Re: x86 CPU RNG support, take 2

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Sun, Dec 27, 2015 at 02:26:36AM +0000, Taylor R Campbell wrote:
> 
> This seems to have several functional changes in it:
> 
> - Add cpu_rng(9) and its rndsource.

This is unusable and untestable without one or the other of the below:

> - Implement cpu_rng(9) on x86 using RDRAND/RDSEED for Intel CPUs.
> - Convert VIA RNG to cpu_rng(9).
> 
> Please split them into separate commits, and separate patches for

So I am somewhat hesitant to do what you ask.

As you noticed the 32-bit code has some brainos and likely won't even
compile.  Unfortunately, I'm travelling and using my laptop for a
build platform, and the tools build currently seems to be broken (mandoc
as a tool splodes on OS X) so I can't test 32-bit anything (I had a
good 64-bit toolchain before the build breakage).

I believe the VIA code has a comment that does point at a lengthy
explanation of why not to use "rep xstorrng".  The summary is that
VIA says not to.

Thanks for all your comments -- I will send another pass at this in
a few days when I'm home and can build/test both platforms.

Thor