zone watermark水位控制

Posted 2022-06-11 Loopers

本节我们来分析下zone的水位控制，在zone那一节中，我们将重点放在了free_area中，故意没有分析zone中的水位控制，本节在重点分析zone中的水位控制。

struct zone 
 
    /* zone watermarks, access with *_wmark_pages(zone) macros */
    unsigned long watermark[NR_WMARK];
 
    unsigned long       managed_pages;
    unsigned long       spanned_pages;
    unsigned long       present_pages;

 
enum zone_watermarks 
    WMARK_MIN,
    WMARK_LOW,
    WMARK_HIGH,
    NR_WMARK
;
 
#define min_wmark_pages(z) (z->watermark[WMARK_MIN])
#define low_wmark_pages(z) (z->watermark[WMARK_LOW])
#define high_wmark_pages(z) (z->watermark[WMARK_HIGH])

每个zone中存在三个水位值。MIN水位，LOW水位，HIGH水位。比如下图的NORMAL_ZONE中存在HIGH,LOW,MIN水位三个值

那HIGH，LOW，MIN三个水位值是如何计算出来的呢？ 先来看如下几个字段的含义

/*
 * spanned_pages is the total pages spanned by the zone, including
 * holes, which is calculated as:
 *  spanned_pages = zone_end_pfn - zone_start_pfn;
 *
 * present_pages is physical pages existing within the zone, which
 * is calculated as:
 *  present_pages = spanned_pages - absent_pages(pages in holes);
 *
 * managed_pages is present pages managed by the buddy system, which
 * is calculated as (reserved_pages includes pages allocated by the
 * bootmem allocator):
 *  managed_pages = present_pages - reserved_pages;
 */
unsigned long       managed_pages;
unsigned long       spanned_pages;
unsigned long       present_pages;

• spanned_pages:  代表的是这个zone中所有的页，包含空洞，计算公式是:   zone_end_pfn - zone_start_pfn 
• present_pages： 代表的是这个zone中可用的所有物理页，计算公式是：spanned_pages-hole_pages
• managed_pages： 代表的是通过buddy管理的所有可用的页，计算公式是：present_pages - reserved_pages
• 三者的关系是： spanned_pages > present_pages > managed_pages

可以看下我手上的机器的zoneinfo信息

root:/ # cat /proc/zoneinfo
pages free     87512
        min      2702
        low      10899
        high     11574
        spanned  2094976     ==>  spanned_pages
        present  1959807     ==>  present_pages
        managed  1911161     ==>  managed_pages
        protection: (0, 0)

知道上述了三个值之后，我们来看源代码分析下HIGH，MIN，LOW水位的值是如何设置的

 

min_free_kbytes

在知道HIGH，MIN，LOW水位是如何获得的，先要知道min_free_kbytes值的含义

kernel/msm-4.19/Documentation/sysctl/vm.txt
 
 
min_free_kbytes:
 
This is used to force the Linux VM to keep a minimum number
of kilobytes free.  The VM uses this number to compute a
watermark[WMARK_MIN] value for each lowmem zone in the system.
Each lowmem zone gets a number of reserved free pages based
proportionally on its size.
 
Some minimal amount of memory is needed to satisfy PF_MEMALLOC
allocations; if you set this to lower than 1024KB, your system will
become subtly broken, and prone to deadlock under high loads.
 
Setting this too high will OOM your machine instantly.

从上面的话得出了以下几点：

• min_free_kbyes代表的是系统保留空闲内存的最低限
• watermark[WMARK_MIN]的值是通过min_free_kbytes计算出来的

我们来看下代码是如何计算的：

unsigned long nr_free_buffer_pages(void)

    return nr_free_zone_pages(gfp_zone(GFP_USER));

此函数计算DMA_ZONE和NORAML_ZONE中超过高水位页的个数

static unsigned long nr_free_zone_pages(int offset)

    struct zoneref *z;
    struct zone *zone;
 
    /* Just pick one node, since fallback list is circular */
    unsigned long sum = 0;
 
    struct zonelist *zonelist = node_zonelist(numa_node_id(), GFP_KERNEL);
 
    for_each_zone_zonelist(zone, z, zonelist, offset) 
        unsigned long size = zone->managed_pages;
        unsigned long high = high_wmark_pages(zone);
        if (size > high)
            sum += size - high;
    
 
    return sum;

对每个zone做计算，将每个zone中超过high水位的值放到sum中。超过高水位的页数计算方法是：managed_pages减去watermark[HIGH]， 这样就可以获取到系统中各个zone超过高水位页的总和

int __meminit init_per_zone_wmark_min(void)

    unsigned long lowmem_kbytes;
    int new_min_free_kbytes;
 
    lowmem_kbytes = nr_free_buffer_pages() * (PAGE_SIZE >> 10);
    new_min_free_kbytes = int_sqrt(lowmem_kbytes * 16);
 
    if (new_min_free_kbytes > user_min_free_kbytes) 
        min_free_kbytes = new_min_free_kbytes;
        if (min_free_kbytes < 128)
            min_free_kbytes = 128;
        if (min_free_kbytes > 65536)
            min_free_kbytes = 65536;
     else 
        pr_warn("min_free_kbytes is not updated to %d because user defined value %d is preferred\\n",
                new_min_free_kbytes, user_min_free_kbytes);
    
    setup_per_zone_wmarks();
    refresh_zone_stat_thresholds();
    setup_per_zone_lowmem_reserve();
    setup_per_zone_inactive_ratio();
    return 0;

