Skip to contentSkip to conversation

riscv: Add numa support for riscv64 platform

RFC PATCH New Greentime Hu 1 participant 5 messages Jan 10, 2020 RISC-V

lore
Share & tools
lore
#0 lore raw

v2 cover riscv: Add numa support for riscv64 platform

riscv: Add numa support for riscv64 platform

This implementation is based on arm64 porting. It is tested with qemu-system-riscv64, SiFive Unleashed board and OmniXtend FPGA platform.

There will be two nodes in /sys/devices/system/node if it is described in dts and CONFIG_NUMA is enabled. We can use numastat/numactl/numademo to see its status.

Changes in v2:
- split this patch to more patches to be more readable
- set cpu->hotplugable to 0 since it is not supported yet
- add more explanation for moving unflatten_device_tree() to paging_init()

Greentime Hu (4):
riscv: Add support pte_protnone and pmd_protnone if
CONFIG_NUMA_BALANCING
riscv: Move unflatten_device_tree() to paging_init() because
riscv_numa_init() needs the dt information.
riscv: Use variable this_cpu instead of smp_processor_id()
riscv: Add numa support for riscv64 platform

 arch/riscv/Kconfig               |  30 ++-
 arch/riscv/include/asm/mmzone.h  |  13 ++
 arch/riscv/include/asm/numa.h    |  46 ++++
 arch/riscv/include/asm/pci.h     |  10 +
 arch/riscv/include/asm/pgtable.h |  20 ++
 arch/riscv/kernel/setup.c        |  26 ++-
 arch/riscv/kernel/smpboot.c      |  20 +-
 arch/riscv/mm/Makefile           |   1 +
 arch/riscv/mm/init.c             |   3 +
 arch/riscv/mm/numa.c             | 372 +++++++++++++++++++++++++++++++
 10 files changed, 536 insertions(+), 5 deletions(-)
 create mode 100644 arch/riscv/include/asm/mmzone.h
 create mode 100644 arch/riscv/include/asm/numa.h
 create mode 100644 arch/riscv/mm/numa.c
Signature · 1 line
2.17.1
#1 lore raw

RFC PATCH v2 1/4 riscv: Add support pte_protnone and pmd_protnone if CONFIG_NUMA_BALANCING

These two functions are used to distinguish between PROT_NONENUMA protections and hinting fault protections.

---
 arch/riscv/include/asm/pgtable.h | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
arch/riscv/include/asm/pgtable.h +20 -0
index 36ae01761352..9650a4ce073e 100644
@@ -166,6 +166,11 @@ static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 	return (unsigned long)pfn_to_virt(pmd_val(pmd) >> _PAGE_PFN_SHIFT);
 }
 
+static inline pte_t pmd_pte(pmd_t pmd)
+{
+	return __pte(pmd_val(pmd));
+}
+
 /* Yields the page frame number (PFN) of a page table entry */
 static inline unsigned long pte_pfn(pte_t pte)
 {
@@ -279,6 +284,21 @@ static inline pte_t pte_mkhuge(pte_t pte)
 	return pte;
 }
 
+#ifdef CONFIG_NUMA_BALANCING
+/*
+ * See the comment in include/asm-generic/pgtable.h
+ */
+static inline int pte_protnone(pte_t pte)
+{
+	return (pte_val(pte) & (_PAGE_PRESENT | _PAGE_PROT_NONE)) == _PAGE_PROT_NONE;
+}
+
+static inline int pmd_protnone(pmd_t pmd)
+{
+	return pte_protnone(pmd_pte(pmd));
+}
+#endif
+
 /* Modify page protection bits */
 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 {
Signature · 1 line
2.17.1
#2 lore raw

RFC PATCH v2 2/4 riscv: Move unflatten_device_tree() to paging_init() because riscv_numa_init() needs the dt information.

