Re: [PATCH 2/3] lsm: introduce security_lsm_manage_policy hook

[John Johansen]

On 5/7/25 08:37, Maxime Bélair wrote:
> On 5/7/25 08:19, Song Liu wrote:
> > On Tue, May 6, 2025 at 7:40 AM Maxime Bélair
> > <maxime.belair@xxxxxxxxxxxxx> wrote:
> > > Define a new LSM hook security_lsm_manage_policy and wire it into the
> > > lsm_manage_policy() syscall so that LSMs can register a unified interface
> > > for policy management. This initial, minimal implementation only supports
> > > the LSM_POLICY_LOAD operation to limit changes.
> > >
> > > Signed-off-by: Maxime Bélair <maxime.belair@xxxxxxxxxxxxx>
> > [...]
> > > diff --git a/security/security.c b/security/security.c
> > > index fb57e8fddd91..256104e338b1 100644
> > > --- a/security/security.c
> > > +++ b/security/security.c
> > > @@ -5883,6 +5883,27 @@ int security_bdev_setintegrity(struct block_device *bdev,
> > >   }
> > >   EXPORT_SYMBOL(security_bdev_setintegrity);
> > >
> > > +/**
> > > + * security_lsm_manage_policy() - Manage the policies of LSMs
> > > + * @lsm_id: id of the lsm to target
> > > + * @op: Operation to perform (one of the LSM_POLICY_XXX values)
> > > + * @buf:  userspace pointer to policy data
> > > + * @size: size of @buf
> > > + * @flags: lsm policy management flags
> > > + *
> > > + * Manage the policies of a LSM. This notably allows to update them even when
> > > + * the lsmfs is unavailable is restricted. Currently, only LSM_POLICY_LOAD is
> > > + * supported.
> > > + *
> > > + * Return: Returns 0 on success, error on failure.
> > > + */
> > > +int security_lsm_manage_policy(u32 lsm_id, u32 op, void __user *buf,
> > > +                              size_t size, u32 flags)
> > > +{
> > > +       return call_int_hook(lsm_manage_policy, lsm_id, op, buf, size, flags);
> >
> > If the LSM doesn't implement this hook, sys_lsm_manage_policy will return 0
> > for any inputs, right? This is gonna be so confusing for users.
>
> Indeed, that was an oversight. It will return -EOPNOTSUPP in the next patch revision.

I think it needs to do more than that. I don't think this should call each LSM, the
infrastructure should filter it and only send it to the LSM identified by the lsm_id