On Feb 4, 2009, at 10:14 AM, Stephen Smalley wrote:
On Wed, 2009-02-04 at 11:12 -0500, Daniel J Walsh wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Stephen Smalley wrote:
On Tue, 2009-02-03 at 16:31 -0600, Xavier Toth wrote:
On Tue, Feb 3, 2009 at 3:49 PM, Xavier Toth <txtoth@xxxxxxxxx>
wrote:
On Tue, Feb 3, 2009 at 3:28 PM, Stephen Smalley
<sds@xxxxxxxxxxxxx> wrote:
On Tue, 2009-02-03 at 13:59 -0600, Xavier Toth wrote:
On Tue, Feb 3, 2009 at 10:38 AM, Stephen Smalley <sds@xxxxxxxxxxxxx
> wrote:
On Tue, 2009-02-03 at 10:28 -0600, Xavier Toth wrote:
I have an app that wasn't working in enforcing but there are
no AVCs.
So I did 'semodule -DB' to see if there were any dontaudit
denials and
restarted the app. The problem is that the app then ran
fine. So I
tried load_policy which had no affect and 'semodule -B'
which makes it
work. Any ideas what could be happening? I've verified with
'semodule
--list' that the module is loaded prior to doing the
'semodule -B'.
- How was the app failing?
This is our security banner app that draw a window across the
top of
the screen with the users MLS range and with the appropriate
background color. When it fails there is just a window with a
gray
background no text or color.
- Did you try running the app in permissive as well?
- Is this reproducible at all or are you unable to reproduce
the
application failure now under any conditions?
- Did the app create/use any transient resources (temporary
files,
system v ipc objects, etc) that could have prevented it from
succeeding
on subsequent execution if they weren't properly cleaned up
on prior
exit?
After further investigation I found that a call to
getseuserbyname for
the login user is returning the user name passed in and
nothing for
the range which would be used in the banner. During our
installation
we don't explicitly map this user to a SELinux user but our
experience
has been that the when there is no mapping the user and range
of the
'__default__' login are returned. Indeed once I rebuild policy
this
appears to be what is happening. How rebuilding and reloading
policy
would affect this is unclear.
If the installed seusers file (i.e. /etc/selinux/$SELINUXTYPE/
seusers)
did not exist originally or was unreadable (e.g. wrong context
or mode),
then getseuserbyname() would behave the way you described.
Rebuilding
policy would have caused the regenerated seusers file to be
installed,
possibly with different context or mode than the original state.
--
Stephen Smalley
National Security Agency
You nailed it for some reason seusers is SystemHigh. Still
investigating ... Thanks
Ted
The plot thickens. semodule -B causes seusers to be relabeled
SystemLow. restorecon /etc/selinux/mls/seusers causes the file to
be
relabeled SystemHigh. Which is right?
I don't think that semodule/libsemanage explicitly set the context
of
files they create, so they would just follow the default policy
behavior
(in this case, inheriting the level of the creating process and
the type
of the parent directory).
restorecon just applies whatever the file_contexts configuration
specifies, which appears to be SystemHigh in the mls policy. I'm
not
sure if that is truly justified.
Since all the login programs have to read this file it should be at
the
SystemLow, So that a login program running at a SystemLow can work.
Makes sense to me, but I assume that someone thought otherwise or it
wouldn't have been marked as SystemHigh in the first place? Does the
LSPP impose a particular restriction on it? There are
other /etc/selinux files in file_contexts as well that are marked with
SystemHigh (or to be precise, s15:c0.c1023) in the -mls policy.
Here is the original rationale:
From: Darrel Goeddel ...
The theory is that these files can contain the lables up to
systemhigh. The
mere existence of the label is classified just like data having that
label.
Since the seusers file contains the clearances, those labels need to
be
protected with the appropriate labels (ain't that fun...). This is
same
reason that the label translation daemon will not translate things for
you if you are not cleared to know about the translation. We are
trying
to prevent someone who is cleared at secret (let's say s7) to be able
to read the seusers file and find out that the label (s10:c0-c100)
is in
use on the system, possibly indicating that the machine is a higher-
value
target.
I think this over classifies the file. Since the contexts are raw, I
don't see a reason for seusers to be SystemHigh. I'll ask our
accreditors, but the answer will not be swiftly forthcoming.
WRT the other files, setrans.conf should be SystemHigh because there
is sensitivity to category names in some communities.
joe
--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to majordomo@xxxxxxxxxxxxx with
the words "unsubscribe selinux" without quotes as the message.
