bin/64150: [PATCH] ls(1) coredumps when started via execve(2) with no argv.

[Tom Rhodes]

On Fri, 12 Mar 2004 17:46:00 +0200
Ruslan Ermilov <ru at freebsd.org> wrote:

> On Fri, Mar 12, 2004 at 06:58:20AM -0600, Jacques A. Vidrine wrote:
> > On Fri, Mar 12, 2004 at 12:15:26PM +0100, Marc Olzheim wrote:
> > > On Fri, Mar 12, 2004 at 01:06:57PM +0200, Ruslan Ermilov wrote:
> > > > And the fact that optind is initially set to 1.  I wonder what
> > > > could be the implications for setuid programs.  There could be
> > > > quite unpredictable results, as the "argv" pointer is incorrectly
> > > > advanced in this case, and at least several setuid programs that
> > > > I've glanced at are vulnerable to this attack.
> > > 
> > > See also: http://www.freebsd.org/cgi/query-pr.cgi?pr=33738
> > 
> > Thanks Ruslan, Marc,
> > 
> > I think kern/33738 is on the money.  I do not see any immediate
> > ramifications, but for peace of mind I believe that exec should fail if
> > the argument array pointer is NULL.
> > 
> > I believe this would be consistent with the relevant standards: POSIX
> > already requires (a) that the first argument ``should point to a
> > filename that is associated with the process being started'' and (b)
> > ``the last member of this array is a null pointer''--- i.e. the array
> > pointer cannot be NULL.
> > 
> As Garrett already pointed out in the PR log, have you considered this?
> 
> http://www.opengroup.org/onlinepubs/007904975/functions/execve.html#tag_03_130_08
> 
> I'm happy with changing our behavior to Strictly Conforming for the
> goods of security, and you?

Will it 'break' anything?

-- 
Tom Rhodes