操作系统与Linux内核概述 / Operating Systems and the Linux Kernel: An Overview

操作系统是计算机系统的核心软件层，它管理硬件资源、提供程序运行环境，并抽象出统一的接口供应用程序使用。从早期的批处理系统到现代的多任务、多用户、支持网络和虚拟化的复杂系统，操作系统的演化历程反映了计算机硬件与软件协同发展的缩影。在这一演化中，Linux内核以其开源、稳定、高效和可定制的特性，成为从嵌入式设备到超级计算机广泛部署的核心。理解Linux内核的源码，不仅是掌握操作系统核心机制的钥匙，也是深入计算机系统本质的必经之路。本章将系统性地介绍操作系统的核心职责与演化脉络，聚焦Linux内核的架构设计、关键子系统及其交互机制，并通过剖析其源码目录结构、编译配置系统（Kconfig）与启动流程（start_kernel），构建对内核整体的认知框架。随后，通过实践案例——搭建内核开发环境、编译自定义内核、使用QEMU运行与调试——将理论联系实际，最终通过关键初始化函数（如setup_arch、rest_init）的源码分析，揭示内核从加载到就绪的完整生命周期。

An operating system (OS) is the foundational software layer that manages computer hardware resources, provides an execution environment for programs, and abstracts a unified interface for applications. From early batch processing systems to modern complex systems supporting multitasking, multi-user access, networking, and virtualization, the evolution of operating systems mirrors the synergistic development of computer hardware and software. Within this evolution, the Linux kernel, with its open-source nature, stability, efficiency, and customizability, has become the core deployed across a spectrum from embedded devices to supercomputers. Understanding the Linux kernel source code is not only a key to mastering core operating system mechanisms but also an essential step toward grasping the essence of computer systems. This chapter provides a systematic introduction to the core responsibilities and evolutionary context of operating systems, focusing on the architectural design, key subsystems, and their interaction mechanisms within the Linux kernel. By analyzing its source directory structure, the compilation configuration system (Kconfig), and the boot process (start_kernel), we establish a cognitive framework for the kernel as a whole. Subsequently, practical cases—setting up a kernel development environment, compiling a custom kernel, running and debugging with QEMU—bridge theory and practice. Finally, through source code analysis of critical initialization functions (such as setup_arch and rest_init), we reveal the complete lifecycle of the kernel from loading to readiness.

---

1.1 操作系统的核心职责与抽象层 / Core Responsibilities and Abstraction Layers of Operating Systems

操作系统的核心职责可概括为四大支柱：进程管理、内存管理、设备管理（I/O与文件系统）以及安全与保护。这些职责通过一系列抽象层实现，将底层硬件的复杂性隐藏，向上层应用程序提供简洁、一致的接口。

The core responsibilities of an operating system can be summarized into four pillars: process management, memory management, device management (I/O and file systems), and security and protection. These responsibilities are realized through a series of abstraction layers that hide the complexity of underlying hardware and provide simple, consistent interfaces to upper-layer applications.

1.1.1 进程管理：从CPU时间片到并发实体 / Process Management: From CPU Time Slices to Concurrent Entities

进程是操作系统进行资源分配和调度的基本单位。传统上，进程包含代码段、数据段、堆、栈以及一组系统资源（如打开的文件、信号量等）。现代操作系统更倾向于区分“进程”作为资源容器，而“线程”作为调度实体，但Linux内核 historically 将线程实现为共享资源的特殊进程（通过`clone()`系统调用创建，共享地址空间等）。

A process is the fundamental unit for resource allocation and scheduling in an operating system. Traditionally, a process comprises a text segment (code), data segment, heap, stack, and a set of system resources (such as open files, semaphores, etc.). Modern operating systems tend to distinguish between "processes" as resource containers and "threads" as schedulable entities, but the Linux kernel historically implements threads as special processes that share resources (created via the `clone()` system call, sharing address space, etc.).

进程管理涉及创建与终止、调度与切换、同步与通信以及状态转换。内核维护一个`task_struct`结构体（在`include/linux/sched.h`中定义）的链表来描述所有进程，称为“进程表”。该结构体包含进程状态（`state`）、优先级（`prio`）、调度策略（`policy`）、地址空间指针（`mm`）、打开文件列表（`files`）等数百个字段，是内核理解进程的“上帝对象”。

Process management involves creation and termination, scheduling and switching, synchronization and communication, and state transitions. The kernel maintains a linked list of `task_struct` structures (defined in `include/linux/sched.h`) to describe all processes, known as the "process table". This structure contains hundreds of fields including process state (`state`), priority (`prio`), scheduling policy (`policy`), address space pointer (`mm`), open file list (`files`), etc., making it the "God object" through which the kernel understands processes.

