mm/demotion: Memory tiers and demotion

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU	300
+#define MEMORY_TIER_DRAM	200
+#define MEMORY_TIER_PMEM	100
+
+#define DEFAULT_MEMORY_TIER	MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID	400
+
+#endif	/* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)		+= memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +68 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+	struct list_head list;
+	int id;
+	nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+	struct list_head *ent;
+	struct memory_tier *tmp_memtier;
+
+	lockdep_assert_held_once(&memory_tier_lock);
+
+	list_for_each(ent, &memory_tiers) {
+		tmp_memtier = list_entry(ent, struct memory_tier, list);
+		if (tmp_memtier->id < memtier->id) {
+			list_add_tail(&memtier->list, ent);
+			return;
+		}
+	}
+	list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+	struct memory_tier *memtier;
+
+	if (tier > MAX_MEMORY_TIER_ID)
+		return ERR_PTR(-EINVAL);
+
+	memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+	if (!memtier)
+		return ERR_PTR(-ENOMEM);
+
+	memtier->id   = tier;
+
+	insert_memory_tier(memtier);
+
+	return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+	struct memory_tier *memtier;
+
+	/*
+	 * Register only default memory tier to hide all empty
+	 * memory tier from sysfs. Since this is early during
+	 * boot, we could avoid holding memtory_tier_lock. But
+	 * keep it simple by holding locks. So we can add lock
+	 * held debug checks in other functions.
+	 */
+	mutex_lock(&memory_tier_lock);
+	memtier = register_memory_tier(default_memtier);

Quoted 185 lines · Aneesh Kumar K.V

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU	300
+#define MEMORY_TIER_DRAM	200
+#define MEMORY_TIER_PMEM	100
+
+#define DEFAULT_MEMORY_TIER	MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID	400
+
+#endif	/* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)		+= memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +77 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+	struct list_head list;
+	int id;
+	nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+	struct list_head *ent;
+	struct memory_tier *tmp_memtier;
+
+	lockdep_assert_held_once(&memory_tier_lock);
+
+	list_for_each(ent, &memory_tiers) {
+		tmp_memtier = list_entry(ent, struct memory_tier, list);
+		if (tmp_memtier->id < memtier->id) {
+			list_add_tail(&memtier->list, ent);
+			return;
+		}
+	}
+	list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+	struct memory_tier *memtier;
+
+	if (tier > MAX_MEMORY_TIER_ID)
+		return ERR_PTR(-EINVAL);
+
+	memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+	if (!memtier)
+		return ERR_PTR(-ENOMEM);
+
+	memtier->id   = tier;
+
+	insert_memory_tier(memtier);
+
+	return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+	struct memory_tier *memtier;
+
+	/*
+	 * Register only default memory tier to hide all empty
+	 * memory tier from sysfs. Since this is early during
+	 * boot, we could avoid holding memtory_tier_lock. But
+	 * keep it simple by holding locks. So we can add lock
+	 * held debug checks in other functions.
+	 */
+	mutex_lock(&memory_tier_lock);
+	memtier = register_memory_tier(default_memtier);
+	if (IS_ERR(memtier))
+		panic("%s() failed to register memory tier: %ld\n",
+		      __func__, PTR_ERR(memtier));
+
+	/* CPU only nodes are not part of memory tiers. */
+	memtier->nodelist = node_states[N_MEMORY];
+	mutex_unlock(&memory_tier_lock);
+	return 0;
+}

mm/memory-tiers.c line 78

Quoted 1 line

mm/memory-tiers.c new file +1 -0

+subsys_initcall(memory_tier_init);

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

- kmem_dax driver will setup distance for PMEM NUMA nodes before online them.

- When a NUMA node is onlined, we will use find_create_memory_tier() to find or create its memory tier and add the NUMA node into the memory tier.

- Or we can add memory type data structure now.

Best Regards,
Huang, Ying

Quoted 187 lines · Ying

Quoted 185 lines · Aneesh Kumar K.V

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU	300
+#define MEMORY_TIER_DRAM	200
+#define MEMORY_TIER_PMEM	100
+
+#define DEFAULT_MEMORY_TIER	MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID	400
+
+#endif	/* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)		+= memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +77 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+	struct list_head list;
+	int id;
+	nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+	struct list_head *ent;
+	struct memory_tier *tmp_memtier;
+
+	lockdep_assert_held_once(&memory_tier_lock);
+
+	list_for_each(ent, &memory_tiers) {
+		tmp_memtier = list_entry(ent, struct memory_tier, list);
+		if (tmp_memtier->id < memtier->id) {
+			list_add_tail(&memtier->list, ent);
+			return;
+		}
+	}
+	list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+	struct memory_tier *memtier;
+
+	if (tier > MAX_MEMORY_TIER_ID)
+		return ERR_PTR(-EINVAL);
+
+	memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+	if (!memtier)
+		return ERR_PTR(-ENOMEM);
+
+	memtier->id   = tier;
+
+	insert_memory_tier(memtier);
+
+	return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+	struct memory_tier *memtier;
+
+	/*
+	 * Register only default memory tier to hide all empty
+	 * memory tier from sysfs. Since this is early during
+	 * boot, we could avoid holding memtory_tier_lock. But
+	 * keep it simple by holding locks. So we can add lock
+	 * held debug checks in other functions.
+	 */
+	mutex_lock(&memory_tier_lock);
+	memtier = register_memory_tier(default_memtier);
+	if (IS_ERR(memtier))
+		panic("%s() failed to register memory tier: %ld\n",
+		      __func__, PTR_ERR(memtier));
+
+	/* CPU only nodes are not part of memory tiers. */
+	memtier->nodelist = node_states[N_MEMORY];
+	mutex_unlock(&memory_tier_lock);
+	return 0;
+}

Quoted 9 lines

mm/memory-tiers.c line 78

Quoted 1 line

mm/memory-tiers.c new file +1 -0

+subsys_initcall(memory_tier_init);

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Sorry I am not able to follow you here. This patchset completely drops exposing memory tiers to userspace via sysfs. Instead it allow creation of memory tiers with specific tierID from within the kernel/device driver. Default tierID is 200 and dax kmem creates memory tier with tierID 100.

Quoted 6 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

Quoted 6 lines

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

I don't follow this. I guess you are trying to have a different design. Would it be much easier if you can write this in the form of a patch?

Quoted 7 lines

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

Quoted 3 lines

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

I was expecting this to be internal to dax kmem. How dax kmem maps "abstract distance" to a memory tier. At this point this patchset is keeping all that for a future patchset.

Quoted 3 lines

- kmem_dax driver will setup distance for PMEM NUMA nodes before online them.

Sure we can do that as part of future patchset ?

Quoted 4 lines

- When a NUMA node is onlined, we will use find_create_memory_tier() to find or create its memory tier and add the NUMA node into the memory tier.

This is what this patchset does. When we online a numa node the kernel find the memory tier for the node (__node_get_memory_tier). If it doesn't exist, we create one. (The new one created is not dynamic as you outlined earlier. But then that can be done in a future patchset). For now I am keeping this simpler.

static int node_set_memory_tier(int node, int tier)

{
	struct memory_tier *memtier;
	int ret = 0;

	mutex_lock(&memory_tier_lock);
	memtier = __node_get_memory_tier(node);
	/*
	 * if node is already part of the tier proceed with the
	 * current tier value, because we might want to establish
	 * new migration paths now. The node might be added to a tier
	 * before it was made part of N_MEMORY, hence estabilish_migration_targets
	 * will have skipped this node.
	 */
	if (!memtier)
		ret = __node_set_memory_tier(node, tier);
	establish_migration_targets();

	mutex_unlock(&memory_tier_lock);

	return ret;
}

Quoted 4 lines

- Or we can add memory type data structure now.

Best Regards,
Huang, Ying

-aneesh

Quoted 8 lines

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Sorry I am not able to follow you here. This patchset completely drops exposing memory tiers to userspace via sysfs. Instead it allow creation of memory tiers with specific tierID from within the kernel/device driver. Default tierID is 200 and dax kmem creates memory tier with tierID 100.

Quoted 6 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

Quoted 6 lines

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

I don't follow this. I guess you are trying to have a different design. Would it be much easier if you can write this in the form of a patch?

Quoted 7 lines

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

Quoted 3 lines

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

I was expecting this to be internal to dax kmem. How dax kmem maps "abstract distance" to a memory tier. At this point this patchset is keeping all that for a future patchset.

Aneesh Kumar K V <[email protected]> writes:

....

Quoted 43 lines

Quoted 8 lines

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Sorry I am not able to follow you here. This patchset completely drops exposing memory tiers to userspace via sysfs. Instead it allow creation of memory tiers with specific tierID from within the kernel/device driver. Default tierID is 200 and dax kmem creates memory tier with tierID 100.

Quoted 6 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

Quoted 6 lines

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

I don't follow this. I guess you are trying to have a different design. Would it be much easier if you can write this in the form of a patch?

Quoted 10 lines

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

Quoted 4 lines

I was expecting this to be internal to dax kmem. How dax kmem maps "abstract distance" to a memory tier. At this point this patchset is keeping all that for a future patchset.

This shows how i was expecting "abstract distance" to be integrated.

On 7/15/22 1:23 PM, Huang, Ying wrote:

Quoted 8 lines

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Sorry I am not able to follow you here. This patchset completely drops exposing memory tiers to userspace via sysfs. Instead it allow creation of memory tiers with specific tierID from within the kernel/device driver. Default tierID is 200 and dax kmem creates memory tier with tierID 100.

Quoted 8 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

Quoted 4 lines

or, just

    pgdat->distance

I don't follow this. I guess you are trying to have a different design. Would it be much easier if you can write this in the form of a patch?

Quoted 1 line · Aneesh Kumar K V

On 7/15/22 1:23 PM, Huang, Ying wrote:

[snip]

Quoted 23 lines

Quoted 8 lines

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Sorry I am not able to follow you here. This patchset completely drops exposing memory tiers to userspace via sysfs. Instead it allow creation of memory tiers with specific tierID from within the kernel/device driver. Default tierID is 200 and dax kmem creates memory tier with tierID 100.

Quoted 8 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

Quoted 7 lines

Quoted 4 lines

or, just

    pgdat->distance

I don't follow this. I guess you are trying to have a different design. Would it be much easier if you can write this in the form of a patch?

Written some pseudo code as follow to show my basic idea.

