From: Joshua Brindle <method@manicmethod.com>
To: Daniel J Walsh <dwalsh@redhat.com>
Cc: selinux@tycho.nsa.gov, Stephen Smalley <sds@tycho.nsa.gov>
Subject: Re: policycoreutils, sepolgen (sepolgen-ifgen) issues on Debian
Date: Wed, 16 Sep 2009 15:52:05 -0400 [thread overview]
Message-ID: <4AB141E5.6040808@manicmethod.com> (raw)
In-Reply-To: <4AB12072.3060004@redhat.com>
Daniel J Walsh wrote:
> On 09/16/2009 01:14 PM, Joshua Brindle wrote:
>>
>> Daniel J Walsh wrote:
>>> On 09/16/2009 11:01 AM, Joshua Brindle wrote:
>>>
>>>> Manoj Srivastava wrote:
>>>>
>>>>> On Mon, Aug 17 2009, Christopher J. PeBenito wrote:
>>>>>
>>>>>
>>>>>
>>>>>> On Fri, 2009-08-14 at 11:50 -0500, Manoj Srivastava wrote:
>>>>>>
>>>>>>
>>>>>>> On Fri, Aug 14 2009, Manoj Srivastava wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> I am running into an issue with sepolgen on Debian. Debian
>>>>>>>> ships
>>>>>>>> more than one version of the refpolicy, a default one, and a
>>>>>>>> MLS enabled one. So, the include files live in either
>>>>>>>> /usr/share/selinux/{default,mls}/include
>>>>>>>>
>>>>>>>> sepolgen (in src/sepolgen/defaults.py) sets
>>>>>>>> refpolicy_devel() to
>>>>>>>> a single location -- and thus, only one version of the security
>>>>>>>> policy
>>>>>>>> may be supported. So, sepolgen-ifgen from policycoreutils can
>>>>>>>> only work
>>>>>>>> with one policy, which may not be the one installed on the target
>>>>>>>> machine. Could this be made configurable, somehow? As far as I
>>>>>>>> can
>>>>>>>> see, sepolgen's python library does not offer any way to set the
>>>>>>>> value.
>>>>>>>>
>>>>>>>> It would be nice if the location of the include directory
>>>>>>>> could
>>>>>>>> be looked for from a PATH like variable setting, to make it
>>>>>>>> easier for
>>>>>>>> distributions to ship more than one policy, or for end users to
>>>>>>>> experiment with other policies without have to overwrite the
>>>>>>>> single
>>>>>>>> default.
>>>>>>>>
>>>>>>>>
>>>>>>> Well, here is a kind of proof-of-concept patch (python is
>>>>>>> not my
>>>>>>> strong suit), and I have only tested in that it allows the
>>>>>>> package to
>>>>>>> compile, and the following code works:
>>>>>>>
>>>>>>>
>>>>>> [...]
>>>>>>
>>>>>>
>>>>>>> def refpolicy_makefile():
>>>>>>> - return refpolicy_devel() + "/Makefile"
>>>>>>> + chooser = PathChoooser("/etc/selinux/sepolgen.conf")
>>>>>>> + return chooser("Makefile")
>>>>>>>
>>>>>>> def headers():
>>>>>>> - return refpolicy_devel() + "/include"
>>>>>>> -
>>>>>>> + chooser = PathChoooser("/etc/selinux/sepolgen.conf")
>>>>>>> + return chooser("include")
>>>>>>> +
>>>>>>>
>>>>>>>
>>>>>> Why are you making another config file rather than just get the policy
>>>>>> name from /etc/selinux/config via selinux_getpolicytype()?
>>>>>>
>>>>>>
>>>>> This will work well for Debian, since the development
>>>>> files are
>>>>> installed under "/usr/share/selinux/" in a subdirectory named
>>>>> after the
>>>>> policy. I was not sure that this convention was followed in other
>>>>> distributions, though. While I am not certain, google implies
>>>>> that in
>>>>> fedora policy type is targeted, but the devel files do not live in
>>>>> /usr/share/selinux/targeted.[0]. Given that, perhaps it is better to
>>>>> let the user provide guidance about how to map the policy type to a
>>>>> directory?
>>>>>
>>>>> Also, I must confess I had forgotten about this call.
>>>>>
>>>>> However, a patch with this is trivial, so an alternate patch
>>>>> follows. (Not sure this will work for fedora, so caveat emptor)
>>>>>
>>>>> manoj
>>>>> [0]
>>>>> http://docs.fedoraproject.org/selinux-user-guide/f11/en-US/chap-Security-Enhanced_Linux-Working_with_SELinux.html
>>>>>
>>>>>
>>>>>
>>>>> --8<---------------cut here---------------start------------->8---
>>>>>
>>>>> If the user installs a policy whose development files do not live under
>>>>> /usr/share/selinux/devel/include, sepolgen wqould not work. Debian, for
>>>>> instance, installs under:
>>>>> /usr/share/selinux/{default,mls}/include
>>>>>
>>>>> This patch uses selinux_getpolicytype() to determine the policy
>>>>> type, and
>>>>> assumes that there is one-on-one correspondence between policytype and
>>>>> the directory the development files live in.