调度器是进程管理的核心。Linux采用完全公平调度器（Completely Fair Scheduler, CFS），其核心思想是模拟一个“理想的、精确的多任务CPU”，每个进程获得公平的CPU时间份额。CFS不再依赖固定的时间片，而是基于虚拟运行时间（vruntime）进行决策。调度实体（`sched_entity`）的vruntime记录该进程已消耗的CPU时间（按优先级加权），调度器总是选择vruntime最小的进程运行。这一机制通过红黑树（`rb_root`）高效维护，确保$O(\log N)$的调度复杂度。

The scheduler is the core of process management. Linux employs the Completely Fair Scheduler (CFS), whose core idea is to simulate an "ideal, perfectly multitasking CPU" where each process receives a fair share of CPU time. CFS no longer relies on fixed time slices but makes decisions based on virtual runtime (vruntime). The vruntime of a scheduling entity (`sched_entity`) records the CPU time consumed by that process (weighted by priority), and the scheduler always selects the process with the smallest vruntime to run. This mechanism is efficiently maintained using a red-black tree (`rb_root`), ensuring $O(\log N)$ scheduling complexity.

实例分析：CFS调度器的核心数据结构 Consider the core data structure of CFS. The `struct cfs_rq` (completely fair runqueue) in `kernel/sched/sched.h` contains: ```c struct cfs_rq { struct rb_root_cached tasks_timeline; // 红黑树根，按vruntime排序 unsigned long nr_running; // 当前运行进程数 u64 min_vruntime; // 最小vruntime，用于新进程公平插入 // ... 其他字段 }; ``` 当需要调度时，`pick_next_task_fair()`函数从红黑树最左侧节点（最小vruntime）取出进程。这种设计避免了传统时间片轮转的固定周期问题，能更好地适应不同负载和进程数量。

Example: Core Data Structure of CFS Scheduler Consider the core data structure of CFS. The `struct cfs_rq` (completely fair runqueue) in `kernel/sched/sched.h` contains: ```c struct cfs_rq { struct rb_root_cached tasks_timeline; // Red-black tree root, sorted by vruntime unsigned long nr_running; // Number of currently running processes u64 min_vruntime; // Minimum vruntime, for fair insertion of new processes // ... other fields }; ``` When scheduling is needed, the `pick_next_task_fair()` function retrieves the process from the leftmost node (smallest vruntime) of the red-black tree. This design avoids the fixed-period issues of traditional time-slice round-robin and adapts better to varying loads and numbers of processes.

1.1.2 内存管理：虚拟地址空间与分页机制 / Memory Management: Virtual Address Space and Paging Mechanism

内存管理负责为进程提供独立的地址空间、物理内存的分配与回收、以及虚拟内存到物理内存的映射。Linux采用虚拟内存机制，每个进程拥有独立的虚拟地址空间（通常32位为4GB，64位更大），通过分页（paging）映射到物理内存。这种抽象带来多重好处：进程间地址空间隔离（安全）、物理内存超量使用（通过交换到磁盘）、以及内存映射文件（mmap）。

Memory management is responsible for providing independent address spaces for processes, allocating and reclaiming physical memory, and mapping virtual memory to physical memory. Linux employs a virtual memory mechanism where each process has its own virtual address space (typically 4GB for 32-bit, larger for 64-bit), mapped to physical memory via paging. This abstraction provides multiple benefits: isolation between process address spaces (security), overcommitment of physical memory (via swapping to disk), and memory-mapped files (mmap).

核心数据结构是页表（Page Table）和内存描述符（`mm_struct`）。`mm_struct`（在`include/linux/mm_types.h`中）描述一个进程的完整地址空间，包含：

• `pgd`：页全局目录指针（顶级页表）
• `mmap`：虚拟内存区域（VMA）链表，描述连续的内存区间（如代码段、数据段、堆、共享库等）
• `rss`：驻留集大小（物理内存页数）
• `total_vm`：虚拟地址空间总页数

The core data structures are the page table and the memory descriptor (`mm_struct`). The `mm_struct` (in `include/linux/mm_types.h`) describes a process's complete address space and contains:

• `pgd`: Page Global Directory pointer (top-level page table)
• `mmap`: Linked list of Virtual Memory Areas (VMAs), describing contiguous memory regions (e.g., code segment, data segment, heap, shared libraries, etc.)
• `rss`: Resident Set Size (number of physical memory pages)
• `total_