb.conf file. We are saying that service can now access it. SELinux was protecting us from some service just using that file or modifying it, which is a good thing! In the samba section we will run the commands to enable smb.conf access and also read/write if desired.

sealert usually targets a log file with the -a option. These usually will give you the messages you need.

# sealert -a /var/log/audit/audit.log

Samba

When you enable Samba, you’ll start getting audit messages.  This is because it doesn’t allow modifying smb.conf which is what you are doing (you’re adding lines to the bottom of the file indicating what shares you want to export). To allow this we need to allow access and also allow read only or read/write.

# semanage fcontext -a -t samba_etc_t /etc/samba/smb.conf
# restorecon -Rv /etc/samba/
restorecon reset /etc/samba/smb.conf context system_u:object_r:tmp_t:s0->system_u:object_r:samba_etc_t:s0

We just set the security context on smb.conf so it is able to be used. Restorecon is necessary because it sets the contexts correctly on the files and directories.

Below is a blanket boolean to allow samba rw/ro access. It would be better to set this for the directories you share rather than all over but for now, we will just set it. Use ro for read only or rw for read/write

 # setsebool -P samba_export_all_ro 1

OR

 # setsebool -P samba_export_all_rw 1

NFS

I haven’t found anything blocking NFS yet.

Docker

These will vary by container so anytime you add a container you should check sealert or grep in /var/log/messages for an error.  In general, this is more likely to come up with persistent volume mapping or exposing external ports. The other area to keep an eye on is sharing volumes between containers. This worked for me but I would always check after enabling.

Importing new pools or adding disks or btrfs

This is going to be the scariest area here. SELinux adds contexts to files and directories and if this gets messed up, you won’t be able to access much. BEFORE you do this, I would likely just go into /etc/selinux/config and set everything to permissive just in case. Do the import, double check the logs then set it back to enforcing.

One thought on “Every time you disable SELinux God kills a kitten

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s