Re: randomness (crypto?) code example wanted please?

[Roy Marples <roy%marples.name@localhost>]

On 25/06/2017 15:19, Robert Elz wrote:
>      Date:        Sun, 25 Jun 2017 13:53:02 +0200
>      From:        Kamil Rytarowski <n54%gmx.com@localhost>
>      Message-ID:  <93fba221-3f6c-07be-8dc8-efd744c83786%gmx.com@localhost>
>
>    | There is a libc function for this kind of tasks: arc4random_uniform(3).
>
> Sorry, I had missed this message when I replied to Martin's (I did see
> your later message).
>
> It looks (from the man page) as if that function fails the "must be able
> to generate the same sequence over and over again" requirement of $RANDOM
>
> If there is some way to prevent the ...
>
> 	arc4random() seeds itself on a regular basis from
> 	the kernel strong random number subsystem described in rnd(4).
>
> mechanism, and set an initial seed, and just generate based upon that
> rather than a true randomness source, I don't see it.

There isn't .... at least not in NetBSD.

>    | I was evaluation at some point whether this could be done differently,
>    | with a dedicated userland random(1) program,
>
> That could be done too.   sh's $RANDOM is not really intended for
> anything of crypto strength I don't think.  I suspect half the
> implementations just return rand(3) results (ie: garbage).

dhcpcd's implementation of IPv4LL has the same limitation, so I use 
random() there and arc4random/arc4random_uniform everywhere else.
https://dev.marples.name/source/dhcpcd/browse/master/src/ipv4ll.c;0d7318eaf52e59fdcff6c472ef8adafd5dbf2069$70

However, I do seed off the interfaces hardware address to avoid random 
clashes of IPv4LL address.
https://dev.marples.name/source/dhcpcd/browse/master/src/ipv4ll.c;0d7318eaf52e59fdcff6c472ef8adafd5dbf2069$359

Roy