VIA padlock (was: Re: GPG with GPUs)

Peter Lebbing peter at
Tue Jun 19 10:55:11 CEST 2012

On 18/06/12 20:39, Werner Koch wrote:
> FWIW, Libgcrypt uses this RNG directly in addition to other sources.

Actually... I just checked, and I see these lines in Libgcrypt,
file random/rndhw.c:

# if defined (__i386__) && SIZEOF_UNSIGNED_LONG == 4 && defined (__GNUC__)
# define USE_PADLOCK
# endif

I'm running x86_64: the VIA Nano is a 64-bits processor. So I think Libgcrypt
doesn't recognise it can use the padlock from the VIA Nano.

It's the same with OpenSSL in Debian. I don't particularly care about support
for the HWRNG (since it's already seeding /dev/random through rngd), but I do
definitely like hardware AES on this modestly powered beast. It runs a TOR node;
I think that is the main user of the hardware AES.

So I patch Debian's OpenSSL with a patch I based on a code repository from
upstream OpenSSL, so I can use the padlock AES under x86_64. I'd submit the
patch to Debian if I could still remember from where it came...


I use the GNU Privacy Guard (GnuPG) in combination with Enigmail.
You can send me encrypted mail if you want some privacy.
My key is available at

More information about the Gnupg-users mailing list