[PATCH] crypto: vmx - CTR: always increment IV as quadword

[Daniel Axtens]

Daniel Axtens <dja at axtens.net> writes:

> The kernel self-tests picked up an issue with CTR mode:
> alg: skcipher: p8_aes_ctr encryption test failed (wrong result) on test vector 3, cfg="uneven misaligned splits, may sleep"
>
> Test vector 3 has an IV of FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFD, so
> after 3 increments it should wrap around to 0.
>
> In the aesp8-ppc code from OpenSSL, there are two paths that
> increment IVs: the bulk (8 at a time) path, and the individual
> path which is used when there are fewer than 8 AES blocks to
> process.
>
> In the bulk path, the IV is incremented with vadduqm: "Vector
> Add Unsigned Quadword Modulo", which does 128-bit addition.
>
> In the individual path, however, the IV is incremented with
> vadduwm: "Vector Add Unsigned Word Modulo", which instead
> does 4 32-bit additions. Thus the IV would instead become
> FFFFFFFFFFFFFFFFFFFFFFFF00000000, throwing off the result.
>
> Use vadduqm.
>
> This was probably a typo originally, what with q and w being
> adjacent. It is a pretty narrow edge case: I am really
> impressed by the quality of the kernel self-tests!
>
> Fixes: 5c380d623ed3 ("crypto: vmx - Add support for VMS instructions by ASM")
> Cc: stable at vger.kernel.org
> Signed-off-by: Daniel Axtens <dja at axtens.net>
>
> ---
>
> I'll pass this along internally to get it into OpenSSL as well.

I passed this along to OpenSSL and got pretty comprehensively schooled:
https://github.com/openssl/openssl/pull/8942

It seems we tweak the openssl code to use a 128-bit counter, whereas
the original code was in fact designed for a 32-bit counter. We must
have changed the vaddu instruction in the bulk path but not in the
individual path, as they're both vadduwm (4x32-bit) upstream.

I think this change is still correct with regards to the kernel,
but I guess it's probably something where I should have done a more
thorough read of the documentation before diving in to the code, and
perhaps we should note it in the code somewhere too. Ah well.

Regards,
Daniel

> ---
>  drivers/crypto/vmx/aesp8-ppc.pl | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/crypto/vmx/aesp8-ppc.pl b/drivers/crypto/vmx/aesp8-ppc.pl
> index de78282b8f44..9c6b5c1d6a1a 100644
> --- a/drivers/crypto/vmx/aesp8-ppc.pl
> +++ b/drivers/crypto/vmx/aesp8-ppc.pl
> @@ -1357,7 +1357,7 @@ Loop_ctr32_enc:
>  	addi		$idx,$idx,16
>  	bdnz		Loop_ctr32_enc
>  
> -	vadduwm		$ivec,$ivec,$one
> +	vadduqm		$ivec,$ivec,$one
>  	 vmr		$dat,$inptail
>  	 lvx		$inptail,0,$inp
>  	 addi		$inp,$inp,16
> -- 
> 2.19.1