为何cpu_idle被作为start_kernel例程的最后调用函数？其功能是什么？

Great question—let’s break this down step by step, since this gets to the heart of how the Linux kernel transitions from boot-up to steady-state operation.

为什么start_kernel最后调用cpu_idle？ #

First, remember what start_kernel does: it’s the entry point for kernel initialization, responsible for setting up all core subsystems—memory management, process scheduler, interrupt handling, device drivers, and more. By the time start_kernel finishes, the kernel is fully prepared to run user-space processes and kernel threads.

The final call to rest_init is where the handoff happens:

• rest_init creates the first user-space process (PID 1, the init or systemd process that boots the rest of the user environment) and the kernel thread manager (PID 2, kthreadd).
• Once these critical processes are up and running, the original boot thread (which started as init_task, PID 0) has finished its initialization work. It needs to transition into a new role: the idle process.

Calling cpu_idle is how this transition happens. PID 0 becomes the system’s fallback process—something the scheduler can run when there are no other tasks waiting for CPU time.

cpu_idle的具体功能 #

The cpu_idle function is the core loop of the PID 0 idle process, and its job is to keep the CPU efficient when it’s not doing useful work:

• Low-power CPU state management: When the scheduler switches to the idle process, cpu_idle triggers the CPU to enter a low-power idle state (like HLT on x86, or WFI/WFE on ARM). In this state, the CPU stops executing instructions and consumes minimal power until an interrupt (hardware event, timer, etc.) wakes it up.
• Fallback for the scheduler: It’s the lowest-priority process in the system—only runs when every other runnable process has been scheduled. As soon as a new task becomes available (e.g., a hardware interrupt triggers a network packet to process), the scheduler will switch away from cpu_idle to that task.
• Idle-time housekeeping: It handles background tasks that only make sense when the CPU is free, like:
  • Tracking system idle time for load averages and performance metrics;
  • Supporting tickless idle (a mode where the periodic timer tick is disabled to save even more power);
  • Coordinating idle states across multiple CPUs in SMP systems.

In short: start_kernel builds the stage, rest_init brings on the main actors, and cpu_idle is the quiet backdrop that keeps things efficient when the stage is empty.

内容的提问来源于stack exchange，提问作者Anthony