[PATCH v5 0/3] Add interfaces for ACPI MRRM table

[PATCH v5 0/3] Add interfaces for ACPI MRRM table

[Tony Luck]

This series based on:
	https://web.git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/log/?h=testing

Memory used to be homogeneous. Then NUMA came along. Later different
types of memory (persistent memory, on-package high bandwidth memory,
CXL attached memory).

Each type of memory has its own performance characteristics, and users
will need to monitor and control access by type.

The MRRM solution is to tag physical address ranges with "region IDs"
so that platform firmware[1] can indicate the type of memory for each
range (with separate tags available for local vs. remote access to
each range). Note that these ranges can include addresses reserved
for future hotplugged memory.

The region IDs will be used to provide separate event counts for each
region for "perf" and for the "resctrl" file system to monitor and
control memory bandwidth in each region.

Users will need to know the address range(s) that are part of each
region. This patch series adds
	/sys/firmware/acpi/memory_ranges/rangeX
directories to provide user space accessible enumeration.

-Tony

[1] MRRM definition allow for future expansion for the OS to assign
these region IDs.

Changes since version 4 here:
https://lore.kernel.org/all/20250429202412.380637-1-tony.luck@intel.com/

*) Dropped patch 1. ACPICA changes have been merged into the linux-pm
   tree.
*) Removed spurious blank file from mrrm_init() (Thanks, Fenghua).

Tony Luck (3):
  ACPI/MRRM: Minimal parse of ACPI MRRM table
  ACPI/MRRM: Add /sys files to describe memory ranges
  ACPI: Add documentation for exposing MRRM data

 include/linux/acpi.h                          |   9 +
 drivers/acpi/acpi_mrrm.c                      | 182 ++++++++++++++++++
 Documentation/ABI/testing/sysfs-firmware-acpi |  21 ++
 arch/x86/Kconfig                              |   1 +
 drivers/acpi/Kconfig                          |   3 +
 drivers/acpi/Makefile                         |   1 +
 6 files changed, 217 insertions(+)
 create mode 100644 drivers/acpi/acpi_mrrm.c


base-repo: git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git
base-branch: testing
base-commit: 70cb3b9a371fe9ff4f50cd7889763abd4ab621dc
-- 
2.48.1

[PATCH v5 1/3] ACPI/MRRM: Minimal parse of ACPI MRRM table

[Tony Luck]

The resctrl file system code needs to know how many region tags
are supported. Parse the ACPI MRRM table and save the max_mem_region
value.

Provide a function for resctrl to collect that value.

Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 include/linux/acpi.h     |  9 +++++++++
 drivers/acpi/acpi_mrrm.c | 42 ++++++++++++++++++++++++++++++++++++++++
 arch/x86/Kconfig         |  1 +
 drivers/acpi/Kconfig     |  3 +++
 drivers/acpi/Makefile    |  1 +
 5 files changed, 56 insertions(+)
 create mode 100644 drivers/acpi/acpi_mrrm.c

diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index fc372bbaa547..f2d8579de539 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -772,6 +772,10 @@ int acpi_get_local_u64_address(acpi_handle handle, u64 *addr);
 int acpi_get_local_address(acpi_handle handle, u32 *addr);
 const char *acpi_get_subsystem_id(acpi_handle handle);
 
+#ifdef CONFIG_ACPI_MRRM
+int acpi_mrrm_max_mem_region(void);
+#endif
+
 #else	/* !CONFIG_ACPI */
 
 #define acpi_disabled 1
@@ -1092,6 +1096,11 @@ static inline acpi_handle acpi_get_processor_handle(int cpu)
 	return NULL;
 }
 
+static inline int acpi_mrrm_max_mem_region(void)
+{
+	return -ENOENT;
+}
+
 #endif	/* !CONFIG_ACPI */
 
 #ifdef CONFIG_ACPI_HMAT
diff --git a/drivers/acpi/acpi_mrrm.c b/drivers/acpi/acpi_mrrm.c
new file mode 100644
index 000000000000..ab8022e58da5
--- /dev/null
+++ b/drivers/acpi/acpi_mrrm.c
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2025, Intel Corporation.
+ *
+ * Memory Range and Region Mapping (MRRM) structure
+ */
+
+#define pr_fmt(fmt) "acpi/mrrm: " fmt
+
+#include <linux/acpi.h>
+#include <linux/init.h>
+
+static int max_mem_region = -ENOENT;
+
+/* Access for use by resctrl file system */
+int acpi_mrrm_max_mem_region(void)
+{
+	return max_mem_region;
+}
+
+static __init int acpi_parse_mrrm(struct acpi_table_header *table)
+{
+	struct acpi_table_mrrm *mrrm;
+
+	mrrm = (struct acpi_table_mrrm *)table;
+	if (!mrrm)
+		return -ENODEV;
+
+	max_mem_region = mrrm->max_mem_region;
+
+	return 0;
+}
+
+static __init int mrrm_init(void)
+{
+	int ret;
+
+	ret = acpi_table_parse(ACPI_SIG_MRRM, acpi_parse_mrrm);
+
+	return ret;
+}
+device_initcall(mrrm_init);
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 4b9f378e05f6..0e43069082df 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -38,6 +38,7 @@ config X86_64
 	select ARCH_HAS_ELFCORE_COMPAT
 	select ZONE_DMA32
 	select EXECMEM if DYNAMIC_FTRACE
+	select ACPI_MRRM if ACPI
 
