cpuset: update top cpuset's mems after adding a node

PATCH In review Miao Xie 3 participants 6 messages Nov 10, 2008 – Nov 11, 2008

Subsystems (3)

Memory management Memory hot(un)plug Control group - cpuset

lore

Share & tools

lore

Miao Xie

author

2008-11-10 06:46 UTC #0 lore raw

PATCH 1/1 cpuset: update top cpuset's mems after adding a node

After adding a node into the machine, top cpuset's mems isn't updated.

By reviewing the code, we found that the update function
cpuset_track_online_nodes()
was invoked after node_states[N_ONLINE] changes. It is wrong because N_ONLINE just means node has pgdat, and if node has/added memory, we use N_HIGH_MEMORY. So, We should invoke the update function after node_states[N_HIGH_MEMORY] changes, just like its commit says.

This patch fixes it. And we use notifier of memory hotplug instead of direct calling of cpuset_track_online_nodes().

Signed-off-by Miao Xie <[email protected]>

Cc Yasunori Goto <[email protected]>

---
 include/linux/cpuset.h |    4 ----
 kernel/cpuset.c        |   19 ++++++++++++++++---
 mm/memory_hotplug.c    |    3 ---
 3 files changed, 16 insertions(+), 10 deletions(-)

include/linux/cpuset.h +0 -4

index 2691926..8e540d3 100644
@@ -74,8 +74,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return current->flags & PF_SPREAD_SLAB;
 }
 
-extern void cpuset_track_online_nodes(void);
-
 extern int current_cpuset_is_being_rebound(void);
 
 extern void rebuild_sched_domains(void);

@@ -151,8 +149,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return 0;
 }
 
-static inline void cpuset_track_online_nodes(void) {}
-
 static inline int current_cpuset_is_being_rebound(void)
 {
 	return 0;

kernel/cpuset.c +16 -3

index 3e00526..909a548 100644
@@ -36,6 +36,7 @@
 #include <linux/list.h>
 #include <linux/mempolicy.h>
 #include <linux/mm.h>
+#include <linux/memory.h>
 #include <linux/module.h>
 #include <linux/mount.h>
 #include <linux/namei.h>

@@ -2011,12 +2012,23 @@ static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
  * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
  * See also the previous routine cpuset_track_online_cpus().
  */
-void cpuset_track_online_nodes(void)
+static int cpuset_track_online_nodes(struct notifier_block *self,
+				unsigned long action, void *arg)
 {
 	cgroup_lock();
-	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
-	scan_for_empty_cpusets(&top_cpuset);
+	switch (action) {
+	case MEM_ONLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		break;
+	case MEM_OFFLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		scan_for_empty_cpusets(&top_cpuset);
+		break;
+	default:
+		break;
+	}
 	cgroup_unlock();
+	return NOTIFY_OK;
 }
 #endif

@@ -2032,6 +2044,7 @@ void __init cpuset_init_smp(void)
 	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
 
 	hotcpu_notifier(cpuset_track_online_cpus, 0);
+	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
 }
 
 /**

mm/memory_hotplug.c +0 -3

index 6837a10..b5b2b15 100644
@@ -22,7 +22,6 @@
 #include <linux/highmem.h>
 #include <linux/vmalloc.h>
 #include <linux/ioport.h>
-#include <linux/cpuset.h>
 #include <linux/delay.h>
 #include <linux/migrate.h>
 #include <linux/page-isolation.h>

@@ -498,8 +497,6 @@ int add_memory(int nid, u64 start, u64 size)
 	/* we online node here. we can't roll back from here. */
 	node_set_online(nid);
 
-	cpuset_track_online_nodes();
-
 	if (new_pgdat) {
 		ret = register_one_node(nid);
 		/*

Signature · 1 line

1.5.3.8

Yasunori Goto

Acked-by on v1 1/1

2008-11-10 06:56 UTC #1 lore raw

Did you test this patch? Was it OK?

If yes,

Acked-by Yasunori Goto <[email protected]>

Thanks.

