[openssl-dev] Proper explicit zeroing in VIA PadLock engine

[Richard Levitte]

In message <20160404144408.GA75125 at thinkpad.swarthmore.edu> on Mon, 4 Apr 2016 10:44:09 -0400, Michael McConville <mmcco at mykolab.com> said:

mmcco> Richard Levitte wrote:
mmcco> > That being said, engines/e_padlock.c has changed quite a bit since, so
mmcco> > if patching is still needed, it needs to be reworked with somewhat
mmcco> > more modern code (that libressl patch applies to OpenSSL 0.9.8, which
mmcco> > is past EOL).
mmcco> 
mmcco> True, but it's still the same one line that needs to be changed.
mmcco> engines/e_padlock.c:779:
mmcco> 
mmcco> > *(volatile unsigned int *)&buf = 0;
mmcco> 
mmcco> I've never worked with OpenSSL before, but the below is what I was
mmcco> imagining.
mmcco> 
mmcco> 
mmcco> diff --git a/engines/e_padlock.c b/engines/e_padlock.c
mmcco> index 96e7483..709c4de 100644
mmcco> --- a/engines/e_padlock.c
mmcco> +++ b/engines/e_padlock.c
mmcco> @@ -776,7 +776,8 @@ static int padlock_rand_bytes(unsigned char *output, int count)
mmcco>          *output++ = (unsigned char)buf;
mmcco>          count--;
mmcco>      }
mmcco> -    *(volatile unsigned int *)&buf = 0;
mmcco> +
mmcco> +    OPENSSL_cleanse(&buf, sizeof(buf));
mmcco>  
mmcco>      return 1;
mmcco>  }

That looks good enough, I'll see to it being inserted.

Cheers,
Richard

-- 
Richard Levitte         levitte at openssl.org
OpenSSL Project         http://www.openssl.org/~levitte/