Add ACPI NUMA support for RISC-V

Add acpi_numa.c file to enable parse NUMA information from ACPI SRAT and SLIT tables. SRAT table provide CPUs(Hart) and memory nodes to proximity domain mapping, while SLIT table provide the distance metrics between proximity domains.

Signed-off-by Haibo Xu <[email protected]>

---
 arch/riscv/include/asm/acpi.h |  15 +++-
 arch/riscv/kernel/Makefile    |   1 +
 arch/riscv/kernel/acpi.c      |   5 --
 arch/riscv/kernel/acpi_numa.c | 131 ++++++++++++++++++++++++++++++++++
 arch/riscv/kernel/setup.c     |   4 +-
 arch/riscv/kernel/smpboot.c   |   2 -
 include/linux/acpi.h          |   6 ++
 7 files changed, 154 insertions(+), 10 deletions(-)
 create mode 100644 arch/riscv/kernel/acpi_numa.c

arch/riscv/include/asm/acpi.h +13 -2

index 7dad0cf9d701..e0a1f84404f3 100644
@@ -61,11 +61,14 @@ static inline void arch_fix_phys_package_id(int num, u32 slot) { }
 
 void acpi_init_rintc_map(void);
 struct acpi_madt_rintc *acpi_cpu_get_madt_rintc(int cpu);
-u32 get_acpi_id_for_cpu(int cpu);
+static inline u32 get_acpi_id_for_cpu(int cpu)
+{
+	return acpi_cpu_get_madt_rintc(cpu)->uid;
+}
+
 int acpi_get_riscv_isa(struct acpi_table_header *table,
 		       unsigned int cpu, const char **isa);
 
-static inline int acpi_numa_get_nid(unsigned int cpu) { return NUMA_NO_NODE; }
 void acpi_get_cbo_block_size(struct acpi_table_header *table, u32 *cbom_size,
 			     u32 *cboz_size, u32 *cbop_size);
 #else

@@ -87,4 +90,12 @@ static inline void acpi_get_cbo_block_size(struct acpi_table_header *table,
 
 #endif /* CONFIG_ACPI */
 
+#ifdef CONFIG_ACPI_NUMA
+int acpi_numa_get_nid(unsigned int cpu);
+void acpi_map_cpus_to_nodes(void);
+#else
+static inline int acpi_numa_get_nid(unsigned int cpu) { return NUMA_NO_NODE; }
+static inline void acpi_map_cpus_to_nodes(void) { }
+#endif /* CONFIG_ACPI_NUMA */
+
 #endif /*_ASM_ACPI_H*/

arch/riscv/kernel/Makefile +1 -0

index f71910718053..5d3e9cf89b76 100644
@@ -105,3 +105,4 @@ obj-$(CONFIG_COMPAT)		+= compat_vdso/
 
 obj-$(CONFIG_64BIT)		+= pi/
 obj-$(CONFIG_ACPI)		+= acpi.o
+obj-$(CONFIG_ACPI_NUMA)	+= acpi_numa.o

arch/riscv/kernel/acpi.c +0 -5

index e619edc8b0cc..040bdbfea2b4 100644
@@ -191,11 +191,6 @@ struct acpi_madt_rintc *acpi_cpu_get_madt_rintc(int cpu)
 	return &cpu_madt_rintc[cpu];
 }
 
-u32 get_acpi_id_for_cpu(int cpu)
-{
-	return acpi_cpu_get_madt_rintc(cpu)->uid;
-}
-
 /*
  * __acpi_map_table() will be called before paging_init(), so early_ioremap()
  * or early_memremap() should be called here to for ACPI table mapping.