 config FORCE_DYNAMIC_FTRACE
 	def_bool y
diff --git a/drivers/acpi/Kconfig b/drivers/acpi/Kconfig
index 7f10aa38269d..7bc40c2735ac 100644
--- a/drivers/acpi/Kconfig
+++ b/drivers/acpi/Kconfig
@@ -576,6 +576,9 @@ config ACPI_FFH
 	  Enable this feature if you want to set up and install the FFH Address
 	  Space handler to handle FFH OpRegion in the firmware.
 
+config ACPI_MRRM
+	bool
+
 source "drivers/acpi/pmic/Kconfig"
 
 config ACPI_VIOT
diff --git a/drivers/acpi/Makefile b/drivers/acpi/Makefile
index 797070fc9a3f..d1b0affb844f 100644
--- a/drivers/acpi/Makefile
+++ b/drivers/acpi/Makefile
@@ -66,6 +66,7 @@ acpi-$(CONFIG_ACPI_WATCHDOG)	+= acpi_watchdog.o
 acpi-$(CONFIG_ACPI_PRMT)	+= prmt.o
 acpi-$(CONFIG_ACPI_PCC)		+= acpi_pcc.o
 acpi-$(CONFIG_ACPI_FFH)		+= acpi_ffh.o
+acpi-$(CONFIG_ACPI_MRRM)	+= acpi_mrrm.o
 
 # Address translation
 acpi-$(CONFIG_ACPI_ADXL)	+= acpi_adxl.o
-- 
2.48.1

[PATCH v5 2/3] ACPI/MRRM: Add /sys files to describe memory ranges

[Tony Luck]

Perf and resctrl users need an enumeration of which memory addresses
are bound to which "region" tag.

Parse the ACPI MRRM table and add /sys entries for each memory range
describing base address, length, NUMA node, and which region tags apply
for same-socket and cross-socket access.

Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 drivers/acpi/acpi_mrrm.c | 142 ++++++++++++++++++++++++++++++++++++++-
 1 file changed, 141 insertions(+), 1 deletion(-)

diff --git a/drivers/acpi/acpi_mrrm.c b/drivers/acpi/acpi_mrrm.c
index ab8022e58da5..2f22f013959a 100644
--- a/drivers/acpi/acpi_mrrm.c
+++ b/drivers/acpi/acpi_mrrm.c
@@ -3,12 +3,16 @@
  * Copyright (c) 2025, Intel Corporation.
  *
  * Memory Range and Region Mapping (MRRM) structure
+ *
+ * Parse and report the platform's MRRM table in /sys.
  */
 
 #define pr_fmt(fmt) "acpi/mrrm: " fmt
 
 #include <linux/acpi.h>
 #include <linux/init.h>
+#include <linux/string.h>
+#include <linux/sysfs.h>
 
 static int max_mem_region = -ENOENT;
 
@@ -18,25 +22,161 @@ int acpi_mrrm_max_mem_region(void)
 	return max_mem_region;
 }
 
+struct mrrm_mem_range_entry {
+	u64 base;
+	u64 length;
+	int node;
+	u8  local_region_id;
+	u8  remote_region_id;
+};
+
+static struct mrrm_mem_range_entry *mrrm_mem_range_entry;
+static u32 mrrm_mem_entry_num;
+
+static int get_node_num(struct mrrm_mem_range_entry *e)
+{
+	unsigned int nid;
+
+	for_each_online_node(nid) {
+		for (int z = 0; z < MAX_NR_ZONES; z++) {
+			struct zone *zone = NODE_DATA(nid)->node_zones + z;
+
+			if (!populated_zone(zone))
+				continue;
+			if (zone_intersects(zone, PHYS_PFN(e->base), PHYS_PFN(e->length)))
+				return zone_to_nid(zone);
+		}
+	}
+
+	return -ENOENT;
+}
+
 static __init int acpi_parse_mrrm(struct acpi_table_header *table)
 {
+	struct acpi_mrrm_mem_range_entry *mre_entry;
 	struct acpi_table_mrrm *mrrm;
+	void *mre, *mrrm_end;
+	int mre_count = 0;
 
 	mrrm = (struct acpi_table_mrrm *)table;
 	if (!mrrm)
 		return -ENODEV;
 
