[RFC PATCH] lsm: add the inode_free_security_rcu() LSM implementation hook

[Paul Moore]

On Mon, Jul 15, 2024 at 9:35 AM Mickaël Salaün <mic at digikod.net> wrote:
> On Wed, Jul 10, 2024 at 12:20:18PM -0400, Paul Moore wrote:
> > On Wed, Jul 10, 2024 at 6:40 AM Mickaël Salaün <mic at digikod.net> wrote:
> > > On Tue, Jul 09, 2024 at 10:40:30PM -0400, Paul Moore wrote:

...

> > > However, I'm wondering if we could backport this patch down to v5.15 .
> > > I guess not, so I'll need to remove this hook implementation for
> > > Landlock, backport it to v5.15, and then you'll need to re-add this
> > > check with this patch.  At least it has been useful to spot this inode
> > > issue, but it could still be useful to spot potential memory leaks with
> > > a negligible performance impact.
> >
> > Yes, it's a bit complicated with the IMA/EVM promotion happening
> > fairly recently.  I'm marking the patch with a stable tag, but
> > considering we're at -rc7 and this needs at least one more respin,
> > testing, ACKs, etc. it might not land in Linus' tree until sometime
> > post v6.11-rc1.  Considering that, I might suggest dropping the
> > Landlock hook in -stable and re-adding it to Linus' tree once this fix
> > lands, but that decision is up to you.
>
> I would prefer to backport the new hook.  I can help with that.  In
> fact, I tried to backport the removal of the hook for Landlock, and it
> requires the same amount of work, so it would be better to work
> together.  That should also ease future backports impacting the same
> part of the code.

Okay, let's get the initial v6.11 LSM PR merged (I just sent it to
Linus) and then I'll post the updated patchset and a proper patch for
review.

> BTW, while trying to backport that to linux-5.15.y I noticed that a fix
> is missing in this branch: the default return value for the
> inode_init_security hook, see commit 6bcdfd2cac55 ("security: Allow all
> LSMs to provide xattrs for inode_init_security hook").

Likely a casualty of a merge conflict; I haven't noticed the stable
kernel folks doing any manual merging of LSM patches that fail an
automated merge.  You can always do the merge and send it to them.

-- 
paul-moore.com