arch/riscv/kernel/acpi_numa.c new file +131 -0

new file mode 100644
index 000000000000..0231482d6946
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * ACPI 6.6 based NUMA setup for RISCV
+ * Lots of code was borrowed from arch/arm64/kernel/acpi_numa.c
+ *
+ * Copyright 2004 Andi Kleen, SuSE Labs.
+ * Copyright (C) 2013-2016, Linaro Ltd.
+ *		Author: Hanjun Guo <[email protected]>
+ * Copyright (C) 2024 Intel Corporation.
+ *
+ * Reads the ACPI SRAT table to figure out what memory belongs to which CPUs.
+ *
+ * Called from acpi_numa_init while reading the SRAT and SLIT tables.
+ * Assumes all memory regions belonging to a single proximity domain
+ * are in one chunk. Holes between them will be included in the node.
+ */
+
+#define pr_fmt(fmt) "ACPI: NUMA: " fmt
+
+#include <linux/acpi.h>
+#include <linux/bitmap.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/memblock.h>
+#include <linux/mmzone.h>
+#include <linux/module.h>
+#include <linux/topology.h>
+
+#include <asm/numa.h>
+
+static int acpi_early_node_map[NR_CPUS] __initdata = { NUMA_NO_NODE };
+
+int __init acpi_numa_get_nid(unsigned int cpu)
+{
+	return acpi_early_node_map[cpu];
+}
+
+static inline int get_cpu_for_acpi_id(u32 uid)
+{
+	int cpu;
+
+	for (cpu = 0; cpu < nr_cpu_ids; cpu++)
+		if (uid == get_acpi_id_for_cpu(cpu))
+			return cpu;
+
+	return -EINVAL;
+}
+
+static int __init acpi_parse_rintc_pxm(union acpi_subtable_headers *header,
+				       const unsigned long end)
+{
+	struct acpi_srat_rintc_affinity *pa;
+	int cpu, pxm, node;
+
+	if (srat_disabled())
+		return -EINVAL;
+
+	pa = (struct acpi_srat_rintc_affinity *)header;
+	if (!pa)
+		return -EINVAL;
+
+	if (!(pa->flags & ACPI_SRAT_RINTC_ENABLED))
+		return 0;
+
+	pxm = pa->proximity_domain;
+	node = pxm_to_node(pxm);
+
+	/*
+	 * If we can't map the UID to a logical cpu this
+	 * means that the UID is not part of possible cpus
+	 * so we do not need a NUMA mapping for it, skip
+	 * the SRAT entry and keep parsing.
+	 */
+	cpu = get_cpu_for_acpi_id(pa->acpi_processor_uid);
+	if (cpu < 0)
+		return 0;
+
+	acpi_early_node_map[cpu] = node;
+	pr_info("SRAT: PXM %d -> HARTID 0x%lx -> Node %d\n", pxm,
+		cpuid_to_hartid_map(cpu), node);
+
+	return 0;
+}
+
+void __init acpi_map_cpus_to_nodes(void)
+{
+	int i;
+
+	/*
+	 * In ACPI, SMP and CPU NUMA information is provided in separate
+	 * static tables, namely the MADT and the SRAT.
+	 *
+	 * Thus, it is simpler to first create the cpu logical map through
+	 * an MADT walk and then map the logical cpus to their node ids
+	 * as separate steps.
+	 */
+	acpi_table_parse_entries(ACPI_SIG_SRAT, sizeof(struct acpi_table_srat),
+				 ACPI_SRAT_TYPE_RINTC_AFFINITY, acpi_parse_rintc_pxm, 0);
+
+	for (i = 0; i < nr_cpu_ids; i++)
+		early_map_cpu_to_node(i, acpi_numa_get_nid(i));
+}
+
+/* Callback for Proximity Domain -> logical node ID mapping */
+void __init acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa)
+{
+	int pxm, node;
+
+	if (srat_disabled())
+		return;
+
+	if (pa->header.length < sizeof(struct acpi_srat_rintc_affinity)) {
+		pr_err("SRAT: Invalid SRAT header length: %d\n", pa->header.length);
+		bad_srat();
+		return;
+	}
+
+	if (!(pa->flags & ACPI_SRAT_RINTC_ENABLED))
+		return;
+
+	pxm = pa->proximity_domain;
+	node = acpi_map_pxm_to_node(pxm);
+
+	if (node == NUMA_NO_NODE) {
+		pr_err("SRAT: Too many proximity domains %d\n", pxm);
+		bad_srat();
+		return;
+	}
+
+	node_set(node, numa_nodes_parsed);
+}

arch/riscv/kernel/setup.c +3 -1

index 4f73c0ae44b2..a2cde65b69e9 100644
@@ -281,8 +281,10 @@ void __init setup_arch(char **cmdline_p)
 	setup_smp();
 #endif
 
-	if (!acpi_disabled)
+	if (!acpi_disabled) {
 		acpi_init_rintc_map();
+		acpi_map_cpus_to_nodes();
+	}
 
 	riscv_init_cbo_blocksizes();
 	riscv_fill_hwcap();

arch/riscv/kernel/smpboot.c +0 -2

index cfbe4b840d42..81a2aa77680c 100644
@@ -100,7 +100,6 @@ static int __init acpi_parse_rintc(union acpi_subtable_headers *header, const un
 	if (hart == cpuid_to_hartid_map(0)) {
 		BUG_ON(found_boot_cpu);
 		found_boot_cpu = true;
-		early_map_cpu_to_node(0, acpi_numa_get_nid(cpu_count));
 		return 0;
 	}
 

@@ -110,7 +109,6 @@ static int __init acpi_parse_rintc(union acpi_subtable_headers *header, const un
 	}
 
 	cpuid_to_hartid_map(cpu_count) = hart;
-	early_map_cpu_to_node(cpu_count, acpi_numa_get_nid(cpu_count));
 	cpu_count++;
 
 	return 0;

include/linux/acpi.h +0 -0

index b7165e52b3c6..f74c62956e07 100644
@@ -269,6 +269,12 @@ acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
 
 int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);

Quoted 15 lines · Haibo Xu

Add acpi_numa.c file to enable parse NUMA information from ACPI SRAT and SLIT tables. SRAT table provide CPUs(Hart) and memory nodes to proximity domain mapping, while SLIT table provide the distance metrics between proximity domains.

Signed-off-by Haibo Xu <[email protected]>

---
 arch/riscv/include/asm/acpi.h |  15 +++-
 arch/riscv/kernel/Makefile    |   1 +
 arch/riscv/kernel/acpi.c      |   5 --
 arch/riscv/kernel/acpi_numa.c | 131 ++++++++++++++++++++++++++++++++++
 arch/riscv/kernel/setup.c     |   4 +-
 arch/riscv/kernel/smpboot.c   |   2 -
 include/linux/acpi.h          |   6 ++
 7 files changed, 154 insertions(+), 10 deletions(-)

Quoted 4 lines

#ifndef PHYS_CPUID_INVALID
typedef u32 phys_cpuid_t;
#define PHYS_CPUID_INVALID (phys_cpuid_t)(-1)
--

This is a large patch spanning across multiple files. Can we split this into multiple smaller patches? Changes look fine to me though.

Reviewed-by Sunil V L <[email protected]>

Add acpi_numa.c file to enable parse NUMA information from ACPI SRAT and SLIT tables. SRAT table provide CPUs(Hart) and memory nodes to proximity domain mapping, while SLIT table provide the distance metrics between proximity domains.

