kernel-hardening - Re: [PATCH v6 8/8] crypto, x86/sha: Eliminate casts on asm implementations

Follow @Openwall on Twitter for new release announcements and other news

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20191122030620.GD32523@sol.localdomain>
Date: Thu, 21 Nov 2019 19:06:20 -0800
From: Eric Biggers <ebiggers@...nel.org>
To: Kees Cook <keescook@...omium.org>
Cc: Herbert Xu <herbert@...dor.apana.org.au>,
	João Moreira <joao.moreira@...el.com>,
	Ard Biesheuvel <ard.biesheuvel@...aro.org>,
	Sami Tolvanen <samitolvanen@...gle.com>,
	Stephan Mueller <smueller@...onox.de>, x86@...nel.org,
	linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org,
	kernel-hardening@...ts.openwall.com
Subject: Re: [PATCH v6 8/8] crypto, x86/sha: Eliminate casts on asm
 implementations

On Thu, Nov 21, 2019 at 05:03:34PM -0800, Kees Cook wrote:
> -asmlinkage void sha1_transform_ssse3(u32 *digest, const char *data,
> -				     unsigned int rounds);
> +asmlinkage void sha1_transform_ssse3(struct sha1_state *digest,
> +				     const u8 *data, int rounds);

Can you please use:

asmlinkage void sha1_transform_ssse3(struct sha1_state *state,
				     const u8 *data, int blocks);

I.e., rename 'digest' => 'state' and 'rounds' => 'blocks'.

(Or alternatively 'sst' instead of 'state' would be okay since that's what
sha{1,256,512}_block_fn uses, but 'state' seems much clearer to me.)

Similarly for the other sha{1,256,512}_transform_*() functions.

'digest' is confusing because it would typically be understood to mean the final
digest, or maybe also the hash chaining state (e.g. u32[5] for SHA-1) if it's
interpreted loosely.  I don't think it would typically be understand to also
include buffered data like struct sha1_state does.

'rounds' is confusing because the parameter is actually the number of blocks,
not the number of rounds -- the latter being fixed for each SHA-* algorithm.
Since sha{1,256,512}_block_fn already call it 'blocks' and this patch has to
update the type of this parameter anyway, let's use that.

>  #ifdef CONFIG_AS_AVX
> -asmlinkage void sha1_transform_avx(u32 *digest, const char *data,
> -				   unsigned int rounds);
> +asmlinkage void sha1_transform_avx(struct sha1_state *digest,
> +				   const u8 *data, int rounds);
>  

This patch is also still missing updates to the corresponding comments in the
assembly files:

	sha1_transform_avx()
	sha1_transform_avx2()
	sha256_transform_avx()
	sha256_transform_rorx()
	sha512_transform_ssse3() [references to D, M, and L remain]
	sha512_transform_avx()
	sha512_transform_rorx()

FWIW, this patch is also independent from 1-7, so it could be sent out
separately if it makes things any easier for you.

Thanks,

- Eric

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.