linux-kernel - Re: [PATCH v2 1/2] crypto: s(h)aving 16 byte constant off arch/x86/crypto/sha1_ni

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

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

Date: Mon, 8 Apr 2024 16:37:22 +0200
From: Ard Biesheuvel <ardb@...nel.org>
To: Stefan Kanthak <stefan.kanthak@...go.de>
Cc: linux-kernel@...r.kernel.org, linux-crypto@...r.kernel.org, 
	tim.c.chen@...ux.intel.com, Eric Biggers <ebiggers@...nel.org>
Subject: Re: [PATCH v2 1/2] crypto: s(h)aving 16 byte constant off arch/x86/crypto/sha1_ni_asm.S

On Mon, 8 Apr 2024 at 16:18, Stefan Kanthak <stefan.kanthak@...go.de> wrote:
>
> Get rid of 16 byte constant, use shift instead of mask
>

This is missing a signoff

> --- -/arch/x86/crypto/sha1_ni_asm.S
> +++ +/arch/x86/crypto/sha1_ni_asm.S
> @@ -104,9 +104,9 @@
>         add             DATA_PTR, NUM_BLKS      /* pointer to end of data */
>
>         /* load initial hash values */
> -       pinsrd          $3, 1*16(DIGEST_PTR), E0
> +       pinsrd          $0, 1*16(DIGEST_PTR), E0
>         movdqu          0*16(DIGEST_PTR), ABCD
> -       pand            UPPER_WORD_MASK(%rip), E0
> +       pslldq          $12, E0
>         pshufd          $0x1B, ABCD, ABCD
>

Wouldn't it be simpler and therefore better to simply zero register E0
before loading E into the top lane?


>         movdqa          PSHUFFLE_BYTE_FLIP_MASK(%rip), SHUF_MASK
> @@ -297,8 +297,3 @@
>  .align 16
>  PSHUFFLE_BYTE_FLIP_MASK:
>         .octa 0x000102030405060708090a0b0c0d0e0f
> -
> -.section       .rodata.cst16.UPPER_WORD_MASK, "aM", @progbits, 16
> -.align 16
> -UPPER_WORD_MASK:
> -       .octa 0xFFFFFFFF000000000000000000000000
>