#define MEMORY_TIER_ADISTANCE_DRAM 150
#define MEMORY_TIER_ADISTANCE_PMEM 250

struct memory_tier {
	/* abstract distance range covered by the memory tier */
	int adistance_start;
	int adistance_len;
	struct list_head list;
	nodemask_t nodemask;
};

/* RCU list of memory tiers */
static LIST_HEAD(memory_tiers);

/* abstract distance of each NUMA node */
int node_adistances[MAX_NUMNODES];

struct memory_tier *find_create_memory_tier(int adistance)

{
	struct memory_tier *tier;

	list_for_each_entry(tier, &memory_tiers, list) {
		if (adistance >= tier->adistance_start &&
		    adistance < tier->adistance_start + tier->adistance_len)
			return tier;
	}
	/* allocate a new memory tier and return */
}

void memory_tier_add_node(int nid)

{
	int adistance;
	struct memory_tier *tier;

	adistance = node_adistances[nid] || MEMORY_TIER_ADISTANCE_DRAM;
	tier = find_create_memory_tier(adistance);
	node_set(nid, &tier->nodemask);
	/* setup demotion data structure, etc */
}

static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
unsigned long action, void *_arg)

{
	struct memory_notify *arg = _arg;
	int nid;

	nid = arg->status_change_nid;
	if (nid < 0)
		return notifier_from_errno(0);

	switch (action) {
	case MEM_ONLINE:
		memory_tier_add_node(nid);
		break;
	}

	return notifier_from_errno(0);
}

/* kmem.c */
static int dev_dax_kmem_probe(struct dev_dax *dev_dax)

{
	node_adistances[dev_dax->target_node] = MEMORY_TIER_ADISTANCE_PMEM;
	/* add_memory_driver_managed() */
}

[snip]

Best Regards,
Huang, Ying

Quoted 107 lines · Ying

Quoted 1 line · Aneesh Kumar K V

On 7/15/22 1:23 PM, Huang, Ying wrote:

[snip]

Quoted 23 lines

Quoted 8 lines

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Sorry I am not able to follow you here. This patchset completely drops exposing memory tiers to userspace via sysfs. Instead it allow creation of memory tiers with specific tierID from within the kernel/device driver. Default tierID is 200 and dax kmem creates memory tier with tierID 100.

Quoted 8 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250. The distance of each NUMA node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

Quoted 7 lines

Quoted 4 lines

or, just

    pgdat->distance

I don't follow this. I guess you are trying to have a different design. Would it be much easier if you can write this in the form of a patch?

Written some pseudo code as follow to show my basic idea.

#define MEMORY_TIER_ADISTANCE_DRAM 150
#define MEMORY_TIER_ADISTANCE_PMEM 250

struct memory_tier {
	/* abstract distance range covered by the memory tier */
	int adistance_start;
	int adistance_len;
	struct list_head list;
	nodemask_t nodemask;
};

/* RCU list of memory tiers */
static LIST_HEAD(memory_tiers);

/* abstract distance of each NUMA node */
int node_adistances[MAX_NUMNODES];

struct memory_tier *find_create_memory_tier(int adistance)

{
	struct memory_tier *tier;

	list_for_each_entry(tier, &memory_tiers, list) {
		if (adistance >= tier->adistance_start &&
		    adistance < tier->adistance_start + tier->adistance_len)
			return tier;
	}
	/* allocate a new memory tier and return */
}

void memory_tier_add_node(int nid)

{
	int adistance;
	struct memory_tier *tier;

	adistance = node_adistances[nid] || MEMORY_TIER_ADISTANCE_DRAM;
	tier = find_create_memory_tier(adistance);
	node_set(nid, &tier->nodemask);
	/* setup demotion data structure, etc */
}

static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
unsigned long action, void *_arg)

{
	struct memory_notify *arg = _arg;
	int nid;

	nid = arg->status_change_nid;
	if (nid < 0)
		return notifier_from_errno(0);

	switch (action) {
	case MEM_ONLINE:
		memory_tier_add_node(nid);
		break;
	}

	return notifier_from_errno(0);
}

/* kmem.c */
static int dev_dax_kmem_probe(struct dev_dax *dev_dax)

{
	node_adistances[dev_dax->target_node] = MEMORY_TIER_ADISTANCE_PMEM;
	/* add_memory_driver_managed() */
}

[snip]

Quoted 2 lines

Best Regards,
Huang, Ying

Implementing that I ended up with the below. The difference is adistance_len is not a memory tier property instead it is a kernel parameter like memory_tier_chunk_size which can
be tuned to create more memory tiers.

How about this? Not yet tested.

struct memory_tier {
	struct list_head list;
	int id;

	int perf_level;
	nodemask_t nodelist;
};

static LIST_HEAD(memory_tiers);
static DEFINE_MUTEX(memory_tier_lock);
static unsigned int default_memtier_perf_level = DEFAULT_MEMORY_TYPE_PERF; core_param(default_memory_tier_perf_level, default_memtier_perf_level, uint, 0644); static unsigned int memtier_perf_chunk_size = 150; core_param(memory_tier_perf_chunk, memtier_perf_chunk_size, uint, 0644);

/*
* performance levels are grouped into memtiers each of chunk size
* memtier_perf_chunk
*/
static struct memory_tier *find_create_memory_tier(unsigned int perf_level)

{
	bool found_slot = false;
	struct list_head *ent;
	struct memory_tier *memtier, *new_memtier;
	static int next_memtier_id = 0;
	/*
	 * zero is special in that it indicates uninitialized
	 * perf level by respective driver. Pick default memory
	 * tier perf level for that.
	 */
	if (!perf_level)
		perf_level = default_memtier_perf_level;

	lockdep_assert_held_once(&memory_tier_lock);

	list_for_each(ent, &memory_tiers) {
		memtier = list_entry(ent, struct memory_tier, list);
		if (perf_level >= memtier->perf_level &&
		    perf_level < memtier->perf_level + memtier_perf_chunk_size)
			return memtier;
		else if (perf_level < memtier->perf_level) {
			found_slot = true;
			break;
		}
	}

	new_memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
	if (!new_memtier)
		return ERR_PTR(-ENOMEM);

	new_memtier->id = next_memtier_id++;
	new_memtier->perf_level = ALIGN_DOWN(perf_level, memtier_perf_chunk_size);
	if (found_slot)
		list_add_tail(&new_memtier->list, ent);
	else
		list_add_tail(&new_memtier->list, &memory_tiers);
	return new_memtier;
}

static int __init memory_tier_init(void)

{
	int node;
	struct memory_tier *memtier;

	/*
	 * Since this is early during  boot, we could avoid
	 * holding memtory_tier_lock. But keep it simple by
	 * holding locks. So we can add lock held debug checks
	 * in other functions.
	 */
	mutex_lock(&memory_tier_lock);
	memtier = find_create_memory_tier(default_memtier_perf_level);
	if (IS_ERR(memtier))
		panic("%s() failed to register memory tier: %ld\n",
		      __func__, PTR_ERR(memtier));

	/* CPU only nodes are not part of memory tiers. */
	memtier->nodelist = node_states[N_MEMORY];

	/*
	 * nodes that are already online and that doesn't
	 * have perf level assigned is assigned a default perf
	 * level.
	 */
	for_each_node_state(node, N_MEMORY) {
		struct node *node_property = node_devices[node];

		if (!node_property->perf_level)
			node_property->perf_level = default_memtier_perf_level;
	}
	mutex_unlock(&memory_tier_lock);
	return 0;
}

subsys_initcall(memory_tier_init);

Quoted 176 lines · Aneesh Kumar K.V

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU  300
+#define MEMORY_TIER_DRAM     200
+#define MEMORY_TIER_PMEM     100
+
+#define DEFAULT_MEMORY_TIER  MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID   400
+
+#endif       /* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)                += memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +68 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+     struct list_head list;
+     int id;
+     nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+     struct list_head *ent;
+     struct memory_tier *tmp_memtier;
+
+     lockdep_assert_held_once(&memory_tier_lock);
+
+     list_for_each(ent, &memory_tiers) {
+             tmp_memtier = list_entry(ent, struct memory_tier, list);
+             if (tmp_memtier->id < memtier->id) {
+                     list_add_tail(&memtier->list, ent);
+                     return;
+             }
+     }
+     list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+     struct memory_tier *memtier;
+
+     if (tier > MAX_MEMORY_TIER_ID)
+             return ERR_PTR(-EINVAL);
+
+     memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+     if (!memtier)
+             return ERR_PTR(-ENOMEM);
+
+     memtier->id   = tier;
+
+     insert_memory_tier(memtier);
+
+     return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+     struct memory_tier *memtier;
+
+     /*
+      * Register only default memory tier to hide all empty
+      * memory tier from sysfs. Since this is early during
+      * boot, we could avoid holding memtory_tier_lock. But
+      * keep it simple by holding locks. So we can add lock
+      * held debug checks in other functions.
+      */
+     mutex_lock(&memory_tier_lock);
+     memtier = register_memory_tier(default_memtier);

mm/memory-tiers.c line 78

Quoted 10 lines

mm/memory-tiers.c new file +10 -0

+     if (IS_ERR(memtier))
+             panic("%s() failed to register memory tier: %ld\n",
+                   __func__, PTR_ERR(memtier));
+
+     /* CPU only nodes are not part of memory tiers. */
+     memtier->nodelist = node_states[N_MEMORY];
+     mutex_unlock(&memory_tier_lock);
+     return 0;
+}
+subsys_initcall(memory_tier_init);

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250.

The distance of each NUMA
node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

- kmem_dax driver will setup distance for PMEM NUMA nodes before online them.

- When a NUMA node is onlined, we will use find_create_memory_tier() to find or create its memory tier and add the NUMA node into the memory tier.

- Or we can add memory type data structure now.

