On Thu, 4 Oct 2018, Kees Cook wrote:
> On Thu, Oct 4, 2018 at 10:49 AM, James Morris <jmorris@xxxxxxxxx> wrote:
> > On Wed, 3 Oct 2018, Kees Cook wrote:
> >> Then someone boots the system with:
> >>
> >> selinux=1 security=selinux
> >>
> >> In what order does selinux get initialized relative to yama?
> >> (apparmor, flagged as a "legacy major", would have been disabled by
> >> the "security=" not matching it.)
> >
> > It doesn't, it needs to be specified in one place.
> >
> > Distros will need to update boot parameter handling for this kernel
> > onwards. Otherwise, we will need to carry this confusing mess forward
> > forever.
>
> Are you saying that you want to overrule Paul and Stephen about
> keeping "selinux=1 secuiryt=selinux" working?
Not overrule, but convince.
At least, deprecate selinux=1 and security=X, but not extend it any
further.
> > In my most recent suggestion, there is no '!' disablement, just
> > enablement. If an LSM is not listed in CONFIG_LSM="", it's not enabled.
>
> And a user would need to specify ALL lsms on the "lsm=" line?
>
Yes, the ones they want enabled.
> What do you think of my latest proposal? It could happily work all
> three ways: old boot params and security= work ("selinux=1
> security=selinux" keeps working), individual LSM enable/disable works
> ("lsm=+loadpin"), and full LSM ordering works
> ("lsm=each,lsm,in,order,here"):
>
> https://lore.kernel.org/lkml/CAGXu5jJJit8bDNvgXaFkuvFPy7NWtJW2oRWFbG-6iWk0+A1qng@xxxxxxxxxxxxxx/
>
I think having something like +yama will still lead to confusion.
Explicitly stating each enabled LSM in order is totally unambiguous.
If people are moving away from the distro defaults, and there is no
high-level interface to manage this, it seems to me there's a deeper
issue with the distro.
--
James Morris
<jmorris@xxxxxxxxx>