In reply to Greentime Hu

It is moved to paging_init() is because that of_numa_init() will use of_numa_parse_cpu_nodes() and of_numa_parse_memory_nodes(). We have to unflatten_device_tree() first then we can call riscv_numa_init(), but riscv_numa_init() shall be called before memblocks_present() because the node information will be used in memblocks_present().

So the calling order will be looked like this. unflatten_device_tree(); //To get dt information for memory and nodes riscv_numa_init(); //It can use of_numa_parse_* and set the nodes information memblocks_present(); //The node information can be used now

---
 arch/riscv/kernel/setup.c | 1 -
 arch/riscv/mm/init.c      | 1 +
 2 files changed, 1 insertion(+), 1 deletion(-)
arch/riscv/kernel/setup.c +0 -1
index 365ff8420bfe..bcd85f502fb2 100644
@@ -67,7 +67,6 @@ void __init setup_arch(char **cmdline_p)
 
 	setup_bootmem();
 	paging_init();
-	unflatten_device_tree();
 	clint_init_boot_cpu();
 
 #ifdef CONFIG_SWIOTLB
arch/riscv/mm/init.c +1 -0
index 69f6678db7f3..fbe69fe47806 100644
@@ -491,6 +491,7 @@ static inline void setup_vm_final(void)
 void __init paging_init(void)
 {
 	setup_vm_final();
+	unflatten_device_tree();
 	memblocks_present();
 	sparse_init();
 	setup_zero_page();
Signature · 1 line
2.17.1
#3 lore raw

RFC PATCH v2 3/4 riscv: Use variable this_cpu instead of smp_processor_id()

In reply to Greentime Hu

To clean up codes in smpboot.c by using a variable this_cpu to prevent too many calls for smp_processor_id()

---
 arch/riscv/kernel/smpboot.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
arch/riscv/kernel/smpboot.c +9 -3
index 8bc01f0ca73b..ceef7c579aeb 100644
@@ -46,6 +46,9 @@ void __init smp_prepare_boot_cpu(void)
 void __init smp_prepare_cpus(unsigned int max_cpus)
 {
 	int cpuid;
+	unsigned int this_cpu;
+
+	this_cpu = smp_processor_id();
 
 	/* This covers non-smp usecase mandated by "nosmp" option */
 	if (max_cpus == 0)
@@ -137,6 +140,9 @@ void __init smp_cpus_done(unsigned int max_cpus)
 asmlinkage __visible void __init smp_callin(void)
 {
 	struct mm_struct *mm = &init_mm;
+	unsigned int this_cpu;
+
+	this_cpu = smp_processor_id();
 
 	if (!IS_ENABLED(CONFIG_RISCV_SBI))
 		clint_clear_ipi(cpuid_to_hartid_map(smp_processor_id()));
@@ -146,9 +152,9 @@ asmlinkage __visible void __init smp_callin(void)
 	current->active_mm = mm;
 
 	trap_init();
-	notify_cpu_starting(smp_processor_id());
-	update_siblings_masks(smp_processor_id());
-	set_cpu_online(smp_processor_id(), 1);
+	notify_cpu_starting(this_cpu);
+	update_siblings_masks(this_cpu);
+	set_cpu_online(this_cpu, true);
 	/*
 	 * Remote TLB flushes are ignored while the CPU is offline, so emit
 	 * a local TLB flush right now just in case.
Signature · 1 line
2.17.1
#4 lore raw

RFC PATCH v2 4/4 riscv: Add numa support for riscv64 platform

In reply to Greentime Hu

This implementation is based on arm64 porting. It is tested with qemu-system-riscv64, SiFive Unleashed board and OmniXtend FPGA platform.

There will be two nodes in /sys/devices/system/node if it is described in dts and CONFIG_NUMA is enabled. We can use numastat/numactl/numademo to see its status.

