Skip to contentSkip to conversation

context trouble with 2.5 bits

DISCUSSION Colin Walters 3 participants 10 messages Jul 8, 2003 – Jul 9, 2003 SELinux security module

lore
Share & tools
lore
#0 lore raw

Hi,

I think I'm pretty close to getting my Debian system running with the SELinux 2.5 bits. However, I can't seem to get either the patched sshd or login to give me the correct context. When I login (using login) as root, I see:

Warning! Could not get new context for tty1, not relabeling.

...followed by a bunch of avc denials about tty1 and /root, and then my context is:

meta:~# id
uid=0(root) gid=0(root) groups=0(root) context:system_u:system_r:user_t

Now, since Debian uses the login in shadow-utils (not the one in util-linux), I had to port over the patch, but I'm fairly sure that's not the problem, because I get the exact same symptoms when logging in over the patched ssh.

Does anyone have any ideas for what could be going wrong?

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.

#1 lore raw
Quoted 1 line · Colin Walters

Warning! Could not get new context for tty1, not relabeling.

Which may mean that the original type for the device node is not one that is expected. Please tell me what "ls --context /dev/tty1" reports.

However this should have nothing to do with the ability to login and run a session in the correct context.

Quoted 5 lines

...followed by a bunch of avc denials about tty1 and /root, and then my context is:

meta:~# id
uid=0(root) gid=0(root) groups=0(root) context:system_u:system_r:user_t

This is wrong. There's a domain_auto_trans() rule to transition to user_t for when using a login program that is not aware of SE Linux. So it looks like the target context is not being set.

Under the new scheme I believe that some file is supposed to be written under /proc for this. Make sure that the right stuff is mounted under /proc, the file is available, etc.

Also while in permissive mode do some tests at the command line, do:

# echo user_u:user_r:user_t > /proc/whatever-it-is
# id

Then id should report a context of user_u:user_r:user_t if the context change is working correctly.

Quoted 6 lines

Now, since Debian uses the login in shadow-utils (not the one in util-linux), I had to port over the patch, but I'm fairly sure that's not the problem, because I get the exact same symptoms when logging in over the patched ssh.

Does anyone have any ideas for what could be going wrong?

Does the NSA supplied patch for 2.5.x do the same things as the patch for 2.4.x? Maybe there's something new in the NSA tree that's necessary and that merely calling execve_secure() is not enough.

Does your /etc/security/default_contexts have the right stuff?

Signature · 10 lines
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#2 lore raw
Quoted 5 lines · Russell Coker
Quoted 1 line · Colin Walters

Warning! Could not get new context for tty1, not relabeling.

Which may mean that the original type for the device node is not one that is expected. Please tell me what "ls --context /dev/tty1" reports.

meta:/var/log# ls --context /dev/tty1
crw-------  root     tty      system_u:object_r:tty_device_t   /dev/tty1
meta:/var/log#
Quoted 2 lines

However this should have nothing to do with the ability to login and run a session in the correct context.

Yeah.

Quoted 3 lines

This is wrong. There's a domain_auto_trans() rule to transition to user_t for when using a login program that is not aware of SE Linux. So it looks like the target context is not being set.

Well this is embarassing, but apparently HAVE_SELINUX wasn't getting defined by my openssh build (the debian/rules file was overriding CFLAGS). Now that I've fixed that, I get:

Jul  8 21:41:52 meta sshd[642]: Accepted password for root from ::ffff:192.168.1.100 port 36286 ssh2
Jul  8 21:41:52 meta sshd[642]: fatal: Failed to get default security context for root.

The same thing happens with a normal user.

Quoted 10 lines

Under the new scheme I believe that some file is supposed to be written under /proc for this. Make sure that the right stuff is mounted under /proc, the file is available, etc.

Also while in permissive mode do some tests at the command line, do:

# echo user_u:user_r:user_t > /proc/whatever-it-is
# id

Then id should report a context of user_u:user_r:user_t if the context change is working correctly.

Here's what I get:

meta:/var/log# cat /proc/self/attr/current  && echo
system_u:system_r:user_t
meta:/var/log# echo root:system_r:sysadm_r > /proc/self/attr/current
meta:/var/log# cat /proc/self/attr/current  && echo
system_u:system_r:user_t
meta:/var/log#
Quoted 3 lines

Does the NSA supplied patch for 2.5.x do the same things as the patch for 2.4.x? Maybe there's something new in the NSA tree that's necessary and that merely calling execve_secure() is not enough.

Well, there is no execve_secure() anymore; now you have to call setexeccon. But their patch does appear to be doing the right things, at least according to their PORTING document.

Quoted 1 line

Does your /etc/security/default_contexts have the right stuff?

It looks like this:

system_r:local_login_t  user_r:user_t staff_r:staff_t
system_r:sshd_t         user_r:user_t staff_r:staff_t
system_r:crond_t        user_r:user_crond_t staff_r:staff_crond_t system_r:system_crond_t

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.

#3 lore raw
Quoted 5 lines · Colin Walters
Quoted 1 line

is expected. Please tell me what "ls --context /dev/tty1" reports.

meta:/var/log# ls --context /dev/tty1
crw-------  root     tty      system_u:object_r:tty_device_t   /dev/tty1
meta:/var/log#

That's what we want to see.

Quoted 5 lines
Jul  8 21:41:52 meta sshd[642]: Accepted password for root from

::ffff:192.168.1.100 port 36286 ssh2 Jul 8 21:41:52 meta sshd[642]: fatal:
Failed to get default security context for root.