+	if (mrrm->flags & ACPI_MRRM_FLAGS_REGION_ASSIGNMENT_OS)
+		return -EOPNOTSUPP;
+
+	mrrm_end = (void *)mrrm + mrrm->header.length - 1;
+	mre = (void *)mrrm + sizeof(struct acpi_table_mrrm);
+	while (mre < mrrm_end) {
+		mre_entry = mre;
+		mre_count++;
+		mre += mre_entry->header.length;
+	}
+	if (!mre_count) {
+		pr_info(FW_BUG "No ranges listed in MRRM table\n");
+		return -EINVAL;
+	}
+
+	mrrm_mem_range_entry = kmalloc_array(mre_count, sizeof(*mrrm_mem_range_entry),
+					     GFP_KERNEL | __GFP_ZERO);
+	if (!mrrm_mem_range_entry)
+		return -ENOMEM;
+
+	mre = (void *)mrrm + sizeof(struct acpi_table_mrrm);
+	while (mre < mrrm_end) {
+		struct mrrm_mem_range_entry *e;
+
+		mre_entry = mre;
+		e = mrrm_mem_range_entry + mrrm_mem_entry_num;
+
+		e->base = mre_entry->addr_base;
+		e->length = mre_entry->addr_len;
+		e->node = get_node_num(e);
+
+		if (mre_entry->region_id_flags & ACPI_MRRM_VALID_REGION_ID_FLAGS_LOCAL)
+			e->local_region_id = mre_entry->local_region_id;
+		else
+			e->local_region_id = -1;
+		if (mre_entry->region_id_flags & ACPI_MRRM_VALID_REGION_ID_FLAGS_REMOTE)
+			e->remote_region_id = mre_entry->remote_region_id;
+		else
+			e->remote_region_id = -1;
+
+		mrrm_mem_entry_num++;
+		mre += mre_entry->header.length;
+	}
+
 	max_mem_region = mrrm->max_mem_region;
 
 	return 0;
 }
 
+#define RANGE_ATTR(name, fmt)						\
+static ssize_t name##_show(struct kobject *kobj,			\
+			  struct kobj_attribute *attr, char *buf)	\
+{									\
+	struct mrrm_mem_range_entry *mre;				\
+	const char *kname = kobject_name(kobj);				\
+	int n, ret;							\
+									\
+	ret = kstrtoint(kname + 5, 10, &n);				\
+	if (ret)							\
+		return ret;						\
+									\
+	mre = mrrm_mem_range_entry + n;					\
+									\
+	return sysfs_emit(buf, fmt, mre->name);				\
+}									\
+static struct kobj_attribute name##_attr = __ATTR_RO(name)
+
+RANGE_ATTR(base, "0x%llx\n");
+RANGE_ATTR(length, "0x%llx\n");
+RANGE_ATTR(node, "%d\n");
+RANGE_ATTR(local_region_id, "%d\n");
+RANGE_ATTR(remote_region_id, "%d\n");
+
+static struct attribute *memory_range_attrs[] = {
+	&base_attr.attr,
+	&length_attr.attr,
+	&node_attr.attr,
+	&local_region_id_attr.attr,
+	&remote_region_id_attr.attr,
+	NULL
+};
+
+ATTRIBUTE_GROUPS(memory_range);
+
+static __init int add_boot_memory_ranges(void)
+{
+	struct kobject *pkobj, *kobj;
+	int ret = -EINVAL;
+	char *name;
+
+	pkobj = kobject_create_and_add("memory_ranges", acpi_kobj);
+
+	for (int i = 0; i < mrrm_mem_entry_num; i++) {
+		name = kasprintf(GFP_KERNEL, "range%d", i);
+		if (!name)
+			break;
+
+		kobj = kobject_create_and_add(name, pkobj);
+
+		ret = sysfs_create_groups(kobj, memory_range_groups);
+		if (ret)
+			return ret;
+	}
+
+	return ret;
+}
+
 static __init int mrrm_init(void)
 {
 	int ret;
 
 	ret = acpi_table_parse(ACPI_SIG_MRRM, acpi_parse_mrrm);
+	if (ret < 0)
+		return ret;
 
-	return ret;
+	return add_boot_memory_ranges();
 }
 device_initcall(mrrm_init);
-- 
2.48.1

[PATCH v5 3/3] ACPI: Add documentation for exposing MRRM data

[Tony Luck]

Initial implementation provides enumeration of the address ranges
NUMA node numbers, and BIOS assigned region IDs for each range.