Quoted 114 lines

After adding a node into the machine, top cpuset's mems isn't updated.

This patch fixes it. And we use notifier of memory hotplug instead of direct calling of cpuset_track_online_nodes().

Signed-off-by Miao Xie <[email protected]>

Cc Yasunori Goto <[email protected]>

---
 include/linux/cpuset.h |    4 ----
 kernel/cpuset.c        |   19 ++++++++++++++++---
 mm/memory_hotplug.c    |    3 ---
 3 files changed, 16 insertions(+), 10 deletions(-)

include/linux/cpuset.h +0 -4

index 2691926..8e540d3 100644
@@ -74,8 +74,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return current->flags & PF_SPREAD_SLAB;
 }
 
-extern void cpuset_track_online_nodes(void);
-
 extern int current_cpuset_is_being_rebound(void);
 
 extern void rebuild_sched_domains(void);

@@ -151,8 +149,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return 0;
 }
 
-static inline void cpuset_track_online_nodes(void) {}
-
 static inline int current_cpuset_is_being_rebound(void)
 {
 	return 0;

kernel/cpuset.c +16 -3

index 3e00526..909a548 100644
@@ -36,6 +36,7 @@
 #include <linux/list.h>
 #include <linux/mempolicy.h>
 #include <linux/mm.h>
+#include <linux/memory.h>
 #include <linux/module.h>
 #include <linux/mount.h>
 #include <linux/namei.h>

@@ -2011,12 +2012,23 @@ static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
  * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
  * See also the previous routine cpuset_track_online_cpus().
  */
-void cpuset_track_online_nodes(void)
+static int cpuset_track_online_nodes(struct notifier_block *self,
+				unsigned long action, void *arg)
 {
 	cgroup_lock();
-	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
-	scan_for_empty_cpusets(&top_cpuset);
+	switch (action) {
+	case MEM_ONLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		break;
+	case MEM_OFFLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		scan_for_empty_cpusets(&top_cpuset);
+		break;
+	default:
+		break;
+	}
 	cgroup_unlock();
+	return NOTIFY_OK;
 }
 #endif

@@ -2032,6 +2044,7 @@ void __init cpuset_init_smp(void)
 	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
 
 	hotcpu_notifier(cpuset_track_online_cpus, 0);
+	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
 }
 
 /**

mm/memory_hotplug.c +0 -3

index 6837a10..b5b2b15 100644
@@ -22,7 +22,6 @@
 #include <linux/highmem.h>
 #include <linux/vmalloc.h>
 #include <linux/ioport.h>
-#include <linux/cpuset.h>
 #include <linux/delay.h>
 #include <linux/migrate.h>
 #include <linux/page-isolation.h>

@@ -498,8 +497,6 @@ int add_memory(int nid, u64 start, u64 size)
 	/* we online node here. we can't roll back from here. */
 	node_set_online(nid);
 
-	cpuset_track_online_nodes();
-
 	if (new_pgdat) {
 		ret = register_one_node(nid);
 		/*

Signature · 1 line

1.5.3.8

Signature · 1 line

Yasunori Goto

Miao Xie

author

2008-11-10 07:42 UTC #2 lore raw

Quoted 1 line · on 2008-11-10 14:56 Yasunori Goto wrote:

Did you test this patch? Was it OK?

Sorry. I forget.
I will do it as soon as possible.

Quoted 122 lines

If yes,

Acked-by Yasunori Goto <[email protected]>

Thanks.

Quoted 114 lines

After adding a node into the machine, top cpuset's mems isn't updated.

This patch fixes it. And we use notifier of memory hotplug instead of direct calling of cpuset_track_online_nodes().