Best Regards,
Huang, Ying

Quoted 197 lines · Ying

Quoted 176 lines · Aneesh Kumar K.V

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU  300
+#define MEMORY_TIER_DRAM     200
+#define MEMORY_TIER_PMEM     100
+
+#define DEFAULT_MEMORY_TIER  MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID   400
+
+#endif       /* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)                += memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +68 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+     struct list_head list;
+     int id;
+     nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+     struct list_head *ent;
+     struct memory_tier *tmp_memtier;
+
+     lockdep_assert_held_once(&memory_tier_lock);
+
+     list_for_each(ent, &memory_tiers) {
+             tmp_memtier = list_entry(ent, struct memory_tier, list);
+             if (tmp_memtier->id < memtier->id) {
+                     list_add_tail(&memtier->list, ent);
+                     return;
+             }
+     }
+     list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+     struct memory_tier *memtier;
+
+     if (tier > MAX_MEMORY_TIER_ID)
+             return ERR_PTR(-EINVAL);
+
+     memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+     if (!memtier)
+             return ERR_PTR(-ENOMEM);
+
+     memtier->id   = tier;
+
+     insert_memory_tier(memtier);
+
+     return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+     struct memory_tier *memtier;
+
+     /*
+      * Register only default memory tier to hide all empty
+      * memory tier from sysfs. Since this is early during
+      * boot, we could avoid holding memtory_tier_lock. But
+      * keep it simple by holding locks. So we can add lock
+      * held debug checks in other functions.
+      */
+     mutex_lock(&memory_tier_lock);
+     memtier = register_memory_tier(default_memtier);

mm/memory-tiers.c line 78

Quoted 10 lines

mm/memory-tiers.c new file +10 -0

+     if (IS_ERR(memtier))
+             panic("%s() failed to register memory tier: %ld\n",
+                   __func__, PTR_ERR(memtier));
+
+     /* CPU only nodes are not part of memory tiers. */
+     memtier->nodelist = node_states[N_MEMORY];
+     mutex_unlock(&memory_tier_lock);
+     return 0;
+}
+subsys_initcall(memory_tier_init);

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Quoted 4 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250.

I agree with this design, though I'd prefer the new attribute to not be named as "distance". This is to both avoid the confusion with the SLIT distance and to avoid the misconception that only the latency matters, but the bandwidth doesn't.

How about we call it "performance level" (perf_level) or something similar instead?

Quoted 8 lines

The distance of each NUMA
node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

I think node_devices[] is a better place to record this new attribute. The HMAT performance data is also listed there.

Quoted 7 lines

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

Quoted 5 lines

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

- kmem_dax driver will setup distance for PMEM NUMA nodes before online them.

This attribute should be a property of the NUMA node based on the device hardware.

For PMEM, it is better to handle at the ACPI level.
For example, we can consider initializing this attribute for a PMEM node in acpi_numa_memory_affinity_init() when the node is non-volatile.

Quoted 3 lines

- When a NUMA node is onlined, we will use find_create_memory_tier() to find or create its memory tier and add the NUMA node into the memory tier.

I think we should create all the memory tiers up-front, just like NUMA nodes, to keep their devices and IDs stable. Similar to offline NUMA nodes, when a memory tier has no online nodes, we can mark it as offline and exclude it from online-related operations (e.g. demotion). A memory tier can be made online when it gets assigned with an online node.

Quoted 1 line

- Or we can add memory type data structure now.

Quoted 203 lines · Wei Xu

Quoted 197 lines · Ying

Quoted 176 lines · Aneesh Kumar K.V

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU  300
+#define MEMORY_TIER_DRAM     200
+#define MEMORY_TIER_PMEM     100
+
+#define DEFAULT_MEMORY_TIER  MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID   400
+
+#endif       /* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)                += memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +68 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+     struct list_head list;
+     int id;
+     nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+     struct list_head *ent;
+     struct memory_tier *tmp_memtier;
+
+     lockdep_assert_held_once(&memory_tier_lock);
+
+     list_for_each(ent, &memory_tiers) {
+             tmp_memtier = list_entry(ent, struct memory_tier, list);
+             if (tmp_memtier->id < memtier->id) {
+                     list_add_tail(&memtier->list, ent);
+                     return;
+             }
+     }
+     list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+     struct memory_tier *memtier;
+
+     if (tier > MAX_MEMORY_TIER_ID)
+             return ERR_PTR(-EINVAL);
+
+     memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+     if (!memtier)
+             return ERR_PTR(-ENOMEM);
+
+     memtier->id   = tier;
+
+     insert_memory_tier(memtier);
+
+     return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+     struct memory_tier *memtier;
+
+     /*
+      * Register only default memory tier to hide all empty
+      * memory tier from sysfs. Since this is early during
+      * boot, we could avoid holding memtory_tier_lock. But
+      * keep it simple by holding locks. So we can add lock
+      * held debug checks in other functions.
+      */
+     mutex_lock(&memory_tier_lock);
+     memtier = register_memory_tier(default_memtier);

mm/memory-tiers.c line 78

Quoted 10 lines

mm/memory-tiers.c new file +10 -0

+     if (IS_ERR(memtier))
+             panic("%s() failed to register memory tier: %ld\n",
+                   __func__, PTR_ERR(memtier));
+
+     /* CPU only nodes are not part of memory tiers. */
+     memtier->nodelist = node_states[N_MEMORY];
+     mutex_unlock(&memory_tier_lock);
+     return 0;
+}
+subsys_initcall(memory_tier_init);

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Quoted 4 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250.

Quoted 7 lines

I agree with this design, though I'd prefer the new attribute to not be named as "distance". This is to both avoid the confusion with the SLIT distance and to avoid the misconception that only the latency matters, but the bandwidth doesn't.

How about we call it "performance level" (perf_level) or something similar instead?

I have no strong opinion on this. Both "distance" or "perf_level" looks OK to me.

Quoted 11 lines

Quoted 8 lines

The distance of each NUMA
node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

I think node_devices[] is a better place to record this new attribute. The HMAT performance data is also listed there.

Firstly, we all agree that we need a place to record this information, per node or per memory type. Personally, I prefer to separate the data and its interface (such as sysfs).

Quoted 7 lines

Quoted 7 lines

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

Quoted 8 lines

Quoted 5 lines

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

- kmem_dax driver will setup distance for PMEM NUMA nodes before online them.

This attribute should be a property of the NUMA node based on the device hardware.

Yes. Or a property of a memory type.

Quoted 4 lines

For PMEM, it is better to handle at the ACPI level.
For example, we can consider initializing this attribute for a PMEM node in acpi_numa_memory_affinity_init() when the node is non-volatile.

The abstract_distance/perf_level may be determined from multiple information sources, e.g., ACPI SLIT/SRAT/HMAT, etc. It should be the responsibility of device drivers (e.g., kmem_dax) to determine the final value of abstract_distance/perf_level based on the information availability/priority and some specific knowledge of the hardware. Yes, ACPI SRAT is valuable to determine the abstract_distance/perf_level. And, it's better for kmem_dax to use it to determine the final value of abstract_distance/perf_level.

To make the first version as simple as possible, I think we can just use some static abstract_distance/perf_level in kmem_dax driver for the NUMA nodes onlined by it. Because we use the driver for PMEM only now. We can enhance the implementation later.

Quoted 10 lines

Quoted 3 lines

- When a NUMA node is onlined, we will use find_create_memory_tier() to find or create its memory tier and add the NUMA node into the memory tier.

I think we should create all the memory tiers up-front, just like NUMA nodes, to keep their devices and IDs stable. Similar to offline NUMA nodes, when a memory tier has no online nodes, we can mark it as offline and exclude it from online-related operations (e.g. demotion). A memory tier can be made online when it gets assigned with an online node.

Each memory tier corresponds to a range of abstract_distance/perf_level. For example, if 1 <= abstract_distance/perf_level <= 500, 5 memory tiers can be defined with abstract_distance/perf_level ranges 1-100, 101-200, 201-300, 301-400, 401-500. We can create these 5 memory tiers up-front of course. In the new design, we may change the ranges at run time according to policy chosen by the users. For example, we may change 5 memory tiers above to 500 memory tiers, with abstract_distance/perf_level ranges 1-1, 2-2, ..., 500-500. This may make memory tier devices and their IDs unstable at some degree. But if we are cautious to customize the ranges, it's possible to make the memory tier devices and their IDs stable in most cases.

Because we may define 500 memory tiers, it's hard to create all memory tiers up-front really. But we can create them all in concept and allocate memory/resources for one when we add the first NUMA node to it.

To make the fist version as simple as possible, I suggest to define 500 memory tiers as above statically.

Quoted 1 line

Quoted 1 line

- Or we can add memory type data structure now.

Best Regards,
Huang, Ying

Quoted 264 lines · Ying

Quoted 203 lines · Wei Xu

Quoted 197 lines · Ying

Quoted 176 lines · Aneesh Kumar K.V

In the current kernel, memory tiers are defined implicitly via a demotion path relationship between NUMA nodes, which is created during the kernel initialization and updated when a NUMA node is hot-added or hot-removed. The current implementation puts all nodes with CPU into the top tier, and builds the tier hierarchy tier-by-tier by establishing the per-node demotion targets based on the distances between nodes.

This current memory tier kernel interface needs to be improved for several important use cases,

The current tier initialization code always initializes each memory-only NUMA node into a lower tier. But a memory-only NUMA node may have a high performance memory device (e.g. a DRAM device attached via CXL.mem or a DRAM-backed memory-only node on a virtual machine) and should be put into a higher tier.

The current tier hierarchy always puts CPU nodes into the top tier. But on a system with HBM or GPU devices, the memory-only NUMA nodes mapping these devices should be in the top tier, and DRAM nodes with CPUs are better to be placed into the next lower tier.

With current kernel higher tier node can only be demoted to selected nodes on the next lower tier as defined by the demotion path, not any other
node from any lower tier. This strict, hard-coded demotion order does not work in all use cases (e.g. some use cases may want to allow cross-socket demotion to another node in the same demotion tier as a fallback when the preferred demotion node is out of space), This demotion order is also inconsistent with the page allocation fallback order when all the nodes in a higher tier are out of space: The page allocation can fall back to any node from
any lower tier, whereas the demotion order doesn't allow that.

The current kernel also don't provide any interfaces for the userspace to learn about the memory tier hierarchy in order to optimize its memory allocations.

This patch series address the above by defining memory tiers explicitly.

This patch introduce explicity memory tiers. The tier ID value of a memory tier is used to derive the demotion order between NUMA nodes.

For example, if we have 3 memtiers: memtier100, memtier200, memiter300 then the memory tier order is: memtier300 -> memtier200 -> memtier100 where memtier300 is the highest tier and memtier100 is the lowest tier.

While reclaim we migrate pages from fast(higher) tiers to slow(lower) tiers when the fast(higher) tier is under memory pressure.

This patchset introduce 3 memory tiers (memtier100, memtier200 and memtier300) which are created by different kernel subsystems. The default memory tier created by the kernel is memtier200. A kernel parameter is provided to override the default memory tier.

