[RFC PATCH 17/29] lsm: introduce an initcall mechanism into the LSM framework

[Paul Moore]

On Wed, Apr 9, 2025 at 5:16 PM Kees Cook <kees at kernel.org> wrote:
> On Wed, Apr 09, 2025 at 02:50:02PM -0400, Paul Moore wrote:
> > Currently the individual LSMs register their own initcalls, and while
> > this should be harmless, it can be wasteful in the case where a LSM
> > is disabled at boot as the initcall will still be executed.  This
> > patch introduces support for managing the initcalls in the LSM
> > framework, and future patches will convert the existing LSMs over to
> > this new mechanism.
> >
> > Only initcall types which are used by the current in-tree LSMs are
> > supported, additional initcall types can easily be added in the future
> > if needed.
> >
> > Signed-off-by: Paul Moore <paul at paul-moore.com>
> > ---
> >  include/linux/lsm_hooks.h | 33 ++++++++++++---
> >  security/lsm_init.c       | 89 +++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 117 insertions(+), 5 deletions(-)
> >
> > diff --git a/include/linux/lsm_hooks.h b/include/linux/lsm_hooks.h
> > index a7ecb0791a0f..0d2c2a017ffc 100644
> > --- a/include/linux/lsm_hooks.h
> > +++ b/include/linux/lsm_hooks.h
> > @@ -148,13 +148,36 @@ enum lsm_order {
> >       LSM_ORDER_LAST = 1,     /* This is only for integrity. */
> >  };
> >
> > +/**
> > + * struct lsm_info - Define an individual LSM for the LSM framework.
> > + * @id: LSM name/ID info
> > + * @order: ordering with respect to other LSMs, optional
> > + * @flags: descriptive flags, optional
> > + * @blobs: LSM blob sharing, optional
> > + * @enabled: controlled by CONFIG_LSM, optional
> > + * @init: LSM specific initialization routine
> > + * @initcall_pure: LSM callback for initcall_pure() setup, optional
> > + * @initcall_early: LSM callback for early_initcall setup, optional
> > + * @initcall_core: LSM callback for core_initcall() setup, optional
> > + * @initcall_subsys: LSM callback for subsys_initcall() setup, optional
> > + * @initcall_fs: LSM callback for fs_initcall setup, optional
> > + * @nitcall_device: LSM callback for device_initcall() setup, optional
> > + * @initcall_late: LSM callback for late_initcall() setup, optional
> > + */
>
> Yay! Proper kerndoc. :)

 ;)

> > +/**
> > + * security_initcall_late - Run the LSM late initcalls
> > + */
> > +static int __init security_initcall_late(void)
> > +{
> > +     int rc;
> > +
> > +     rc = lsm_initcall(late);
> > +     lsm_pr_dbg("all enabled LSMs fully activated\n");
> > +
> > +     return rc;
> > +}
> > +late_initcall(security_initcall_late);
>
> You'd need a new place for the lsm_pr_dbg, but these are all just
> copy/paste. These could be macro-ified too?

If we didn't want to move the other LSM framework initcalls into these
initcalls (yes, I prefer it this way), or add the LSM_STARTED_ADD
event at the end, I would tend to agree with you.  Let's leave it
as-is for now, if something changes in the future wrt to any of things
above we can revisit this.

I'm also somewhat hopeful that this work will bring attention to the
different initcall types/levels that are in use by various LSMs, I
suspect there are a few LSMs which are currently using multiple
initcall types that could be consolidated into one.  That's not
something I wanted to tackle in this patchset, but if we could reduce
the number of initcall types that we use in the LSM subsystem as a
whole this may not really be an issue of any significance ...

-- 
paul-moore.com