The same thing happens with a normal user.

What is the line in the "users" file for root? Is there a line in the "users" file for the non-root user you tried? If not is there a line for user_u?

Quoted 7 lines
Quoted 6 lines

Also while in permissive mode do some tests at the command line, do:

# echo user_u:user_r:user_t > /proc/whatever-it-is
# id

Then id should report a context of user_u:user_r:user_t if the context change is working correctly.

Quoted 8 lines

Here's what I get:

meta:/var/log# cat /proc/self/attr/current  && echo
system_u:system_r:user_t
meta:/var/log# echo root:system_r:sysadm_r > /proc/self/attr/current
meta:/var/log# cat /proc/self/attr/current  && echo
system_u:system_r:user_t
meta:/var/log#

Sorry, you misunderstood my message.

I mean to echo a value to the file that specifies the context for the child process (you can't change the context of a running process, not even in permissive mode).

Quoted 8 lines
Quoted 1 line

Does your /etc/security/default_contexts have the right stuff?

It looks like this:

system_r:local_login_t  user_r:user_t staff_r:staff_t
system_r:sshd_t         user_r:user_t staff_r:staff_t
system_r:crond_t        user_r:user_crond_t staff_r:staff_crond_t

system_r:system_crond_t

Have you checked that sshd is running as system_u:system_r:sshd_t and that the "users" file permits the contexts user_r:user_t or staff_r:staff_t? If you configured "root" to only be permitted the context sysadm_r:sysadm_t then it will be impossible to login as root via ssh (I have intentionally configured it this way on my laptop).

Signature · 10 lines
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#4 lore raw
Quoted 2 lines

Have you checked that sshd is running as system_u:system_r:sshd_t

Ah hah! That was exactly the problem. After fixing that it works fine.

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.

#5 lore raw
In reply to Russell Coker
Quoted 6 lines · Russell Coker

Also while in permissive mode do some tests at the command line, do:

# echo user_u:user_r:user_t > /proc/whatever-it-is
# id

Then id should report a context of user_u:user_r:user_t if the context change is working correctly.

You can't change your context on-the-fly like this, since we only permit context transitions upon exec, and you cannot effectively test /proc/self/attr/exec from the shell, as the attribute is per-process and automatically cleared upon exec. You can use the 'execcon' utility from libselinux or 'runcon' (formerly runas) from the patched coreutils to try it, e.g. 'execcon root:sysadm_r:sysadm_t /bin/bash'.

Signature · 8 lines
Stephen Smalley <[email protected]>
National Security Agency


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#6 lore raw
In reply to Colin Walters
Quoted 6 lines · Colin Walters

I think I'm pretty close to getting my Debian system running with the SELinux 2.5 bits. However, I can't seem to get either the patched sshd or login to give me the correct context. When I login (using login) as root, I see:

Warning! Could not get new context for tty1, not relabeling.

Did you remember to mount selinuxfs on /selinux, as described in the selinux-doc README? Otherwise, the daemons cannot obtain security policy decisions from the kernel, such as the new label for the tty.

Quoted 5 lines

...followed by a bunch of avc denials about tty1 and /root, and then my context is:

meta:~# id
uid=0(root) gid=0(root) groups=0(root) context:system_u:system_r:user_t

Verify that execcon or runcon work for you, e.g.
execcon root:sysadm_r:sysadm_t bash

If so, then you need to verify that your login patch is properly calling setexeccon() with a legitimate user context prior to exec'ing the shell and that the call is not failing.

Signature · 8 lines
Stephen Smalley <[email protected]>
National Security Agency


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#7 lore raw
Quoted 3 lines · Stephen Smalley

Did you remember to mount selinuxfs on /selinux, as described in the selinux-doc README? Otherwise, the daemons cannot obtain security policy decisions from the kernel, such as the new label for the tty.

Yes, I've done that.

Quoted 2 lines

Verify that execcon or runcon work for you, e.g.
execcon root:sysadm_r:sysadm_t bash

It seems to work fine; my context is correct afterwards.

Quoted 3 lines

If so, then you need to verify that your login patch is properly calling setexeccon() with a legitimate user context prior to exec'ing the shell and that the call is not failing.

Ok, thanks for the hint, I'll double check that.

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.

#8 lore raw
Quoted 7 lines · Colin Walters
Quoted 3 lines · Stephen Smalley

Did you remember to mount selinuxfs on /selinux, as described in the selinux-doc README? Otherwise, the daemons cannot obtain security policy decisions from the kernel, such as the new label for the tty.

Yes, I've done that.

Try running the compute_relabel utility from libselinux, e.g.:

$ compute_relabel root:sysadm_r:sysadm_t system_u:object_r:tty_device_t chr_file

It should display something like:
root:object_r:sysadm_tty_device_t

If that works, then check what login is actually passing to security_compute_relabel, e.g. is the user context a valid context? is the tty context that is being passed correct?

Signature · 8 lines
Stephen Smalley <[email protected]>
National Security Agency


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
#9 lore raw
In reply to Colin Walters
Quoted 6 lines · Colin Walters

Hi,

I think I'm pretty close to getting my Debian system running with the SELinux 2.5 bits. However, I can't seem to get either the patched sshd or login to give me the correct context. When I login (using login) as root, I see:

Also, make sure that the login process is running in the correct domain, i.e. local_login_t. You need to relabel /bin/login to login_exec_t after updating it.

Signature · 8 lines
Stephen Smalley <[email protected]>
National Security Agency


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.