Signed-off-by Haibo Xu <[email protected]>

---
  arch/riscv/include/asm/acpi.h |  15 +++-
  arch/riscv/kernel/Makefile    |   1 +
  arch/riscv/kernel/acpi.c      |   5 --
  arch/riscv/kernel/acpi_numa.c | 131 ++++++++++++++++++++++++++++++++++
  arch/riscv/kernel/setup.c     |   4 +-
  arch/riscv/kernel/smpboot.c   |   2 -
  include/linux/acpi.h          |   6 ++
  7 files changed, 154 insertions(+), 10 deletions(-)

create mode 100644 arch/riscv/kernel/acpi_numa.c

arch/riscv/include/asm/acpi.h +13 -2

index 7dad0cf9d701..e0a1f84404f3 100644
@@ -61,11 +61,14 @@ static inline void arch_fix_phys_package_id(int num, u32 slot) { }
 
  void acpi_init_rintc_map(void);
  struct acpi_madt_rintc *acpi_cpu_get_madt_rintc(int cpu);
-u32 get_acpi_id_for_cpu(int cpu);
+static inline u32 get_acpi_id_for_cpu(int cpu)
+{
+	return acpi_cpu_get_madt_rintc(cpu)->uid;
+}
+
  int acpi_get_riscv_isa(struct acpi_table_header *table,
  		       unsigned int cpu, const char **isa);
 
-static inline int acpi_numa_get_nid(unsigned int cpu) { return NUMA_NO_NODE; }
  void acpi_get_cbo_block_size(struct acpi_table_header *table, u32 *cbom_size,
  			     u32 *cboz_size, u32 *cbop_size);
  #else

@@ -87,4 +90,12 @@ static inline void acpi_get_cbo_block_size(struct acpi_table_header *table,
 
  #endif /* CONFIG_ACPI */
 
+#ifdef CONFIG_ACPI_NUMA
+int acpi_numa_get_nid(unsigned int cpu);
+void acpi_map_cpus_to_nodes(void);
+#else
+static inline int acpi_numa_get_nid(unsigned int cpu) { return NUMA_NO_NODE; }
+static inline void acpi_map_cpus_to_nodes(void) { }
+#endif /* CONFIG_ACPI_NUMA */
+
  #endif /*_ASM_ACPI_H*/

arch/riscv/kernel/Makefile +1 -0

index f71910718053..5d3e9cf89b76 100644
@@ -105,3 +105,4 @@ obj-$(CONFIG_COMPAT)		+= compat_vdso/
 
  obj-$(CONFIG_64BIT)		+= pi/
  obj-$(CONFIG_ACPI)		+= acpi.o
+obj-$(CONFIG_ACPI_NUMA)	+= acpi_numa.o

arch/riscv/kernel/acpi.c +0 -5

index e619edc8b0cc..040bdbfea2b4 100644
@@ -191,11 +191,6 @@ struct acpi_madt_rintc *acpi_cpu_get_madt_rintc(int cpu)
  	return &cpu_madt_rintc[cpu];
  }
 
-u32 get_acpi_id_for_cpu(int cpu)
-{
-	return acpi_cpu_get_madt_rintc(cpu)->uid;
-}
-
  /*
   * __acpi_map_table() will be called before paging_init(), so early_ioremap()
   * or early_memremap() should be called here to for ACPI table mapping.

arch/riscv/kernel/acpi_numa.c new file +121 -0

new file mode 100644
index 000000000000..0231482d6946
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * ACPI 6.6 based NUMA setup for RISCV
+ * Lots of code was borrowed from arch/arm64/kernel/acpi_numa.c
+ *
+ * Copyright 2004 Andi Kleen, SuSE Labs.
+ * Copyright (C) 2013-2016, Linaro Ltd.
+ *		Author: Hanjun Guo <[email protected]>
+ * Copyright (C) 2024 Intel Corporation.
+ *
+ * Reads the ACPI SRAT table to figure out what memory belongs to which CPUs.
+ *
+ * Called from acpi_numa_init while reading the SRAT and SLIT tables.
+ * Assumes all memory regions belonging to a single proximity domain
+ * are in one chunk. Holes between them will be included in the node.
+ */
+
+#define pr_fmt(fmt) "ACPI: NUMA: " fmt
+
+#include <linux/acpi.h>
+#include <linux/bitmap.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/memblock.h>
+#include <linux/mmzone.h>
+#include <linux/module.h>
+#include <linux/topology.h>
+
+#include <asm/numa.h>
+
+static int acpi_early_node_map[NR_CPUS] __initdata = { NUMA_NO_NODE };
+
+int __init acpi_numa_get_nid(unsigned int cpu)
+{
+	return acpi_early_node_map[cpu];
+}
+
+static inline int get_cpu_for_acpi_id(u32 uid)
+{
+	int cpu;
+
+	for (cpu = 0; cpu < nr_cpu_ids; cpu++)
+		if (uid == get_acpi_id_for_cpu(cpu))
+			return cpu;
+
+	return -EINVAL;
+}
+
+static int __init acpi_parse_rintc_pxm(union acpi_subtable_headers *header,
+				       const unsigned long end)
+{
+	struct acpi_srat_rintc_affinity *pa;
+	int cpu, pxm, node;
+
+	if (srat_disabled())
+		return -EINVAL;
+
+	pa = (struct acpi_srat_rintc_affinity *)header;
+	if (!pa)
+		return -EINVAL;
+
+	if (!(pa->flags & ACPI_SRAT_RINTC_ENABLED))
+		return 0;
+
+	pxm = pa->proximity_domain;
+	node = pxm_to_node(pxm);
+
+	/*
+	 * If we can't map the UID to a logical cpu this
+	 * means that the UID is not part of possible cpus
+	 * so we do not need a NUMA mapping for it, skip
+	 * the SRAT entry and keep parsing.
+	 */
+	cpu = get_cpu_for_acpi_id(pa->acpi_processor_uid);
+	if (cpu < 0)
+		return 0;
+
+	acpi_early_node_map[cpu] = node;
+	pr_info("SRAT: PXM %d -> HARTID 0x%lx -> Node %d\n", pxm,
+		cpuid_to_hartid_map(cpu), node);
+
+	return 0;
+}
+
+void __init acpi_map_cpus_to_nodes(void)
+{
+	int i;
+
+	/*
+	 * In ACPI, SMP and CPU NUMA information is provided in separate
+	 * static tables, namely the MADT and the SRAT.
+	 *
+	 * Thus, it is simpler to first create the cpu logical map through
+	 * an MADT walk and then map the logical cpus to their node ids
+	 * as separate steps.
+	 */
+	acpi_table_parse_entries(ACPI_SIG_SRAT, sizeof(struct acpi_table_srat),
+				 ACPI_SRAT_TYPE_RINTC_AFFINITY, acpi_parse_rintc_pxm, 0);
+
+	for (i = 0; i < nr_cpu_ids; i++)
+		early_map_cpu_to_node(i, acpi_numa_get_nid(i));
+}
+
+/* Callback for Proximity Domain -> logical node ID mapping */
+void __init acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa)
+{
+	int pxm, node;
+
+	if (srat_disabled())
+		return;
+
+	if (pa->header.length < sizeof(struct acpi_srat_rintc_affinity)) {
+		pr_err("SRAT: Invalid SRAT header length: %d\n", pa->header.length);
+		bad_srat();
+		return;
+	}
+
+	if (!(pa->flags & ACPI_SRAT_RINTC_ENABLED))
+		return;
+
+	pxm = pa->proximity_domain;

