Re: OpenBSD kernel holes ...

[noir () uberhax0r net]

i will be releasing a paper regarding kmem allocator (heap) overflows in
kernel space and exploit for patch 005 will be in its content.

buf = malloc(user_controled_size);
vn_rdwr(UIO_READ, ..., user_buf, user_controlled_size, ...);

these types of vulnerabilities are %100 exploitable!
check kern_malloc.c line 178

                if (size > MAXALLOCSAVE)
                        allocsize = round_page(size);

this might hint you or not ...

i have only release the stack based exploit since there is nothing new in
the technique but the heap technique deserves more explanation and
attention than an exploit post ...

 - noir


On Tue, 18 Nov 2003, Steve Tornio wrote:

> <snip>
>
> > from http://www.wideopenbsd.org/errata.html
> >
> > All architectures
> >
> >      005: RELIABILITY FIX: November 4, 2003
> >      It is possible for a local user to cause a system panic by
> > executing
> >      a specially crafted binary with an invalid header.
> >      A source code patch exists which remedies the problem.
> >
> >
> > reliability ??? ehh ;-P yeah yeah right!
>
> um, that's the wrong errata entry.  For 3.4 -
>
> 006: SECURITY FIX: November 17, 2003
>   It may be possible for a local user to overrun the stack in
> compat_ibcs2(8).
>   ProPolice catches this, turning a potential privilege escalation into
> a denial of service. iBCS2 emulation does not need to be enabled via
> sysctl(8) for this to happen.
>   A source code patch exists which remedies the problem.
>
> Taken from http://www.openbsd.org/security.html#34