[PATCH] selinux: reorder hooks to make runtime disable less broken

[Casey Schaufler]

On 12/9/2019 5:58 AM, Stephen Smalley wrote:
> On 12/9/19 8:21 AM, Stephen Smalley wrote:
>> On 12/9/19 2:57 AM, Ondrej Mosnacek wrote:
>>> Commit b1d9e6b0646d ("LSM: Switch to lists of hooks") switched the LSM
>>> infrastructure to use per-hook lists, which meant that removing the
>>> hooks for a given module was no longer atomic. Even though the commit
>>> clearly documents that modules implementing runtime revmoval of hooks
>>> (only SELinux attempts this madness) need to take special precautions to
>>> avoid race conditions, SELinux has never addressed this.
>>>
>>> By inserting an artificial delay between the loop iterations of
>>> security_delete_hooks() (I used 100 ms), booting to a state where
>>> SELinux is enabled, but policy is not yet loaded, and running these
>>> commands:
>>>
>>>      while true; do ping -c 1 <some IP>; done &
>>>      echo -n 1 >/sys/fs/selinux/disable
>>>      kill %1
>>>      wait
>>>
>>> ...I was able to trigger NULL pointer dereferences in various places. I
>>> also have a report of someone getting panics on a stock RHEL-8 kernel
>>> after setting SELINUX=disabled in /etc/selinux/config and rebooting
>>> (without adding "selinux=0" to kernel command-line).
>>>
>>> Reordering the SELinux hooks such that those that allocate structures
>>> are removed last seems to prevent these panics. It is very much possible
>>> that this doesn't make the runtime disable completely race-free, but at
>>> least it makes the operation much less fragile.
>>>
>>> An alternative (and safer) solution would be to add NULL checks to each
>>> hook, but doing this just to support the runtime disable hack doesn't
>>> seem to be worth the effort...
>>
>> Personally, I would prefer to just get rid of runtime disable altogether; it also precludes making the hooks read-only after initialization.  IMHO, selinux=0 is the proper way to disable SELinux if necessary.  I believe there is an open bugzilla on Fedora related to this issue, since runtime disable was originally introduced for Fedora.
>
> Also, if we have to retain this support, it seems like this ought to be fixed in the LSM framework especially since it was a change there that broke the SELinux implementation.

Agreed, mostly. Deleting an LSM is fundamentally something the infrastructure
should handle *if* we allow it. Should we decide at some point to allow loadable
modules, as Tetsuo has advocated from time to time, we would need a general
solution. We don't have a general solution now because only SELinux wants it.
The previous implementation was under #ifdef for SELinux. At the time I understood
that there was no interest in investing in it. The implementation passed tests
at the time.

I propose that until such time as someone decides to seriously investigate
loadable security modules* the sole user of the deletion mechanism is
welcome to invest whatever they like in their special case, and I will be
happy to lend whatever assistance I can.

---
* I do not plan to propose an implementation of loadable modules.
  I leave that as an exercise for the next generation.