arch/riscv/kernel/setup.c +3 -1

index 4f73c0ae44b2..a2cde65b69e9 100644
@@ -281,8 +281,10 @@ void __init setup_arch(char **cmdline_p)
  	setup_smp();
  #endif
 
-	if (!acpi_disabled)
+	if (!acpi_disabled) {
  		acpi_init_rintc_map();
+		acpi_map_cpus_to_nodes();
+	}
 
  	riscv_init_cbo_blocksizes();
  	riscv_fill_hwcap();

arch/riscv/kernel/smpboot.c +0 -2

index cfbe4b840d42..81a2aa77680c 100644
@@ -100,7 +100,6 @@ static int __init acpi_parse_rintc(union acpi_subtable_headers *header, const un
  	if (hart == cpuid_to_hartid_map(0)) {
  		BUG_ON(found_boot_cpu);
  		found_boot_cpu = true;
-		early_map_cpu_to_node(0, acpi_numa_get_nid(cpu_count));
  		return 0;
  	}
 

@@ -110,7 +109,6 @@ static int __init acpi_parse_rintc(union acpi_subtable_headers *header, const un
  	}
 
  	cpuid_to_hartid_map(cpu_count) = hart;
-	early_map_cpu_to_node(cpu_count, acpi_numa_get_nid(cpu_count));
  	cpu_count++;
 
  	return 0;

include/linux/acpi.h +6 -0

index b7165e52b3c6..f74c62956e07 100644
@@ -269,6 +269,12 @@ acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
 
  int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
 
+#ifdef CONFIG_RISCV
+void acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa);
+#else
+static inline void acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa) { }
+#endif
+
  #ifndef PHYS_CPUID_INVALID
  typedef u32 phys_cpuid_t;
  #define PHYS_CPUID_INVALID (phys_cpuid_t)(-1)

Hi Haibo,

Quoted 212 lines · Haibo Xu

Add acpi_numa.c file to enable parse NUMA information from ACPI SRAT and SLIT tables. SRAT table provide CPUs(Hart) and memory nodes to proximity domain mapping, while SLIT table provide the distance metrics between proximity domains.

Signed-off-by Haibo Xu <[email protected]>

---
  arch/riscv/include/asm/acpi.h |  15 +++-
  arch/riscv/kernel/Makefile    |   1 +
  arch/riscv/kernel/acpi.c      |   5 --
  arch/riscv/kernel/acpi_numa.c | 131 ++++++++++++++++++++++++++++++++++
  arch/riscv/kernel/setup.c     |   4 +-
  arch/riscv/kernel/smpboot.c   |   2 -
  include/linux/acpi.h          |   6 ++
  7 files changed, 154 insertions(+), 10 deletions(-)

create mode 100644 arch/riscv/kernel/acpi_numa.c

arch/riscv/include/asm/acpi.h +13 -2

index 7dad0cf9d701..e0a1f84404f3 100644
@@ -61,11 +61,14 @@ static inline void arch_fix_phys_package_id(int num, u32 slot) { }
 
  void acpi_init_rintc_map(void);
  struct acpi_madt_rintc *acpi_cpu_get_madt_rintc(int cpu);
-u32 get_acpi_id_for_cpu(int cpu);
+static inline u32 get_acpi_id_for_cpu(int cpu)
+{
+     return acpi_cpu_get_madt_rintc(cpu)->uid;
+}
+
  int acpi_get_riscv_isa(struct acpi_table_header *table,
                     unsigned int cpu, const char **isa);
 
-static inline int acpi_numa_get_nid(unsigned int cpu) { return NUMA_NO_NODE; }
  void acpi_get_cbo_block_size(struct acpi_table_header *table, u32 *cbom_size,
                           u32 *cboz_size, u32 *cbop_size);
  #else

@@ -87,4 +90,12 @@ static inline void acpi_get_cbo_block_size(struct acpi_table_header *table,
 
  #endif /* CONFIG_ACPI */
 
+#ifdef CONFIG_ACPI_NUMA
+int acpi_numa_get_nid(unsigned int cpu);
+void acpi_map_cpus_to_nodes(void);
+#else
+static inline int acpi_numa_get_nid(unsigned int cpu) { return NUMA_NO_NODE; }
+static inline void acpi_map_cpus_to_nodes(void) { }
+#endif /* CONFIG_ACPI_NUMA */
+
  #endif /*_ASM_ACPI_H*/

arch/riscv/kernel/Makefile +1 -0

index f71910718053..5d3e9cf89b76 100644
@@ -105,3 +105,4 @@ obj-$(CONFIG_COMPAT)              += compat_vdso/
 
  obj-$(CONFIG_64BIT)         += pi/
  obj-$(CONFIG_ACPI)          += acpi.o
+obj-$(CONFIG_ACPI_NUMA)      += acpi_numa.o

arch/riscv/kernel/acpi.c +0 -5

index e619edc8b0cc..040bdbfea2b4 100644
@@ -191,11 +191,6 @@ struct acpi_madt_rintc *acpi_cpu_get_madt_rintc(int cpu)
      return &cpu_madt_rintc[cpu];
  }
 