Link https://lore.kernel.org/linux-mm/CAAPL-u9Wv+nH1VOZTj=9p9S70Y3Qz3+63EkqncRDdHfubsrjfw@mail.gmail.com open here

Link https://lore.kernel.org/linux-mm/[email protected] open here

Signed-off-by Jagdish Gediya <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h | 15 +++++++
 mm/Makefile                  |  1 +
 mm/memory-tiers.c            | 78 ++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+)
 create mode 100644 include/linux/memory-tiers.h
 create mode 100644 mm/memory-tiers.c

include/linux/memory-tiers.h new file +15 -0

new file mode 100644
index 000000000000..a81dbc20e0d1
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MEMORY_TIERS_H
+#define _LINUX_MEMORY_TIERS_H
+
+#ifdef CONFIG_NUMA
+
+#define MEMORY_TIER_HBM_GPU  300
+#define MEMORY_TIER_DRAM     200
+#define MEMORY_TIER_PMEM     100
+
+#define DEFAULT_MEMORY_TIER  MEMORY_TIER_DRAM
+#define MAX_MEMORY_TIER_ID   400
+
+#endif       /* CONFIG_NUMA */
+#endif  /* _LINUX_MEMORY_TIERS_H */

mm/Makefile +1 -0

index 6f9ffa968a1a..d30acebc2164 100644
@@ -92,6 +92,7 @@ obj-$(CONFIG_KFENCE) += kfence/
 obj-$(CONFIG_FAILSLAB) += failslab.o
 obj-$(CONFIG_MEMTEST)                += memtest.o
 obj-$(CONFIG_MIGRATION) += migrate.o
+obj-$(CONFIG_NUMA) += memory-tiers.o
 obj-$(CONFIG_DEVICE_MIGRATION) += migrate_device.o
 obj-$(CONFIG_TRANSPARENT_HUGEPAGE) += huge_memory.o khugepaged.o
 obj-$(CONFIG_PAGE_COUNTER) += page_counter.o

mm/memory-tiers.c new file +68 -0

new file mode 100644
index 000000000000..011877b6dbb9
@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <linux/nodemask.h>
+#include <linux/slab.h>
+#include <linux/lockdep.h>
+#include <linux/moduleparam.h>
+#include <linux/memory-tiers.h>
+
+struct memory_tier {
+     struct list_head list;
+     int id;
+     nodemask_t nodelist;
+};
+
+static DEFINE_MUTEX(memory_tier_lock);
+static LIST_HEAD(memory_tiers);
+
+static void insert_memory_tier(struct memory_tier *memtier)
+{
+     struct list_head *ent;
+     struct memory_tier *tmp_memtier;
+
+     lockdep_assert_held_once(&memory_tier_lock);
+
+     list_for_each(ent, &memory_tiers) {
+             tmp_memtier = list_entry(ent, struct memory_tier, list);
+             if (tmp_memtier->id < memtier->id) {
+                     list_add_tail(&memtier->list, ent);
+                     return;
+             }
+     }
+     list_add_tail(&memtier->list, &memory_tiers);
+}
+
+static struct memory_tier *register_memory_tier(unsigned int tier)
+{
+     struct memory_tier *memtier;
+
+     if (tier > MAX_MEMORY_TIER_ID)
+             return ERR_PTR(-EINVAL);
+
+     memtier = kzalloc(sizeof(struct memory_tier), GFP_KERNEL);
+     if (!memtier)
+             return ERR_PTR(-ENOMEM);
+
+     memtier->id   = tier;
+
+     insert_memory_tier(memtier);
+
+     return memtier;
+}
+
+static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
+core_param(default_memory_tier, default_memtier, uint, 0644);
+
+static int __init memory_tier_init(void)
+{
+     struct memory_tier *memtier;
+
+     /*
+      * Register only default memory tier to hide all empty
+      * memory tier from sysfs. Since this is early during
+      * boot, we could avoid holding memtory_tier_lock. But
+      * keep it simple by holding locks. So we can add lock
+      * held debug checks in other functions.
+      */
+     mutex_lock(&memory_tier_lock);
+     memtier = register_memory_tier(default_memtier);

mm/memory-tiers.c line 78

Quoted 10 lines

mm/memory-tiers.c new file +10 -0

+     if (IS_ERR(memtier))
+             panic("%s() failed to register memory tier: %ld\n",
+                   __func__, PTR_ERR(memtier));
+
+     /* CPU only nodes are not part of memory tiers. */
+     memtier->nodelist = node_states[N_MEMORY];
+     mutex_unlock(&memory_tier_lock);
+     return 0;
+}
+subsys_initcall(memory_tier_init);

You dropped the original sysfs interface patches from the series, but the kernel internal implementation is still for the original sysfs interface. For example, memory tier ID is for the original sysfs interface, not for the new proposed sysfs interface. So I suggest you to implement with the new interface in mind. What do you think about the following design?

Quoted 4 lines

- Each NUMA node belongs to a memory type, and each memory type corresponds to a "abstract distance", so each NUMA node corresonds to a "distance". For simplicity, we can start with static distances, for example, DRAM (default): 150, PMEM: 250.

Quoted 7 lines

I agree with this design, though I'd prefer the new attribute to not be named as "distance". This is to both avoid the confusion with the SLIT distance and to avoid the misconception that only the latency matters, but the bandwidth doesn't.

How about we call it "performance level" (perf_level) or something similar instead?

I have no strong opinion on this. Both "distance" or "perf_level" looks OK to me.

Quoted 11 lines

Quoted 8 lines

The distance of each NUMA
node can be recorded in a global array,

    int node_distances[MAX_NUMNODES];

or, just

    pgdat->distance

I think node_devices[] is a better place to record this new attribute. The HMAT performance data is also listed there.

Firstly, we all agree that we need a place to record this information, per node or per memory type. Personally, I prefer to separate the data and its interface (such as sysfs).

Quoted 7 lines

Quoted 7 lines

- Each memory tier corresponds to a range of distance, for example, 0-100, 100-200, 200-300, >300, we can start with static ranges too.

- The core API of memory tier could be

    struct memory_tier *find_create_memory_tier(int distance);

Quoted 8 lines

Quoted 5 lines

it will find the memory tier which covers "distance" in the memory tier list, or create a new memory tier if not found.

- kmem_dax driver will setup distance for PMEM NUMA nodes before online them.

This attribute should be a property of the NUMA node based on the device hardware.

Yes. Or a property of a memory type.

Quoted 4 lines

For PMEM, it is better to handle at the ACPI level.
For example, we can consider initializing this attribute for a PMEM node in acpi_numa_memory_affinity_init() when the node is non-volatile.

The abstract_distance/perf_level may be determined from multiple information sources, e.g., ACPI SLIT/SRAT/HMAT, etc. It should be the responsibility of device drivers (e.g., kmem_dax) to determine the final value of abstract_distance/perf_level based on the information availability/priority and some specific knowledge of the hardware. Yes, ACPI SRAT is valuable to determine the abstract_distance/perf_level. And, it's better for kmem_dax to use it to determine the final value of abstract_distance/perf_level.

Quoted 4 lines

To make the first version as simple as possible, I think we can just use some static abstract_distance/perf_level in kmem_dax driver for the NUMA nodes onlined by it. Because we use the driver for PMEM only now. We can enhance the implementation later.

I agree. Ideally I think all this should be derived from ACPI tables, etc. However I think it will take a while for both FW and SW to make that information available and correct. Letting drivers initialise that for now at least should aid development in determining how performance levels should be set from multiple information sources, especially if there is no way of overriding it from userspace.

If the new NUMA node onlined doesn't have a memory tier assigned, the kernel adds the NUMA node to default memory tier.

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 mm/memory-tiers.c | 68 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 68 insertions(+)

mm/memory-tiers.c +7 -0

index 79347d4ab05e..5706ad647136 100644
@@ -5,6 +5,7 @@
 #include <linux/slab.h>
 #include <linux/lockdep.h>
 #include <linux/moduleparam.h>
+#include <linux/memory.h>
 #include <linux/memory-tiers.h>
 
 struct memory_tier {

@@ -130,8 +131,73 @@ int node_create_and_set_memory_tier(int node, int tier)
 }
 EXPORT_SYMBOL_GPL(node_create_and_set_memory_tier);
 
+static int __node_set_memory_tier(int node, int tier)
+{
+	int ret = 0;
+	struct memory_tier *memtier;
+
+	memtier = __get_memory_tier_from_id(tier);

mm/memory-tiers.c +51 -0

+{
+	struct memory_tier *memtier;
+	int ret = 0;
+
+	mutex_lock(&memory_tier_lock);
+	memtier = __node_get_memory_tier(node);
+	if (!memtier)
+		ret = __node_set_memory_tier(node, tier);
+
+	mutex_unlock(&memory_tier_lock);
+
+	return ret;
+}
+
 static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
 core_param(default_memory_tier, default_memtier, uint, 0644);
+/*
+ * This runs whether reclaim-based migration is enabled or not,
+ * which ensures that the user can turn reclaim-based migration
+ * at any time without needing to recalculate migration targets.
+ */
+static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
+						 unsigned long action, void *_arg)
+{
+	struct memory_notify *arg = _arg;
+
+	/*
+	 * Only update the node migration order when a node is
+	 * changing status, like online->offline.
+	 */
+	if (arg->status_change_nid < 0)
+		return notifier_from_errno(0);
+
+	switch (action) {
+	case MEM_ONLINE:
+		/*
+		 * We ignore the error here, if the node already have the tier
+		 * registered, we will continue to use that for the new memory
+		 * we are adding here.
+		 */
+		node_set_memory_tier(arg->status_change_nid, default_memtier);
+		break;
+	}
+
+	return notifier_from_errno(0);
+}
+
+static void __init migrate_on_reclaim_init(void)
+{
+	hotplug_memory_notifier(migrate_on_reclaim_callback, 100);
+}
 
 static int __init memory_tier_init(void)
 {

@@ -153,6 +219,8 @@ static int __init memory_tier_init(void)
 	/* CPU only nodes are not part of memory tiers. */
 	memtier->nodelist = node_states[N_MEMORY];
 	mutex_unlock(&memory_tier_lock);
+
+	migrate_on_reclaim_init();
 	return 0;
 }
 subsys_initcall(memory_tier_init);

Quoted 30 lines · Aneesh Kumar K.V

If the new NUMA node onlined doesn't have a memory tier assigned, the kernel adds the NUMA node to default memory tier.

