> try yourself with OpenOffice.org 2.1:
> http://www.milw0rm.com/sploits/12122006-djtest.doc

This looks like it's an integer overflow that's then crashing on when
memset tries to write lots of zeros onto the heap.

Breakpoint 2, WW8PLCF::GeneratePLCF (this=0xb12a36e8, pSt=0xabae6cc8,
nPN=0, ncpN=587202560)
at /usr/src/debug/OOD680_m5/sw/source/filter/ww8/ww8scan.cxx:2299
2299 nIMax = ncpN;
Current language: auto; currently c++
(gdb) list
2294
2295 void WW8PLCF::GeneratePLCF( SvStream* pSt, long nPN, long ncpN )
2296 {
2297 ASSERT(!this, "Not a bug, but I (cmc) want to see this .doc as
an example");
2298 ASSERT( nIMax < (long)ncpN, "Pcl.Fkp: Warum ist PLCF zu gross
?" );

2299 nIMax = ncpN;
2300 long nSiz = 6 * nIMax + 4;
2301 pPLCF_PosArray = new INT32[ ( nSiz + 3 ) / 4 ]; // Pointer auf
Pos-Array
2302 memset( pPLCF_PosArray, 0, (size_t)nSiz );

(gdb) print ncpN
$1 = 587202560
(gdb) print nSiz
$2 = -771751932
(gdb) print *pPLCF_PosArray
$3 = 0

The memset at line 2302 tries to write 3523215364 zeros onto the heap at
the location of pPLCF_PosArray.

I don't see this exploitable beyond a DoS given the codepath here. If
someone else could take a look and agree or disagree with me I would
appreciate it.

--
JB

[ reply ]