CVE-2004-2135 and CVE-2004-2136

CVE-2004-2135 and CVE-2004-2136

[Michael Gilbert]

Hi,

CVE-2004-2135 [0] and CVE-2004-2136 [1] were disclosed over five years
ago; however, i have been unable to verify whether they have been fixed
or not in the latest kernels.  The paper [2] indicates that the
disclosing team developed a 3,000 line patch, but I have been unable to
track this down to determine whether or not it has been applied.  There
is also this more current write-up [3], which is rather revealing.  Have
these issues slipped through the cracks?  Thanks upfront for any info on
the matter.

Best wishes,
Mike

[0] http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2004-2135
[1] http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2004-2136
[2] http://mareichelt.de/pub/notmine/diskenc.pdf
[3] http://mareichelt.de/pub/texts.cryptoloop.php?alt_styles=2

Re: CVE-2004-2135 and CVE-2004-2136

[Michael Gilbert]

On Thu, 3 Dec 2009 17:53:21 +0100 markus reichelt wrote:
> * Michael Gilbert wrote:
> 
> > CVE-2004-2135 [0] and CVE-2004-2136 [1] were disclosed over five
> > years ago; however, i have been unable to verify whether they have
> > been fixed or not in the latest kernels.
> 
> They are not, and both CVE-entries are still marked as 'candidate'.
> Add the fact that (crypto)loop code has had a poor relation with the
> vanilla kernel in favour of dm-crypt, and is also unmaintained,
> nobody really cares.
> 
> Did you receive any other replies to your mail? Just curious.

None so far.  Would it make sense to remove cryptoloop at this point
since it obviously has some major design/implementation weaknesses?
Apropos, there are two other full-disk encryption options (dm-crypt and
loop-aes) that, importantly, are not affected by these issues. Is it
really a good idea to have three different implementations of
essentially the same functionality?

Best wishes,
Mike

Re: CVE-2004-2135 and CVE-2004-2136

[Michael Gilbert]

On Thu, 3 Dec 2009 20:53:11 -0500 Michael Gilbert wrote:

> On Thu, 3 Dec 2009 17:53:21 +0100 markus reichelt wrote:
> > * Michael Gilbert wrote:
> > 
> > > CVE-2004-2135 [0] and CVE-2004-2136 [1] were disclosed over five
> > > years ago; however, i have been unable to verify whether they have
> > > been fixed or not in the latest kernels.
> > 
> > They are not, and both CVE-entries are still marked as 'candidate'.
> > Add the fact that (crypto)loop code has had a poor relation with the
> > vanilla kernel in favour of dm-crypt, and is also unmaintained,
> > nobody really cares.
> > 
> > Did you receive any other replies to your mail? Just curious.
> 
> None so far.  Would it make sense to remove cryptoloop at this point
> since it obviously has some major design/implementation weaknesses?
> Apropos, there are two other full-disk encryption options (dm-crypt and
> loop-aes) that, importantly, are not affected by these issues. Is it
> really a good idea to have three different implementations of
> essentially the same functionality?

For anyone that is interested, I have found that the dm-crypt issue,
CVE-2004-2136, was addressed in the 2.6.10 kernel release. The
cryptoloop issue, CVE-2004-2135, still exists.  

So, should cryptoloop be removed from the kernel?  I've heard elsewhere
that it is already deprecated, so perhaps now is the time to finally get
rid of it.

Best wishes,
Mike