内核解读之内存管理（2）内存管理三级架构之内存结点node

文章目录

• • 0、概述
  • 1、内存节点node
  • 2、NUMA的初始化

0、概述

结合NUMA的架构，Linux抽象出了三级内存管理架构：内存节点node、内存区域zone和物理页框page。

在NUMA模型中，每个CPU都有自己的本地内存节点（memory node），而且还可以通过QPI总线访问其他CPU下挂的内存节点，只是访问本地内存要比访问其他CPU下的内存的速度高许多，一般经过一次QPI要增加30%的访问时延。

内存节点node是为了解决多处理器内存访问竞争的问题，而内存区域zone是为了解决32位系统内核只有1G的虚拟地址空间，无法管理大于1G的物理内存这个问题，zone按照用途划分成几种类型，比如低端内存区，高端内存区，DMA内存区等。

图 内核内存管理三级架构

1、内存节点node

NUMA结构下, 每个处理器与一个本地内存直接相连, 而不同处理器之间则通过总线进行进一步的连接。linux内核把物理内存按照CPU节点划分为不同的node, 每个node作为某个cpu结点的本地内存, 而作为其他CPU节点的远程内存, 而UMA结构下, 则任务系统中只存在一个内存node, 这样对于UMA结构来说, 内核把内存当成只有一个内存node节点的伪NUMA。因此，NUMA和UMA的管理方式便一致了。

Linux内核中使用数据结构pg_data_t来表示内存节点node，我们把它叫做结点描述符。如常用的ARM架构为UMA架构。对于UMA架构只有一个内存节点，对于NUMA架构有多个内存节点。

在numa.h中有如下定义：

#ifdef CONFIG_NODES_SHIFT#define NODES_SHIFT     CONFIG_NODES_SHIFT#else#define NODES_SHIFT     0#endif#define MAX_NUMNODES    (1 << NODES_SHIFT)

CONFIG_NODES_SHIFT是由用户配置的内存节点的数目。可以看到，对于UMA架构，MAX_NUMNODES等于1。

include/linux/mmzone.h:

