slab: Convert to hotplug state machine

Install the callbacks via the state machine.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: linux-mm@kvack.org
Cc: rt@linutronix.de
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Christoph Lameter <cl@linux.com>
Link: http://lkml.kernel.org/r/20160823125319.abeapfjapf2kfezp@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

include/linux/cpuhotplug.h

@@ -22,6 +22,7 @@ enum cpuhp_state {
 	CPUHP_X2APIC_PREPARE,
 	CPUHP_SMPCFD_PREPARE,
 	CPUHP_RELAY_PREPARE,
+	CPUHP_SLAB_PREPARE,
 	CPUHP_RCUTREE_PREP,
 	CPUHP_NOTIFY_PREPARE,
 	CPUHP_TIMERS_DEAD,

include/linux/slab.h

@@ -650,4 +650,12 @@ static inline void *kzalloc_node(size_t size, gfp_t flags, int node)
 unsigned int kmem_cache_size(struct kmem_cache *s);
 void __init kmem_cache_init_late(void);
 
+#if defined(CONFIG_SMP) && defined(CONFIG_SLAB)
+int slab_prepare_cpu(unsigned int cpu);
+int slab_dead_cpu(unsigned int cpu);
+#else
+#define slab_prepare_cpu	NULL
+#define slab_dead_cpu		NULL
+#endif
+
 #endif	/* _LINUX_SLAB_H */

kernel/cpu.c

@@ -24,6 +24,7 @@
 #include <linux/irq.h>
 #include <linux/smpboot.h>
 #include <linux/relay.h>
+#include <linux/slab.h>
 
 #include <trace/events/power.h>
 #define CREATE_TRACE_POINTS
@@ -1278,6 +1279,11 @@ static struct cpuhp_step cpuhp_bp_states[] = {
 		.startup.single		= relay_prepare_cpu,
 		.teardown.single	= NULL,
 	},
+	[CPUHP_SLAB_PREPARE] = {
+		.name			= "slab:prepare",
+		.startup.single		= slab_prepare_cpu,
+		.teardown.single	= slab_dead_cpu,
+	},
 	[CPUHP_RCUTREE_PREP] = {
 		.name			= "RCU/tree:prepare",
 		.startup.single		= rcutree_prepare_cpu,

mm/slab.c

@@ -886,6 +886,7 @@ static int init_cache_node(struct kmem_cache *cachep, int node, gfp_t gfp)
 	return 0;
 }
 
+#if (defined(CONFIG_NUMA) && defined(CONFIG_MEMORY_HOTPLUG)) || defined(CONFIG_SMP)
 /*
  * Allocates and initializes node for a node on each slab cache, used for
  * either memory or cpu hotplug.  If memory is being hot-added, the kmem_cache_node
@@ -908,6 +909,7 @@ static int init_cache_node_node(int node)
 
 	return 0;
 }
+#endif
 
 static int setup_kmem_cache_node(struct kmem_cache *cachep,
 				int node, gfp_t gfp, bool force_change)
@@ -975,6 +977,8 @@ fail:
 	return ret;
 }
 
+#ifdef CONFIG_SMP
+
 static void cpuup_canceled(long cpu)
 {
 	struct kmem_cache *cachep;
@@ -1075,65 +1079,54 @@ bad:
 	return -ENOMEM;
 }
 
-static int cpuup_callback(struct notifier_block *nfb,
-				    unsigned long action, void *hcpu)
+int slab_prepare_cpu(unsigned int cpu)
 {
-	long cpu = (long)hcpu;
-	int err = 0;
+	int err;
 
