Re: [PATCH 0/3] LoadPin: Enable loading from trusted dm-verity devices

From: Kees Cook <keescook@chromium.org>
To: Matthias Kaehlcke <mka@chromium.org>
Cc: Alasdair Kergon <agk@redhat.com>,
	Mike Snitzer <snitzer@kernel.org>,
	James Morris <jmorris@namei.org>,
	"Serge E . Hallyn" <serge@hallyn.com>,
	linux-raid@vger.kernel.org, linux-kernel@vger.kernel.org,
	Song Liu <song@kernel.org>,
	linux-security-module@vger.kernel.org,
	Douglas Anderson <dianders@chromium.org>,
	dm-devel@redhat.com
Subject: Re: [PATCH 0/3] LoadPin: Enable loading from trusted dm-verity devices
Date: Mon, 18 Apr 2022 19:33:55 -0700	[thread overview]
Message-ID: <202204181931.A618DFF2C@keescook> (raw)
In-Reply-To: <Yl3pj72hM/Bo+Kf5@google.com>

On Mon, Apr 18, 2022 at 03:43:27PM -0700, Matthias Kaehlcke wrote:
> Hi Kees,
> 
> On Mon, Apr 18, 2022 at 03:14:14PM -0700, Kees Cook wrote:
> > [oops, resending to actual CC list]
> > 
> > On Mon, Apr 18, 2022 at 02:15:56PM -0700, Matthias Kaehlcke wrote:
> > > This series extends LoadPin to allow loading of kernel files
> > > from trusted dm-verity devices. It adds the concept of
> > > trusted verity devices to LoadPin. Userspace can use the
> > > new systl file 'loadpin/trusted_verity_root_digests' to
> > > provide LoadPin with a list of root digests from dm-verity
> > > devices that LoadPin should consider as trusted. When a
> > > kernel file is read LoadPin first checks (as usual) whether
> > > the file is located on the pinned root, if so the file can
> > > be loaded. Otherwise, if the verity extension is enabled,
> > > LoadPin determines whether the file is located on a verity
> > > backed device and whether the root digest of that device
> > > is in the list of trusted digests. The file can be loaded
> > > if the verity device has a trusted root digest.
> > > 
> > > The list of trusted root digests can only be written once
> > > (typically at boot time), to limit the possiblity of
> > > attackers setting up rogue verity devices and marking them
> > > as trusted.
> 
> 
> > Thanks for working all this out! Where does the list of trusted
> > roothashes come from? I assume some chain of trust exists. Is the list
> > maybe already stored on the rootfs?
> 
> Yes, at least the content of the list comes from the rootfs. The
> userspace part is still TBD (also pending on the evolution of this
> patchset), having the list pre-formatted in a single file on the
> rootfs should be fine.

Ah-ha, that's perfect.

> > It'd be nice if there was some way to pass the trust chain to LoadPin
> > more directly.
> 
> I imagine you envision LoadPin reading the file itself, instead of
> just processing the content. That should be doable. One option would
> be to pass the path of the file with the hashes through the sysctl
> file and use kernel_read_file_from_path() to read it if the path is
> in the pinned root (or maybe even in any trusted file system ;-)

It could be a boot param or a Kconfig too. But yeah, having LoadPin able
to use itself to validate the file path would be much nicer.

-- 
Kees Cook