[PATCH] cpu/hotplug: Serialize callback invocations proper

[Sebastian Andrzej Siewior]

The setup/remove_state/instance() functions in the hotplug core code are
serialized against concurrent CPU hotplug, but unfortunately not serialized
against themself.

As a consequence a concurrent invocation of these function results in
corruption of the callback machinery because two instances try to invoke
callbacks on remote cpus at the same time. This results in missing callback
invocations and initiator threads waiting forever on the completion.

The obvious solution to replace get_cpu_online() with cpu_hotplug_begin()
is not possible because at least one callsite calls into these functions
from a get_online_cpu() locked region.

Extend the protection scope of the cpuhp_state_mutex from solely protecting
the state arrays to cover the callback invocation machinery as well.

Reported-by: Bart Van Assche <Bart.VanAssche@xxxxxxxxxxx>
Fixes: 5b7aa87e0482 ("cpu/hotplug: Implement setup/removal interface")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
---
 kernel/cpu.c |   28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)

--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1335,26 +1335,21 @@ static int cpuhp_store_callbacks(enum cp
 	struct cpuhp_step *sp;
 	int ret = 0;
 
-	mutex_lock(&cpuhp_state_mutex);
-
 	if (state == CPUHP_AP_ONLINE_DYN || state == CPUHP_BP_PREPARE_DYN) {
 		ret = cpuhp_reserve_state(state);
 		if (ret < 0)
-			goto out;
+			return ret;
 		state = ret;
 	}
 	sp = cpuhp_get_step(state);
-	if (name && sp->name) {
-		ret = -EBUSY;
-		goto out;
-	}
+	if (name && sp->name)
+		return -EBUSY;
+
 	sp->startup.single = startup;
 	sp->teardown.single = teardown;
 	sp->name = name;
 	sp->multi_instance = multi_instance;
 	INIT_HLIST_HEAD(&sp->list);
-out:
-	mutex_unlock(&cpuhp_state_mutex);
 	return ret;
 }
 
@@ -1428,6 +1423,7 @@ int __cpuhp_state_add_instance(enum cpuh
 		return -EINVAL;
 
 	get_online_cpus();
+	mutex_lock(&cpuhp_state_mutex);
 
 	if (!invoke || !sp->startup.multi)
 		goto add_node;
@@ -1447,16 +1443,14 @@ int __cpuhp_state_add_instance(enum cpuh
 		if (ret) {
 			if (sp->teardown.multi)
 				cpuhp_rollback_install(cpu, state, node);
-			goto err;
+			goto unlock;
 		}
 	}
 add_node:
 	ret = 0;
-	mutex_lock(&cpuhp_state_mutex);
 	hlist_add_head(node, &sp->list);
+unlock:
 	mutex_unlock(&cpuhp_state_mutex);
-
-err:
 	put_online_cpus();
 	return ret;
 }
@@ -1491,6 +1485,7 @@ int __cpuhp_setup_state(enum cpuhp_state
 		return -EINVAL;
 
 	get_online_cpus();
+	mutex_lock(&cpuhp_state_mutex);
 
 	ret = cpuhp_store_callbacks(state, name, startup, teardown,
 				    multi_instance);
@@ -1524,6 +1519,7 @@ int __cpuhp_setup_state(enum cpuhp_state
 		}
 	}
 out:
+	mutex_unlock(&cpuhp_state_mutex);
 	put_online_cpus();
 	/*
 	 * If the requested state is CPUHP_AP_ONLINE_DYN, return the
@@ -1547,6 +1543,8 @@ int __cpuhp_state_remove_instance(enum c
 		return -EINVAL;
 
 	get_online_cpus();
+	mutex_lock(&cpuhp_state_mutex);
+
 	if (!invoke || !cpuhp_get_teardown_cb(state))
 		goto remove;
 	/*
@@ -1563,7 +1561,6 @@ int __cpuhp_state_remove_instance(enum c
 	}
 
 remove:
-	mutex_lock(&cpuhp_state_mutex);
 	hlist_del(node);
 	mutex_unlock(&cpuhp_state_mutex);
 	put_online_cpus();
@@ -1571,6 +1568,7 @@ int __cpuhp_state_remove_instance(enum c
 	return 0;
 }
 EXPORT_SYMBOL_GPL(__cpuhp_state_remove_instance);
+
 /**
  * __cpuhp_remove_state - Remove the callbacks for an hotplug machine state
  * @state:	The state to remove
@@ -1589,6 +1587,7 @@ void __cpuhp_remove_state(enum cpuhp_sta
 
 	get_online_cpus();
 
+	mutex_lock(&cpuhp_state_mutex);
 	if (sp->multi_instance) {
 		WARN(!hlist_empty(&sp->list),
 		     "Error: Removing state %d which has instances left.\n",
@@ -1613,6 +1612,7 @@ void __cpuhp_remove_state(enum cpuhp_sta
 	}
 remove:
 	cpuhp_store_callbacks(state, NULL, NULL, NULL, false);
+	mutex_unlock(&cpuhp_state_mutex);
 	put_online_cpus();
 }
 EXPORT_SYMBOL(__cpuhp_remove_state);