Subject: Re: openssl (libcrypto) repaired for non-32bit arch
To: None <eeh@netbsd.org>
From: None <itojun@iijlab.net>
List: tech-security
Date: 10/02/2000 08:14:32
  by mail.netbsd.org with SMTP; 1 Oct 2000 23:14:34 -0000
	by coconut.itojun.org (8.9.3+3.2W/3.7W) with ESMTP id IAA14787;
	Mon, 2 Oct 2000 08:14:33 +0900 (JST)
To: eeh@netbsd.org
cc: tech-security@netbsd.org
In-reply-to: eeh's message of 01 Oct 2000 15:19:02 GMT.
      <20001001151902.4812.qmail@mail.netbsd.org>
Subject: Re: openssl (libcrypto) repaired for non-32bit arch
From: itojun@iijlab.net
Date: Mon, 02 Oct 2000 08:14:32 +0900
Message-ID: <14785.970442072@coconut.itojun.org>


>NetBSD/sparc64, of course.
>
>It seems to have something to do with the acquisition of sufficient entropy
>for the random number generator.  Probably some floating point bug in the
>compiler.  Still looking into it.
>
>BTW, why is it that we require /dev/random to have working crypto?  The 
>random number generator in libcrypto should not refuse to work if 
>there's no rnd device.  (It's not like rnd is such a good source of
>entropy.  Most of its sources are clocked so that data is not really
>random.)

	is it in md_rand.c:ssleay_rand_initialize() where you see a hangup?
	this is where we try to read 16 bytes /dev/random using fread().
	or somewhere else?

itojun