-	switch (action) {
-	case CPU_UP_PREPARE:
-	case CPU_UP_PREPARE_FROZEN:
-		mutex_lock(&slab_mutex);
-		err = cpuup_prepare(cpu);
-		mutex_unlock(&slab_mutex);
-		break;
-	case CPU_ONLINE:
-	case CPU_ONLINE_FROZEN:
-		start_cpu_timer(cpu);
-		break;
-#ifdef CONFIG_HOTPLUG_CPU
-  	case CPU_DOWN_PREPARE:
-  	case CPU_DOWN_PREPARE_FROZEN:
-		/*
-		 * Shutdown cache reaper. Note that the slab_mutex is
-		 * held so that if cache_reap() is invoked it cannot do
-		 * anything expensive but will only modify reap_work
-		 * and reschedule the timer.
-		*/
-		cancel_delayed_work_sync(&per_cpu(slab_reap_work, cpu));
-		/* Now the cache_reaper is guaranteed to be not running. */
-		per_cpu(slab_reap_work, cpu).work.func = NULL;
-  		break;
-  	case CPU_DOWN_FAILED:
-  	case CPU_DOWN_FAILED_FROZEN:
-		start_cpu_timer(cpu);
-  		break;
-	case CPU_DEAD:
-	case CPU_DEAD_FROZEN:
-		/*
-		 * Even if all the cpus of a node are down, we don't free the
-		 * kmem_cache_node of any cache. This to avoid a race between
-		 * cpu_down, and a kmalloc allocation from another cpu for
-		 * memory from the node of the cpu going down.  The node
-		 * structure is usually allocated from kmem_cache_create() and
-		 * gets destroyed at kmem_cache_destroy().
-		 */
-		/* fall through */
-#endif
-	case CPU_UP_CANCELED:
-	case CPU_UP_CANCELED_FROZEN:
-		mutex_lock(&slab_mutex);
-		cpuup_canceled(cpu);
-		mutex_unlock(&slab_mutex);
-		break;
-	}
-	return notifier_from_errno(err);
+	mutex_lock(&slab_mutex);
+	err = cpuup_prepare(cpu);
+	mutex_unlock(&slab_mutex);
+	return err;
 }
 
-static struct notifier_block cpucache_notifier = {
-	&cpuup_callback, NULL, 0
-};
+/*
+ * This is called for a failed online attempt and for a successful
+ * offline.
+ *
+ * Even if all the cpus of a node are down, we don't free the
+ * kmem_list3 of any cache. This to avoid a race between cpu_down, and
+ * a kmalloc allocation from another cpu for memory from the node of
+ * the cpu going down.  The list3 structure is usually allocated from
+ * kmem_cache_create() and gets destroyed at kmem_cache_destroy().
+ */
+int slab_dead_cpu(unsigned int cpu)
+{
+	mutex_lock(&slab_mutex);
+	cpuup_canceled(cpu);
+	mutex_unlock(&slab_mutex);
+	return 0;
+}
+#endif
+
+static int slab_online_cpu(unsigned int cpu)
+{
+	start_cpu_timer(cpu);
+	return 0;
+}
+
+static int slab_offline_cpu(unsigned int cpu)
+{
+	/*
+	 * Shutdown cache reaper. Note that the slab_mutex is held so
+	 * that if cache_reap() is invoked it cannot do anything
+	 * expensive but will only modify reap_work and reschedule the
+	 * timer.
+	 */
+	cancel_delayed_work_sync(&per_cpu(slab_reap_work, cpu));
+	/* Now the cache_reaper is guaranteed to be not running. */
+	per_cpu(slab_reap_work, cpu).work.func = NULL;
+	return 0;
+}
 
 #if defined(CONFIG_NUMA) && defined(CONFIG_MEMORY_HOTPLUG)
 /*
@@ -1336,12 +1329,6 @@ void __init kmem_cache_init_late(void)
 	/* Done! */
 	slab_state = FULL;
 
-	/*
-	 * Register a cpu startup notifier callback that initializes
-	 * cpu_cache_get for all new cpus
-	 */
-	register_cpu_notifier(&cpucache_notifier);
-
 #ifdef CONFIG_NUMA
 	/*
 	 * Register a memory hotplug callback that initializes and frees
@@ -1358,13 +1345,14 @@ void __init kmem_cache_init_late(void)
 
 static int __init cpucache_init(void)
 {
-	int cpu;
+	int ret;
 
 	/*
 	 * Register the timers that return unneeded pages to the page allocator
 	 */
-	for_each_online_cpu(cpu)
-		start_cpu_timer(cpu);
+	ret = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "SLAB online",
+				slab_online_cpu, slab_offline_cpu);
+	WARN_ON(ret < 0);
 
 	/* Done! */
 	slab_state = FULL;