>>>>>
>>>>> Signed-off-by: Manoj Srivastava<srivasta@debian.org>
>>>>> ---
>>>>> src/sepolgen/defaults.py | 4 +++-
>>>>> src/sepolgen/module.py | 2 +-
>>>>> 2 files changed, 4 insertions(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/src/sepolgen/defaults.py b/src/sepolgen/defaults.py
>>>>> index 45ce61a..85e5fb0 100644
>>>>> --- a/src/sepolgen/defaults.py
>>>>> +++ b/src/sepolgen/defaults.py
>>>>> @@ -21,6 +21,8 @@
>>>>> Various default settings, including file and directory locations.
>>>>> """
>>>>>
>>>>> +import selinux
>>>>> +
>>>>> def data_dir():
>>>>> return "/var/lib/sepolgen"
>>>>>
>>>>> @@ -31,7 +33,7 @@ def interface_info():
>>>>> return data_dir() + "/interface_info"
>>>>>
>>>>> def refpolicy_devel():
>>>>> - return "/usr/share/selinux/devel"
>>>>> + return "/usr/share/selinux/" + selinux.selinux_getpolicytype()[1]
>>>>>
>>>>> def refpolicy_makefile():
>>>>> return refpolicy_devel() + "/Makefile"
>>>>> diff --git a/src/sepolgen/module.py b/src/sepolgen/module.py
>>>>> index edd24c6..355c9b8 100644
>>>>> --- a/src/sepolgen/module.py
>>>>> +++ b/src/sepolgen/module.py
>>>>> @@ -120,7 +120,7 @@ class ModuleCompiler:
>>>>> self.semodule_package = "/usr/bin/semodule_package"
>>>>> self.output = output
>>>>> self.last_output = ""
>>>>> - self.refpol_makefile = "/usr/share/selinux/devel/Makefile"
>>>>> + self.refpol_makefile = "/usr/share/selinux/" +
>>>>> selinux.selinux_getpolicytype()[1] + "/Makefile"
>>>>> self.make = "/usr/bin/make"
>>>>>
>>>>> def o(self, str):
>>>>>
>>>>>
>>>> This will break Fedora/RHEL AFAIK. I don't necessarily like that RH has
>>>> interface files in /usr/share/selinux/devel rather than
>>>> /usr/share/selinux/<policy>/devel or similar but we can't break them.
>>>>
>>>> Dan, any chance you could change the location of the interface files?
>>>>
>>>>
>>>>
>>> We could carry a patch although I don't think anyone is shipping
>>> different interfaces for different policies.
>>>
>>>
>> I'm not willing to break upstream behavior and force you to carry a
>> patch for something that previously worked.
>>
>> Perhaps not for distro shipped policies but for custom policies I know
>> interfaces are changed and if the developers on those end systems want
>> to use sepolgen for interface matching they have to over write the
>> distro shipped interface files.
>>
>>> We could add a link in each policy types back to the devel environment.
>>> Or do /usr/share/selinux/POLICYTYPE/devel/Makefile and on RHEL and
>>> Fedora systems
>>> have /usr/share/selinux/POLICYTYPE/devel -> /usr/share/selinux/devel/
>>>
>>>
>> wasn't this done in the past? I remember a symlink being there but can't
>> remember why it was removed (unless I'm misremembering)
>>
> It was there in the past, but "devel" was a separate package and in some cases we ended up with a dangling link.
> devel has since been moved into the selinux-policy package so it would not be a problem any longer.
>
So you'd need to update the policy package before we pull this in
upstream. Just let me know when that has been done.
--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to majordomo@tycho.nsa.gov with
the words "unsubscribe selinux" without quotes as the message.
next prev parent reply other threads:[~2009-09-16 19:52 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-08-14 14:59 policycoreutils, sepolgen (sepolgen-ifgen) issues on Debian Manoj Srivastava
2009-08-14 16:50 ` Manoj Srivastava
2009-08-16 14:52 ` Manoj Srivastava
2009-08-17 12:36 ` Christopher J. PeBenito
2009-08-17 15:40 ` Manoj Srivastava
2009-09-16 15:01 ` Joshua Brindle
2009-09-16 16:44 ` Manoj Srivastava
2009-09-16 20:56 ` Joshua Brindle
2009-09-18 21:56 ` Manoj Srivastava
2009-09-16 17:10 ` Daniel J Walsh
2009-09-16 17:14 ` Joshua Brindle
2009-09-16 17:29 ` Daniel J Walsh
2009-09-16 19:52 ` Joshua Brindle [this message]
2009-09-16 20:23 ` Daniel J Walsh
2010-05-13 19:11 ` Stephen Smalley
2010-05-13 19:15 ` Daniel J Walsh
2010-05-13 19:28 ` Stephen Smalley
2010-05-13 19:53 ` Daniel J Walsh
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4AB141E5.6040808@manicmethod.com \
--to=method@manicmethod.com \
--cc=dwalsh@redhat.com \
--cc=sds@tycho.nsa.gov \
--cc=selinux@tycho.nsa.gov \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.