Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 14 Feb 2003 19:38:07 -0800
From:      David Schultz <dschultz@uclink.Berkeley.EDU>
To:        "Andrey A. Chernov" <ache@nagual.pp.ru>
Cc:        Nate Lawson <nate@root.org>, Juli Mallett <jmallett@FreeBSD.ORG>, Kirk McKusick <mckusick@FreeBSD.ORG>, src-committers@FreeBSD.ORG, cvs-src@FreeBSD.ORG, cvs-all@FreeBSD.ORG
Subject:   Re: cvs commit: src/sbin/newfs mkfs.c src/sys/ufs/ffs ffs_alloc.c ffs_vfsops.c
Message-ID:  <20030215033807.GB3750@HAL9000.homeunix.com>
In-Reply-To: <20030215022542.GA62285@nagual.pp.ru>
References:  <20030214221503.GA59673@nagual.pp.ru> <Pine.BSF.4.21.0302141625390.39134-100000@root.org> <20030215022542.GA62285@nagual.pp.ru>

next in thread | previous in thread | raw e-mail | index | archive | help
Thus spake Andrey A. Chernov <ache@nagual.pp.ru>:
> On Fri, Feb 14, 2003 at 16:31:10 -0800, Nate Lawson wrote:
> > On Sat, 15 Feb 2003, Andrey A. Chernov wrote:
> > > On Fri, Feb 14, 2003 at 15:33:28 -0600, Juli Mallett wrote:
> > > > Are the sequences for it also repeatable in the newfs case for e.g. the
> > > > regression tests, where it is used unseeded?
> > > 
> > > Obvious answer - no.
> > > 
> > > BTW, this fix really fixes nothing for FreeBSD case because we already use
> > > srandomdev() (for non-regression case).
> > 
> > Speaking totally informally, arc4 is more secure for this purpose than an
> > LFSR despite seeding the latter via srandomdev().
> 
> It isn't big news, arc4random() internal state size is 256 bytes while
> default random() state size is 124 bytes, but I don't think that anybody
> can feel that difference inside short range needed for mkfs. The complaint
> is about broken regression test case, not about srandom()->arc4random()
> transition taken alone.

The issue isn't how random the data ``feels'' according to ad hoc
tests; the issue is predictability.  NFS relies upon generation
numbers being unguessable to prevent unauthenticated clients from
spoofing file handles.  Even if random() is a pretty good PRNG,
arc4random() gives people peace of mind because it uses a real
cryptographic hash.  If you could guess arc4random() values
without knowing the entropy pool, that would likely constitute an
attack on RC4 itself.

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-src" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030215033807.GB3750>