-u32 get_acpi_id_for_cpu(int cpu)
-{
-     return acpi_cpu_get_madt_rintc(cpu)->uid;
-}
-
  /*
   * __acpi_map_table() will be called before paging_init(), so early_ioremap()
   * or early_memremap() should be called here to for ACPI table mapping.

arch/riscv/kernel/acpi_numa.c new file +130 -0

new file mode 100644
index 000000000000..0231482d6946
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * ACPI 6.6 based NUMA setup for RISCV
+ * Lots of code was borrowed from arch/arm64/kernel/acpi_numa.c
+ *
+ * Copyright 2004 Andi Kleen, SuSE Labs.
+ * Copyright (C) 2013-2016, Linaro Ltd.
+ *           Author: Hanjun Guo <[email protected]>
+ * Copyright (C) 2024 Intel Corporation.
+ *
+ * Reads the ACPI SRAT table to figure out what memory belongs to which CPUs.
+ *
+ * Called from acpi_numa_init while reading the SRAT and SLIT tables.
+ * Assumes all memory regions belonging to a single proximity domain
+ * are in one chunk. Holes between them will be included in the node.
+ */
+
+#define pr_fmt(fmt) "ACPI: NUMA: " fmt
+
+#include <linux/acpi.h>
+#include <linux/bitmap.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/memblock.h>
+#include <linux/mmzone.h>
+#include <linux/module.h>
+#include <linux/topology.h>
+
+#include <asm/numa.h>
+
+static int acpi_early_node_map[NR_CPUS] __initdata = { NUMA_NO_NODE };
+
+int __init acpi_numa_get_nid(unsigned int cpu)
+{
+     return acpi_early_node_map[cpu];
+}
+
+static inline int get_cpu_for_acpi_id(u32 uid)
+{
+     int cpu;
+
+     for (cpu = 0; cpu < nr_cpu_ids; cpu++)
+             if (uid == get_acpi_id_for_cpu(cpu))
+                     return cpu;
+
+     return -EINVAL;
+}
+
+static int __init acpi_parse_rintc_pxm(union acpi_subtable_headers *header,
+                                    const unsigned long end)
+{
+     struct acpi_srat_rintc_affinity *pa;
+     int cpu, pxm, node;
+
+     if (srat_disabled())
+             return -EINVAL;
+
+     pa = (struct acpi_srat_rintc_affinity *)header;
+     if (!pa)
+             return -EINVAL;
+
+     if (!(pa->flags & ACPI_SRAT_RINTC_ENABLED))
+             return 0;
+
+     pxm = pa->proximity_domain;
+     node = pxm_to_node(pxm);
+
+     /*
+      * If we can't map the UID to a logical cpu this
+      * means that the UID is not part of possible cpus
+      * so we do not need a NUMA mapping for it, skip
+      * the SRAT entry and keep parsing.
+      */
+     cpu = get_cpu_for_acpi_id(pa->acpi_processor_uid);
+     if (cpu < 0)
+             return 0;
+
+     acpi_early_node_map[cpu] = node;
+     pr_info("SRAT: PXM %d -> HARTID 0x%lx -> Node %d\n", pxm,
+             cpuid_to_hartid_map(cpu), node);
+
+     return 0;
+}
+
+void __init acpi_map_cpus_to_nodes(void)
+{
+     int i;
+
+     /*
+      * In ACPI, SMP and CPU NUMA information is provided in separate
+      * static tables, namely the MADT and the SRAT.
+      *
+      * Thus, it is simpler to first create the cpu logical map through
+      * an MADT walk and then map the logical cpus to their node ids
+      * as separate steps.
+      */
+     acpi_table_parse_entries(ACPI_SIG_SRAT, sizeof(struct acpi_table_srat),
+                              ACPI_SRAT_TYPE_RINTC_AFFINITY, acpi_parse_rintc_pxm, 0);
+
+     for (i = 0; i < nr_cpu_ids; i++)
+             early_map_cpu_to_node(i, acpi_numa_get_nid(i));
+}
+
+/* Callback for Proximity Domain -> logical node ID mapping */
+void __init acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa)
+{
+     int pxm, node;
+
+     if (srat_disabled())
+             return;
+
+     if (pa->header.length < sizeof(struct acpi_srat_rintc_affinity)) {
+             pr_err("SRAT: Invalid SRAT header length: %d\n", pa->header.length);
+             bad_srat();
+             return;
+     }
+
+     if (!(pa->flags & ACPI_SRAT_RINTC_ENABLED))
+             return;
+
+     pxm = pa->proximity_domain;
+     node = acpi_map_pxm_to_node(pxm);
+
+     if (node == NUMA_NO_NODE) {
+             pr_err("SRAT: Too many proximity domains %d\n", pxm);
+             bad_srat();
+             return;
+     }
+
+     node_set(node, numa_nodes_parsed);

arch/riscv/kernel/acpi_numa.c line 131

Quoted 1 line

arch/riscv/kernel/acpi_numa.c new file +1 -0

+}

What is riscv specific in the parsing of those tables? Can't we try to merge this into generic ACPI code? I know that's a burden to try and factorize code with other architectures instead of reusing it, but it showed numerous times that duplicating was even worse (I have the NAPOT code in mind).