/* * On NUMA machines, each NUMA node would have a pg_data_t to describe * it's memory layout. On UMA machines there is a single pglist_data which * describes the whole memory. * * Memory statistics and page replacement data structures are maintained on a * per-zone basis. */typedef struct pglist_data {/* * node_zones contains just the zones for THIS node. Not all of the * zones may be populated, but it is the full list. It is referenced by * this node's node_zonelists as well as other node's node_zonelists. */struct zone node_zones[MAX_NR_ZONES];    /* node_zones是一个数组，包含节点中各内存区(ZONE_DMA, ZONE_DMA32, ZONE_NORMAL...)的描述符。*//* * node_zonelists contains references to all zones in all nodes. * Generally the first zones will be references to this node's * node_zones. */struct zonelist node_zonelists[MAX_ZONELISTS]; /* 指定了节点的备用zone列表 */   /*struct zonelist {struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];}该结构包含了类型为 struct zoneref 的一个备用列表，由于该备用列表必须包括所有结点的所有内存域，因此由 MAX_NUMNODES * MAX_NZ_ZONES 项组成，外加一个用于标记列表结束的空指针。/* Maximum number of zones on a zonelist */#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)/* * This struct contains information about a zone in a zonelist. It is stored * here to avoid dereferences into large structures and lookups of tables */struct zoneref {struct zone *zone;/* Pointer to actual zone */int zone_idx;/* zone_idx(zoneref->zone) */};*/        int nr_zones; /* number of populated zones in this node */ /*指示了节点中zone的数目*/#ifdef CONFIG_FLATMEM/* means !SPARSEMEM */struct page *node_mem_map; /* 平铺式内存模式下，物理page数组，linux为每个物理页分配了一个struct page的管理结构体，并形成了一个结构体数组，node_mem_map即为数组的指针；pfn_to_page和page_to_pfn都借助该数组实 */#ifdef CONFIG_PAGE_EXTENSIONstruct page_ext *node_page_ext;#endif#endif#if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)/* * Must be held any time you expect node_start_pfn, * node_present_pages, node_spanned_pages or nr_zones to stay constant. * Also synchronizes pgdat->first_deferred_pfn during deferred page * init. * * pgdat_resize_lock() and pgdat_resize_unlock() are provided to * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG * or CONFIG_DEFERRED_STRUCT_PAGE_INIT. * * Nests above zone->lock and zone->span_seqlock */spinlock_t node_size_lock;#endifunsigned long node_start_pfn;   /* 节点第一页帧逻辑编号 */unsigned long node_present_pages; /* total number of physical pages */ /* 节点中物理页帧总数目 */unsigned long node_spanned_pages; /* total size of physical page     range, including holes */   /* 按照平铺计算的节点物理页帧总数目。由于空洞的存在可能不等于node_present_pages，应该是大于等于node_present_pages。*/int node_id; /*结点id*/wait_queue_head_t kswapd_wait; /* kswapd页换出守护进程使用的等待队列 */wait_queue_head_t pfmemalloc_wait;/* workqueues for throttling reclaim for different reasons. */wait_queue_head_t reclaim_wait[NR_VMSCAN_THROTTLE];atomic_t nr_writeback_throttled;/* nr of writeback-throttled tasks */unsigned long nr_reclaim_start;/* nr pages written while throttled * when throttling started. */#ifdef CONFIG_MEMORY_HOTPLUGstruct mutex kswapd_lock;#endifstruct task_struct *kswapd;/* Protected by kswapd_lock */ /* 指针指向kswapd内核线程的进程描述符 */    /* 每个结点都有一个内核进程kswapd，它的作用就是将进程或内核持有的，但是不常用的页交换到磁盘上，以腾出更多可用内存。不信你可以ps看一下。*/int kswapd_order; /* 需要释放的区域的长度，以页阶为单位 */enum zone_type kswapd_highest_zoneidx;int kswapd_failures;/* Number of 'reclaimed == 0' runs */#ifdef CONFIG_COMPACTIONint kcompactd_max_order;enum zone_type kcompactd_highest_zoneidx;wait_queue_head_t kcompactd_wait;struct task_struct *kcompactd;bool proactive_compact_trigger;#endif/* * This is a per-node reserve of pages that are not available * to userspace allocations. */unsigned longtotalreserve_pages;    /* 结点page回收相关 */#ifdef CONFIG_NUMA/* * node reclaim becomes active if more unmapped pages exist. */unsigned longmin_unmapped_pages;unsigned longmin_slab_pages;#endif /* CONFIG_NUMA *//* Write-intensive fields used by page reclaim */CACHELINE_PADDING(_pad1_);#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT/* * If memory initialisation on large machines is deferred then this * is the first PFN that needs to be initialised. */unsigned long first_deferred_pfn;#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */#ifdef CONFIG_TRANSPARENT_HUGEPAGEstruct deferred_split deferred_split_queue;#endif#ifdef CONFIG_NUMA_BALANCING/* start time in ms of current promote rate limit period */unsigned int nbp_rl_start;/* number of promote candidate pages at start time of current rate limit period */unsigned long nbp_rl_nr_cand;/* promote threshold in ms */unsigned int nbp_threshold;/* start time in ms of current promote threshold adjustment period */unsigned int nbp_th_start;/* * number of promote candidate pages at stat time of current promote * threshold adjustment period */unsigned long nbp_th_nr_cand;#endif/* Fields commonly accessed by the page reclaim scanner *//* * NOTE: THIS IS UNUSED IF MEMCG IS ENABLED. * * Use mem_cgroup_lruvec() to look up lruvecs. */struct lruvec__lruvec;unsigned longflags; /* 结点标记 */#ifdef CONFIG_LRU_GEN/* kswap mm walk data */struct lru_gen_mm_walkmm_walk;#endifCACHELINE_PADDING(_pad2_);/* Per-node vmstats */struct per_cpu_nodestat __percpu *per_cpu_nodestats;atomic_long_tvm_stat[NR_VM_NODE_STAT_ITEMS];#ifdef CONFIG_NUMAstruct memory_tier __rcu *memtier;#endif} pg_data_t;

在UMA结构的机器中, 只有一个node结点contig_page_data, 此时NODE_DATA直接指向了全局的contig_page_data, 而与node的编号nid无关,其中全局唯一的内存node结点contig_page_data定义在include/linux/mmzone.h:。

#ifndef CONFIG_NUMAextern struct pglist_data contig_page_data;static inline struct pglist_data *NODE_DATA(int nid){return &contig_page_data;}#else /* CONFIG_NUMA */#include #endif /* !CONFIG_NUMA */

定义了NUMA时，NUMA和平台相关，截取arm64的定义如下：

arch/arm64/include/asm/mmzone.h:

/* SPDX-License-Identifier: GPL-2.0 */#ifndef __ASM_MMZONE_H#define __ASM_MMZONE_H#ifdef CONFIG_NUMA#include extern struct pglist_data *node_data[];#define NODE_DATA(nid)(node_data[(nid)])#endif /* CONFIG_NUMA */#endif /* __ASM_MMZONE_H */