• lowmem_kbytes： 代表的意思是lowmem中超过高水位的页的总和，这里的单位是kbytes, 这就是lowmem中超过high水位的页乘以4这就是lowmem_kbytes
• new_min_free_kbytes = sqrt(lowmem_kbytes * 16) = √(lowmem_kbytes * 16)
• min_free_kbytes最小不能小于128K，最大不能超过65536K
• 到这里位置我们已经得出了min_free_kbytes的值

假设当前只有一个NORAML_ZONE，根据上面的公式managed_pages - watermark[WMARK_HIGH]的值就是这个NORMAL_ZONE中空闲的页面，然后经过int_sqrt可以计算出系统中可用空闲页面的最低限

举例： 

root:/ # cat /proc/zoneinfo
Node 0, zone   Normal
 pages free     82075
        min      2702
        low      10899
        high     11574
        spanned  2094976
        present  1959807
        managed  1911161
        protection: (0, 0)
 
 
root:/ # cat /proc/sys/vm/min_free_kbytes
10811

根据上面的数据，我们来推到下min_free_kbytes

lowmem_kbytes = （1911161 - 11574 ）*4= 1899587 *4 = 7598348

min_free_kbytes = int_sqrt(7598348 * 16) =  11026 ~= 10811

 

建立各个zone的水位值

原生的kernel代码，在kernel.org上下载的4.4的代码。

static void __setup_per_zone_wmarks(void)

    unsigned long pages_min = min_free_kbytes >> (PAGE_SHIFT - 10);
    unsigned long lowmem_pages = 0;
    struct zone *zone;
    unsigned long flags;
 
    /* Calculate total number of !ZONE_HIGHMEM pages */
    for_each_zone(zone) 
        if (!is_highmem(zone))
            lowmem_pages += zone->managed_pages;
    
 
    for_each_zone(zone) 
        u64 tmp;
 
        spin_lock_irqsave(&zone->lock, flags);
        tmp = (u64)pages_min * zone->managed_pages;
        do_div(tmp, lowmem_pages);
        if (is_highmem(zone)) 
            /*
             * __GFP_HIGH and PF_MEMALLOC allocations usually don't
             * need highmem pages, so cap pages_min to a small
             * value here.
             *
             * The WMARK_HIGH-WMARK_LOW and (WMARK_LOW-WMARK_MIN)
             * deltas control asynch page reclaim, and so should
             * not be capped for highmem.
             */
            unsigned long min_pages;
 
            min_pages = zone->managed_pages / 1024;
            min_pages = clamp(min_pages, SWAP_CLUSTER_MAX, 128UL);
            zone->watermark[WMARK_MIN] = min_pages;
         else 
            /*
             * If it's a lowmem zone, reserve a number of pages
             * proportionate to the zone's size.
             */
            zone->watermark[WMARK_MIN] = tmp;
        
 
        zone->watermark[WMARK_LOW]  = min_wmark_pages(zone) + (tmp >> 2);
        zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp >> 1);
 
        __mod_zone_page_state(zone, NR_ALLOC_BATCH,
            high_wmark_pages(zone) - low_wmark_pages(zone) -
            atomic_long_read(&zone->vm_stat[NR_ALLOC_BATCH]));
 
        spin_unlock_irqrestore(&zone->lock, flags);
    
 
    /* update totalreserve_pages */
    calculate_totalreserve_pages();

• pages_min = min_free_kbytes >> (PAGE_SHIFT - 10);   //将最小保留的内存转化为以page为单位，最小预留的空闲页
• lowmem_pages： 代表的是除过HIGHMEM_ZONE的所有zone的managed_pages
• tmp = (u64)pages_min * zone→managed_pages; //用每个zone的manage_present的页数乘以pages_min的值
• do_div(tmp, lowmem_pages);  //取商，得到的值就是min的值
• zone->watermark[WMARK_MIN] = tmp;  //设置MIN水位的值
• zone->watermark[WMARK_LOW]  = min_wmark_pages(zone) + (tmp >> 2);  //LOW水位的值=MIN+MIN/4=125%MIN
• zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp >> 1);  //HIGH水位的值=MIN + MIN/2 = 150%MIN

举例说明：

root@root:~$ cat /proc/zoneinfo      //ubuntu机器
Node 0, zone      DMA 
pages free     3973
        min      16
        low      20
        high     24
        spanned  4095
        present  3998
        managed  3973

LOW = MIN + MIN/4 = 16 + 16/4 = 16 + 4 =20

HIGH = MIN + MIN/2 = 16 + 16/2 = 16+8 = 24

 