Thanks,

Alex

Quoted 53 lines

arch/riscv/kernel/setup.c +3 -1

index 4f73c0ae44b2..a2cde65b69e9 100644
@@ -281,8 +281,10 @@ void __init setup_arch(char **cmdline_p)
      setup_smp();
  #endif
 
-     if (!acpi_disabled)
+     if (!acpi_disabled) {
              acpi_init_rintc_map();
+             acpi_map_cpus_to_nodes();
+     }
 
      riscv_init_cbo_blocksizes();
      riscv_fill_hwcap();

arch/riscv/kernel/smpboot.c +0 -2

index cfbe4b840d42..81a2aa77680c 100644
@@ -100,7 +100,6 @@ static int __init acpi_parse_rintc(union acpi_subtable_headers *header, const un
      if (hart == cpuid_to_hartid_map(0)) {
              BUG_ON(found_boot_cpu);
              found_boot_cpu = true;
-             early_map_cpu_to_node(0, acpi_numa_get_nid(cpu_count));
              return 0;
      }
 

@@ -110,7 +109,6 @@ static int __init acpi_parse_rintc(union acpi_subtable_headers *header, const un
      }
 
      cpuid_to_hartid_map(cpu_count) = hart;
-     early_map_cpu_to_node(cpu_count, acpi_numa_get_nid(cpu_count));
      cpu_count++;
 
      return 0;

include/linux/acpi.h +6 -0

index b7165e52b3c6..f74c62956e07 100644
@@ -269,6 +269,12 @@ acpi_numa_gicc_affinity_init(struct acpi_srat_gicc_affinity *pa) { }
 
  int acpi_numa_memory_affinity_init (struct acpi_srat_mem_affinity *ma);
 
+#ifdef CONFIG_RISCV
+void acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa);
+#else
+static inline void acpi_numa_rintc_affinity_init(struct acpi_srat_rintc_affinity *pa) { }
+#endif
+
  #ifndef PHYS_CPUID_INVALID
  typedef u32 phys_cpuid_t;
  #define PHYS_CPUID_INVALID (phys_cpuid_t)(-1)

Add RINTC affinity structure handler during parsing SRAT table.

Signed-off-by Haibo Xu <[email protected]>

---
 drivers/acpi/numa/srat.c | 32 +++++++++++++++++++++++++++++++-
 1 file changed, 31 insertions(+), 1 deletion(-)

drivers/acpi/numa/srat.c +29 -1

index 0214518fc582..1946431c0eef 100644
@@ -165,6 +165,19 @@ acpi_table_print_srat_entry(struct acpi_subtable_header *header)
 		}
 	}
 	break;
+
+	case ACPI_SRAT_TYPE_RINTC_AFFINITY:
+		{
+			struct acpi_srat_rintc_affinity *p =
+			    (struct acpi_srat_rintc_affinity *)header;
+			pr_debug("SRAT Processor (acpi id[0x%04x]) in proximity domain %d %s\n",
+				 p->acpi_processor_uid,
+				 p->proximity_domain,
+				 (p->flags & ACPI_SRAT_RINTC_ENABLED) ?
+				 "enabled" : "disabled");
+		}
+		break;
+
 	default:
 		pr_warn("Found unsupported SRAT entry (type = 0x%x)\n",
 			header->type);

@@ -448,6 +461,21 @@ acpi_parse_gi_affinity(union acpi_subtable_headers *header,
 }
 #endif /* defined(CONFIG_X86) || defined (CONFIG_ARM64) */
 
+static int __init
+acpi_parse_rintc_affinity(union acpi_subtable_headers *header,
+			  const unsigned long end)
+{
+	struct acpi_srat_rintc_affinity *rintc_affinity;
+
+	rintc_affinity = (struct acpi_srat_rintc_affinity *)header;
+	acpi_table_print_srat_entry(&header->common);
+
+	/* let architecture-dependent part to do it */
+	acpi_numa_rintc_affinity_init(rintc_affinity);
+
+	return 0;
+}
+
 static int __initdata parsed_numa_memblks;
 
 static int __init

@@ -501,7 +529,7 @@ int __init acpi_numa_init(void)
 
 	/* SRAT: System Resource Affinity Table */
 	if (!acpi_table_parse(ACPI_SIG_SRAT, acpi_parse_srat)) {
-		struct acpi_subtable_proc srat_proc[4];
+		struct acpi_subtable_proc srat_proc[5];
 
 		memset(srat_proc, 0, sizeof(srat_proc));
 		srat_proc[0].id = ACPI_SRAT_TYPE_CPU_AFFINITY;

The acpi_numa_slit_init(), acpi_numa_memory_affinity_init() and acpi_parse_cfmws() functions are common enough to be used on platforms that support ACPI_NUMA(x86/arm64/loongarch). Remove the condition to avoid long defined(CONFIG_ARCH) check when new platform(riscv) support was enabled.

Suggested-by Sunil V L <[email protected]>

Signed-off-by Haibo Xu <[email protected]>

---
 drivers/acpi/numa/srat.c | 8 --------
 1 file changed, 8 deletions(-)

drivers/acpi/numa/srat.c +0 -1

index 1946431c0eef..938c4adb7ec4 100644
@@ -219,7 +219,6 @@ int __init srat_disabled(void)
 	return acpi_numa < 0;
 }
 
-#if defined(CONFIG_X86) || defined(CONFIG_ARM64) || defined(CONFIG_LOONGARCH)
 /*
  * Callback for SLIT parsing.  pxm_to_node() returns NUMA_NO_NODE for
  * I/O localities since SRAT does not list them.  I/O localities are

@@ -351,13 +350,6 @@ static int __init acpi_parse_cfmws(union acpi_subtable_headers *header,
 	(*fake_pxm)++;
 	return 0;
 }

x86/arm64/loongarch would select ACPI_NUMA by default and riscv would do the same thing, so the dependency is no longer needed since these are the four architectures that support ACPI.

