[dm-devel] [RFC PATCH 0/3] crypto: switch to shash for ESSIV generation

[Ard Biesheuvel]

On Mon, 17 Jun 2019 at 10:52, Gilad Ben-Yossef <gilad at benyossef.com> wrote:
>
> On Sun, Jun 16, 2019 at 11:44 PM Eric Biggers <ebiggers at kernel.org> wrote:
> >
> > [+Cc dm-devel and linux-fscrypt]
> >
> > On Fri, Jun 14, 2019 at 10:34:01AM +0200, Ard Biesheuvel wrote:
> > > This series is presented as an RFC for a couple of reasons:
> > > - it is only build tested
> > > - it is unclear whether this is the right way to move away from the use of
> > >   bare ciphers in non-crypto code
> > > - we haven't really discussed whether moving away from the use of bare ciphers
> > >   in non-crypto code is a goal we agree on
> > >
> > > This series creates an ESSIV shash template that takes a (cipher,hash) tuple,
> > > where the digest size of the hash must be a valid key length for the cipher.
> > > The setkey() operation takes the hash of the input key, and sets into the
> > > cipher as the encryption key. Digest operations accept input up to the
> > > block size of the cipher, and perform a single block encryption operation to
> > > produce the ESSIV output.
> ...
> > I agree that moving away from bare block ciphers is generally a good idea.  For
> > fscrypt I'm fine with moving ESSIV into the crypto API, though I'm not sure a
> > shash template is the best approach.  Did you also consider making it a skcipher
> > template so that users can do e.g. "essiv(cbc(aes),sha256,aes)"?  That would
> > simplify things much more on the fscrypt side, since then all the ESSIV-related
> > code would go away entirely except for changing the string "cbc(aes)" to
> > "essiv(cbc(aes),sha256,aes)".
>
> I will also add that going the skcipher route rather than shash will
> allow hardware tfm providers like CryptoCell that can do the ESSIV
> part in hardware implement that as a single API call and/or hardware
> invocation flow.
> For those system that benefit from hardware providers this can be beneficial.
>

Ah yes, thanks for reminding me. There was some debate in the past
about this, but I don't remember the details.

I think implementing essiv() as a skcipher template is indeed going to
be the best approach, I will look into that.