RE: [PATCH v4 3/3] crypto: aesgcm - Provide minimal library implementation

From: "Elliott, Robert (Servers)" <elliott@hpe.com>
To: Ard Biesheuvel <ardb@kernel.org>,
	"linux-crypto@vger.kernel.org" <linux-crypto@vger.kernel.org>
Cc: "herbert@gondor.apana.org.au" <herbert@gondor.apana.org.au>,
	"ebiggers@kernel.org" <ebiggers@kernel.org>,
	"keescook@chromium.org" <keescook@chromium.org>,
	"jason@zx2c4.com" <jason@zx2c4.com>,
	"nikunj@amd.com" <nikunj@amd.com>
Subject: RE: [PATCH v4 3/3] crypto: aesgcm - Provide minimal library implementation
Date: Mon, 24 Oct 2022 16:21:52 +0000	[thread overview]
Message-ID: <MW5PR84MB1842E8E6C99E4C9F25A1E95EAB2E9@MW5PR84MB1842.NAMPRD84.PROD.OUTLOOK.COM> (raw)
In-Reply-To: <20221024063052.109148-4-ardb@kernel.org>

> diff --git a/lib/crypto/aesgcm.c b/lib/crypto/aesgcm.c
> +static int __init libaesgcm_init(void)
...
> +		if (aesgcm_expandkey(&ctx, aesgcm_tv[i].key, aesgcm_tv[i].klen,
> +				     aesgcm_tv[i].clen - plen)) {
> +			WARN(1, "aesgcm_expandkey() failed on vector %d\n", i);
> +			return -EINVAL;
> +		}

Other self-test errors in lib/crypto (blake2s, chacha20poly1305,
and curve25519) are reported with pr_err. Could you add that in
addition to the WARNs?

Also, they return -ENODEV rather than -EINVAL on self-test failures,
like this:

static int __init blake2s_mod_init(void)
{
        if (!IS_ENABLED(CONFIG_CRYPTO_MANAGER_DISABLE_TESTS) &&
            WARN_ON(!blake2s_selftest()))
                return -ENODEV;
        return 0;
}