On Sat, Oct 25, 2025 at 06:58:00PM +0200, Thorsten Blum wrote:
> On 22. Oct 2025, Thorsten Blum wrote:
> > Index: sys/crypto/set_key.c
> > ===================================================================
> > RCS file: /cvs/src/sys/crypto/set_key.c,v
> > retrieving revision 1.5
> > diff -u -p -r1.5 set_key.c
> > --- sys/crypto/set_key.c	12 Mar 2021 10:22:46 -0000	1.5
> > +++ sys/crypto/set_key.c	21 Oct 2025 11:38:57 -0000
> > @@ -74,7 +74,7 @@ check_parity(des_cblock (*key))
> > 	return (1);
> > }
> > 
> > -/* Weak and semi week keys as take from
> > +/* Weak and semi-weak keys as taken from
> >  * %A D.W. Davies
> >  * %A W.L. Price
> >  * %T Security for Computer Networks
> 
> While browsing the book referenced above (available at [1]), I noticed
> that some of the weak key values differ from those in the book (page 65,
> figure 3.17). The patch below shows the different values. Not sure if
> this is intentional or a bug.
> 
> Additionally, it appears that 'weak_keys' and des_is_weak_key() may no
> longer be used. des_is_weak_key() is only called from des_set_key() when
> 'int des_check_key' is true, but I couldn't find any code in /usr/src
> setting this variable.
> 
> Should the values be fixed, or is the code obsolete/unused and about to
> be removed anyway?

I have committed your two diffs. I agree that this is currently
unreachable. I'm not sure if we still need DES in the kernel at all...

You rediscovered a very old bug. These keys were fixed in SSLeay 0.9.1b
end of December 1998 and in libcrypto when OpenSSL 0.9.4 was imported.