[patch V2 59/67] smp: Convert core to hotplug state machine

[Anna-Maria Gleixner]

From: Richard Weinberger <richard@xxxxxx>

Install the callbacks via the state machine. They are installed at runtime so
smpcfd_prepare_cpu() needs to be invoked by the boot-CPU.

Signed-off-by: Richard Weinberger <richard@xxxxxx>
[ Added the dropped CPU dying case back in. ]
Signed-off-by: Richard Cochran <rcochran@xxxxxxxxxxxxx>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
Cc: Davidlohr Bueso <dave@stgolabs>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Mel Gorman <mgorman@xxxxxxxxxxxxxxxxxxx>
Cc: Oleg Nesterov <oleg@xxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Rasmus Villemoes <linux@xxxxxxxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Signed-off-by: Anna-Maria Gleixner <anna-maria@xxxxxxxxxxxxx>
---
 include/linux/cpuhotplug.h |    2 +
 include/linux/smp.h        |    5 ++
 kernel/cpu.c               |    9 +++++
 kernel/smp.c               |   79 ++++++++++++++++++---------------------------
 4 files changed, 48 insertions(+), 47 deletions(-)

--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -19,6 +19,7 @@ enum cpuhp_state {
 	CPUHP_HRTIMERS_PREPARE,
 	CPUHP_PROFILE_PREPARE,
 	CPUHP_X2APIC_PREPARE,
+	CPUHP_SMPCFD_PREPARE,
 	CPUHP_TIMERS_DEAD,
 	CPUHP_NOTIFY_PREPARE,
 	CPUHP_BRINGUP_CPU,
@@ -58,6 +59,7 @@ enum cpuhp_state {
 	CPUHP_AP_ARM_CORESIGHT4_STARTING,
 	CPUHP_AP_ARM64_ISNDEP_STARTING,
 	CPUHP_AP_LEDTRIG_STARTING,
+	CPUHP_AP_SMPCFD_DYING,
 	CPUHP_AP_X86_TBOOT_DYING,
 	CPUHP_AP_NOTIFY_STARTING,
 	CPUHP_AP_ONLINE,
--- a/include/linux/smp.h
+++ b/include/linux/smp.h
@@ -196,4 +196,9 @@ extern void arch_enable_nonboot_cpus_end
 
 void smp_setup_processor_id(void);
 
+/* SMP core functions */
+int smpcfd_prepare_cpu(unsigned int cpu);
+int smpcfd_dead_cpu(unsigned int cpu);
+int smpcfd_dying_cpu(unsigned int cpu);
+
 #endif /* __LINUX_SMP_H */
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1195,6 +1195,11 @@ static struct cpuhp_step cpuhp_bp_states
 		.startup = hrtimers_prepare_cpu,
 		.teardown = hrtimers_dead_cpu,
 	},
+	[CPUHP_SMPCFD_PREPARE] = {
+		.name = "SMPCFD prepare",
+		.startup = smpcfd_prepare_cpu,
+		.teardown = smpcfd_dead_cpu,
+	},
 	[CPUHP_TIMERS_DEAD] = {
 		.name = "timers dead",
 		.startup = NULL,
@@ -1218,6 +1223,10 @@ static struct cpuhp_step cpuhp_bp_states
 		.teardown		= NULL,
 		.cant_stop		= true,
 	},
+	[CPUHP_AP_SMPCFD_DYING] = {
+		.startup = NULL,
+		.teardown = smpcfd_dying_cpu,
+	},
 	/*
 	 * Handled on controll processor until the plugged processor manages
 	 * this itself.
--- a/kernel/smp.c
+++ b/kernel/smp.c
@@ -33,69 +33,54 @@ static DEFINE_PER_CPU_SHARED_ALIGNED(str
 
 static void flush_smp_call_function_queue(bool warn_cpu_offline);
 
-static int
-hotplug_cfd(struct notifier_block *nfb, unsigned long action, void *hcpu)
+int smpcfd_prepare_cpu(unsigned int cpu)
 {
-	long cpu = (long)hcpu;
 	struct call_function_data *cfd = &per_cpu(cfd_data, cpu);
 
-	switch (action) {
-	case CPU_UP_PREPARE:
-	case CPU_UP_PREPARE_FROZEN:
-		if (!zalloc_cpumask_var_node(&cfd->cpumask, GFP_KERNEL,
-				cpu_to_node(cpu)))
-			return notifier_from_errno(-ENOMEM);
-		cfd->csd = alloc_percpu(struct call_single_data);
-		if (!cfd->csd) {
-			free_cpumask_var(cfd->cpumask);
-			return notifier_from_errno(-ENOMEM);
-		}
-		break;
-
-#ifdef CONFIG_HOTPLUG_CPU
-	case CPU_UP_CANCELED:
-	case CPU_UP_CANCELED_FROZEN:
-		/* Fall-through to the CPU_DEAD[_FROZEN] case. */
-
-	case CPU_DEAD:
-	case CPU_DEAD_FROZEN:
+	if (!zalloc_cpumask_var_node(&cfd->cpumask, GFP_KERNEL,
+				     cpu_to_node(cpu)))
+		return -ENOMEM;
+	cfd->csd = alloc_percpu(struct call_single_data);
+	if (!cfd->csd) {
 		free_cpumask_var(cfd->cpumask);
-		free_percpu(cfd->csd);
-		break;
+		return -ENOMEM;
+	}
 
-	case CPU_DYING:
-	case CPU_DYING_FROZEN:
-		/*
-		 * The IPIs for the smp-call-function callbacks queued by other
-		 * CPUs might arrive late, either due to hardware latencies or
-		 * because this CPU disabled interrupts (inside stop-machine)
-		 * before the IPIs were sent. So flush out any pending callbacks
-		 * explicitly (without waiting for the IPIs to arrive), to
-		 * ensure that the outgoing CPU doesn't go offline with work
-		 * still pending.
-		 */
-		flush_smp_call_function_queue(false);
-		break;
-#endif
-	};
+	return 0;
+}
 
-	return NOTIFY_OK;
+int smpcfd_dead_cpu(unsigned int cpu)
+{
+	struct call_function_data *cfd = &per_cpu(cfd_data, cpu);
+
+	free_cpumask_var(cfd->cpumask);
+	free_percpu(cfd->csd);
+	return 0;
 }
 
-static struct notifier_block hotplug_cfd_notifier = {
-	.notifier_call		= hotplug_cfd,
-};
+int smpcfd_dying_cpu(unsigned int cpu)
+{
+	/*
+	 * The IPIs for the smp-call-function callbacks queued by other
+	 * CPUs might arrive late, either due to hardware latencies or
+	 * because this CPU disabled interrupts (inside stop-machine)
+	 * before the IPIs were sent. So flush out any pending callbacks
+	 * explicitly (without waiting for the IPIs to arrive), to
+	 * ensure that the outgoing CPU doesn't go offline with work
+	 * still pending.
+	 */
+	flush_smp_call_function_queue(false);
+	return 0;
+}
 
 void __init call_function_init(void)
 {
-	void *cpu = (void *)(long)smp_processor_id();
 	int i;
 
 	for_each_possible_cpu(i)
 		init_llist_head(&per_cpu(call_single_queue, i));
 
-	hotplug_cfd(&hotplug_cfd_notifier, CPU_UP_PREPARE, cpu);
-	register_cpu_notifier(&hotplug_cfd_notifier);
+	smpcfd_prepare_cpu(smp_processor_id());
 }
 
 /*