Changes in v2:
- split this patch to more patches to be more readable
- set cpu->hotplugable to 0 since it is not supported yet
- add more explanation for moving unflatten_device_tree() to paging_init()

---
 arch/riscv/Kconfig              |  30 ++-
 arch/riscv/include/asm/mmzone.h |  13 ++
 arch/riscv/include/asm/numa.h   |  46 ++++
 arch/riscv/include/asm/pci.h    |  10 +
 arch/riscv/kernel/setup.c       |  25 +++
 arch/riscv/kernel/smpboot.c     |   8 +
 arch/riscv/mm/Makefile          |   1 +
 arch/riscv/mm/init.c            |   2 +
 arch/riscv/mm/numa.c            | 372 ++++++++++++++++++++++++++++++++
 9 files changed, 506 insertions(+), 1 deletion(-)
 create mode 100644 arch/riscv/include/asm/mmzone.h
 create mode 100644 arch/riscv/include/asm/numa.h
 create mode 100644 arch/riscv/mm/numa.c
arch/riscv/Kconfig +29 -1
index d8efbaa78d67..5531913edab2 100644
@@ -22,7 +22,6 @@ config RISCV
 	select CLONE_BACKWARDS
 	select COMMON_CLK
 	select GENERIC_CLOCKEVENTS
-	select GENERIC_CPU_DEVICES
 	select GENERIC_IRQ_SHOW
 	select GENERIC_PCI_IOMAP
 	select GENERIC_SCHED_CLOCK
@@ -253,6 +252,35 @@ config TUNE_GENERIC
 	bool "generic"
 
 endchoice
+# Common NUMA Features
+config NUMA
+	bool "Numa Memory Allocation and Scheduler Support"
+	select OF_NUMA
+	select ARCH_SUPPORTS_NUMA_BALANCING
+	depends on SPARSEMEM
+	help
+	  Enable NUMA (Non-Uniform Memory Access) support.
+
+	  The kernel will try to allocate memory used by a CPU on the
+	  local memory of the CPU and add some more
+	  NUMA awareness to the kernel.
+
+config NODES_SHIFT
+	int "Maximum NUMA Nodes (as a power of 2)"
+	range 1 10
+	default "2"
+	depends on NEED_MULTIPLE_NODES
+	help
+	  Specify the maximum number of NUMA Nodes available on the target
+	  system.  Increases memory reserved to accommodate various tables.
+
+config USE_PERCPU_NUMA_NODE_ID
+	def_bool y
+	depends on NUMA
+
+config NEED_PER_CPU_EMBED_FIRST_CHUNK
+	def_bool y
+	depends on NUMA
 
 config RISCV_ISA_C
 	bool "Emit compressed instructions when building Linux"
arch/riscv/include/asm/mmzone.h new file +13 -0
new file mode 100644
index 000000000000..fa17e01d9ab2
@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __ASM_MMZONE_H
+#define __ASM_MMZONE_H
+
+#ifdef CONFIG_NUMA
+
+#include <asm/numa.h>
+
+extern struct pglist_data *node_data[];
+#define NODE_DATA(nid)		(node_data[(nid)])
+
+#endif /* CONFIG_NUMA */
+#endif /* __ASM_MMZONE_H */
arch/riscv/include/asm/numa.h new file +46 -0

46 lines not on this page · raw message

arch/riscv/include/asm/pci.h +10 -0
index 1c473a1bd986..d6a0e59638c0 100644
@@ -32,6 +32,16 @@ static inline int pci_proc_domain(struct pci_bus *bus)
 	/* always show the domain in /proc */
 	return 1;
 }
+
+#ifdef	CONFIG_NUMA
+int pcibus_to_node(struct pci_bus *bus);
+#ifndef cpumask_of_pcibus
+#define cpumask_of_pcibus(bus)	(pcibus_to_node(bus) == -1 ?		\
+				 cpu_all_mask :				\
+				 cpumask_of_node(pcibus_to_node(bus)))
+#endif
+#endif	/* CONFIG_NUMA */
+
 #endif  /* CONFIG_PCI */
 
