[PATCH 03/16] slab: Convert to hotplug state machine

[PATCH 03/16] slab: Convert to hotplug state machine

[Sebastian Andrzej Siewior]

From: Richard Weinberger <richard@nod.at>

Install the callbacks via the state machine.

Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
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>
---
 include/linux/cpuhotplug.h |   1 +
 include/linux/slab.h       |   8 ++++
 kernel/cpu.c               |   7 ++-
 mm/slab.c                  | 117 ++++++++++++++++++++-------------------------
 4 files changed, 68 insertions(+), 65 deletions(-)

diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index 4c79f40fcebc..c2cf14953abc 100644
--- a/include/linux/cpuhotplug.h
+++ b/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,
diff --git a/include/linux/slab.h b/include/linux/slab.h
index 4293808d8cfb..084b12bad198 100644
--- a/include/linux/slab.h
+++ b/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 */
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 2a11381d5997..82bf61c67316 100644
--- a/kernel/cpu.c
+++ b/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
@@ -1273,6 +1274,11 @@ static struct cpuhp_step cpuhp_bp_states[] = {
 		.startup = relay_prepare_cpu,
 		.teardown = NULL,
 	},
+	[CPUHP_SLAB_PREPARE] = {
+		.name = "SLAB prepare",
+		.startup = slab_prepare_cpu,
+		.teardown = slab_dead_cpu,
+	},
 	[CPUHP_RCUTREE_PREP] = {
 		.name = "RCU-tree prepare",
 		.startup = rcutree_prepare_cpu,
@@ -1376,7 +1382,6 @@ static struct cpuhp_step cpuhp_ap_states[] = {
 		.startup = rcutree_online_cpu,
 		.teardown = rcutree_offline_cpu,
 	},
-
 	/*
 	 * Online/down_prepare notifiers. Will be removed once the notifiers
 	 * are converted to states.
diff --git a/mm/slab.c b/mm/slab.c
index 0eb6691ae6fc..e8d465069b87 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -887,6 +887,7 @@ 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
@@ -909,6 +910,7 @@ 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)
@@ -976,6 +978,8 @@ static int setup_kmem_cache_node(struct kmem_cache *cachep,
 	return ret;
 }
 
+#ifdef CONFIG_SMP
+
 static void cpuup_canceled(long cpu)
 {
 	struct kmem_cache *cachep;
@@ -1076,65 +1080,55 @@ static int cpuup_prepare(long cpu)
 	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)
+{
+	pr_err("%s(%d) %d\n", __func__, __LINE__, 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)
 /*
@@ -1337,12 +1331,6 @@ 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
@@ -1359,13 +1347,14 @@ 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;
-- 
2.9.3

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 04/16] slub: Convert to hotplug state machine

[Sebastian Andrzej Siewior]

Install the callbacks via the state machine.

Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 include/linux/cpuhotplug.h |  1 +
 mm/slub.c                  | 65 +++++++++++++++-------------------------------
 2 files changed, 22 insertions(+), 44 deletions(-)

diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index c2cf14953abc..82ee32107dff 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -15,6 +15,7 @@ enum cpuhp_state {
 	CPUHP_X86_HPET_DEAD,
 	CPUHP_X86_APB_DEAD,
 	CPUHP_VIRT_NET_DEAD,
+	CPUHP_SLUB_DEAD,
 	CPUHP_WORKQUEUE_PREP,
 	CPUHP_POWER_NUMA_PREPARE,
 	CPUHP_HRTIMERS_PREPARE,
diff --git a/mm/slub.c b/mm/slub.c
index 30e5a0f0e191..1db4d5304ceb 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -194,10 +194,6 @@ static inline bool kmem_cache_has_cpu_partial(struct kmem_cache *s)
 #define __OBJECT_POISON		0x80000000UL /* Poison object */
 #define __CMPXCHG_DOUBLE	0x40000000UL /* Use cmpxchg_double */
 
-#ifdef CONFIG_SMP
-static struct notifier_block slab_notifier;
-#endif
-
 /*
  * Tracking user of a slab.
  */
@@ -2288,6 +2284,25 @@ static void flush_all(struct kmem_cache *s)
 }
 
