_______________________________________________________________________ Rapid7, LLC Security Advisory Visit http://www.rapid7.com/ to download NeXpose, SC Magazine Winner of Best Vulnerability Management product. _______________________________________________________________________ Rapid7 Advisory R7-0025 Buffer Overflow in NVIDIA Binary Graphics Driver For Linux Published: Oct 16, 2006 Revision: 1.0 http://www.rapid7.com/advisories/R7-0025.jsp 1. Affected system(s): KNOWN VULNERABLE: o NVIDIA Driver For Linux v8774 o NVIDIA Driver For Linux v8762 PROBABLY VULNERABLE: o NVIDIA Driver for FreeBSD o NVIDIA Driver for Solaris o Earlier versions KNOWN FIXED: o None 2. Summary The NVIDIA Binary Graphics Driver for Linux is vulnerable to a buffer overflow that allows an attacker to run arbitrary code as root. This bug can be exploited both locally or remotely (via a remote X client or an X client which visits a malicious web page). A working proof-of-concept root exploit is included with this advisory. The NVIDIA drivers for Solaris and FreeBSD are also likely to be vulnerable. 3. Vendor status and information NVIDIA Corporation http://www.nvidia.com There have been multiple public reports of this NVIDIA bug on the NVNews forum [1,2] and elsewhere, dating back to 2004 [3]. NVIDIA's first public acknowledgement of this bug was on July 7th, 2006. In a public posting [1] on the NVNews forum, an NVIDIA employee reported having reproduced the problem, assigned it bug ID 239065, and promised a fix would be forthcoming. As of the publication date, the latest NVIDIA binary driver is still vulnerable. Furthermore, it is our opinion that NVIDIA's binary driver remains an unacceptable security risk based on the large numbers of reproducible, unfixed crashes that have been reported in public forums and bug databases. This number does not include bugs reported directly to NVIDIA. 1. http://www.nvnews.net/vbulletin/showthread.php?p=931048 (Jul 2006) 2. http://www.nvnews.net/vbulletin/showthread.php?t=76493 (Sep 2006) 3. https://bugs.freedesktop.org/show_bug.cgi?id=2129 (Dec 2004) 4. http://lists.freedesktop.org/archives/xorg/2005-January/005642.html 5. http://forums.gentoo.org/viewtopic.php?t=282107 (Jan 2005) 6. https://bugs.eclipse.org/bugs/show_bug.cgi?id=87299 (Mar 2005) 7. http://www.nvnews.net/vbulletin/showthread.php?t=76206 (Sep 2006) 4. Solution Disable the binary blob driver and use the open-source "nv" driver that is included by default with X. 5. Detailed analysis There are two NVIDIA graphics drivers for Linux: a closed-source binary blob driver provided by NVIDIA (which provides acceleration) and an open-source driver (which lacks acceleration). NVIDIA's binary blob driver contains an error in its accelerated rendering of glyphs (text character data) that can be exploited to write arbitrary data to anywhere in memory. The open-source driver is not vulnerable. The XRender extension provides a client function named XRenderCompositeString8 which tells the X server to render glyphs onto the screen. This request is processed by the server's ProcRenderCompositeGlpyhs function. This function pulls the glyphs out of the render request, constructs a glyph list, and then calls into the graphics driver via a registered callback function. The NVIDIA binary blob driver registers a function named _nv000373X. This function calculates a bounding BoxRec of the total area occupied by the glyph data. It then uses Xalloc to allocate a buffer large enough to hold the data by multiplying width * height. This buffer is then passed to another internal function called _nv000053X. The _nv000053X function iterates over the glyph list and copies glyph data into the buffer using each glyph's accumulated width, xOff, height, and yOff values to calculate the destination position in the buffer. The NVIDIA binary blob driver does not check this calculation against the size of the allocated buffer. As a result, a short sequence of user-supplied glyphs can be used to trick the function into writing to an arbitrary location in memory. It is important to note that glyph data is supplied to the X server by the X client. Any remote X client can gain root privileges on the X server using the proof of concept program attached. It is also trivial to exploit this vulnerability as a DoS by causing an existing X client program (such as Firefox) to render a long text string. It may be possible to use Flash movies, Java applets, or embedded web fonts to supply the custom glyph data necessary for reliable remote code execution. A simple HTML page containing an INPUT field with a long value is sufficient to demonstrate the DoS. 5. Credit This vulnerability was discovered by Derek Abdine of Rapid7. Special thanks to Marc Bevand for his assistance. 6. Contact Information Rapid7, LLC Email: advisory@rapid7.com Web: http://www.rapid7.com Phone: +1 (310) 316-1235 7. Disclaimer and Copyright Rapid7, LLC is not responsible for the misuse of the information provided in our security advisories. These advisories are a service to the professional security community. There are NO WARRANTIES with regard to this information. Any application or distribution of this information constitutes acceptance AS IS, at the user's own risk. This information is subject to change without notice. This advisory Copyright (C) 2006 Rapid7, LLC. Permission is hereby granted to redistribute this advisory, providing that no changes are made and that the copyright notices and disclaimers remain intact.