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 mm/memory-tiers.c | 68 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 68 insertions(+)

mm/memory-tiers.c +7 -0

index 79347d4ab05e..5706ad647136 100644
@@ -5,6 +5,7 @@
 #include <linux/slab.h>
 #include <linux/lockdep.h>
 #include <linux/moduleparam.h>
+#include <linux/memory.h>
 #include <linux/memory-tiers.h>
 
 struct memory_tier {

@@ -130,8 +131,73 @@ int node_create_and_set_memory_tier(int node, int tier)
 }
 EXPORT_SYMBOL_GPL(node_create_and_set_memory_tier);
 
+static int __node_set_memory_tier(int node, int tier)
+{
+	int ret = 0;
+	struct memory_tier *memtier;
+
+	memtier = __get_memory_tier_from_id(tier);

mm/memory-tiers.c int node_create_and_set_memory_tier(int node, int tier) · line 149

Quoted 10 lines

mm/memory-tiers.c +10 -0

+	if (!memtier) {
+		ret = -EINVAL;
+		goto out;
+	}
+	node_set(node, memtier->nodelist);
+out:
+	return ret;
+}
+
+static int node_set_memory_tier(int node, int tier)

Minor comment, but I don't like the name of this function as it doesn't always set the node to the given tier.

Something like this would make it clearer the tier value is only used if the node isn't already assigned to a tier:

static int init_node_memory_tier(int node, int default_tier)

This patch switch the demotion target building logic to use memory tiers instead of NUMA distance. All N_MEMORY NUMA nodes will be placed in the default tier 200 and additional memory tiers will be added by drivers like dax kmem.

This patch builds the demotion target for a NUMA node by looking at all memory tiers below the tier to which the NUMA node belongs. The closest node in the immediately following memory tier is used as a demotion target.

Since we are now only building demotion target for N_MEMORY NUMA nodes the CPU hotplug calls are removed in this patch.

A new memory tier can be inserted into the tier hierarchy for a new set of nodes without affecting the node assignment of any existing memtier, provided that there is enough gap in the tier ID values for the new memtier.

The absolute value of tier ID of a memtier doesn't necessarily carry any meaning. Its value relative to other memtiers decides the level of this memtier in the tier hierarchy.

For now, This patch supports hardcoded tier ID values which are 300, 200 and 100 for memory tiers.

Suggested-by Wei Xu <[email protected]>

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/memory-tiers.h |  13 ++
 include/linux/migrate.h      |  13 --
 mm/memory-tiers.c            | 227 ++++++++++++++++++++
 mm/migrate.c                 | 394 -----------------------------------
 mm/vmstat.c                  |   4 -
 5 files changed, 240 insertions(+), 411 deletions(-)

include/linux/memory-tiers.h +13 -0

index 9d36ff13c954..3234301c2537 100644
@@ -15,6 +15,14 @@
 
 extern bool numa_demotion_enabled;
 int node_create_and_set_memory_tier(int node, int tier);
+#ifdef CONFIG_MIGRATION
+int next_demotion_node(int node);
+#else
+static inline int next_demotion_node(int node)
+{
+	return NUMA_NO_NODE;
+}
+#endif
 
 #else
 

@@ -23,5 +31,10 @@ static inline int node_create_and_set_memory_tier(int node, int tier)
 {
 	return 0;
 }
+
+static inline int next_demotion_node(int node)
+{
+	return NUMA_NO_NODE;
+}
 #endif	/* CONFIG_NUMA */
 #endif  /* _LINUX_MEMORY_TIERS_H */

include/linux/migrate.h +0 -13

index 43e737215f33..93fab62e6548 100644
@@ -75,19 +75,6 @@ static inline int migrate_huge_page_move_mapping(struct address_space *mapping,
 
 #endif /* CONFIG_MIGRATION */
 
-#if defined(CONFIG_MIGRATION) && defined(CONFIG_NUMA)
-extern void set_migration_target_nodes(void);
-extern void migrate_on_reclaim_init(void);
-extern int next_demotion_node(int node);
-#else
-static inline void set_migration_target_nodes(void) {}
-static inline void migrate_on_reclaim_init(void) {}
-static inline int next_demotion_node(int node)
-{
-        return NUMA_NO_NODE;
-}
-#endif
-
 #ifdef CONFIG_COMPACTION
 extern int PageMovable(struct page *page);
 extern void __SetPageMovable(struct page *page, struct address_space *mapping);

mm/memory-tiers.c +155 -0

index 5706ad647136..e951f54ce56c 100644
@@ -6,16 +6,85 @@
 #include <linux/lockdep.h>
 #include <linux/moduleparam.h>
 #include <linux/memory.h>
+#include <linux/random.h>
 #include <linux/memory-tiers.h>
 
+#include "internal.h"
+
 struct memory_tier {
 	struct list_head list;
 	int id;
 	nodemask_t nodelist;
 };
 
+struct demotion_nodes {
+	nodemask_t preferred;
+};
+
+static void establish_migration_targets(void);
 static DEFINE_MUTEX(memory_tier_lock);
 static LIST_HEAD(memory_tiers);
+/*
+ * node_demotion[] examples:
+ *
+ * Example 1:
+ *
+ * Node 0 & 1 are CPU + DRAM nodes, node 2 & 3 are PMEM nodes.
+ *
+ * node distances:
+ * node   0    1    2    3
+ *    0  10   20   30   40
+ *    1  20   10   40   30
+ *    2  30   40   10   40
+ *    3  40   30   40   10
+ *
+ * memory_tiers[0] = <empty>
+ * memory_tiers[1] = 0-1
+ * memory_tiers[2] = 2-3
+ *
+ * node_demotion[0].preferred = 2
+ * node_demotion[1].preferred = 3
+ * node_demotion[2].preferred = <empty>
+ * node_demotion[3].preferred = <empty>
+ *
+ * Example 2:
+ *
+ * Node 0 & 1 are CPU + DRAM nodes, node 2 is memory-only DRAM node.
+ *
+ * node distances:
+ * node   0    1    2
+ *    0  10   20   30
+ *    1  20   10   30
+ *    2  30   30   10
+ *
+ * memory_tiers[0] = <empty>
+ * memory_tiers[1] = 0-2
+ * memory_tiers[2] = <empty>
+ *
+ * node_demotion[0].preferred = <empty>
+ * node_demotion[1].preferred = <empty>
+ * node_demotion[2].preferred = <empty>
+ *
+ * Example 3:
+ *
+ * Node 0 is CPU + DRAM nodes, Node 1 is HBM node, node 2 is PMEM node.
+ *
+ * node distances:
+ * node   0    1    2
+ *    0  10   20   30
+ *    1  20   10   40
+ *    2  30   40   10
+ *
+ * memory_tiers[0] = 1
+ * memory_tiers[1] = 0
+ * memory_tiers[2] = 2
+ *
+ * node_demotion[0].preferred = 2
+ * node_demotion[1].preferred = 0
+ * node_demotion[2].preferred = <empty>
+ *
+ */
+static struct demotion_nodes *node_demotion __read_mostly;
 
 static void insert_memory_tier(struct memory_tier *memtier)
 {

@@ -108,6 +177,7 @@ int node_create_and_set_memory_tier(int node, int tier)
 	current_tier = __node_get_memory_tier(node);
 	if (!current_tier) {
 		ret = __node_create_and_set_memory_tier(node, tier);
+		establish_migration_targets();
 		goto out;
 	}
 

@@ -124,6 +194,8 @@ int node_create_and_set_memory_tier(int node, int tier)
 	}
 	if (nodes_empty(current_tier->nodelist))
 		unregister_memory_tier(current_tier);
+
+	establish_migration_targets();
 out:
 	mutex_unlock(&memory_tier_lock);
 

@@ -153,14 +225,152 @@ static int node_set_memory_tier(int node, int tier)
 
 	mutex_lock(&memory_tier_lock);
 	memtier = __node_get_memory_tier(node);
+	/*
+	 * if node is already part of the tier proceed with the
+	 * current tier value, because we might want to establish
+	 * new migration paths now. The node might be added to a tier
+	 * before it was made part of N_MEMORY, hence estabilish_migration_targets
+	 * will have skipped this node.
+	 */
 	if (!memtier)
 		ret = __node_set_memory_tier(node, tier);
+	establish_migration_targets();
 
 	mutex_unlock(&memory_tier_lock);
 
 	return ret;
 }
 
+#ifdef CONFIG_MIGRATION
+/**
+ * next_demotion_node() - Get the next node in the demotion path
+ * @node: The starting node to lookup the next node
+ *
+ * Return: node id for next memory node in the demotion path hierarchy
+ * from @node; NUMA_NO_NODE if @node is terminal.  This does not keep
+ * @node online or guarantee that it *continues* to be the next demotion
+ * target.
+ */
+int next_demotion_node(int node)
+{
+	struct demotion_nodes *nd;
+	int target;
+
+	if (!node_demotion)
+		return NUMA_NO_NODE;
+
+	nd = &node_demotion[node];
+
+	/*
+	 * node_demotion[] is updated without excluding this
+	 * function from running.
+	 *
+	 * Make sure to use RCU over entire code blocks if
+	 * node_demotion[] reads need to be consistent.
+	 */
+	rcu_read_lock();
+	/*
+	 * If there are multiple target nodes, just select one
+	 * target node randomly.
+	 *
+	 * In addition, we can also use round-robin to select
+	 * target node, but we should introduce another variable
+	 * for node_demotion[] to record last selected target node,
+	 * that may cause cache ping-pong due to the changing of
+	 * last target node. Or introducing per-cpu data to avoid
+	 * caching issue, which seems more complicated. So selecting
+	 * target node randomly seems better until now.
+	 */
+	target = node_random(&nd->preferred);
+	rcu_read_unlock();
+
+	return target;
+}
+
+/* Disable reclaim-based migration. */
+static void __disable_all_migrate_targets(void)
+{
+	int node;
+
+	for_each_node_state(node, N_MEMORY)
+		node_demotion[node].preferred = NODE_MASK_NONE;
+}
+
+static void disable_all_migrate_targets(void)
+{
+	__disable_all_migrate_targets();
+
+	/*
+	 * Ensure that the "disable" is visible across the system.
+	 * Readers will see either a combination of before+disable
+	 * state or disable+after.  They will never see before and
+	 * after state together.
+	 */
+	synchronize_rcu();
+}
+#else
+static void disable_all_migrate_targets(void) {}
+#endif
+
+/*
+ * Find an automatic demotion target for all memory
+ * nodes. Failing here is OK.  It might just indicate
+ * being at the end of a chain.