 /*
+ * Use the cpu notifier to insure that the cpu slabs are flushed when
+ * necessary.
+ */
+static int slub_cpu_dead(unsigned int cpu)
+{
+	struct kmem_cache *s;
+	unsigned long flags;
+
+	mutex_lock(&slab_mutex);
+	list_for_each_entry(s, &slab_caches, list) {
+		local_irq_save(flags);
+		__flush_cpu_slab(s, cpu);
+		local_irq_restore(flags);
+	}
+	mutex_unlock(&slab_mutex);
+	return 0;
+}
+
+/*
  * Check if the objects in a per cpu structure fit numa
  * locality expectations.
  */
@@ -4127,9 +4142,8 @@ void __init kmem_cache_init(void)
 	/* Setup random freelists for each cache */
 	init_freelist_randomization();
 
-#ifdef CONFIG_SMP
-	register_cpu_notifier(&slab_notifier);
-#endif
+	cpuhp_setup_state_nocalls(CPUHP_SLUB_DEAD, "SLUB_DEAD", NULL,
+				  slub_cpu_dead);
 
 	pr_info("SLUB: HWalign=%d, Order=%d-%d, MinObjects=%d, CPUs=%d, Nodes=%d\n",
 		cache_line_size(),
@@ -4193,43 +4207,6 @@ int __kmem_cache_create(struct kmem_cache *s, unsigned long flags)
 	return err;
 }
 
-#ifdef CONFIG_SMP
-/*
- * Use the cpu notifier to insure that the cpu slabs are flushed when
- * necessary.
- */
-static int slab_cpuup_callback(struct notifier_block *nfb,
-		unsigned long action, void *hcpu)
-{
-	long cpu = (long)hcpu;
-	struct kmem_cache *s;
-	unsigned long flags;
-
-	switch (action) {
-	case CPU_UP_CANCELED:
-	case CPU_UP_CANCELED_FROZEN:
-	case CPU_DEAD:
-	case CPU_DEAD_FROZEN:
-		mutex_lock(&slab_mutex);
-		list_for_each_entry(s, &slab_caches, list) {
-			local_irq_save(flags);
-			__flush_cpu_slab(s, cpu);
-			local_irq_restore(flags);
-		}
-		mutex_unlock(&slab_mutex);
-		break;
-	default:
-		break;
-	}
-	return NOTIFY_OK;
-}
-
-static struct notifier_block slab_notifier = {
-	.notifier_call = slab_cpuup_callback
-};
-
-#endif
-
 void *__kmalloc_track_caller(size_t size, gfp_t gfpflags, unsigned long caller)
 {
 	struct kmem_cache *s;
-- 
2.9.3

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 05/16] mm: writeback: Convert to hotplug state machine

[Sebastian Andrzej Siewior]

Install the callbacks via the state machine and let the core invoke
the callbacks on the already online CPUs.

Cc: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@fb.com>
Cc: linux-mm@kvack.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 include/linux/cpuhotplug.h |  1 +
 mm/page-writeback.c        | 26 +++++++-------------------
 2 files changed, 8 insertions(+), 19 deletions(-)

diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index 82ee32107dff..854e59a426d4 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -16,6 +16,7 @@ enum cpuhp_state {
 	CPUHP_X86_APB_DEAD,
 	CPUHP_VIRT_NET_DEAD,
 	CPUHP_SLUB_DEAD,
+	CPUHP_MM_WRITEBACK_DEAD,
 	CPUHP_WORKQUEUE_PREP,
 	CPUHP_POWER_NUMA_PREPARE,
 	CPUHP_HRTIMERS_PREPARE,
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 82e72524db55..6d3d0e87f309 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2050,26 +2050,12 @@ void writeback_set_ratelimit(void)
 		ratelimit_pages = 16;
 }
 
-static int
-ratelimit_handler(struct notifier_block *self, unsigned long action,
-		  void *hcpu)
+static int page_writeback_cpu_online(unsigned int cpu)
 {
-
-	switch (action & ~CPU_TASKS_FROZEN) {
-	case CPU_ONLINE:
-	case CPU_DEAD:
-		writeback_set_ratelimit();
-		return NOTIFY_OK;
-	default:
-		return NOTIFY_DONE;
-	}
+	writeback_set_ratelimit();
+	return 0;
 }
 
