On 03/16/2015 11:20 AM, Higgs, Stephen wrote:
>>>>>>> Hello all,
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> If there is a more appropriate forum for this question please let me
>> know:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> I have a system that uses confined users by default and some files
>>>>>>> are managed by a puppet server. When I run (via run_init) the
>>>>>>> puppet startup script, I get the following avc log:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> avc: denied { relabelto } for pid=30707 comm="puppet" name="crl.pem"
>>>>>>> dev=dm-1 ino=527257 scontext=system_u:system_r:puppet_t:s0
>>>>>>> tcontext=system_u:object_r:puppet_var_lib_t:s0:c0.c1023
>>>>>>> tclass=file
>>>>>>>
>>>>>>> I added "typeattribute puppet_t can_change_object_identity" and
>>>>>>> appropriate "allow" statements to the puppet_t type after reading
>>>>>>> the constraints in the targeted policy. However, it was the
>>>>>>> category "s0:c0.c1023" that was also preventing puppet from
>>>>>>> relabeling the crl.pem file.
>>>>>>>
>>>>>>> I was able to fix this by manually relabeling the file to "s0"
>>>>>>> instead of "s0:c0.c1023". My question is, how *should* I handle
>>>>>>> this so puppet can handle the relabel of the category?
>>>>>>
>>>>>> It requires an appropriate attribute for the mcs or mls constraint
>>>>>> that is blocking access. Which attribute depends on your policy;
>>>>>> MCS in particular has changed a lot over time in Fedora and RHEL.
>>>>>> What distro &
>>>> version?
>>>>>>
>>>>>
>>>>> I'm using CentOS / RedHat 6.6, targeted reference policy 24.
>>>>
>>>> Hmmm...looking at selinux-policy-3.7.19-260.el6.src.rpm,
>>>> serefpolicy-3.719/policy/mcs has this:
>>>>
>>>> # New filesystem object labels must be dominated by the relabeling
>>>> subject # clearance, also the objects are single-level.
>>>> mlsconstrain file { create relabelto }
>>>> (( h1 dom h2 ) and ( l2 eq h2 ));
>>>>
>>>> So no attributes are exempted from that constraint; your only option
>>>> is to run puppet ranged (i.e. as
>>>> system_u:system_r:puppet_t:s0-s0:c0.c1023)
>>>> so that its high level dominates any potential file level.
>>>>
>>>> You should be able to do that with a range_transition rule, e.g.
>>>> range_transition initrc_t puppet_exec_t:process s0 - s0:c0.c0123;
>>>> (assuming that the puppet entrypoint is labeled with puppet_exec_t).
>>>
>>> Thanks Stephen, this makes sense to me, but I can't get that statement to
>> compile in my policy module:
>>>
>>> Compiling targeted puppet module
>>> /usr/bin/checkmodule: loading policy configuration from tmp/puppet.tmp
>>> puppet.te":14:ERROR 'unknown level s0-s0 used in range_transition
>> definition' at token ';' on line 1041:
>>> range_transition initrc_t puppet_exec_t:process s0-s0:c0.c1023;
>>> #init_ranged_daemon_domain(puppet_t,puppet_exec_t,s0-s0:c0.c1023);
>>> /usr/bin/checkmodule: error(s) encountered while parsing configuration
>>> make: *** [tmp/puppet.mod] Error 1
>>>
>>> I did try checkmodule as well, and I tried using the
>> init_ranged_daemon_domain macro. Here is the policy module that I am
>> trying to compile:
>>>
>>> module puppet 1.2;
>>> require {
>>> type puppet_t;
>>> type puppet_exec_t;
>>> type initrc_t;
>>> attribute can_change_object_identity;
>>> class process { transition };
>>> }
>>> typeattribute puppet_t can_change_object_identity;
>>> #init_ranged_daemon_domain(puppet_t,puppet_exec_t,s0-s0:c0.c1023);
>>> range_transition initrc_t puppet_exec_t:process s0-s0:c0.c1023;
>>>
>>> I feel like I'm close, but perhaps I'm missing how to import the level
>> definitions?
>>
>> As Dominick suggested, whitespace unfortunately matters for the MLS range
>> specification - you need whitespace around the - (dash).
>> checkpolicy scanner issue introduced when IDENTIFIER was expanded to include
>> dash characters to support usage in filesystem type names and user names
>> IIRC. Should probably refactor that.
>>
>
> Thanks everybody for your input, the format
>
> ifdef(`enable_mcs',`
> init_ranged_daemon_domain(puppet_t, puppet_exec_t, s0 - s0:c0.c1023)
> ')
>
> did the trick, and compiled with the devel makefile. For posterity, note that it did not compile with checkmodule, the spaces around the dash in the range level was required, and the ifdef format was also required.
>
> Thanks again,
If you used the original range_transition rule I specified, including
the whitespace, then it should have compiled with checkmodule, but to
use the macroized version suggested by Miroslav, you have to build with
the devel Makefile which applies m4 and includes the interface files
that define the macros.
_______________________________________________
Selinux mailing list
Selinux@xxxxxxxxxxxxx
To unsubscribe, send email to Selinux-leave@xxxxxxxxxxxxx.
To get help, send an email containing "help" to Selinux-request@xxxxxxxxxxxxx.
