cpu/hotplug: Add CPU state tracking and synchronization

The CPU state tracking and synchronization mechanism in smpboot.c is
completely independent of the hotplug code and all logic around it is
implemented in architecture specific code.

Except for the state reporting of the AP there is absolutely nothing
architecture specific and the sychronization and decision functions can be
moved into the generic hotplug core code.

Provide an integrated variant and add the core synchronization and decision
points. This comes in two flavours:

  1) DEAD state synchronization

     Updated by the architecture code once the AP reaches the point where
     it is ready to be torn down by the control CPU, e.g. by removing power
     or clocks or tear down via the hypervisor.

     The control CPU waits for this state to be reached with a timeout. If
     the state is reached an architecture specific cleanup function is
     invoked.

  2) Full state synchronization

     This extends #1 with AP alive synchronization. This is new
     functionality, which allows to replace architecture specific wait
     mechanims, e.g. cpumasks, completely.

     It also prevents that an AP which is in a limbo state can be brought
     up again. This can happen when an AP failed to report dead state
     during a previous off-line operation.

The dead synchronization is what most architectures use. Only x86 makes a
bringup decision based on that state at the moment.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Michael Kelley <mikelley@microsoft.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Helge Deller <deller@gmx.de> # parisc
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com> # Steam Deck
Link: https://lore.kernel.org/r/20230512205256.476305035@linutronix.de

1 files changed, 192 insertions, 1 deletions

diff --git a/kernel/cpu.c b/kernel/cpu.c
index df8f137f0271..64b624291316 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -17,6 +17,7 @@
 #include <linux/cpu.h>
 #include <linux/oom.h>
 #include <linux/rcupdate.h>
+#include <linux/delay.h>
 #include <linux/export.h>
 #include <linux/bug.h>
 #include <linux/kthread.h>
@@ -59,6 +60,7 @@
  * @last:	For multi-instance rollback, remember how far we got
  * @cb_state:	The state for a single callback (install/uninstall)
  * @result:	Result of the operation
+ * @ap_sync_state:	State for AP synchronization
  * @done_up:	Signal completion to the issuer of the task for cpu-up
  * @done_down:	Signal completion to the issuer of the task for cpu-down
  */
@@ -76,6 +78,7 @@ struct cpuhp_cpu_state {
 	struct hlist_node	*last;
 	enum cpuhp_state	cb_state;
 	int			result;
+	atomic_t		ap_sync_state;
 	struct completion	done_up;
 	struct completion	done_down;
 #endif
@@ -276,6 +279,182 @@ static bool cpuhp_is_atomic_state(enum cpuhp_state state)
 	return CPUHP_AP_IDLE_DEAD <= state && state < CPUHP_AP_ONLINE;
 }
 
