Re: [RFC v2 1/3] security: Add safe, dynamic (runtime-loadable) hook support

[Sargun Dhillon]

On Sun, Dec 10, 2017 at 11:28 PM, James Morris
<james.l.morris@xxxxxxxxxx> wrote:
> On Fri, 8 Dec 2017, Sargun Dhillon wrote:
>
>> The primary purpose of this patchset is to facilitate the development of
>> out-of-tree minor LSMs.
>
> This is not a valid use-case for the mainline kernel: we don't add
> features for out of tree code.
>
> Please reconsider your rationale :-)
>
>
> - James
> --
> James Morris
> <james.l.morris@xxxxxxxxxx>
>
Even a la livepatch? One of our primary usecases is for a safer
version of livepatch, with simpler consistency guarantees. The model
is to be able to generate a module that targets a specific CVE, and
roll it out via DKMS, or similar to a heterogenous fleet.

The other one is to be able to dynamically compile LSMs a la
systemtap, using a DSL. One of the biggest problems we've seen is that
the capability of LSMs is limited by the policy. Changing the policy
requires in-kernel changes, whereas a compiler in userspace is
significantly easier to modify.