mm/memory-tiers.c +62 -0

+		return;
+
+	disable_all_migrate_targets();
+
+	for_each_node_state(node, N_MEMORY) {
+		best_distance = -1;
+		nd = &node_demotion[node];
+
+		memtier = __node_get_memory_tier(node);
+		if (!memtier || list_is_last(&memtier->list, &memory_tiers))
+			continue;
+		/*
+		 * Get the next memtier to find the  demotion node list.
+		 */
+		memtier = list_next_entry(memtier, list);
+
+		/*
+		 * find_next_best_node, use 'used' nodemask as a skip list.
+		 * Add all memory nodes except the selected memory tier
+		 * nodelist to skip list so that we find the best node from the
+		 * memtier nodelist.
+		 */
+		nodes_andnot(used, node_states[N_MEMORY], memtier->nodelist);
+
+		/*
+		 * Find all the nodes in the memory tier node list of same best distance.
+		 * add them to the preferred mask. We randomly select between nodes
+		 * in the preferred mask when allocating pages during demotion.
+		 */
+		do {
+			target = find_next_best_node(node, &used);
+			if (target == NUMA_NO_NODE)
+				break;
+
+			distance = node_distance(node, target);
+			if (distance == best_distance || best_distance == -1) {
+				best_distance = distance;
+				node_set(target, nd->preferred);
+			} else {
+				break;
+			}
+		} while (1);
+	}
+}
+
 static unsigned int default_memtier = DEFAULT_MEMORY_TIER;
 core_param(default_memory_tier, default_memtier, uint, 0644);
 /*

@@ -181,6 +391,17 @@ static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
 		return notifier_from_errno(0);
 
 	switch (action) {
+	case MEM_OFFLINE:
+		/*
+		 * In case we are moving out of N_MEMORY. Keep the node
+		 * in the memory tier so that when we bring memory online,
+		 * they appear in the right memory tier. We still need
+		 * to rebuild the demotion order.
+		 */
+		mutex_lock(&memory_tier_lock);
+		establish_migration_targets();
+		mutex_unlock(&memory_tier_lock);
+		break;
 	case MEM_ONLINE:
 		/*
 		 * We ignore the error here, if the node already have the tier

@@ -196,6 +417,12 @@ static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
 
 static void __init migrate_on_reclaim_init(void)
 {
+
+	if (IS_ENABLED(CONFIG_MIGRATION)) {
+		node_demotion = kcalloc(MAX_NUMNODES, sizeof(struct demotion_nodes),
+					GFP_KERNEL);
+		WARN_ON(!node_demotion);
+	}
 	hotplug_memory_notifier(migrate_on_reclaim_callback, 100);
 }
 

mm/migrate.c +0 -394

index fce7d4a9e940..c758c9c21d7d 100644
@@ -2117,398 +2117,4 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
 	return 0;
 }
 #endif /* CONFIG_NUMA_BALANCING */
-
-/*
- * node_demotion[] example:
- *
- * Consider a system with two sockets.  Each socket has
- * three classes of memory attached: fast, medium and slow.
- * Each memory class is placed in its own NUMA node.  The
- * CPUs are placed in the node with the "fast" memory.  The
- * 6 NUMA nodes (0-5) might be split among the sockets like
- * this:
- *
- *	Socket A: 0, 1, 2
- *	Socket B: 3, 4, 5
- *
- * When Node 0 fills up, its memory should be migrated to
- * Node 1.  When Node 1 fills up, it should be migrated to
- * Node 2.  The migration path start on the nodes with the
- * processors (since allocations default to this node) and
- * fast memory, progress through medium and end with the
- * slow memory:
- *
- *	0 -> 1 -> 2 -> stop
- *	3 -> 4 -> 5 -> stop
- *
- * This is represented in the node_demotion[] like this:
- *
- *	{  nr=1, nodes[0]=1 }, // Node 0 migrates to 1
- *	{  nr=1, nodes[0]=2 }, // Node 1 migrates to 2
- *	{  nr=0, nodes[0]=-1 }, // Node 2 does not migrate
- *	{  nr=1, nodes[0]=4 }, // Node 3 migrates to 4
- *	{  nr=1, nodes[0]=5 }, // Node 4 migrates to 5
- *	{  nr=0, nodes[0]=-1 }, // Node 5 does not migrate
- *
- * Moreover some systems may have multiple slow memory nodes.
- * Suppose a system has one socket with 3 memory nodes, node 0
- * is fast memory type, and node 1/2 both are slow memory
- * type, and the distance between fast memory node and slow
- * memory node is same. So the migration path should be:
- *
- *	0 -> 1/2 -> stop
- *
- * This is represented in the node_demotion[] like this:
- *	{ nr=2, {nodes[0]=1, nodes[1]=2} }, // Node 0 migrates to node 1 and node 2
- *	{ nr=0, nodes[0]=-1, }, // Node 1 dose not migrate
- *	{ nr=0, nodes[0]=-1, }, // Node 2 does not migrate
- */
-
-/*
- * Writes to this array occur without locking.  Cycles are
- * not allowed: Node X demotes to Y which demotes to X...
- *
- * If multiple reads are performed, a single rcu_read_lock()
- * must be held over all reads to ensure that no cycles are
- * observed.
- */
-#define DEFAULT_DEMOTION_TARGET_NODES 15
-
-#if MAX_NUMNODES < DEFAULT_DEMOTION_TARGET_NODES
-#define DEMOTION_TARGET_NODES	(MAX_NUMNODES - 1)
-#else
-#define DEMOTION_TARGET_NODES	DEFAULT_DEMOTION_TARGET_NODES
-#endif
-
-struct demotion_nodes {
-	unsigned short nr;
-	short nodes[DEMOTION_TARGET_NODES];
-};
-
-static struct demotion_nodes *node_demotion __read_mostly;
-
-/**
- * next_demotion_node() - Get the next node in the demotion path
- * @node: The starting node to lookup the next node
- *
- * Return: node id for next memory node in the demotion path hierarchy
- * from @node; NUMA_NO_NODE if @node is terminal.  This does not keep
- * @node online or guarantee that it *continues* to be the next demotion
- * target.
- */
-int next_demotion_node(int node)
-{
-	struct demotion_nodes *nd;
-	unsigned short target_nr, index;
-	int target;
-
-	if (!node_demotion)
-		return NUMA_NO_NODE;
-
-	nd = &node_demotion[node];
-
-	/*
-	 * node_demotion[] is updated without excluding this
-	 * function from running.  RCU doesn't provide any
-	 * compiler barriers, so the READ_ONCE() is required
-	 * to avoid compiler reordering or read merging.
-	 *
-	 * Make sure to use RCU over entire code blocks if
-	 * node_demotion[] reads need to be consistent.
-	 */
-	rcu_read_lock();
-	target_nr = READ_ONCE(nd->nr);
-
-	switch (target_nr) {
-	case 0:
-		target = NUMA_NO_NODE;
-		goto out;
-	case 1:
-		index = 0;
-		break;
-	default:
-		/*
-		 * If there are multiple target nodes, just select one
-		 * target node randomly.
-		 *
-		 * In addition, we can also use round-robin to select
-		 * target node, but we should introduce another variable
-		 * for node_demotion[] to record last selected target node,
-		 * that may cause cache ping-pong due to the changing of
-		 * last target node. Or introducing per-cpu data to avoid
-		 * caching issue, which seems more complicated. So selecting
-		 * target node randomly seems better until now.
-		 */
-		index = get_random_int() % target_nr;
-		break;
-	}
-
-	target = READ_ONCE(nd->nodes[index]);
-
-out:
-	rcu_read_unlock();
-	return target;
-}
-
-/* Disable reclaim-based migration. */
-static void __disable_all_migrate_targets(void)
-{
-	int node, i;
-
-	if (!node_demotion)
-		return;
-
-	for_each_online_node(node) {
-		node_demotion[node].nr = 0;
-		for (i = 0; i < DEMOTION_TARGET_NODES; i++)
-			node_demotion[node].nodes[i] = NUMA_NO_NODE;
-	}
-}
-
-static void disable_all_migrate_targets(void)
-{
-	__disable_all_migrate_targets();
-
-	/*
-	 * Ensure that the "disable" is visible across the system.
-	 * Readers will see either a combination of before+disable
-	 * state or disable+after.  They will never see before and
-	 * after state together.
-	 *
-	 * The before+after state together might have cycles and
-	 * could cause readers to do things like loop until this
-	 * function finishes.  This ensures they can only see a
-	 * single "bad" read and would, for instance, only loop
-	 * once.
-	 */
-	synchronize_rcu();
-}
-
-/*
- * Find an automatic demotion target for 'node'.
- * Failing here is OK.  It might just indicate
- * being at the end of a chain.
- */
-static int establish_migrate_target(int node, nodemask_t *used,
-				    int best_distance)
-{
-	int migration_target, index, val;
-	struct demotion_nodes *nd;
-
-	if (!node_demotion)
-		return NUMA_NO_NODE;
-
-	nd = &node_demotion[node];
-
-	migration_target = find_next_best_node(node, used);
-	if (migration_target == NUMA_NO_NODE)
-		return NUMA_NO_NODE;
-
-	/*
-	 * If the node has been set a migration target node before,
-	 * which means it's the best distance between them. Still
-	 * check if this node can be demoted to other target nodes
-	 * if they have a same best distance.
-	 */
-	if (best_distance != -1) {
-		val = node_distance(node, migration_target);
-		if (val > best_distance)
-			goto out_clear;
-	}
-
-	index = nd->nr;
-	if (WARN_ONCE(index >= DEMOTION_TARGET_NODES,
-		      "Exceeds maximum demotion target nodes\n"))
-		goto out_clear;
-
-	nd->nodes[index] = migration_target;
-	nd->nr++;
-
-	return migration_target;
-out_clear:
-	node_clear(migration_target, *used);
-	return NUMA_NO_NODE;
-}
-
-/*
- * When memory fills up on a node, memory contents can be
- * automatically migrated to another node instead of
- * discarded at reclaim.
- *
- * Establish a "migration path" which will start at nodes
- * with CPUs and will follow the priorities used to build the
- * page allocator zonelists.
- *
- * The difference here is that cycles must be avoided.  If
- * node0 migrates to node1, then neither node1, nor anything
- * node1 migrates to can migrate to node0. Also one node can
- * be migrated to multiple nodes if the target nodes all have
- * a same best-distance against the source node.
- *
- * This function can run simultaneously with readers of
- * node_demotion[].  However, it can not run simultaneously
- * with itself.  Exclusion is provided by memory hotplug events
- * being single-threaded.
- */
-static void __set_migration_target_nodes(void)
-{
-	nodemask_t next_pass;
-	nodemask_t this_pass;
-	nodemask_t used_targets = NODE_MASK_NONE;
-	int node, best_distance;
-
-	/*
-	 * Avoid any oddities like cycles that could occur
-	 * from changes in the topology.  This will leave
-	 * a momentary gap when migration is disabled.
-	 */
-	disable_all_migrate_targets();
-
-	/*
-	 * Allocations go close to CPUs, first.  Assume that
-	 * the migration path starts at the nodes with CPUs.
-	 */
-	next_pass = node_states[N_CPU];
-again:
-	this_pass = next_pass;
-	next_pass = NODE_MASK_NONE;
-	/*
-	 * To avoid cycles in the migration "graph", ensure
-	 * that migration sources are not future targets by
-	 * setting them in 'used_targets'.  Do this only
-	 * once per pass so that multiple source nodes can
-	 * share a target node.
-	 *
-	 * 'used_targets' will become unavailable in future
-	 * passes.  This limits some opportunities for
-	 * multiple source nodes to share a destination.
-	 */
-	nodes_or(used_targets, used_targets, this_pass);
-
-	for_each_node_mask(node, this_pass) {
-		best_distance = -1;
-
-		/*
-		 * Try to set up the migration path for the node, and the target
-		 * migration nodes can be multiple, so doing a loop to find all
-		 * the target nodes if they all have a best node distance.
-		 */
-		do {
-			int target_node =
-				establish_migrate_target(node, &used_targets,
-							 best_distance);
-
-			if (target_node == NUMA_NO_NODE)
-				break;
-
-			if (best_distance == -1)
-				best_distance = node_distance(node, target_node);
-
-			/*
-			 * Visit targets from this pass in the next pass.
-			 * Eventually, every node will have been part of
-			 * a pass, and will become set in 'used_targets'.
-			 */
-			node_set(target_node, next_pass);
-		} while (1);
-	}
-	/*
-	 * 'next_pass' contains nodes which became migration
-	 * targets in this pass.  Make additional passes until
-	 * no more migrations targets are available.
-	 */
-	if (!nodes_empty(next_pass))
-		goto again;
-}
-
-/*
- * For callers that do not hold get_online_mems() already.
- */
-void set_migration_target_nodes(void)
-{
-	get_online_mems();
-	__set_migration_target_nodes();
-	put_online_mems();
-}
-
-/*
- * This leaves migrate-on-reclaim transiently disabled between
- * the MEM_GOING_OFFLINE and MEM_OFFLINE events.  This runs
- * whether reclaim-based migration is enabled or not, which
- * ensures that the user can turn reclaim-based migration at
- * any time without needing to recalculate migration targets.
- *
- * These callbacks already hold get_online_mems().  That is why
- * __set_migration_target_nodes() can be used as opposed to
- * set_migration_target_nodes().
- */
-#ifdef CONFIG_MEMORY_HOTPLUG
-static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
-						 unsigned long action, void *_arg)
-{
-	struct memory_notify *arg = _arg;
-
-	/*
-	 * Only update the node migration order when a node is
-	 * changing status, like online->offline.  This avoids
-	 * the overhead of synchronize_rcu() in most cases.
-	 */
-	if (arg->status_change_nid < 0)
-		return notifier_from_errno(0);
-
-	switch (action) {
-	case MEM_GOING_OFFLINE:
-		/*
-		 * Make sure there are not transient states where
-		 * an offline node is a migration target.  This
-		 * will leave migration disabled until the offline
-		 * completes and the MEM_OFFLINE case below runs.
-		 */
-		disable_all_migrate_targets();
-		break;
-	case MEM_OFFLINE:
-	case MEM_ONLINE:
-		/*
-		 * Recalculate the target nodes once the node
-		 * reaches its final state (online or offline).
-		 */
-		__set_migration_target_nodes();
-		break;
-	case MEM_CANCEL_OFFLINE:
-		/*
-		 * MEM_GOING_OFFLINE disabled all the migration
-		 * targets.  Reenable them.
-		 */
-		__set_migration_target_nodes();
-		break;
-	case MEM_GOING_ONLINE:
-	case MEM_CANCEL_ONLINE:
-		break;
-	}
-
-	return notifier_from_errno(0);
-}
-#endif
-
-void __init migrate_on_reclaim_init(void)
-{
-	node_demotion = kcalloc(nr_node_ids,
-				sizeof(struct demotion_nodes),
-				GFP_KERNEL);
-	WARN_ON(!node_demotion);
-#ifdef CONFIG_MEMORY_HOTPLUG
-	hotplug_memory_notifier(migrate_on_reclaim_callback, 100);
-#endif
-	/*
-	 * At this point, all numa nodes with memory/CPus have their state
-	 * properly set, so we can build the demotion order now.
-	 * Let us hold the cpu_hotplug lock just, as we could possibily have
-	 * CPU hotplug events during boot.
-	 */
-	cpus_read_lock();
-	set_migration_target_nodes();
-	cpus_read_unlock();
-}
 #endif /* CONFIG_NUMA */
-
-

