Re: [PATCH v2] fscrypt: support trusted keys

From: Mimi Zohar <zohar@linux.ibm.com>
To: Ahmad Fatoum <a.fatoum@pengutronix.de>,
	Eric Biggers <ebiggers@kernel.org>,
	Jarkko Sakkinen <jarkko@kernel.org>
Cc: "Theodore Y. Ts'o" <tytso@mit.edu>,
	Jaegeuk Kim <jaegeuk@kernel.org>,
	kernel@pengutronix.de, James Morris <jmorris@namei.org>,
	"Serge E. Hallyn" <serge@hallyn.com>,
	James Bottomley <jejb@linux.ibm.com>,
	Sumit Garg <sumit.garg@linaro.org>,
	David Howells <dhowells@redhat.com>,
	linux-fscrypt@vger.kernel.org, linux-crypto@vger.kernel.org,
	linux-integrity@vger.kernel.org,
	linux-security-module@vger.kernel.org, keyrings@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] fscrypt: support trusted keys
Date: Tue, 17 Aug 2021 09:55:13 -0400	[thread overview]
Message-ID: <285cb263d9c1c16f3918c98dd36074ef16568e6d.camel@linux.ibm.com> (raw)
In-Reply-To: <a6eb6f38-b9f4-c59c-4181-2049f181e67d@pengutronix.de>

On Tue, 2021-08-17 at 15:04 +0200, Ahmad Fatoum wrote:
> Hi,
> 
> On 12.08.21 02:54, Mimi Zohar wrote:
> > On Wed, 2021-08-11 at 10:16 -0700, Eric Biggers wrote:
> > 
> >> Neither of you actually answered my question, which is whether the support for
> >> trusted keys in dm-crypt is a mistake.  I think you're saying that it is?  That
> >> would imply that fscrypt shouldn't support trusted keys, but rather encrypted
> >> keys -- which conflicts with Ahmad's patch which is adding support for trusted
> >> keys.  Note that your reasoning for this is not documented at all in the
> >> trusted-encrypted keys documentation; it needs to be (email threads don't really
> >> matter), otherwise how would anyone know when/how to use this feature?
> > 
> > True, but all of the trusted-encrypted key examples in the
> > documentation are "encrypted" type keys, encrypted/decrypted based on a
> > "trusted" type key.  There are no examples of using the "trusted" key
> > type directly.  Before claiming that adding "trusted" key support in
> > dm-crypt was a mistake, we should ask Ahmad why he felt dm-crypt needed
> > to directly support "trusted" type keys.
> 
> I wanted to persist the dm-crypt key as a sealed blob. With encrypted keys,
> I would have to persist and unseal two blobs (load trusted key blob, load
> encrypted key blob rooted to trusted key) with no extra benefit.
> 
> I thus added direct support for trusted keys. Jarkko even commented on the
> thread, but didn't voice objection to the approach (or agreement for that
> matter), so I assumed the approach is fine.
> 
> I can see the utility of using a single trusted key for TPMs, but for CAAM,
> I see none and having an encrypted key for every trusted key just makes
> it more cumbersome.
> 
> In v1 here, I added encrypted key support as well, but dropped it for v2,
> because I am not in a position to justify its use. Now that you and Eric
> discussed it, should I send v3 with support for both encrypted and trusted
> keys like with dm-crypt or how should we proceed?

With some applications, the indirection is important.   It allows the
"encrypted" key type to be updated/re-encypted based on a new "trusted"
key, without affecting the on disk encrypted key usage.

As much as I expected, directly using "trusted" keys is a result of the
new trusted key sources.  I have no opinion as to whether this is/isn't
a valid usecase.

thanks,

Mimi