-static struct notifier_block ratelimit_nb = {
-	.notifier_call	= ratelimit_handler,
-	.next		= NULL,
-};
-
 /*
  * Called early on to tune the page writeback dirty limits.
  *
@@ -2092,8 +2078,10 @@ void __init page_writeback_init(void)
 {
 	BUG_ON(wb_domain_init(&global_wb_domain, GFP_KERNEL));
 
-	writeback_set_ratelimit();
-	register_cpu_notifier(&ratelimit_nb);
+	cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "MM_WRITEBACK_ONLINE",
+			  page_writeback_cpu_online, NULL);
+	cpuhp_setup_state(CPUHP_MM_WRITEBACK_DEAD, "MM_WRITEBACK_DEAD", NULL,
+			  page_writeback_cpu_online);
 }
 
 /**
-- 
2.9.3

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 03/16] slab: Convert to hotplug state machine

[Sebastian Andrzej Siewior]

On 2016-08-18 14:57:18 [+0200], To linux-kernel@vger.kernel.org wrote:
> diff --git a/mm/slab.c b/mm/slab.c
> index 0eb6691ae6fc..e8d465069b87 100644
> --- a/mm/slab.c
> +++ b/mm/slab.c
…
> +static int slab_online_cpu(unsigned int cpu)
> +{
> +	pr_err("%s(%d) %d\n", __func__, __LINE__, cpu);

as the careful reader will notice, it has been double checked whether or
not this callback is invoked. This output needs to go.

> +	start_cpu_timer(cpu);
> +	return 0;
> +}

Sebastian

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 03/16 v2] slab: Convert to hotplug state machine

[Sebastian Andrzej Siewior]

Install the callbacks via the state machine.

Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
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>
---
v1…v2: remove debug pr_err()

 include/linux/cpuhotplug.h |   1 +
 include/linux/slab.h       |   8 ++++
 kernel/cpu.c               |   7 ++-
 mm/slab.c                  | 116 ++++++++++++++++++++-------------------------
 4 files changed, 67 insertions(+), 65 deletions(-)

diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index 4c79f40fcebc..c2cf14953abc 100644
--- a/include/linux/cpuhotplug.h
+++ b/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,
diff --git a/include/linux/slab.h b/include/linux/slab.h
index 4293808d8cfb..084b12bad198 100644
--- a/include/linux/slab.h
+++ b/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 */
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 2a11381d5997..82bf61c67316 100644
--- a/kernel/cpu.c
+++ b/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
@@ -1273,6 +1274,11 @@ static struct cpuhp_step cpuhp_bp_states[] = {
 		.startup = relay_prepare_cpu,
 		.teardown = NULL,
 	},
+	[CPUHP_SLAB_PREPARE] = {
+		.name = "SLAB prepare",
+		.startup = slab_prepare_cpu,
+		.teardown = slab_dead_cpu,
+	},
 	[CPUHP_RCUTREE_PREP] = {
 		.name = "RCU-tree prepare",
 		.startup = rcutree_prepare_cpu,
@@ -1376,7 +1382,6 @@ static struct cpuhp_step cpuhp_ap_states[] = {
 		.startup = rcutree_online_cpu,
 		.teardown = rcutree_offline_cpu,
 	},
-
 	/*
 	 * Online/down_prepare notifiers. Will be removed once the notifiers
 	 * are converted to states.
diff --git a/mm/slab.c b/mm/slab.c
index 0eb6691ae6fc..1cecd9fe23e3 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -887,6 +887,7 @@ 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
@@ -909,6 +910,7 @@ 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)
@@ -976,6 +978,8 @@ static int setup_kmem_cache_node(struct kmem_cache *cachep,
 	return ret;
 }
 
+#ifdef CONFIG_SMP
+
 static void cpuup_canceled(long cpu)
 {
 	struct kmem_cache *cachep;
@@ -1076,65 +1080,54 @@ static int cpuup_prepare(long cpu)
 	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)
 /*
@@ -1337,12 +1330,6 @@ 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
@@ -1359,13 +1346,14 @@ 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;
-- 
2.9.3

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>