Re: RND_TYPE_RNG question concerning rate of random bits production

[Frank Zerangue <frank.zerangue%gmail.com@localhost>]

Taylor. That is much appreciated. I was on the 6.1.2 release.

Thanks --

On Jan 27, 2014, at 1:17 PM, Taylor R Campbell wrote:

>   Date: Mon, 27 Jan 2014 13:05:58 -0600
>   From: Frank Zerangue <frank.zerangue%gmail.com@localhost>
> 
>   I looked at hifn, amdpm, and bcm28335 RNG drivers and they all seem
>   to do the same. They setup a callout to periodically call there RNG
>   reader to prime the random pool with more bits; but I don't see how
>   the polling is based on entropy needed. Could you clarify for me
>   what I have missed in regard to your suggestion below.
> 
> Perhaps you're not looking at the code in HEAD.  See, e.g.,
> 
> http://nxr.netbsd.org/xref/src/sys/arch/arm/broadcom/bcm2835_rng.c#128
> 
> which sets up a callback so that when the system thinks it needs some
> more entropy, it will call bcmrng_get_cb, which schedules a softintr
> to actually read data from the hardware and feed it to the system
> entropy pool.
> 
> This API is not yet stable -- at least, some months ago, Thor and I
> discussed changing it to render the softintr indirection and locking
> dance unnecessary -- but it will work for now.

--
Frank Zerangue
frank.zerangue%gmail.com@localhost