mm/vmstat.c +0 -4

index 373d2730fcf2..35c6ff97cf29 100644
@@ -28,7 +28,6 @@
 #include <linux/mm_inline.h>
 #include <linux/page_ext.h>
 #include <linux/page_owner.h>
-#include <linux/migrate.h>
 
 #include "internal.h"
 

@@ -2060,7 +2059,6 @@ static int vmstat_cpu_online(unsigned int cpu)
 
 	if (!node_state(cpu_to_node(cpu), N_CPU)) {
 		node_set_state(cpu_to_node(cpu), N_CPU);
-		set_migration_target_nodes();
 	}
 
 	return 0;

@@ -2085,7 +2083,6 @@ static int vmstat_cpu_dead(unsigned int cpu)
 		return 0;
 
 	node_clear_state(node, N_CPU);
-	set_migration_target_nodes();
 
 	return 0;
 }

@@ -2118,7 +2115,6 @@ void __init init_mm_internals(void)
 
 	start_shepherd_timer();
 #endif
-	migrate_on_reclaim_init();
 #ifdef CONFIG_PROC_FS
 	proc_create_seq("buddyinfo", 0444, NULL, &fragmentation_op);
 	proc_create_seq("pagetypeinfo", 0400, NULL, &pagetypeinfo_op);

+ */

mm/memory-tiers.c static int node_set_memory_tier(int node, int tier) · line 328

Quoted 9 lines

mm/memory-tiers.c +9 -0

+static void establish_migration_targets(void)
+{
+	struct memory_tier *memtier;
+	struct demotion_nodes *nd;
+	int target = NUMA_NO_NODE, node;
+	int distance, best_distance;
+	nodemask_t used;
+
+	if (!node_demotion || !IS_ENABLED(CONFIG_MIGRATION))

Does it make sense to include the memory tiering/demotion code if CONFIG_MIGRATION isn't enabled? From what I can tell none of the information established here is used if CONFIG_MIGRATION isn't enabled, so it would be better to remove the IS_ENABLED checks and not include the code at all.

Quoted 38 lines

mm/memory-tiers.c +37 -0

+		return;
+
+	disable_all_migrate_targets();
+
+	for_each_node_state(node, N_MEMORY) {
+		best_distance = -1;
+		nd = &node_demotion[node];
+
+		memtier = __node_get_memory_tier(node);
+		if (!memtier || list_is_last(&memtier->list, &memory_tiers))
+			continue;
+		/*
+		 * Get the next memtier to find the  demotion node list.
+		 */
+		memtier = list_next_entry(memtier, list);
+
+		/*
+		 * find_next_best_node, use 'used' nodemask as a skip list.
+		 * Add all memory nodes except the selected memory tier
+		 * nodelist to skip list so that we find the best node from the
+		 * memtier nodelist.
+		 */
+		nodes_andnot(used, node_states[N_MEMORY], memtier->nodelist);
+
+		/*
+		 * Find all the nodes in the memory tier node list of same best distance.
+		 * add them to the preferred mask. We randomly select between nodes
+		 * in the preferred mask when allocating pages during demotion.
+		 */
+		do {
+			target = find_next_best_node(node, &used);
+			if (target == NUMA_NO_NODE)
+				break;
+
+			distance = node_distance(node, target);
+			if (distance == best_distance || best_distance == -1) {
+				best_distance = distance;

mm/memory-tiers.c static int node_set_memory_tier(int node, int tier) · line 373

Quoted 8 lines

mm/memory-tiers.c +8 -0

+				node_set(target, nd->preferred);
+			} else {
+				break;
+			}
+		} while (1);
+	}
+}
+

Alistair Popple <[email protected]> writes:
....

Quoted 16 lines

+ */

mm/memory-tiers.c static int node_set_memory_tier(int node, int tier) · line 328

Quoted 9 lines

mm/memory-tiers.c +9 -0

+static void establish_migration_targets(void)
+{
+	struct memory_tier *memtier;
+	struct demotion_nodes *nd;
+	int target = NUMA_NO_NODE, node;
+	int distance, best_distance;
+	nodemask_t used;
+
+	if (!node_demotion || !IS_ENABLED(CONFIG_MIGRATION))

Does it make sense to include the memory tiering/demotion code if CONFIG_MIGRATION isn't enabled? From what I can tell none of the information established here is used if CONFIG_MIGRATION isn't enabled, so it would be better to remove the IS_ENABLED checks and not include the code at all.

We use the same function/codepath for updating top_tier details. We would want to get node_is_toptier() to work even with CONFIG_MIGRATION disabled?

Quoted 38 lines

Quoted 38 lines

mm/memory-tiers.c +37 -0

+		return;
+
+	disable_all_migrate_targets();
+
+	for_each_node_state(node, N_MEMORY) {
+		best_distance = -1;
+		nd = &node_demotion[node];
+
+		memtier = __node_get_memory_tier(node);
+		if (!memtier || list_is_last(&memtier->list, &memory_tiers))
+			continue;
+		/*
+		 * Get the next memtier to find the  demotion node list.
+		 */
+		memtier = list_next_entry(memtier, list);
+
+		/*
+		 * find_next_best_node, use 'used' nodemask as a skip list.
+		 * Add all memory nodes except the selected memory tier
+		 * nodelist to skip list so that we find the best node from the
+		 * memtier nodelist.
+		 */
+		nodes_andnot(used, node_states[N_MEMORY], memtier->nodelist);
+
+		/*
+		 * Find all the nodes in the memory tier node list of same best distance.
+		 * add them to the preferred mask. We randomly select between nodes
+		 * in the preferred mask when allocating pages during demotion.
+		 */
+		do {
+			target = find_next_best_node(node, &used);
+			if (target == NUMA_NO_NODE)
+				break;
+
+			distance = node_distance(node, target);
+			if (distance == best_distance || best_distance == -1) {
+				best_distance = distance;

Quoted 8 lines

mm/memory-tiers.c static int node_set_memory_tier(int node, int tier) · line 373

Quoted 8 lines

mm/memory-tiers.c +8 -0

+				node_set(target, nd->preferred);
+			} else {
+				break;
+			}
+		} while (1);
+	}
+}
+

.....

