[lkml]   [2003]   [Oct]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[RFC] frandom - fast random generator module

Frandom is the faster version of the well-known /dev/urandom random
number generator. Not instead of, but rather as a supplement, when
pseudorandom data is needed at high rate. Few tests so far show that
frandom is 10-50 times faster than urandom.

The project's home page:

The module works on 2.2, 2.4 and 2.6 kernels. A few straightforward
#ifdef's handle compatability (easy to remove to match common coding style).


(1) Frandom is a handy source of bulk random data.
(2) It is *not* intended for encryption and security-related applications.
(3) frandom is intended for (scientific) simulations, wiping the disk,
stress tests on algorithms and so on.
(4) It is more of an /dev/zero than /dev/random

Quality of random numbers

(1) The module has been tested for random number quality with the
"diehard" set of tests, and passed them all. This indicates that the
bytes are random enough for most scientific purposes.
(2) Additional tests results are welcomed.
(3) The core of frandom is based upon RC4. frandom is exactly RC4, minus
the XOR operation with the data. So if frandom doesn't generate good
random numbers, I would wonder why RC4 is considered safe.
(4) The random generator is seeded with 256 bytes of the kernel's
get_random_bytes() for every file opened on /dev/frandom. This is
equivalent to a 2048-bit random key on RC4.
(5) I don't see frandom fit for crypto purposes, mainly because the
module was naively written. I won't fall off my chair if it turns out to
be crypto-safe, but I wouldn't trust it either. Not yet, anyhow.
(6) Those who read the source and feel that such a simple algorithm
can't create good random: That's exactly the beauty of RC4: It's simple
and it works.

frandom and the linux kernel tree

(1) Occasionally, people complain that /dev/urandom is too slow, wishing
for something faster.
(2) Other argue that a random generator can be written in user space.
(3) I agree with both. And I use /dev/zero a lot. I know how to write a
zero-generating application in user space.
(4) The module is small: 6kB of source code as a standalone module, and
2.3 kB of kernel memory.

Test results and comments will be appreciated.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:50    [W:0.138 / U:52.204 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site