[PULL REQUEST] Lock down patches

[Matthew Garrett]

On Thu, Feb 28, 2019 at 4:05 PM Mimi Zohar <zohar at linux.ibm.com> wrote:
>
> On Thu, 2019-02-28 at 15:13 -0800, Matthew Garrett wrote:
> > On Thu, Feb 28, 2019 at 2:20 PM Mimi Zohar <zohar at linux.ibm.com> wrote:
> > > Where/when was this latest version of the patches posted?
> >
> > They should have followed this, but git-send-email choked on some
> > reviewed-by: lines so I'm just trying to sort that out.
>
> I'm a little perplexed as to why you would send a pull request, before
> re-posting the patches with the changes for review.

They should be there now. There's no substantive change to the
patches, other than having dropped a few from the series.

> > It's a little more complicated than this. We can't just rely on IMA
> > appraisal - it has to be based on digital signatures, and the existing
> > patch only made that implicit by enabling the secure_boot policy.
>
> Right, which is the reason the IMA architecture specific policy
> requires file signatures. [1][2]

The current patches seem to require ima signatures - shouldn't this
allow ima digests as long as there's an evm signature?

> > I
> > think we do want to integrate these, but there's a few things we need
> > to take into account:
> >
> > 1) An integrated solution can't depend on xattrs, both because of the
> > lagging support for distributing those signatures but also because we
> > need to support filesystems that don't support xattrs
>
> That's not a valid reason for preventing systems that do use IMA for
> verifying the kexec kernel image signature or kernel module signatures
> from enabling "lock down".  This just means that there needs to be
> some coordination between the different signature verification
> methods. [1][2]

I agree, but the current form of the integration makes it impossible
for anyone using an IMA-enabled kernel (but not using IMA) to do
anything unless they have IMA signatures. It's a problem we need to
solve, I just don't think it's a problem we need to solve before
merging the patchset.

> > 2) An integrated solution can't depend on the current secure_boot
> > policy because that requires signed IMA policy updates, but
> > distributions have no way of knowing what IMA policy end users require
>
> Both the "CONFIG_IMA_APPRAISE_REQUIRE_KEXEC_SIGS" and the IMA
> architecture policy rules persist after loading a custom policy.
>  Neither of them require loading or signing a custom policy.

The previous version of the lockdown patchset sets the secure_boot
policy when lockdown is enabled, which does require that any custom
policy be signed.

> > In any case, I do agree that we should aim to make this more
> > reasonable - having orthogonal signing code doesn't benefit anyone.
> > Once there's solid agreement on that we can extend this support.
> >
>
> Having multiple signature verification methods is going to be around
> for a while.  The solution is to coordinate the signature verification
> methods, without requiring both types of signatures. [1][2]

Agree, and once we have a solution to this we should integrate that
with lockdown. I don't think merging this first makes that any harder.
Importantly, this version of the patchset doesn't enable lockdown
automatically unless explicitly configured to do so, which means you
can build a lockdown kernel without interfering with IMA.