Suggested-by Arnd Bergmann <[email protected]>

Suggested-by Sunil V L <[email protected]>

Signed-off-by Haibo Xu <[email protected]>

---
 drivers/acpi/numa/Kconfig | 1 -
 1 file changed, 1 deletion(-)

drivers/acpi/numa/Kconfig +0 -0

index 849c2bd820b9..2bf47ad1ec9b 100644

Quoted 16 lines · Haibo Xu

x86/arm64/loongarch would select ACPI_NUMA by default and riscv would do the same thing, so the dependency is no longer needed since these are the four architectures that support ACPI.

Suggested-by Arnd Bergmann <[email protected]>

Suggested-by Sunil V L <[email protected]>

Signed-off-by Haibo Xu <[email protected]>

---
 drivers/acpi/numa/Kconfig | 1 -
 1 file changed, 1 deletion(-)

drivers/acpi/numa/Kconfig +0 -0

index 849c2bd820b9..2bf47ad1ec9b 100644

drivers/acpi/numa/Kconfig line 5

Quoted 5 lines

drivers/acpi/numa/Kconfig +0 -1

@@ -2,7 +2,6 @@
 config ACPI_NUMA
      bool "NUMA support"
      depends on NUMA
-     depends on (X86 || ARM64 || LOONGARCH)
      default y if ARM64

Can we remove the prompt as well and make this a hidden option? I think this is now always selected when it can be used anyway.

If we make it

      def_bool NUMA && !X86

then the select statements except for the X86_64_ACPI_NUMA can also go away.

Arnd

Quoted 4 lines · Arnd Bergmann

On Thu, Mar 7, 2024, at 09:47, Haibo Xu wrote:

If we make it

Quoted 5 lines

      def_bool NUMA && !X86

then the select statements except for the X86_64_ACPI_NUMA can also go away.

Good idea!
Shall we include the ACPI in the def_bool definition?

Quoted 11 lines · Haibo Xu

Quoted 9 lines · Arnd Bergmann

On Thu, Mar 7, 2024, at 09:47, Haibo Xu wrote:

If we make it

      def_bool NUMA && !X86

then the select statements except for the X86_64_ACPI_NUMA can also go away.

Quoted 3 lines

Good idea!
Shall we include the ACPI in the def_bool definition?

No need, because this is inside of an 'if ACPI' block.

Arnd

Enable ACPI based NUMA for RISCV in Kconfig.

Signed-off-by Haibo Xu <[email protected]>

---
 arch/riscv/Kconfig | 1 +
 1 file changed, 1 insertion(+)

arch/riscv/Kconfig +0 -0

index 0bfcfec67ed5..0fb55f166701 100644

arch/riscv/Kconfig +0 -0

 	select OF_NUMA
 	select USE_PERCPU_NUMA_NODE_ID
 	help

Signature · 1 line

2.34.1

Hi Haibo,

Quoted 7 lines · Haibo Xu

Enable ACPI based NUMA for RISCV in Kconfig.

Signed-off-by Haibo Xu <[email protected]>

---
 arch/riscv/Kconfig | 1 +
 1 file changed, 1 insertion(+)

arch/riscv/Kconfig config NUMA · line 450

Quoted 9 lines

arch/riscv/Kconfig +1 -0

index 0bfcfec67ed5..0fb55f166701 100644
@@ -447,6 +447,7 @@ config NUMA
      select HAVE_SETUP_PER_CPU_AREA
      select NEED_PER_CPU_EMBED_FIRST_CHUNK
      select NEED_PER_CPU_PAGE_FIRST_CHUNK
+     select ACPI_NUMA if ACPI

If the firmware didn't provide the SRAT/SLIT on ACPI based systems, then there will be a message "Failed to initialise from firmware" from arch_acpi_numa_init(). This is not specific to RISC-V. But I am wondering why should it be pr_info instead of pr_debug.

Thanks,
Sunil

Quoted 6 lines

arch/riscv/Kconfig +0 -1

      select OF_NUMA
      select USE_PERCPU_NUMA_NODE_ID
      help
--

2.34.1

Quoted 5 lines

If the firmware didn't provide the SRAT/SLIT on ACPI based systems, then there will be a message "Failed to initialise from firmware" from arch_acpi_numa_init(). This is not specific to RISC-V. But I am wondering why should it be pr_info instead of pr_debug.

My understanding is maybe it just wants to expose explicit logs to avoid any potential bugs from FW or Kernel.

Quoted 6 lines

Quoted 5 lines

If the firmware didn't provide the SRAT/SLIT on ACPI based systems, then there will be a message "Failed to initialise from firmware" from arch_acpi_numa_init(). This is not specific to RISC-V. But I am wondering why should it be pr_info instead of pr_debug.

Quoted 2 lines

My understanding is maybe it just wants to expose explicit logs to avoid any potential bugs from FW or Kernel.

There are lots of ACPI enabled systems that aren't NUMA (single socket servers, desktops, laptops). Making this "pr_info()" would just add noise to the boot on all of those.

Quoted 9 lines · Tony

Quoted 8 lines

Quoted 5 lines

If the firmware didn't provide the SRAT/SLIT on ACPI based systems, then there will be a message "Failed to initialise from firmware" from arch_acpi_numa_init(). This is not specific to RISC-V. But I am wondering why should it be pr_info instead of pr_debug.

My understanding is maybe it just wants to expose explicit logs to avoid any potential bugs from FW or Kernel.

Quoted 4 lines

There are lots of ACPI enabled systems that aren't NUMA (single socket servers, desktops, laptops). Making this "pr_info()" would just add noise to the boot on all of those.

