nook (nk) n. A small corner, alcove, or recess, especially one in a large room. A hidden or secluded spot. A lightweight kernel protection domain for preventing device drivers from crashing an operating system (new).

Overview

Despite decades of research in extensible operating system technology, extensions such as device drivers remain a significant cause of system failures. In Windows XP, for example, drivers account for 85% of recently reported failures. 

Nooks is a reliability subsystem that seeks to greatly enhance OS reliability by isolating the OS from driver failures. The Nooks approach is practical: rather than guaranteeing complete fault tolerance through a new (and incompatible) OS or driver architecture, our goal is to prevent the vast majority of driver-caused crashes with little or no change to existing driver and system code. To achieve this, Nooks isolates drivers within lightweight protection domains inside the kernel address space, where hardware and software prevent them from corrupting the kernel. Nooks also tracks a driver's use of kernel resources to hasten automatic clean-up during recovery.

More recently, we have extended Nooks with shadow drivers. A shadow driver is a kernel agent that (1) conceals a driver failure from its clients, including the operating system and applications, and (2) transparently restores the driver back to a functioning state. In this way, applications and the operating system are unaware that the driver failed, and hence continue executing correctly themselves.

People

Publications

• Michael Swift, Muthukaruppan Annamalai, Brian N. Bershad, and Henry M. Levy. Recovering Device Drivers. ACM Trans. on Computer Systems 24(4), November 2006.
• Michael M. Swift, Muthukaruppan Annamalai, Brian N. Bershad, Henry M. Levy. Recovering Device Drivers (pdf, 550k), Proceedings of the 6th ACM/USENIX Symposium on Operating Systems Design and Implementation, San Francisco, CA, Dec. 2004. Best paper award.
  
• Michael M. Swift, Brian N. Bershad, and Henry M. Levy. Improving the Reliability of Commodity Operating Systems (pdf, 400k),  to appear in ACM Transactions on Computer Systems, 22(4), Nov. 2004.
  
• Michael M. Swift, Brian N. Bershad, and Henry M. Levy. Improving the Reliability of Commodity Operating Systems (pdf, 300k),  in Proceedings of the 19th ACM Symposium on Operating Systems Principles, Bolton Landing, NY, Oct. 2003. Best paper award.
• Michael M. Swift, Steven Martin, Henry M. Levy, and Susan J. Eggers.
Nooks: an architecture for reliable device drivers (pdf, 162k) in Proceedings of the Tenth ACM SIGOPS European Workshop, Saint-Emilion, France, Sept. 2002. 

Presentations

• Recovering Device Drivers, or Cleaning Up Nooks talk in UW class CSE551: Graduate Operating Systems (pdf)
  
• Shadow Drivers: Transparent Recovery for Kernel Extensions poster at UW industrial Affiliates 2004 (pdf)
  
• Improving the Reliability of Commodity Operating Systems talk at SOSP 2003, October 2003 (pdf)
• Nooks poster at UW industrial Affiliates 2003 (pdf)
  
• Nooks: an architecture for reliable device drivers talk at ACM SIGOPS worksop, September 2002(ppt)
  
• Nooks: an architecture for reliable device drivers talk at UW Networking and Systems Retreat, June 2002 (ppt)

Lessons Learned

• How to profile the Linux Kernel: use kernrprof, which may cause double-faults when used with Nooks but provides call graphs, or oprofile, which does interrupt-based sampling.

• How to mesaure microarchitectural events on the Pentium 4 with Linux: Use Brink-Abyss, but be sure to set the duration knob long enough to complete your experiment.

• What architectural state must be writable for a task gate to execute in Ring 0 on the Pentium: the GDT and TSS of the current task must be writable with the current page table, so that these can be updated before switching to the new task.

Software Downloads

• Software Fault Injection for Linux This patch applies to the linux.2.4.18 kernel and requires the kdb-v2.1.2-2.4.18-common-1 and kdb-v2.1-2.4.18-i386.1 patches. It is enabled by turning on kernel debugging and enabling the "Load all symbols for debugging" and "Automatically inject faults" options when configuring the kernel.
  This tool was created by porting the fault injector tool used in the Rio file cache project and is describe in the paper The Systematic Improvement of Fault Tolerance in the Rio File Cache.
  • Linux 2.4.18 Kernel patch (updated 1/12/2005)
  • Fault injection tool

• Nooks source code for the Linux 2.4.18 kernel. This tarball includes a patch against the Linux 2.4.18 kernel (available from kernel.org here ). We've tested our kernel with RedHat Linux 7.3, 9, and Enterprise Linux 3. It is known not to work with the Fedora Core 2 distribution.
  Also included are a suite of usermode tools:
  • nooks-agent: the recover agent, to be placed in /sbin
  • systest: the test tool for manually creating and manipulating nooks
  • modutils: new module loading tools to replace insmod and modprobe
  Nooks source

Grant Support

• This work is supported in part by NSF grant CCR-0326546.

Driver Reliabily Links

• Slate Magazine: Blue Screen of Death
• Ananova: Computer Crashes cost small businesses "billions"
• Computerworld: Intel's Centrino may crash Windows when used with VPNs
• ATI technologies: VPU Recovery White Paper
• Microsoft: Fault Resilient Driver Hardening for Longhorn Server
• Microsoft: An Introduction to Driver Quality
• Microsoft: Windows Error Reporting

Last modified 9/30/2004