Re: VIA Padlock on AMD64

[Taylor R Campbell <riastradh%NetBSD.org@localhost>]

> Date: Sun, 19 Jul 2020 14:16:25 +0300
> From: Andrius V <vezhlys%gmail.com@localhost>
> 
> I just noticed that padlock driver was made to compile on amd64
> recently. I believe it should be included in modules Makefile
> (/sys/modules/Makefile) as well (as per my git diff in previous
> messages).

I guess we could do that.  It's not really that useful -- the only
application it's good for is in-kernel IPsec.  OpenSSL should already
be able to take advantage of the CPU support in userland -- no kernel
driver needed.

(I also don't have any 64-bit VIA hardware to test it on, which is why
I added it only to amd64/ALL to make compile-testing easier.)

> > I created a patch to show the intention. It builds and "padlock"
> > attaches to the CPU. On another hand, I failed to find any application
> > which would be taking advantage of padlock (including i386). Neither
> > openssl (tried speed tests), nor openssh (some commands I found in
> > internet)  showed any difference with padlock module or without it.
> > Maybe some patching is needed but it seems hardware acceleration is
> > ignored with these apps. If somebody know a good way to test it on
> > NetBSD, I would be glad to hear it.

Try comparing

openssl speed aes-256-cbc
openssl speed -evp aes-256-cbc

and particularly note the user time spent.

> > On the side note, same goes to viadrmums module (it's i386 only now)
> > but at least on VX900 I ended up with the crash, so I guess it may be
> > incompatible with amd64 (though it builds successfully). Will try to
> > test VX800 later on.

I haven't tried viadrmums in a while.  What was the crash?