+/* Synchronization state management */
+enum cpuhp_sync_state {
+	SYNC_STATE_DEAD,
+	SYNC_STATE_KICKED,
+	SYNC_STATE_SHOULD_DIE,
+	SYNC_STATE_ALIVE,
+	SYNC_STATE_SHOULD_ONLINE,
+	SYNC_STATE_ONLINE,
+};
+
+#ifdef CONFIG_HOTPLUG_CORE_SYNC
+/**
+ * cpuhp_ap_update_sync_state - Update synchronization state during bringup/teardown
+ * @state:	The synchronization state to set
+ *
+ * No synchronization point. Just update of the synchronization state, but implies
+ * a full barrier so that the AP changes are visible before the control CPU proceeds.
+ */
+static inline void cpuhp_ap_update_sync_state(enum cpuhp_sync_state state)
+{
+	atomic_t *st = this_cpu_ptr(&cpuhp_state.ap_sync_state);
+
+	(void)atomic_xchg(st, state);
+}
+
+void __weak arch_cpuhp_sync_state_poll(void) { cpu_relax(); }
+
+static bool cpuhp_wait_for_sync_state(unsigned int cpu, enum cpuhp_sync_state state,
+				      enum cpuhp_sync_state next_state)
+{
+	atomic_t *st = per_cpu_ptr(&cpuhp_state.ap_sync_state, cpu);
+	ktime_t now, end, start = ktime_get();
+	int sync;
+
+	end = start + 10ULL * NSEC_PER_SEC;
+
+	sync = atomic_read(st);
+	while (1) {
+		if (sync == state) {
+			if (!atomic_try_cmpxchg(st, &sync, next_state))
+				continue;
+			return true;
+		}
+
+		now = ktime_get();
+		if (now > end) {
+			/* Timeout. Leave the state unchanged */
+			return false;
+		} else if (now - start < NSEC_PER_MSEC) {
+			/* Poll for one millisecond */
+			arch_cpuhp_sync_state_poll();
+		} else {
+			usleep_range_state(USEC_PER_MSEC, 2 * USEC_PER_MSEC, TASK_UNINTERRUPTIBLE);
+		}
+		sync = atomic_read(st);
+	}
+	return true;
+}
+#else  /* CONFIG_HOTPLUG_CORE_SYNC */
+static inline void cpuhp_ap_update_sync_state(enum cpuhp_sync_state state) { }
+#endif /* !CONFIG_HOTPLUG_CORE_SYNC */
+
+#ifdef CONFIG_HOTPLUG_CORE_SYNC_DEAD
+/**
+ * cpuhp_ap_report_dead - Update synchronization state to DEAD
+ *
+ * No synchronization point. Just update of the synchronization state.
+ */
+void cpuhp_ap_report_dead(void)
+{
+	cpuhp_ap_update_sync_state(SYNC_STATE_DEAD);
+}
+
+void __weak arch_cpuhp_cleanup_dead_cpu(unsigned int cpu) { }
+
+/*
+ * Late CPU shutdown synchronization point. Cannot use cpuhp_state::done_down
+ * because the AP cannot issue complete() at this stage.
+ */
+static void cpuhp_bp_sync_dead(unsigned int cpu)
+{
+	atomic_t *st = per_cpu_ptr(&cpuhp_state.ap_sync_state, cpu);
+	int sync = atomic_read(st);
+
+	do {
+		/* CPU can have reported dead already. Don't overwrite that! */
+		if (sync == SYNC_STATE_DEAD)
+			break;
+	} while (!atomic_try_cmpxchg(st, &sync, SYNC_STATE_SHOULD_DIE));
+
+	if (cpuhp_wait_for_sync_state(cpu, SYNC_STATE_DEAD, SYNC_STATE_DEAD)) {
+		/* CPU reached dead state. Invoke the cleanup function */
+		arch_cpuhp_cleanup_dead_cpu(cpu);
+		return;
+	}
+
+	/* No further action possible. Emit message and give up. */
+	pr_err("CPU%u failed to report dead state\n", cpu);
+}
+#else /* CONFIG_HOTPLUG_CORE_SYNC_DEAD */
+static inline void cpuhp_bp_sync_dead(unsigned int cpu) { }
+#endif /* !CONFIG_HOTPLUG_CORE_SYNC_DEAD */
+
+#ifdef CONFIG_HOTPLUG_CORE_SYNC_FULL
+/**
+ * cpuhp_ap_sync_alive - Synchronize AP with the control CPU once it is alive
+ *
+ * Updates the AP synchronization state to SYNC_STATE_ALIVE and waits
+ * for the BP to release it.
+ */
+void cpuhp_ap_sync_alive(void)
+{
+	atomic_t *st = this_cpu_ptr(&cpuhp_state.ap_sync_state);
+
+	cpuhp_ap_update_sync_state(SYNC_STATE_ALIVE);
+
+	/* Wait for the control CPU to release it. */
+	while (atomic_read(st) != SYNC_STATE_SHOULD_ONLINE)
+		cpu_relax();
+}
+
+static bool cpuhp_can_boot_ap(unsigned int cpu)
+{
+	atomic_t *st = per_cpu_ptr(&cpuhp_state.ap_sync_state, cpu);
+	int sync = atomic_read(st);
+
+again:
+	switch (sync) {
+	case SYNC_STATE_DEAD:
+		/* CPU is properly dead */
+		break;
+	case SYNC_STATE_KICKED:
+		/* CPU did not come up in previous attempt */
+		break;
+	case SYNC_STATE_ALIVE:
+		/* CPU is stuck cpuhp_ap_sync_alive(). */
+		break;
+	default:
+		/* CPU failed to report online or dead and is in limbo state. */
+		return false;
+	}
+
+	/* Prepare for booting */
+	if (!atomic_try_cmpxchg(st, &sync, SYNC_STATE_KICKED))
+		goto again;
+
+	return true;
+}
+
+void __weak arch_cpuhp_cleanup_kick_cpu(unsigned int cpu) { }
+
+/*
+ * Early CPU bringup synchronization point. Cannot use cpuhp_state::done_up
+ * because the AP cannot issue complete() so early in the bringup.
+ */
+static int cpuhp_bp_sync_alive(unsigned int cpu)
+{
+	int ret = 0;
+
+	if (!IS_ENABLED(CONFIG_HOTPLUG_CORE_SYNC_FULL))
+		return 0;
+
+	if (!cpuhp_wait_for_sync_state(cpu, SYNC_STATE_ALIVE, SYNC_STATE_SHOULD_ONLINE)) {
+		pr_err("CPU%u failed to report alive state\n", cpu);
+		ret = -EIO;
+	}
+
+	/* Let the architecture cleanup the kick alive mechanics. */
+	arch_cpuhp_cleanup_kick_cpu(cpu);
+	return ret;
+}
+#else /* CONFIG_HOTPLUG_CORE_SYNC_FULL */
+static inline int cpuhp_bp_sync_alive(unsigned int cpu) { return 0; }
+static inline bool cpuhp_can_boot_ap(unsigned int cpu) { return true; }
+#endif /* !CONFIG_HOTPLUG_CORE_SYNC_FULL */
+
 /* Serializes the updates to cpu_online_mask, cpu_present_mask */
 static DEFINE_MUTEX(cpu_add_remove_lock);
 bool cpuhp_tasks_frozen;
@@ -588,6 +767,9 @@ static int bringup_cpu(unsigned int cpu)
 	struct task_struct *idle = idle_thread_get(cpu);
 	int ret;
 
