All of lore.kernel.org
 help / color / mirror / Atom feed
From: Stephen Smalley <sds@tycho.nsa.gov>
To: Nicolas Iooss <nicolas.iooss@m4x.org>
Cc: selinux@vger.kernel.org
Subject: Re: [PATCH] libsemanage: set selinux policy root to match semanage root or storename
Date: Mon, 19 Nov 2018 08:52:50 -0500	[thread overview]
Message-ID: <42f12202-0298-cbc7-2ce9-7cf6e7020049@tycho.nsa.gov> (raw)
In-Reply-To: <a29de7d8-7efb-f8ce-1c29-01dd4c43bb73@tycho.nsa.gov>

On 11/8/18 9:20 AM, Stephen Smalley wrote:
> On 11/7/18 3:45 PM, Nicolas Iooss wrote:
>> On Tue, Nov 6, 2018 at 8:18 PM Stephen Smalley <sds@tycho.nsa.gov> wrote:
>>>
>>> As reported in #109, semodule -p /path/to/policyroot -s minimum -n -B
>>> tries to use /etc/selinux/targeted/booleans.subs_dist.  This is because
>>> it invokes the libselinux selinux_boolean_sub() interface, which uses
>>> the active/installed policy files rather than the libsemanage ones.
>>>
>>> To fix, we need to set the selinux policy root when either the semanage
>>> root or the semanage storename is set.  When setting the semanage root,
>>> we need to prepend the semanage root to the selinux policy root.  When
>>> setting the semanage storename, we need to replace the last component
>>> of the selinux policy root with the new storename.
>>>
>>> Test:
>>> strace semodule -p ~/policy-root -s minimum -n -B
>>>
>>> Before:
>>> openat(AT_FDCWD, "/etc/selinux/targeted/booleans.subs_dist", 
>>> O_RDONLY|O_CLOEXEC) = 5
>>>
>>> After:
>>> openat(AT_FDCWD, 
>>> "/home/sds/policy-root/etc/selinux/minimum/booleans.subs_dist", 
>>> O_RDONLY|O_CLOEXEC) = 5
>>>
>>> Fixes https://github.com/SELinuxProject/selinux/issues/109
>>>
>>> Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
>>> ---
>>>   libsemanage/src/handle.c | 29 ++++++++++++++++++++++++++++-
>>>   1 file changed, 28 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/libsemanage/src/handle.c b/libsemanage/src/handle.c
>>> index a6567bd4..c163e553 100644
>>> --- a/libsemanage/src/handle.c
>>> +++ b/libsemanage/src/handle.c
>>> @@ -43,8 +43,21 @@ static char *private_semanage_root = NULL;
>>>
>>>   int semanage_set_root(const char *root)
>>>   {
>>> +       char *new_selinux_root = NULL;
>>> +
>>> +       asprintf(&new_selinux_root, "%s%s", root, 
>>> selinux_policy_root());
>>> +       if (!new_selinux_root)
>>> +               return -1;
>>
>> https://travis-ci.org/SELinuxProject/selinux/builds/451528669 failed
>> because the return value of asprintf needs to be checked instead of
>> new_selinux_root. http://man7.org/linux/man-pages/man3/asprintf.3.html
>> states:
>>
>>      If memory allocation wasn't possible, or some other error occurs,
>> these functions will return -1, and the contents of strp are
>> undefined.
>>
>> [...]
>>
>>> +
>>> +       char *newroot = NULL;
>>> +       asprintf(&newroot, "%s%s", root, storename);
>>> +       assert(newroot);
>>
>> Same here.
> 
> Unfortunately there are more fundamental problems with the patch as 
> well, e.g.
> 
> 1) Mutating selinux_policy_root breaks the test of whether we are 
> installing to the active store in semanage_install_final_tmp(), which 
> will cause policy reloads to be triggered when performing an operation 
> on a different store, a different root, or both.
> 
> 2) semanage_select_store() is supposed to only modify the per-handle 
> state.  Changing selinux policy root is global state and would affect 
> any other handles.
> 
> 3) Multiple semanage_set_root() calls would yield the wrong result.
> 
> 4) An allocation failure for private_semanage_root would leave the 
> policy root modified.
> 
> Not sure of the best approach here.  Could possibly just modify the 
> policy root around the selinux_boolean_sub() call and leave the rest 
> unchanged...

I guess the underlying bug here is that booleans.subs_dist is not itself 
managed via libsemanage.  If it was managed and therefore lived within 
the policy store, then libsemanage could access the appropriate 
booleans.subs_dist file without using the libselinux interface at all, 
and thus would not need to switch the selinux policy root. Moving 
booleans.subs_dist to being a managed file however would be a more 
substantial change.  The short term fix might be to switch the selinux 
policy root around the selinux_boolean_sub() call, with a longer term 
fix of taking booleans.subs_dist to being a managed file.


      reply	other threads:[~2018-11-19 13:50 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-11-06 19:20 [PATCH] libsemanage: set selinux policy root to match semanage root or storename Stephen Smalley
2018-11-07 20:45 ` Nicolas Iooss
2018-11-08 14:20   ` Stephen Smalley
2018-11-19 13:52     ` Stephen Smalley [this message]

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=42f12202-0298-cbc7-2ce9-7cf6e7020049@tycho.nsa.gov \
    --to=sds@tycho.nsa.gov \
    --cc=nicolas.iooss@m4x.org \
    --cc=selinux@vger.kernel.org \
    /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.