Android机器上水位的计算

上面我们详细计算了ubuntu机器上MIN, LOW, HIGH水位的计算方法，我们这里详细描述下android机器上水位的计算，同时也需要了解下android上为啥会出现不一样的算法

这个图画的非常的清楚很直观

• 我们分配页第一次尝试是从LOW水位开始分配的，当所剩余的空闲页小于LOW水位的时候，则会唤醒Kswapd内核线程进行内存回收
• 如果回收内存效果很显著，当空闲页大于HIGH水位的时候，则会停止Kswapd内核线程回收
• 如果回收内存效果不明显，当空闲内存直接小于MIN水位的时候，则会进行直接的内存回收(Direct-reclaim)，这样空闲内存就会逐渐增大
• 当回收效果依然不明显的时候，则会启动OOM杀死进程

 

明白了上述的背后原理后，我们接着分析下android版本上的水位计算方法

比如上面ubunut的水位，其中min=16, low=20, high=24

比如当前空闲内存是在LOW水位以下MIN以上，这时候后台会启动Kswaped内核线程在进程内存回收，假设这时候突然有一个很大的进程需要很大的内存请求，这样一来Kswaped回收速度赶不上分配速度，内存一下掉到了MIN水位，这样直接就进行了直接回收，直接回收很影响系统的性能的。这样看来linux原生的代码涉及MIN-LOW之间的间隙太小，很容易导致进入直接回收的情况的。所以在android的版本上增加了一个变量：extra_free_kbytes

static void __setup_per_zone_wmarks(void)

    unsigned long pages_min = min_free_kbytes >> (PAGE_SHIFT - 10);
    unsigned long pages_low = extra_free_kbytes >> (PAGE_SHIFT - 10);
    unsigned long lowmem_pages = 0;
    struct zone *zone;
    unsigned long flags;
 
    /* Calculate total number of !ZONE_HIGHMEM pages */
    for_each_zone(zone) 
        if (!is_highmem(zone))
            lowmem_pages += zone->managed_pages;
    
 
    for_each_zone(zone) 
        u64 min, low;
 
        spin_lock_irqsave(&zone->lock, flags);
        min = (u64)pages_min * zone->managed_pages;
        do_div(min, lowmem_pages);
        low = (u64)pages_low * zone->managed_pages;
        do_div(low, vm_total_pages);
 
        if (is_highmem(zone)) 
            /*
             * __GFP_HIGH and PF_MEMALLOC allocations usually don't
             * need highmem pages, so cap pages_min to a small
             * value here.
             *
             * The WMARK_HIGH-WMARK_LOW and (WMARK_LOW-WMARK_MIN)
             * deltas control asynch page reclaim, and so should
             * not be capped for highmem.
             */
            unsigned long min_pages;
 
            min_pages = zone->managed_pages / 1024;
            min_pages = clamp(min_pages, SWAP_CLUSTER_MAX, 128UL);
            zone->_watermark[WMARK_MIN] = min_pages;
         else 
            /*
             * If it's a lowmem zone, reserve a number of pages
             * proportionate to the zone's size.
             */
            zone->_watermark[WMARK_MIN] = min;
        
 
        /*
         * Set the kswapd watermarks distance according to the
         * scale factor in proportion to available memory, but
         * ensure a minimum size on small systems.
         */
        min = max_t(u64, min >> 2,
                mult_frac(zone->managed_pages,
                      watermark_scale_factor, 10000));
 
        zone->watermark_boost = 0;
        zone->_watermark[WMARK_LOW]  = min_wmark_pages(zone) +
                    low + min;
        zone->_watermark[WMARK_HIGH] = min_wmark_pages(zone) +
                    low + min * 2;
 
        spin_unlock_irqrestore(&zone->lock, flags);
    
 
    /* update totalreserve_pages */
    calculate_totalreserve_pages();

extra_free_kbytes此值就是用来增加MIN和LOW之间的间隙的。我们再来看下android版本下的水位计算

• lowmem_pages += zone→managed_pages;    //所有lowmem zone的managed_pages总和
• min = (u64)pages_min * zone→managed_pages;  
• do_div(min, lowmem_pages);      //min水位的的计算方法一样
• low = (u64)pages_low * zone→managed_pages;
• do_div(low, vm_total_pages);       //low水位的值是通过extra_free_kbytes计算的
• zone->_watermark[WMARK_MIN] = min;    //min水位
• zone->_watermark[WMARK_LOW]  = min_wmark_pages(zone) + low + min = 2min + low
• zone->_watermark[WMARK_HIGH] = min_wmark_pages(zone) + low + 2*min = 3min + low

 

android手机举例说明：

root:/ # cat /proc/zoneinfo
Node 0, zone   Normal
  pages free     56198
        min      2702
        low      10899
        high     11574
        spanned  2094976
        present  1959807
        managed  1911161
        protection: (0, 0)
 
 
root:/ # cat /proc/sys/vm/extra_free_kbytes
30375

可以看到min和low之间的距离 已经差的很大了

开发者涨薪指南 48位大咖的思考法则、工作方式、逻辑体系