Received-SPF: pass (google.com: best guess record for domain of linux-crypto-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 2 Jul 2019 09:06:14 -0700
From:   Eric Biggers <ebiggers@kernel.org>
To:     Tom Lendacky <thomas.lendacky@amd.com>,
        Gary Hook <gary.hook@amd.com>
Cc:     linux-crypto@vger.kernel.org, Cfir Cohen <cfir@google.com>,
        David Rientjes <rientjes@google.com>
Subject: gcm-aes-ccp self-tests failure
Message-ID: <20190702160614.GC895@sol.localdomain>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.12.1 (2019-06-15)
Sender: linux-crypto-owner@vger.kernel.org
Precedence: bulk

Hi Tom and Gary,

On latest cryptodev tree, I'm seeing the following self-test failure after I
built a kernel with the AMD CCP driver and crypto self-tests enabled, and booted
it on system with a Ryzen processor ("Threadripper 1950X"):

[    4.378985] alg: aead: gcm-aes-ccp encryption test failed (wrong result) on test vector 2, cfg="two even aligned splits"

i.e., in some cases the AES-GCM implementation produces the wrong ciphertext
and/or authentication tag.

Is this is a known issue?  When will it be fixed?

The potentially relevant bits of my Kconfig are:

	CONFIG_CRYPTO_AES=y
	CONFIG_CRYPTO_GCM=y
	CONFIG_CRYPTO_DEV_CCP=y
	CONFIG_CRYPTO_DEV_CCP_DD=y
	CONFIG_CRYPTO_DEV_SP_CCP=y
	CONFIG_CRYPTO_DEV_CCP_CRYPTO=y
	CONFIG_CRYPTO_DEV_SP_PSP=y
	# CONFIG_CRYPTO_MANAGER_DISABLE_TESTS is not set
	CONFIG_DEBUG_KERNEL=y
	CONFIG_CRYPTO_MANAGER_EXTRA_TESTS=y

- Eric