sched/core: Initialize the idle task with preemption disabled

As pointed out by commit

de9b8f5dcbd9 ("sched: Fix crash trying to dequeue/enqueue the idle thread")

init_idle() can and will be invoked more than once on the same idle
task. At boot time, it is invoked for the boot CPU thread by
sched_init(). Then smp_init() creates the threads for all the secondary
CPUs and invokes init_idle() on them.

As the hotplug machinery brings the secondaries to life, it will issue
calls to idle_thread_get(), which itself invokes init_idle() yet again.
In this case it's invoked twice more per secondary: at _cpu_up(), and at
bringup_cpu().

Given smp_init() already initializes the idle tasks for all *possible*
CPUs, no further initialization should be required. Now, removing
init_idle() from idle_thread_get() exposes some interesting expectations
with regards to the idle task's preempt_count: the secondary startup always
issues a preempt_disable(), requiring some reset of the preempt count to 0
between hot-unplug and hotplug, which is currently served by
idle_thread_get() -> idle_init().

Given the idle task is supposed to have preemption disabled once and never
see it re-enabled, it seems that what we actually want is to initialize its
preempt_count to PREEMPT_DISABLED and leave it there. Do that, and remove
init_idle() from idle_thread_get().

Secondary startups were patched via coccinelle:

@begone@
@@

-preempt_disable();
...
cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);

Signed-off-by: Valentin Schneider <[e-mail]>
Signed-off-by: Ingo Molnar <[e-mail]>
Acked-by: Peter Zijlstra <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

arch/alpha/kernel/smp.c

···
 	DBGS(("smp_callin: commencing CPU %d current %p active_mm %p\n",
 	      cpuid, current, current->active_mm));
 
-	preempt_disable();
 	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
 }
 

arch/arc/kernel/smp.c

···
 	pr_info("## CPU%u LIVE ##: Executing Code...\n", cpu);
 
 	local_irq_enable();
-	preempt_disable();
 	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
 }
 

arch/arm/kernel/smp.c

···
 #endif
 	pr_debug("CPU%u: Booted secondary processor\n", cpu);
 
-	preempt_disable();
 	trace_hardirqs_off();
 
 	/*

arch/arm64/include/asm/preempt.h

···
 } while (0)
 
 #define init_idle_preempt_count(p, cpu) do { \
-	task_thread_info(p)->preempt_count = PREEMPT_ENABLED; \
+	task_thread_info(p)->preempt_count = PREEMPT_DISABLED; \
 } while (0)
 
 static inline void set_preempt_need_resched(void)

arch/arm64/kernel/smp.c

···
 		init_gic_priority_masking();
 
 	rcu_cpu_starting(cpu);
-	preempt_disable();
 	trace_hardirqs_off();
 
 	/*

arch/csky/kernel/smp.c

···
 	pr_info("CPU%u Online: %s...\n", cpu, __func__);
 
 	local_irq_enable();
-	preempt_disable();
 	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
 }
 

arch/ia64/kernel/smpboot.c

···
 #endif
 	efi_map_pal_code();
 	cpu_init();
-	preempt_disable();
 	smp_callin();
 
 	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);

arch/mips/kernel/smp.c

···
 	 */
 
 	calibrate_delay();
-	preempt_disable();
 	cpu = smp_processor_id();
 	cpu_data[cpu].udelay_val = loops_per_jiffy;
 

arch/openrisc/kernel/smp.c

···
 	set_cpu_online(cpu, true);
 
 	local_irq_enable();
-
-	preempt_disable();
 	/*
 	 * OK, it's off to the idle thread for us
 	 */

arch/parisc/kernel/smp.c

···
 #endif
 
 	smp_cpu_init(slave_id);
-	preempt_disable();
 
 	flush_cache_all_local(); /* start with known state */
 	flush_tlb_all_local(NULL);

arch/powerpc/kernel/smp.c

···
 	smp_store_cpu_info(cpu);
 	set_dec(tb_ticks_per_jiffy);
 	rcu_cpu_starting(cpu);
-	preempt_disable();
 	cpu_callin_map[cpu] = 1;
 
 	if (smp_ops->setup_cpu)

arch/riscv/kernel/smpboot.c

···
 	 * Disable preemption before enabling interrupts, so we don't try to
 	 * schedule a CPU that hasn't actually started yet.
 	 */
-	preempt_disable();
 	local_irq_enable();
 	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
 }

arch/s390/include/asm/preempt.h

···
 #define init_task_preempt_count(p)	do { } while (0)
 
