CVE-2011-2896: Buffer Overflow

GIF image file format readers in various open source projects are based on the GIF decoder implementation written by David Koblas. This implementation contains a bug in the LZW decompressor, causing it to in correctly handle compressed streams that contain code words that were not yet added to the decompression table. LZW decompression has a special case (a KwKwK string) when code word may match the first free entry in the decompression table. The implementation used in this GIF reading code allows code words not only matching, but also exceeding the first free entry.

This problem is identical to a bug found in BSD compress (CVE-2011-2895, bug #727624), but given the unclear relationship between BSD compress and GIF decoder code bases, separate CVE is used here.

Several projects refer to pbmplus as the source form where GIF reading code was taken:

http://www.acme.com/software/pbmplus/

In pbmplus version of the code, the flaw can be found in LWZReadByte():

if (code >= maxcode) { sp++ = firstcode; code = oldcode; }

This allows creating a loop in the decompression table, which leads to an "infinite" loop:

while (code >= clearcode) { sp++ = table[1][code]; if (code == table[0][code]) pmerror("circular table entry BIG ERROR"); code = table[0][code]; }

where:

#define MAXLWZBITS 12 static int table[2][(1<< MAXLWZBITS)]; static int stack[(1<<(MAXLWZBITS))2], sp; sp = stack;

This results in stack[] buffer overflow. If table[][] is located above stack[], stack[] overflow may further modify decoding table and break infinite loop.