Signed-off-by: Tony Luck <tony.luck@intel.com>
---
 Documentation/ABI/testing/sysfs-firmware-acpi | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-firmware-acpi b/Documentation/ABI/testing/sysfs-firmware-acpi
index 5249ad5a96d9..fffba38f9ce1 100644
--- a/Documentation/ABI/testing/sysfs-firmware-acpi
+++ b/Documentation/ABI/testing/sysfs-firmware-acpi
@@ -248,3 +248,24 @@ Description:
 		  # cat ff_pwr_btn
 		  7	enabled
 
+What:		/sys/firmware/acpi/memory_ranges/rangeX
+Date:		February 2025
+Contact:	Tony Luck <tony.luck@intel.com>
+Description:
+		On systems with the ACPI MRRM table reports the
+		parameters for each range.
+
+		base: Starting system physical address.
+
+		length: Length of this range in bytes.
+
+		node: NUMA node that this range belongs to. Negative numbers
+		indicate that the node number could not be determined (e.g
+		for an address range that is reserved for future hot add of
+		memory).
+
+		local_region_id: ID associated with access by agents
+		local to this range of addresses.
+
+		remote_region_id: ID associated with access by agents
+		non-local to this range of addresses.
-- 
2.48.1

Re: [PATCH v5 0/3] Add interfaces for ACPI MRRM table

[Rafael J. Wysocki]

On Mon, May 5, 2025 at 7:38 PM Tony Luck <tony.luck@intel.com> wrote:
>
> This series based on:
>         https://web.git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/log/?h=testing
>
> Memory used to be homogeneous. Then NUMA came along. Later different
> types of memory (persistent memory, on-package high bandwidth memory,
> CXL attached memory).
>
> Each type of memory has its own performance characteristics, and users
> will need to monitor and control access by type.
>
> The MRRM solution is to tag physical address ranges with "region IDs"
> so that platform firmware[1] can indicate the type of memory for each
> range (with separate tags available for local vs. remote access to
> each range). Note that these ranges can include addresses reserved
> for future hotplugged memory.
>
> The region IDs will be used to provide separate event counts for each
> region for "perf" and for the "resctrl" file system to monitor and
> control memory bandwidth in each region.
>
> Users will need to know the address range(s) that are part of each
> region. This patch series adds
>         /sys/firmware/acpi/memory_ranges/rangeX
> directories to provide user space accessible enumeration.
>
> -Tony
>
> [1] MRRM definition allow for future expansion for the OS to assign
> these region IDs.
>
> Changes since version 4 here:
> https://lore.kernel.org/all/20250429202412.380637-1-tony.luck@intel.com/
>
> *) Dropped patch 1. ACPICA changes have been merged into the linux-pm
>    tree.
> *) Removed spurious blank file from mrrm_init() (Thanks, Fenghua).
>
> Tony Luck (3):
>   ACPI/MRRM: Minimal parse of ACPI MRRM table
>   ACPI/MRRM: Add /sys files to describe memory ranges
>   ACPI: Add documentation for exposing MRRM data
>
>  include/linux/acpi.h                          |   9 +
>  drivers/acpi/acpi_mrrm.c                      | 182 ++++++++++++++++++
>  Documentation/ABI/testing/sysfs-firmware-acpi |  21 ++
>  arch/x86/Kconfig                              |   1 +
>  drivers/acpi/Kconfig                          |   3 +
>  drivers/acpi/Makefile                         |   1 +
>  6 files changed, 217 insertions(+)
>  create mode 100644 drivers/acpi/acpi_mrrm.c
>

All patches applied as 6.16 material, thanks!

RE: [PATCH v5 1/3] ACPI/MRRM: Minimal parse of ACPI MRRM table

[Luck, Tony]

> diff --git a/drivers/acpi/acpi_mrrm.c b/drivers/acpi/acpi_mrrm.c
> new file mode 100644
> index 000000000000..ab8022e58da5
> --- /dev/null
> +++ b/drivers/acpi/acpi_mrrm.c
> @@ -0,0 +1,42 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (c) 2025, Intel Corporation.
> + *
> + * Memory Range and Region Mapping (MRRM) structure
> + */
> +
> +#define pr_fmt(fmt) "acpi/mrrm: " fmt
> +
> +#include <linux/acpi.h>
> +#include <linux/init.h>
> +
> +static int max_mem_region = -ENOENT;
> This should be initialized to 1 instead of -ENOENT(-2).
> Per the spec, sec 5.1, if platform does not define MRRM, it says, SW may assume that only one memory region is defined.
>
> +
> +/* Access for use by resctrl file system */
> +int acpi_mrrm_max_mem_region(void)
> +{
> +     return max_mem_region;
> +}
> In which case this function will return a minimum of 1 region as max_mem_region of -2 has no meaning.
> When resctrl file system calls this function, it expects at least one region exists.

[Replying to HTML e-mail that will have been dropped from LKML and other lists]

Anil,

Rafael has this queued for next merge window. Please send a patch to
fix the issues you found.

Thanks

-Tony