arm64_nmi_cpu_stop(9)

✨ NAME ✨

okay so bestie, welcome to arm64_nmi_cpu_stop(9) — the Kernel API Manual page you didn't know you needed 🎀 — and here's the whole vibe in one line: arm64_nmi_cpu_stop — stop the local CPU after it is told to stop. literally that's it, that's the function 💅

---

🌸 SYNOPSIS 🌸

void __noreturn arm64_nmi_cpu_stop (struct pt_regs *regs , bool
die_on_crash );

no cap, it returns void, it's __noreturn (like, it does NOT come back, don't wait up 😭), and it takes exactly two things. let's meet them 🫶

---

💜 ARGUMENTS 💜

• regs — the register state you wanna record in the vmcore on a crash stop. OR, plot twist, pass NULL for panic_smp_self_stop, because that path has no interrupted context to save. it's giving "nothing to see here" 🙈
• die_on_crash — on the kdump crash path, this tells the CPU to power itself off via PSCI CPU_OFF (so a capture kernel can come reclaim it) instead of just parking it. powered off vs. parked, iconic difference 👑

---

🔥 DESCRIPTION 🔥

so this function is the single point that every arm64 stop path funnels through 🌀 — it keeps all the bookkeeping in one place so nobody's out here duplicating chores:

• mask interrupts
• save the crash context
• mark offline
• mask SDEI
• optionally power off

one function, all the errands, main character energy fr fr 💅

and the paths that funnel into it are:

• the regular IPI_CPU_STOP and pseudo-NMI IPI_CPU_STOP_NMI handlers;
• panic_smp_self_stop, which is a CPU just parking itself on a parallel panic (she's doing her own thing, respectfully);
• the SDEI cross-CPU NMI handler (drivers/firmware/arm_sdei_nmi.c), which reaches CPUs the stop IPIs could not. lowkey the outreach team of this whole operation 📡

now the important bit, bestie, listen up 👀 — the IPI stop handlers pass die_on_crash as true. the SDEI handler and panic_smp_self_stop pass false and only park. and for SDEI that's required, not just being conservative:

it runs inside an SDEI event that is deliberately never completed (completing it has firmware resume the wedged context), and a CPU_OFF from that not-yet-completed context wedges EL3 on some firmware — a documented follow-up. 😬💀

so yeah, that's a whole "do not do this" situation. parking also matches this path's own fallback for when CPU_OFF isn't available anyway, so it's consistent AND safe. slay 🌈

---

🦋 SEE ALSO 🦋

• Kernel file ./arch/arm64/kernel/smp.c — go read it, iykyk 📖✨

---

and that's the page, bestie 🎀 this cutie came from ./arch/arm64/kernel, dated October 2026, page name arm64_nmi_cpu_stop(9) — go forth and stop those CPUs safely, mwah 💖🫶