LoongArch: Add Non-Uniform Memory Access (NUMA) support

Add Non-Uniform Memory Access (NUMA) support for LoongArch. LoongArch
has 48-bit physical address, but the HyperTransport I/O bus only support
40-bit address, so we need a custom phys_to_dma() and dma_to_phys() to
extract the 4-bit node id (bit 44~47) from Loongson-3's 48-bit physical
address space and embed it into 40-bit. In the 40-bit dma address, node
id offset can be read from the LS7A_DMA_CFG register.

Reviewed-by: WANG Xuerui <git@xen0n.name>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>

1 files changed, 21 insertions, 0 deletions

diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
index 85ef258d36d2..80657bf83b05 100644
--- a/arch/loongarch/Kconfig
+++ b/arch/loongarch/Kconfig
@@ -7,6 +7,7 @@ config LOONGARCH
 	select ARCH_ENABLE_MEMORY_HOTPLUG
 	select ARCH_ENABLE_MEMORY_HOTREMOVE
 	select ARCH_HAS_ACPI_TABLE_UPGRADE	if ACPI
+	select ARCH_HAS_PHYS_TO_DMA
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_INLINE_READ_LOCK if !PREEMPTION
@@ -41,6 +42,7 @@ config LOONGARCH
 	select ARCH_SUPPORTS_ACPI
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select ARCH_SUPPORTS_HUGETLBFS
+	select ARCH_SUPPORTS_NUMA_BALANCING
 	select ARCH_USE_BUILTIN_BSWAP
 	select ARCH_USE_CMPXCHG_LOCKREF
 	select ARCH_USE_QUEUED_RWLOCKS
@@ -91,12 +93,15 @@ config LOONGARCH
 	select HAVE_PERF_EVENTS
 	select HAVE_REGS_AND_STACK_ACCESS_API
 	select HAVE_RSEQ
+	select HAVE_SETUP_PER_CPU_AREA if NUMA
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_TIF_NOHZ
 	select HAVE_VIRT_CPU_ACCOUNTING_GEN if !SMP
 	select IRQ_FORCED_THREADING
 	select IRQ_LOONGARCH_CPU
 	select MODULES_USE_ELF_RELA if MODULES
+	select NEED_PER_CPU_EMBED_FIRST_CHUNK
+	select NEED_PER_CPU_PAGE_FIRST_CHUNK
 	select OF
 	select OF_EARLY_FLATTREE
 	select PERF_USE_VMALLOC
@@ -105,6 +110,7 @@ config LOONGARCH
 	select SYSCTL_EXCEPTION_TRACE
 	select SWIOTLB
 	select TRACE_IRQFLAGS_SUPPORT
+	select USE_PERCPU_NUMA_NODE_ID
 	select ZONE_DMA32
 
 config 32BIT
@@ -335,6 +341,20 @@ config NR_CPUS
 	  This allows you to specify the maximum number of CPUs which this
 	  kernel will support.
 
+config NUMA
+	bool "NUMA Support"
+	select ACPI_NUMA if ACPI
+	help
+	  Say Y to compile the kernel with NUMA (Non-Uniform Memory Access)
+	  support.  This option improves performance on systems with more
+	  than one NUMA node; on single node systems it is generally better
+	  to leave it disabled.
+
+config NODES_SHIFT
+	int
+	default "6"
+	depends on NUMA
+
 config FORCE_MAX_ZONEORDER
 	int "Maximum zone order"
 	range 14 64 if PAGE_SIZE_64KB
@@ -381,6 +401,7 @@ config ARCH_SELECT_MEMORY_MODEL
 
 config ARCH_FLATMEM_ENABLE
 	def_bool y
+	depends on !NUMA
 
 config ARCH_SPARSEMEM_ENABLE
 	def_bool y