 #define init_idle_preempt_count(p, cpu)	do { \
-	S390_lowcore.preempt_count = PREEMPT_ENABLED; \
+	S390_lowcore.preempt_count = PREEMPT_DISABLED; \
 } while (0)
 
 static inline void set_preempt_need_resched(void)
···
 #define init_task_preempt_count(p)	do { } while (0)
 
 #define init_idle_preempt_count(p, cpu)	do { \
-	S390_lowcore.preempt_count = PREEMPT_ENABLED; \
+	S390_lowcore.preempt_count = PREEMPT_DISABLED; \
 } while (0)
 
 static inline void set_preempt_need_resched(void)

arch/s390/kernel/smp.c

···
 	restore_access_regs(S390_lowcore.access_regs_save_area);
 	cpu_init();
 	rcu_cpu_starting(cpu);
-	preempt_disable();
 	init_cpu_timer();
 	vtime_init();
 	vdso_getcpu_init();

arch/sh/kernel/smp.c

···
 
 	per_cpu_trap_init();
 
-	preempt_disable();
-
 	notify_cpu_starting(cpu);
 
 	local_irq_enable();

arch/sparc/kernel/smp_32.c

···
 	 */
 	arch_cpu_pre_starting(arg);
 
-	preempt_disable();
 	cpu = smp_processor_id();
 
 	notify_cpu_starting(cpu);

arch/sparc/kernel/smp_64.c

···
 
 	set_cpu_online(cpuid, true);
 
-	/* idle thread is expected to have preempt disabled */
-	preempt_disable();
-
 	local_irq_enable();
 
 	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);

arch/x86/include/asm/preempt.h

···
 #define init_task_preempt_count(p) do { } while (0)
 
 #define init_idle_preempt_count(p, cpu) do { \
-	per_cpu(__preempt_count, (cpu)) = PREEMPT_ENABLED; \
+	per_cpu(__preempt_count, (cpu)) = PREEMPT_DISABLED; \
 } while (0)
 
 /*

arch/x86/kernel/smpboot.c

···
 	cpu_init();
 	rcu_cpu_starting(raw_smp_processor_id());
 	x86_cpuinit.early_percpu_clock_init();
-	preempt_disable();
 	smp_callin();
 
 	enable_start_cpu0 = 0;

arch/xtensa/kernel/smp.c

···
 	cpumask_set_cpu(cpu, mm_cpumask(mm));
 	enter_lazy_tlb(mm, current);
 
-	preempt_disable();
 	trace_hardirqs_off();
 
 	calibrate_delay();

include/asm-generic/preempt.h

···
 } while (0)
 
 #define init_idle_preempt_count(p, cpu) do { \
-	task_thread_info(p)->preempt_count = PREEMPT_ENABLED; \
+	task_thread_info(p)->preempt_count = PREEMPT_DISABLED; \
 } while (0)
 
 static __always_inline void set_preempt_need_resched(void)

init/main.c

···
 	 * time - but meanwhile we still have a functioning scheduler.
 	 */
 	sched_init();
-	/*
-	 * Disable preemption - early bootup scheduling is extremely
-	 * fragile until we cpu_idle() for the first time.
-	 */
-	preempt_disable();
+
 	if (WARN(!irqs_disabled(),
 		 "Interrupts were enabled *very* early, fixing it\n"))
 		local_irq_disable();

kernel/fork.c

···
 	}
 }
 
-struct task_struct *fork_idle(int cpu)
+struct task_struct * __init fork_idle(int cpu)
 {
 	struct task_struct *task;
 	struct kernel_clone_args args = {

kernel/sched/core.c

···
  * NOTE: this function does not set the idle thread's NEED_RESCHED
  * flag, to make booting more robust.
  */
-void init_idle(struct task_struct *idle, int cpu)
+void __init init_idle(struct task_struct *idle, int cpu)
 {
 	struct rq *rq = cpu_rq(cpu);
 	unsigned long flags;

kernel/smpboot.c

···
 
 	if (!tsk)
 		return ERR_PTR(-ENOMEM);
-	init_idle(tsk, cpu);
 	return tsk;
 }