-aneesh

Also update different helpes to use NODE_DATA()->memtier. Since node specific memtier can change based on the reassignment of NUMA node to a different memory tiers, accessing NODE_DATA()->memtier needs to happen under an rcu read lock or memory_tier_lock.

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/mmzone.h |  3 ++
 mm/memory-tiers.c      | 64 +++++++++++++++++++++++++++++++-----------
 2 files changed, 50 insertions(+), 17 deletions(-)

include/linux/mmzone.h +3 -0

index aab70355d64f..353812495a70 100644
@@ -928,6 +928,9 @@ typedef struct pglist_data {
 	/* Per-node vmstats */
 	struct per_cpu_nodestat __percpu *per_cpu_nodestats;
 	atomic_long_t		vm_stat[NR_VM_NODE_STAT_ITEMS];
+#ifdef CONFIG_NUMA
+	struct memory_tier __rcu *memtier;
+#endif
 } pg_data_t;
 
 #define node_present_pages(nid)	(NODE_DATA(nid)->node_present_pages)

mm/memory-tiers.c +25 -7

index e951f54ce56c..bab4700bf58d 100644
@@ -7,6 +7,7 @@
 #include <linux/moduleparam.h>
 #include <linux/memory.h>
 #include <linux/random.h>
+#include <linux/rcupdate.h>
 #include <linux/memory-tiers.h>
 
 #include "internal.h"

@@ -124,18 +125,23 @@ static struct memory_tier *register_memory_tier(unsigned int tier)
 static void unregister_memory_tier(struct memory_tier *memtier)
 {
 	list_del(&memtier->list);
-	kfree(memtier);
+	kfree_rcu(memtier);
 }
 
 static struct memory_tier *__node_get_memory_tier(int node)
 {
-	struct memory_tier *memtier;
+	pg_data_t *pgdat;
 
-	list_for_each_entry(memtier, &memory_tiers, list) {
-		if (node_isset(node, memtier->nodelist))
-			return memtier;
-	}
-	return NULL;
+	pgdat = NODE_DATA(node);
+	if (!pgdat)
+		return NULL;
+	/*
+	 * Since we hold memory_tier_lock, we can avoid
+	 * RCU read locks when accessing the details. No
+	 * parallel updates are possible here.
+	 */
+	return rcu_dereference_check(pgdat->memtier,
+				     lockdep_is_held(&memory_tier_lock));
 }
 
 static struct memory_tier *__get_memory_tier_from_id(int id)

@@ -149,6 +155,33 @@ static struct memory_tier *__get_memory_tier_from_id(int id)
 	return NULL;
 }
 
+/*
+ * Called with memory_tier_lock. Hence the device references cannot
+ * be dropped during this function.
+ */
+static void memtier_node_set(int node, struct memory_tier *memtier)
+{
+	pg_data_t *pgdat;
+	struct memory_tier *current_memtier;
+
+	pgdat = NODE_DATA(node);
+	if (!pgdat)
+		return;

mm/memory-tiers.c +8 -9

+	synchronize_rcu();
+	node_set(node, memtier->nodelist);
+	rcu_assign_pointer(pgdat->memtier, memtier);
+}
+
 static int __node_create_and_set_memory_tier(int node, int tier)
 {
 	int ret = 0;

@@ -162,7 +195,7 @@ static int __node_create_and_set_memory_tier(int node, int tier)
 			goto out;
 		}
 	}
-	node_set(node, memtier->nodelist);
+	memtier_node_set(node, memtier);
 out:
 	return ret;
 }

@@ -184,14 +217,7 @@ int node_create_and_set_memory_tier(int node, int tier)
 	if (current_tier->id == tier)
 		goto out;
 
-	node_clear(node, current_tier->nodelist);
-
 	ret = __node_create_and_set_memory_tier(node, tier);
-	if (ret) {
-		/* reset it back to older tier */
-		node_set(node, current_tier->nodelist);
-		goto out;
-	}
 	if (nodes_empty(current_tier->nodelist))
 		unregister_memory_tier(current_tier);
 

@@ -213,7 +239,7 @@ static int __node_set_memory_tier(int node, int tier)
 		ret = -EINVAL;
 		goto out;
 	}
-	node_set(node, memtier->nodelist);
+	memtier_node_set(node, memtier);
 out:
 	return ret;
 }

@@ -428,6 +454,7 @@ static void __init migrate_on_reclaim_init(void)
 
 static int __init memory_tier_init(void)
 {
+	int node;
 	struct memory_tier *memtier;
 
 	/*

mm/memory-tiers.c +1 -0

+	}
 	mutex_unlock(&memory_tier_lock);
 
 	migrate_on_reclaim_init();

Quoted 91 lines · Aneesh Kumar K.V

Also update different helpes to use NODE_DATA()->memtier. Since node specific memtier can change based on the reassignment of NUMA node to a different memory tiers, accessing NODE_DATA()->memtier needs to happen under an rcu read lock or memory_tier_lock.

Signed-off-by Aneesh Kumar K.V <[email protected]>

---
 include/linux/mmzone.h |  3 ++
 mm/memory-tiers.c      | 64 +++++++++++++++++++++++++++++++-----------
 2 files changed, 50 insertions(+), 17 deletions(-)

include/linux/mmzone.h +3 -0

index aab70355d64f..353812495a70 100644
@@ -928,6 +928,9 @@ typedef struct pglist_data {
 	/* Per-node vmstats */
 	struct per_cpu_nodestat __percpu *per_cpu_nodestats;
 	atomic_long_t		vm_stat[NR_VM_NODE_STAT_ITEMS];
+#ifdef CONFIG_NUMA
+	struct memory_tier __rcu *memtier;
+#endif
 } pg_data_t;
 
 #define node_present_pages(nid)	(NODE_DATA(nid)->node_present_pages)

mm/memory-tiers.c +32 -7

index e951f54ce56c..bab4700bf58d 100644
@@ -7,6 +7,7 @@
 #include <linux/moduleparam.h>
 #include <linux/memory.h>
 #include <linux/random.h>
+#include <linux/rcupdate.h>
 #include <linux/memory-tiers.h>
 
 #include "internal.h"

@@ -124,18 +125,23 @@ static struct memory_tier *register_memory_tier(unsigned int tier)
 static void unregister_memory_tier(struct memory_tier *memtier)
 {
 	list_del(&memtier->list);
-	kfree(memtier);
+	kfree_rcu(memtier);
 }
 
 static struct memory_tier *__node_get_memory_tier(int node)
 {
-	struct memory_tier *memtier;
+	pg_data_t *pgdat;
 
-	list_for_each_entry(memtier, &memory_tiers, list) {
-		if (node_isset(node, memtier->nodelist))
-			return memtier;
-	}
-	return NULL;
+	pgdat = NODE_DATA(node);
+	if (!pgdat)
+		return NULL;
+	/*
+	 * Since we hold memory_tier_lock, we can avoid
+	 * RCU read locks when accessing the details. No
+	 * parallel updates are possible here.
+	 */
+	return rcu_dereference_check(pgdat->memtier,
+				     lockdep_is_held(&memory_tier_lock));
 }
 
 static struct memory_tier *__get_memory_tier_from_id(int id)

@@ -149,6 +155,33 @@ static struct memory_tier *__get_memory_tier_from_id(int id)
 	return NULL;
 }
 
+/*
+ * Called with memory_tier_lock. Hence the device references cannot
+ * be dropped during this function.
+ */
+static void memtier_node_set(int node, struct memory_tier *memtier)
+{
+	pg_data_t *pgdat;
+	struct memory_tier *current_memtier;
+
+	pgdat = NODE_DATA(node);
+	if (!pgdat)
+		return;
+	/*
+	 * Make sure we mark the memtier NULL before we assign the new memory tier
+	 * to the NUMA node. This make sure that anybody looking at NODE_DATA
+	 * finds a NULL memtier or the one which is still valid.
+	 */
+	current_memtier = rcu_dereference_check(pgdat->memtier,
+						lockdep_is_held(&memory_tier_lock));

mm/memory-tiers.c static struct memory_tier *__get_memory_tier_from_id(int id) · line 179

Quoted 3 lines

mm/memory-tiers.c +3 -0

+	rcu_assign_pointer(pgdat->memtier, NULL);
+	if (current_memtier)
+		node_clear(node, current_memtier->nodelist);

It seems odd to me that you would update the current memtier prior to the synchronize_rcu(). I suppose it's really memory_tier_lock that protects the details like ->nodelist, but is there any reason not do the update after anyway?

Quoted 53 lines

mm/memory-tiers.c +8 -9

+	synchronize_rcu();
+	node_set(node, memtier->nodelist);
+	rcu_assign_pointer(pgdat->memtier, memtier);
+}
+
 static int __node_create_and_set_memory_tier(int node, int tier)
 {
 	int ret = 0;

@@ -162,7 +195,7 @@ static int __node_create_and_set_memory_tier(int node, int tier)
 			goto out;
 		}
 	}
-	node_set(node, memtier->nodelist);
+	memtier_node_set(node, memtier);
 out:
 	return ret;
 }

@@ -184,14 +217,7 @@ int node_create_and_set_memory_tier(int node, int tier)
 	if (current_tier->id == tier)
 		goto out;
 
-	node_clear(node, current_tier->nodelist);
-
 	ret = __node_create_and_set_memory_tier(node, tier);
-	if (ret) {
-		/* reset it back to older tier */
-		node_set(node, current_tier->nodelist);
-		goto out;
-	}
 	if (nodes_empty(current_tier->nodelist))
 		unregister_memory_tier(current_tier);
 

@@ -213,7 +239,7 @@ static int __node_set_memory_tier(int node, int tier)
 		ret = -EINVAL;
 		goto out;
 	}
-	node_set(node, memtier->nodelist);
+	memtier_node_set(node, memtier);
 out:
 	return ret;
 }

@@ -428,6 +454,7 @@ static void __init migrate_on_reclaim_init(void)
 
 static int __init memory_tier_init(void)
 {
+	int node;
 	struct memory_tier *memtier;
 
 	/*

@@ -444,7 +471,10 @@ static int __init memory_tier_init(void)
 		      __func__, PTR_ERR(memtier));