Bowman I. Concrete Architecture of the Linux Kernel

University of Waterloo, 1998 — 34 p.
The objective of this report is to describe the concrete (as-built) architecture of the Linux kernel. A concrete architecture description of the Linux kernel serves as a high-level specification for developers to modify and extend the existing kernel source code.
We used a reverse engineering tool (Portable Bookshelf) to extract the detailed design information out of the Linux kernel source. The biggest challenge we had to overcome was to cluster the enormous volume of extracted information into subsystems. We clustered the extracted information based on our domain-specific knowledge about modern operating systems; the Software Landscape visualization tool was used in an iterative process to view and refine this clustering.
Although the Portable Bookshelf tool provided detailed design information and some initial clustering based on file names, we found that this was not sufficient to form a concrete architecture description. The extraction tool didn’t accurately extract dependencies. It missed some dependencies, and asserted some that did not exist. The tool didn’t automatically provide a structure that helped understand the system. We verified the extracted design facts and refined the clustering using our conceptual model of the Linux kernel.
We concluded that the Portable Bookshelf tool by itself cannot accurately describe the concrete architecture of the Linux kernel. We remedied this limitation by using domain-specific knowledge to describe the concrete architecture.