mirror of
https://github.com/ElliotKillick/operating-system-design-review
synced 2026-06-06 15:34:33 +00:00
Elliot Killick
8f9257c1da
From June to now, most of my time on this project was spent writing "The Rise of Microsoft". Researching and writing that thing took an inordinate amount of time. Also, I took a break to work primarily on another unreleased project I have before switching back to this one. I am proud of what I got here though and I feel like it is all coming together. My plans for the whitepapers are already well drafted and I know they will make an impact. I have learned a ton from working on this project and it has definitely given me something to idly think about and work on in my spare time. Finishing "The Problem with How Windows Uses Threads" was something I thought I wanted done before this release, but oh well. The main README is at about 50K words now, plus all the other smaller documents and you are looking at about a small novel's worth of technical writing. And technical writing takes multiple times longer than typical story writing. Anyway, this project is exciting for me - people who I show it to also really like it - and I am even more excited about what is to come!
LoadLibrary Initialization Circular Dependency Experiment
We evaluate the loader's behavior when encountering a circular dependency during module initialization.
See the glibc equivalent experiment.
Result
Start: DLL 1 Init
Start: DLL 2 Init
DLL 1, func
End: DLL 2 Init
End: DLL 1 Init
Conclusion: As expected, the Windows loader has no choice but to initialize DLL 2 while DLL 1 could be partially initialized due to a circular dependency. The PE executable format exposes all initialization and finalization tasks as merged into one externally callable EntryPoint and arbitrary code cannot be interrupted or preempted, so the Windows loader's current behavior is the best it can be under the conditions.