Re: [PATCH] arm64: SHA-224/SHA-256 using ARMv8 Crypto Extensions

[Marek Vasut <marex@denx.de>]

On Thursday, March 27, 2014 at 02:23:41 PM, Ard Biesheuvel wrote:
> On 24 March 2014 21:36, Marek Vasut <marex@denx.de> wrote:
> > On Thursday, March 20, 2014 at 03:48:06 PM, Ard Biesheuvel wrote:
> >> This patch adds support for the SHA-224 and SHA-256 hash algorithms
> >> using the NEON based SHA-256 instructions that were introduced in ARM
> >> v8.
> >> 
> >> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> >> ---
> > 
> > [...]
> > 
> >> + * Copyright (c) Alan Smithee.
> > 
> > Email contact is missing here.
> > 
> > [...]
> 
> Actually, this is mostly copied from the original sha1_generic.c. In
> fact, my current v2 (which I will post shortly) has been reworked to
> such an extent that I am contemplating dropping this attribution
> altogether.

OK

> >> +static int sha224_init(struct shash_desc *desc)
> >> +{
> >> +     struct sha256_state *sctx = shash_desc_ctx(desc);
> >> +
> >> +     *sctx = (struct sha256_state){
> > 
> > This cast is interesting, I don't quite understand it. Can you please
> > explain that to me ?
> 
> http://gcc.gnu.org/onlinedocs/gcc/Compound-Literals.html

I have to wonder how many people will stumble across this and will wonder about 
the same honestly. Sure, it's a valid construct, but it's quite strange in my 
opinion. On the other hand, if noone else is against writing it like so, I won't 
push it ...

> >> +             .state = {
> >> +                     SHA224_H0, SHA224_H1, SHA224_H2, SHA224_H3,
> >> +                     SHA224_H4, SHA224_H5, SHA224_H6, SHA224_H7,
> >> +             }
> >> +     };
> >> +     return 0;
> >> +}
> > 
> > [...]
> > 
> >> +static int sha224_final(struct shash_desc *desc, u8 *out)
> >> +{
> >> +     struct sha256_state *sctx = shash_desc_ctx(desc);
> >> +     __be32 *dst = (__be32 *)out;
> >> +     int i;
> >> +
> >> +     sha2_final(desc);
> >> +
> >> +     for (i = 0; i < SHA224_DIGEST_SIZE / sizeof(*dst); i++)
> >> +             dst[i] = cpu_to_be32(sctx->state[i]);
> > 
> > Won't this cause unaligned access if *dst is not aligned to 32 bytes ?
> 
> arm64 does not care about that, but I agree it would be better (and
> more explicit) to use put_unaligned() here, and leave it up to the
> architecture to allow it or work around it.
> WIll update that in v2.

OK, thanks!

> Thanks for the review.

Best regards,
Marek Vasut