Ask HN: Recommended resources to learn the Linux kernel and OS theory?

This sounds like a bit of an oxymoron. What is a modular-monolithic kernel?

For learning the basics about how the kernels of Unix-like systems are designed, I recommend the following books:

- The Design of the UNIX Operating System (1986) by Maurice Bach. You could also compare this book to the source code of Version 7 Unix, which is freely available under a BSD-style license. Many of the basic design principles of Unix-like operating systems have not changed dramatically over the years, and so studying the design and implementation of older, smaller implementations of Unix-like operating systems can aid with figuring out the code base of Linux, which is huge.

- UNIX Internals: The New Frontier (1996) by Uresh Vahalia. This discusses some more modern features that weren't part of AT&T Unix.

- The Design and Implementation of the FreeBSD Operating System (2014) by Marshall Kirk McKusick et al. I've met "Kirk" in person a few times at USENIX FAST conferences; he's not only a great writer, but he is also very friendly and approachable. This is a well-written text describing FreeBSD 11, including modern features such as ZFS.

There are some good Linux kernel books (I'm partial to "Linux Kernel Development" by Robert Love, which helped me out nine years ago when I was writing a device-mapper device), but what I've found during my experience working with the Linux kernel is that the Linux kernel changes much faster than the books describing it get updated. Nevertheless, the best thing to do is to get a solid understanding of how Unix works, then get a basic understanding of the Linux kernel, then start diving into the source code, which is ultimately the definitive documentation of the Linux kernel.

When I was working on that device-mapper device, I found that perusing the Linux kernel mailing list was excellent for learning about how relevant portions of the kernel changed over time.

Good luck and have fun! I haven't done much kernel-level work lately, but I miss working at that level!

The big metric is motion-to-photon latency, which is a measure (in milliseconds) of how quickly a user's head motion is reflected on the display. Anything greater than ~10ms is considered insufficient, as you've missed a whole frame of rendering for VR headsets that run at 90hz refresh rate.

Current XR applications need to wait for the USB or Bluetooth stack to return the head pose before modifying the view, then wait for the graphics stack to reflect that change. There are many ring transitions along the way. This leads to an absolute minimum motion-to-photon latency that can't be broken just because of OS overhead. This eats into your frame budget, as every ms of context-switching out of the IO drivers leaves you fewer ms to render a frame in time to get it back into the kernel for context-switching back to the graphics driver.

Udacity also hosts the GT Advanced Operating Systems course: https://www.udacity.com/course/advanced-operating-systems--u.... AOS is not a "how to" course, but rather a tour of the challenges operating systems may face (distribution, fault tolerance, scale) and of academic papers attempting to solve them (mostly from the 90s and 00s). As such, the material might not be a good fit for the original post, but this commenter found it fascinating.

The reading list is here: https://www.udacity.com/wiki/ud156-readings