Signed-off-by Miao Xie <[email protected]>

Cc Yasunori Goto <[email protected]>

---
 include/linux/cpuset.h |    4 ----
 kernel/cpuset.c        |   19 ++++++++++++++++---
 mm/memory_hotplug.c    |    3 ---
 3 files changed, 16 insertions(+), 10 deletions(-)

include/linux/cpuset.h +0 -4

index 2691926..8e540d3 100644
@@ -74,8 +74,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return current->flags & PF_SPREAD_SLAB;
 }
 
-extern void cpuset_track_online_nodes(void);
-
 extern int current_cpuset_is_being_rebound(void);
 
 extern void rebuild_sched_domains(void);

@@ -151,8 +149,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return 0;
 }
 
-static inline void cpuset_track_online_nodes(void) {}
-
 static inline int current_cpuset_is_being_rebound(void)
 {
 	return 0;

kernel/cpuset.c +16 -3

index 3e00526..909a548 100644
@@ -36,6 +36,7 @@
 #include <linux/list.h>
 #include <linux/mempolicy.h>
 #include <linux/mm.h>
+#include <linux/memory.h>
 #include <linux/module.h>
 #include <linux/mount.h>
 #include <linux/namei.h>

@@ -2011,12 +2012,23 @@ static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
  * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
  * See also the previous routine cpuset_track_online_cpus().
  */
-void cpuset_track_online_nodes(void)
+static int cpuset_track_online_nodes(struct notifier_block *self,
+				unsigned long action, void *arg)
 {
 	cgroup_lock();
-	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
-	scan_for_empty_cpusets(&top_cpuset);
+	switch (action) {
+	case MEM_ONLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		break;
+	case MEM_OFFLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		scan_for_empty_cpusets(&top_cpuset);
+		break;
+	default:
+		break;
+	}
 	cgroup_unlock();
+	return NOTIFY_OK;
 }
 #endif

@@ -2032,6 +2044,7 @@ void __init cpuset_init_smp(void)
 	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
 
 	hotcpu_notifier(cpuset_track_online_cpus, 0);
+	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
 }
 
 /**

mm/memory_hotplug.c +0 -3

index 6837a10..b5b2b15 100644
@@ -22,7 +22,6 @@
 #include <linux/highmem.h>
 #include <linux/vmalloc.h>
 #include <linux/ioport.h>
-#include <linux/cpuset.h>
 #include <linux/delay.h>
 #include <linux/migrate.h>
 #include <linux/page-isolation.h>

@@ -498,8 +497,6 @@ int add_memory(int nid, u64 start, u64 size)
 	/* we online node here. we can't roll back from here. */
 	node_set_online(nid);
 
-	cpuset_track_online_nodes();
-
 	if (new_pgdat) {
 		ret = register_one_node(nid);
 		/*

Signature · 1 line

1.5.3.8

Miao Xie

author

2008-11-10 12:33 UTC #3 lore raw

In reply to Yasunori Goto

Quoted 1 line · on 2008-11-10 14:56 Yasunori Goto wrote:

Did you test this patch? Was it OK?

I did it on IA64 just now. Updating top cpuset's mems is fine.
But I forgot to update the mems_allowed of every task in top cpuset group. It is difficult to fix it, because I don't know whether updating a memory-bound task's mem_allowed is good or not.

cpuset_track_online_cpus() has the same problem.

Quoted 121 lines

If yes,

Acked-by Yasunori Goto <[email protected]>

Thanks.

Quoted 114 lines

After adding a node into the machine, top cpuset's mems isn't updated.

This patch fixes it. And we use notifier of memory hotplug instead of direct calling of cpuset_track_online_nodes().

Signed-off-by Miao Xie <[email protected]>

Cc Yasunori Goto <[email protected]>

---
 include/linux/cpuset.h |    4 ----
 kernel/cpuset.c        |   19 ++++++++++++++++---
 mm/memory_hotplug.c    |    3 ---
 3 files changed, 16 insertions(+), 10 deletions(-)

include/linux/cpuset.h +0 -4

index 2691926..8e540d3 100644
@@ -74,8 +74,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return current->flags & PF_SPREAD_SLAB;
 }
 
-extern void cpuset_track_online_nodes(void);
-
 extern int current_cpuset_is_being_rebound(void);
 
 extern void rebuild_sched_domains(void);

@@ -151,8 +149,6 @@ static inline int cpuset_do_slab_mem_spread(void)
 	return 0;
 }
 
