Date:   Wed, 5 Jan 2022 14:09:20 -0800
From:   Eric Biggers <ebiggers@kernel.org>
To:     Ard Biesheuvel <ardb@kernel.org>
Cc:     Herbert Xu <herbert@gondor.apana.org.au>,
        "Jason A. Donenfeld" <Jason@zx2c4.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Masahiro Yamada <masahiroy@kernel.org>,
        Linux Kbuild mailing list <linux-kbuild@vger.kernel.org>,
        Linux Crypto Mailing List <linux-crypto@vger.kernel.org>
Subject: Re: [PATCH v7] lib/crypto: blake2s: include as built-in
Message-ID: <YdYXEBhc7vuY+GQW@gmail.com>
References: <CAHmME9oPcEjRnqDesPNKJNOsT+i9vmWRxy9c62t+Xu9Nohsi2A@mail.gmail.com>
 <20220103123152.1043972-1-Jason@zx2c4.com>
 <YdOhMPwL9sXllm8X@gondor.apana.org.au>
 <CAMj1kXFhygHnB12g9MD0wMo_deZ6xd7FMEzbrvEvKVtqYdskAQ@mail.gmail.com>
 <YdYTUCR2GM/6CXmF@gmail.com>
 <CAMj1kXHS8Gr9wbbbzvX+7u7qk0bAhgTpQw-bEB5Y_gmrcJ8iGw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAMj1kXHS8Gr9wbbbzvX+7u7qk0bAhgTpQw-bEB5Y_gmrcJ8iGw@mail.gmail.com>
Precedence: bulk

On Wed, Jan 05, 2022 at 11:01:04PM +0100, Ard Biesheuvel wrote:
> On Wed, 5 Jan 2022 at 22:53, Eric Biggers <ebiggers@kernel.org> wrote:
> >
> > On Tue, Jan 04, 2022 at 06:02:52PM +0100, Ard Biesheuvel wrote:
> > > The only downside here is that the ARM/x86 accelerated shashes and the
> > > generic shash now use the same core transform, right?
> >
> > I don't see how this is the case, given that crypto/blake2s_generic.c still uses
> > blake2s_compress_generic(), not blake2s_compress().
> >
> 
> Ah ok, I stand corrected then.
> 
> So what are your thoughts on this? Should we keep the shashes while
> they have no users?

I don't know.  Removing unused stuff is good per se, but I wouldn't have
expected this to be something that is being considered here.  It's not like this
is a "controversial" algorithm, blake2b is already supported, and there could be
users of it already (dm-integrity, dm-verity, AF_ALG, etc.).  If this is going
to happen, then the acceptance criteria for new algorithms need to get *much*
stricter, so that algorithms aren't constantly being added and removed.

- Eric