RecoverOS

This code may not look so suitable in many places, and will be refactored sooner or later. New warehouse recover-os. Wait for me to come back to fill in the pit. Refactored version: https://github.com/yiezias/recover-os

Introduction

This project is based on the AMD64 architecture (or x86_64 architecture) version of the operating system implemented by "Operating System Truth Restore". The main ideas and code are consistent with the original book.
The operating system implemented by "Operating System Truth Restore" runs on a 32-bit x86 architecture processor, but now we have already entered the 64-bit era. So I copy Refer to the code in "Operating System Truth Restore" to implement its 64-bit version.

Quick use

step

• Install newer versions of gcc, nasm, bochs, GNU make and other programs under x64 Linux.
• Clone this repository and enter the source code directory.
• Then $ make run can compile and start the bochs emulator.
  This system was successfully tested in bochs 2.7, gcc 12.2.0, nasm 2.15.05, make 4.3 environments. Different versions of some software should not have much impact. The operation effect is as follows:

Example

This operating system has built-in 2048, Snake and other mini games.
Write code with recover:
(The terminal does not implement input and output redirection, and the cat below is different from the cat under Linux)
(Just show the function of writing text files)

2048 mini game:

Greedy snake (the frame drop is serious, so I won’t let the animations go):

About "Operating System Truth Restore"

The most easy-to-understand related books I have read, detailed and interesting. Even if it seems a bit long-winded, the author still needs to try different ways to let readers understand (where can I find such a friendly book). Among the many operating system-related books, I finally chose this one The only one that can understand Book implementation of operating system.

Different from the original book

Mainly the difference between AMD64 and x86 architectures

Function Call Convention

Under the AMD64 architecture, functions generally use registers to pass parameters, and the function calling convention under x86 has no reference value. But it is easy to understand itself. Write a C function in detail and compile it into assembly and then you will understand it by writing it in GCC. This knowledge is needed in many places during the development process.

Protected mode and long mode

Long mode can be understood as 64-bit feature Protected mode, the processor needs to switch from real mode to protected mode and then to long mode.

• The segmentation mechanism of the protection mode is very cumbersome, and it has been greatly simplified in the long mode.
• However, the paging mechanism in the long mode is more complex, and only the four-level page table is used in this system. However, the author wrote very well in the book about the secondary page table. He understood the secondary page table and the fourth page table is not that difficult.
• The interrupt exception parts of the two are basically the same. It should be noted that the tss and ist mechanisms. Due to the change of function calling conventions, stack overwrite occurs when using the author's code when the process switches. Through the ist mechanism, the interrupt stack can be changed every time the process switch is changed, thereby solving this problem. For details, you can read the relevant code.

The most convenient way to obtain information is of course a search engine, but search engines often cannot find the content they want, and the truly authoritative information is still AMD or Intel documents.

Quick system calls

This system completely abandons the method of interrupting the implementation of system calls and uses syscall sysret fast system call instruction to implement it. For details, please refer to AMD or Intel documents, and the content is not complicated.

File system

The author's file system doesn't feel very well written. On the one hand, the file system is already quite complicated, but the author often uses nearly hundreds of lines of functions, and some code can be reused, but it needs to be implemented repeatedly, which is very confusing overall. This system file system is equivalent to a major change in the author, but it should not be difficult to understand based on the author's book and code. The file system of this system does not implement directory-level functions, but does not affect the implementation of other functions of the system, so it is omitted.

Application layer

The author's shell is implemented in the kernel. This system uses the shell as an application, and the init process uses the execv system call to load the root directory shell. Many built-in commands of the author system are also directly implemented as applications in this system. But the code level is similar. There are also special mini-games that can call the operating system interface to implement applications independently, and this system is basically completed.