Exactly. But this is an existing pr_info message across architectures. My suggestion is to add one more patch in this series to convert this to pr_debug unless someone has strong reason to keep it pr_info.

Thanks,
Sunil

This patch series enable RISC-V ACPI NUMA support which was based on the recently approved ACPI ECR[1].

Patch 1/6 is generated from the acpica PR[2] and should be merged through the acpica project. Due to this dependency, other 5 patches can only be merged after the corresponding ACPICA patch was pulled into linux.

Patch 2/6 add RISC-V specific acpi_numa.c file to parse NUMA information from SRAT and SLIT ACPI tables.
Patch 3/6 add the common SRAT RINTC affinity structure handler. Patch 4/6 remove the '#if defined(CONFIG_ARCH)' condition to make some NUMA related parse functions common for all current architectures that support ACPI_NUMA
Patch 5/6 remove ARCH depends option in ACPI_NUMA Kconfig which was no longer needed since all the current architectures that support ACPI would select ACPI_NUMA by default.
Patch 6/6 add corresponding ACPI_NUMA config for RISC-V Kconfig.

Based-on: https://github.com/linux-riscv/linux-riscv/tree/for-next

[1] https://drive.google.com/file/d/1YTdDx2IPm5IeZjAW932EYU-tUtgS08tX/view?usp=sharing
[2] https://github.com/acpica/acpica/pull/926

Testing:
Since the ACPI AIA/PLIC support patch set is still under upstream review, hence it is tested using the poll based HVC SBI console and RAM disk.
1) Build latest Qemu with the following patch backported
https://lore.kernel.org/all/[email protected]/
https://github.com/vlsunil/qemu/commit/42bd4eeefd5d4410a68f02d54fee406d8a1269b0

2) Build latest EDK-II
https://github.com/tianocore/edk2/blob/master/OvmfPkg/RiscVVirt/README.md

3) Build Linux with the following configs enabled
CONFIG_RISCV_SBI_V01=y
CONFIG_SERIAL_EARLYCON_RISCV_SBI=y
CONFIG_HVC_RISCV_SBI=y
CONFIG_NUMA=y
CONFIG_ACPI_NUMA=y

4) Build buildroot rootfs.cpio

5) Launch the Qemu machine

   qemu-system-riscv64 -nographic \
   -machine virt,pflash0=pflash0,pflash1=pflash1 -smp 4 -m 8G \
   -blockdev node-name=pflash0,driver=file,read-only=on,filename=RISCV_VIRT_CODE.fd \
   -blockdev node-name=pflash1,driver=file,filename=RISCV_VIRT_VARS.fd \
   -object memory-backend-ram,size=4G,id=m0 \
   -object memory-backend-ram,size=4G,id=m1 \
   -numa node,memdev=m0,cpus=0-1,nodeid=0 \
   -numa node,memdev=m1,cpus=2-3,nodeid=1 \
   -numa dist,src=0,dst=1,val=30 \
   -kernel linux/arch/riscv/boot/Image \
   -initrd buildroot/output/images/rootfs.cpio \
   -append "root=/dev/ram ro console=hvc0 earlycon=sbi"

[    0.000000] ACPI: SRAT: Node 0 PXM 0 [mem 0x80000000-0x17fffffff]
[    0.000000] ACPI: SRAT: Node 1 PXM 1 [mem 0x180000000-0x27fffffff]
[    0.000000] NUMA: NODE_DATA [mem 0x17fe3bc40-0x17fe3cfff]
[    0.000000] NUMA: NODE_DATA [mem 0x27fff4c40-0x27fff5fff]
...
[    0.000000] ACPI: NUMA: SRAT: PXM 0 -> HARTID 0x0 -> Node 0
[    0.000000] ACPI: NUMA: SRAT: PXM 0 -> HARTID 0x1 -> Node 0
[    0.000000] ACPI: NUMA: SRAT: PXM 1 -> HARTID 0x2 -> Node 1
[    0.000000] ACPI: NUMA: SRAT: PXM 1 -> HARTID 0x3 -> Node 1

---
Changes since v1
- Switch the order of patch 2/4 and 3/4 - Per Sunil's suggestion
- Add a new patch 4/6 to make some NUMA related parse functions common for all the architectures that support ACPI NUMA - Per Sunil's suggestion
- Add a new patch 5/6 to remove ARCH depends option in ACPI_NUMA Kconfig since all the current architectures that support ACPI would select ACPI_NUMA by default - Per Arnd and Sunil's suggestion
- Other minor fix for code format - Per Sunil's comments

Haibo Xu (6):
ACPICA: SRAT: Add RISC-V RINTC affinity structure
ACPI: RISCV: Add NUMA support based on SRAT and SLIT
ACPI: NUMA: Add handler for SRAT RINTC affinity structure
ACPI: NUMA: Make some NUMA related parse functions common
ACPI: NUMA: Remove ARCH depends option in ACPI_NUMA Kconfig
ACPI: RISCV: Enable ACPI based NUMA

 arch/riscv/Kconfig            |   1 +
 arch/riscv/include/asm/acpi.h |  15 +++-
 arch/riscv/kernel/Makefile    |   1 +
 arch/riscv/kernel/acpi.c      |   5 --
 arch/riscv/kernel/acpi_numa.c | 131 ++++++++++++++++++++++++++++++++++
 arch/riscv/kernel/setup.c     |   4 +-
 arch/riscv/kernel/smpboot.c   |   2 -
 drivers/acpi/numa/Kconfig     |   1 -
 drivers/acpi/numa/srat.c      |  40 ++++++++---
 include/acpi/actbl3.h         |  18 ++++-
 include/linux/acpi.h          |   6 ++
 11 files changed, 203 insertions(+), 21 deletions(-)
 create mode 100644 arch/riscv/kernel/acpi_numa.c

--
2.34.1