定义了一个全局指针数组node_data用于存放系统的所有node，宏NODE_DATA用于取得指定id的node。

定义了一个全局静态位图数组用于存放各种状态下的node：

/* * Array of node states. */nodemask_t node_states[NR_NODE_STATES] __read_mostly = {[N_POSSIBLE] = NODE_MASK_ALL,[N_ONLINE] = { { [0] = 1UL } },#ifndef CONFIG_NUMA[N_NORMAL_MEMORY] = { { [0] = 1UL } },#ifdef CONFIG_HIGHMEM[N_HIGH_MEMORY] = { { [0] = 1UL } },#endif[N_MEMORY] = { { [0] = 1UL } },[N_CPU] = { { [0] = 1UL } },#endif/* NUMA */};

/* * Bitmasks that are kept for all the nodes. */enum node_states {N_POSSIBLE,/* The node could become online at some point */N_ONLINE,/* The node is online */N_NORMAL_MEMORY,/* The node has regular memory */#ifdef CONFIG_HIGHMEMN_HIGH_MEMORY,/* The node has regular or high memory */#elseN_HIGH_MEMORY = N_NORMAL_MEMORY,#endifN_MEMORY,/* The node has memory(regular, high, movable) */N_CPU,/* The node has one or more cpus */N_GENERIC_INITIATOR,/* The node has one or more Generic Initiators */NR_NODE_STATES};

N_POSSIBLE是假定所有的node都可能存在。

N_ONLINE标记在线的结点。

N_MEMORY标记有内存的结点。

上面的位图数组里，都假定至少有一个node。

结合状态位图数组，提供了遍历特定状态结点的宏：

如果没有启用NUMA：
相当于只有一个结点，循环一次即结束：

#define for_each_node_state(node, __state) \for ( (node) = 0; (node) == 0; (node) = 1)

如果启用了NUMA：

#define for_each_node_state(__node, __state) \for_each_node_mask((__node), node_states[__state])

for_each_node_mask就是去遍历状态位图，挨个取出置位的node：

#if MAX_NUMNODES > 1#define for_each_node_mask(node, mask)    \for ((node) = first_node(mask);    \     (node >= 0) && (node) < MAX_NUMNODES;    \     (node) = next_node((node), (mask)))#else /* MAX_NUMNODES == 1 */#define for_each_node_mask(node, mask)                                  \for ((node) = 0; (node) < 1 && !nodes_empty(mask); (node)++)#endif /* MAX_NUMNODES */

for_each_node是遍历N_POSSIBLE位图，就是遍历可能存在的结点，就是根据配置，所有结点都遍历一遍。

for_each_online_node是遍历N_ONLINE位图，就是遍历在线的结点。

#define for_each_node(node)   for_each_node_state(node, N_POSSIBLE)#define for_each_online_node(node) for_each_node_state(node, N_ONLINE)

2、NUMA的初始化

--->setup_arch  --->bootmem_init    --->arch_numa_init

如果没有启用NUMA，arch_numa_init是个空函数

如果启用了NUMA:

/** * arch_numa_init() - Initialize NUMA * * Try each configured NUMA initialization method until one succeeds. The * last fallback is dummy single node config encompassing whole memory. */void __init arch_numa_init(void){if (!numa_off) {if (!acpi_disabled && !numa_init(arch_acpi_numa_init))return;if (acpi_disabled && !numa_init(of_numa_init))return;}numa_init(dummy_numa_init); // 失败时才走这里，关闭numa}

在没有开启acpi时，就是从设备树读取node信息，然后进行初始化。

of_numa_init：

1、从设备树解析出每个cpu所属的node_id，初始化node全局位图numa_nodes_parsed。

2、从设备树解析出每段内存所属的node_id，注册到全局静态早期内存管理器memblock。

3、从设备树解析出两两结点间的距离，初始化全局静态node距离表numa_distance。

numa_init会调用of_numa_init，将从设备树读取到的结果初始化全局结点状态位图，比如哪些结点在线，以及每个结点的内存起始PFN。

static int __init numa_init(int (*init_func)(void)){int ret;nodes_clear(numa_nodes_parsed);nodes_clear(node_possible_map);nodes_clear(node_online_map);ret = numa_alloc_distance();if (ret < 0)return ret;ret = init_func();if (ret < 0)goto out_free_distance;if (nodes_empty(numa_nodes_parsed)) {pr_info("No NUMA configuration found\n");ret = -EINVAL;goto out_free_distance;}ret = numa_register_nodes();if (ret < 0)goto out_free_distance;setup_node_to_cpumask_map();return 0;out_free_distance:numa_free_distance();return ret;}