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

From: Eric Biggers <ebiggers@kernel.org>
To: Ard Biesheuvel <ardb@kernel.org>
Cc: linux-crypto@vger.kernel.org, herbert@gondor.apana.org.au,
	keescook@chromium.org, jason@zx2c4.com, nikunj@amd.com
Subject: Re: [PATCH v4 3/3] crypto: aesgcm - Provide minimal library implementation
Date: Mon, 24 Oct 2022 22:32:28 -0700	[thread overview]
Message-ID: <Y1d07G+jIeGron7E@sol.localdomain> (raw)
In-Reply-To: <20221024063052.109148-4-ardb@kernel.org>

On Mon, Oct 24, 2022 at 08:30:52AM +0200, Ard Biesheuvel wrote:
> The former concern is addressed trivially, given that the function call
> API uses 32-bit signed types for the input lengths. It is still up to
> the caller to avoid IV reuse in general, but this is not something we
> can police at the implementation level.

This doesn't seem to have been any note left about this in the code itself.
Sizes are usually size_t, so if another type is used intentionally, that should
be carefully documented.

Also, does it really need to be signed?

> +int __must_check aesgcm_decrypt(const struct aesgcm_ctx *ctx, u8 *dst,
> +				const u8 *src, int crypt_len, const u8 *assoc,
> +				int assoc_len, const u8 iv[GCM_AES_IV_SIZE],
> +				const u8 *authtag);

This returns 0 or -EBADMSG, which is inconsistent with
chacha20poly1305_decrypt() which returns a bool.  It would be nice if the
different algorithms would use consistent conventions.

-  Eric