+	if (!cpuhp_can_boot_ap(cpu))
+		return -EAGAIN;
+
 	/*
 	 * Reset stale stack state from the last time this CPU was online.
 	 */
@@ -610,6 +792,10 @@ static int bringup_cpu(unsigned int cpu)
 	if (ret)
 		goto out_unlock;
 
+	ret = cpuhp_bp_sync_alive(cpu);
+	if (ret)
+		goto out_unlock;
+
 	ret = bringup_wait_for_ap_online(cpu);
 	if (ret)
 		goto out_unlock;
@@ -1113,6 +1299,8 @@ static int takedown_cpu(unsigned int cpu)
 	/* This actually kills the CPU. */
 	__cpu_die(cpu);
 
+	cpuhp_bp_sync_dead(cpu);
+
 	tick_cleanup_dead_cpu(cpu);
 	rcutree_migrate_callbacks(cpu);
 	return 0;
@@ -1359,8 +1547,10 @@ void cpuhp_online_idle(enum cpuhp_state state)
 	if (state != CPUHP_AP_ONLINE_IDLE)
 		return;
 
+	cpuhp_ap_update_sync_state(SYNC_STATE_ONLINE);
+
 	/*
-	 * Unpart the stopper thread before we start the idle loop (and start
+	 * Unpark the stopper thread before we start the idle loop (and start
 	 * scheduling); this ensures the stopper task is always available.
 	 */
 	stop_machine_unpark(smp_processor_id());
@@ -2737,6 +2927,7 @@ void __init boot_cpu_hotplug_init(void)
 {
 #ifdef CONFIG_SMP
 	cpumask_set_cpu(smp_processor_id(), &cpus_booted_once_mask);
+	atomic_set(this_cpu_ptr(&cpuhp_state.ap_sync_state), SYNC_STATE_ONLINE);
 #endif
 	this_cpu_write(cpuhp_state.state, CPUHP_ONLINE);
 	this_cpu_write(cpuhp_state.target, CPUHP_ONLINE);