-static inline void cpuset_track_online_nodes(void) {}
-
 static inline int current_cpuset_is_being_rebound(void)
 {
 	return 0;

kernel/cpuset.c +16 -3

index 3e00526..909a548 100644
@@ -36,6 +36,7 @@
 #include <linux/list.h>
 #include <linux/mempolicy.h>
 #include <linux/mm.h>
+#include <linux/memory.h>
 #include <linux/module.h>
 #include <linux/mount.h>
 #include <linux/namei.h>

@@ -2011,12 +2012,23 @@ static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
  * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
  * See also the previous routine cpuset_track_online_cpus().
  */
-void cpuset_track_online_nodes(void)
+static int cpuset_track_online_nodes(struct notifier_block *self,
+				unsigned long action, void *arg)
 {
 	cgroup_lock();
-	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
-	scan_for_empty_cpusets(&top_cpuset);
+	switch (action) {
+	case MEM_ONLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		break;
+	case MEM_OFFLINE:
+		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
+		scan_for_empty_cpusets(&top_cpuset);
+		break;
+	default:
+		break;
+	}
 	cgroup_unlock();
+	return NOTIFY_OK;
 }
 #endif

@@ -2032,6 +2044,7 @@ void __init cpuset_init_smp(void)
 	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
 
 	hotcpu_notifier(cpuset_track_online_cpus, 0);
+	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
 }
 
 /**

mm/memory_hotplug.c +0 -3

index 6837a10..b5b2b15 100644
@@ -22,7 +22,6 @@
 #include <linux/highmem.h>
 #include <linux/vmalloc.h>
 #include <linux/ioport.h>
-#include <linux/cpuset.h>
 #include <linux/delay.h>
 #include <linux/migrate.h>
 #include <linux/page-isolation.h>

@@ -498,8 +497,6 @@ int add_memory(int nid, u64 start, u64 size)
 	/* we online node here. we can't roll back from here. */
 	node_set_online(nid);
 
-	cpuset_track_online_nodes();
-
 	if (new_pgdat) {
 		ret = register_one_node(nid);
 		/*

Signature · 1 line

1.5.3.8

David Rientjes

2008-11-10 20:27 UTC #4 lore raw

Quoted 8 lines · Miao Xie

Quoted 1 line · on 2008-11-10 14:56 Yasunori Goto wrote:
Did you test this patch? Was it OK?
I did it on IA64 just now. Updating top cpuset's mems is fine.
But I forgot to update the mems_allowed of every task in top cpuset group. It is difficult to fix it, because I don't know whether updating a memory-bound task's mem_allowed is good or not.

There's a fix pending in the mmotm tree that calls cpuset_update_task_memory_state() in the slow path of the page allocator, so if one of these tasks fails to allocate memory, its cached mems_allowed nodemask will be updated to include the new node(s).

The patch is available at
http://userweb.kernel.org/~akpm/mmotm/broken-out/cpusets-update-mems-allowed-in-page-allocator.patch

Yasunori Goto

Acked-by

2008-11-11 01:48 UTC #5 lore raw

Quoted 11 lines · Miao Xie

Quoted 9 lines
Quoted 1 line · on 2008-11-10 14:56 Yasunori Goto wrote:
Did you test this patch? Was it OK?
I did it on IA64 just now. Updating top cpuset's mems is fine.
But I forgot to update the mems_allowed of every task in top cpuset group. It is difficult to fix it, because I don't know whether updating a memory-bound task's mem_allowed is good or not.

Quoted 7 lines

There's a fix pending in the mmotm tree that calls cpuset_update_task_memory_state() in the slow path of the page allocator, so if one of these tasks fails to allocate memory, its cached mems_allowed nodemask will be updated to include the new node(s).
The patch is available at
http://userweb.kernel.org/~akpm/mmotm/broken-out/cpusets-update-mems-allowed-in-page-allocator.patch

Ah, OK. This seems reasonable.

Then Miao-san's patch is fine.

Acked-by Yasunori Goto <[email protected]>

Thanks.

Signature · 1 line

Yasunori Goto