 #endif  /* _ASM_RISCV_PCI_H */
arch/riscv/kernel/setup.c +25 -0
index bcd85f502fb2..fd44eaf2051d 100644
@@ -54,6 +54,31 @@ void __init parse_dtb(void)
 #endif
 }
 
+static DEFINE_PER_CPU(struct cpu, cpu_devices);
+
+static int __init topology_init(void)
+{
+	int i, ret;
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+	for_each_online_node(i)
+		register_one_node(i);
+#endif
+
+	for_each_possible_cpu(i) {
+		struct cpu *cpu = &per_cpu(cpu_devices, i);
+
+		cpu->hotpluggable = 0;
+		ret = register_cpu(cpu, i);
+		if (unlikely(ret))
+			pr_warn("Warning: %s: register_cpu %d failed (%d)\n",
+			       __func__, i, ret);
+	}
+
+	return 0;
+}
+subsys_initcall(topology_init);
+
 void __init setup_arch(char **cmdline_p)
 {
 	init_mm.start_code = (unsigned long) _stext;
arch/riscv/kernel/smpboot.c +8 -0
index ceef7c579aeb..9e911774f4b8 100644
@@ -27,6 +27,7 @@
 #include <asm/clint.h>
 #include <asm/irq.h>
 #include <asm/mmu_context.h>
+#include <asm/numa.h>
 #include <asm/tlbflush.h>
 #include <asm/sections.h>
 #include <asm/sbi.h>
@@ -49,6 +50,8 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 	unsigned int this_cpu;
 
 	this_cpu = smp_processor_id();
+	numa_store_cpu_info(this_cpu);
+	numa_add_cpu(this_cpu);
 
 	/* This covers non-smp usecase mandated by "nosmp" option */
 	if (max_cpus == 0)
@@ -58,6 +61,7 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 		if (cpuid == smp_processor_id())
 			continue;
 		set_cpu_present(cpuid, true);
+		numa_store_cpu_info(cpuid);
 	}
 }
 
@@ -76,6 +80,7 @@ void __init setup_smp(void)
 		if (hart == cpuid_to_hartid_map(0)) {
 			BUG_ON(found_boot_cpu);
 			found_boot_cpu = 1;
+			early_map_cpu_to_node(0, of_node_to_nid(dn));
 			continue;
 		}
 		if (cpuid >= NR_CPUS) {
@@ -85,6 +90,7 @@ void __init setup_smp(void)
 		}
 
 		cpuid_to_hartid_map(cpuid) = hart;
+		early_map_cpu_to_node(cpuid, of_node_to_nid(dn));
 		cpuid++;
 	}
 
@@ -153,6 +159,8 @@ asmlinkage __visible void __init smp_callin(void)
 
 	trap_init();
 	notify_cpu_starting(this_cpu);
+	numa_store_cpu_info(this_cpu);
+	numa_add_cpu(this_cpu);
 	update_siblings_masks(this_cpu);
 	set_cpu_online(this_cpu, true);
 	/*
arch/riscv/mm/Makefile +1 -0
index a1bd95c8047a..bada5c6d1a45 100644
@@ -15,3 +15,4 @@ ifeq ($(CONFIG_MMU),y)
 obj-$(CONFIG_SMP) += tlbflush.o
 endif
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
+obj-$(CONFIG_NUMA)         += numa.o
arch/riscv/mm/init.c +2 -0
index fbe69fe47806..589f3ef505c7 100644
@@ -18,6 +18,7 @@
 #include <asm/sections.h>
 #include <asm/pgtable.h>
 #include <asm/io.h>
+#include <asm/numa.h>
 
 #include "../kernel/head.h"
 
@@ -492,6 +493,7 @@ void __init paging_init(void)
 {
 	setup_vm_final();
 	unflatten_device_tree();
+	riscv_numa_init();
 	memblocks_present();
 	sparse_init();
 	setup_zero_page();
arch/riscv/mm/numa.c new file +372 -0

372 lines not on this page · raw message

Signature · 1 line
2.17.1