注：本文分析基于3.10.0-693.el7内核版本，即CentOS 7.4

1、关于drop_caches

通常在内存不足时，我们习惯通过echo 3 > /proc/sys/vm/drop_caches 的方式手动清理系统缓存，

[root@localhost  ~]# free -m

              total        used        free      shared  buff/cache   available

Mem:           7822        3436        2068          40        2317        3997

Swap:             0           0           0

[root@localhost  ~]# echo 3 > /proc/sys/vm/drop_caches

[root@localhost ~]# free -m

              total        used        free      shared  buff/cache   available

Mem:           7822        3433        4036          40         352        4037

Swap:             0           0           0

对于数字3的含义，我们可以通过内核文档了解其具体含义，

To free pagecache:

	echo 1 > /proc/sys/vm/drop_caches

To free reclaimable slab objects (includes dentries and inodes):

	echo 2 > /proc/sys/vm/drop_caches

To free slab objects and pagecache:

	echo 3 > /proc/sys/vm/drop_caches

2、释放pagecache

在之前我们知道当内存低于某个阈值时，会触发脏页回写，提交回写work到对应BDI设备上，由BDI writebacke进程回写脏页释放内存。这和drop_caches中的echo 1类似，都是释放脏页，因此其最后路径是一致的。

int drop_caches_sysctl_handler(ctl_table *table, int write,

	void __user *buffer, size_t *length, loff_t *ppos)

{

	int ret;
ret <span class="token operator">=</span> <span class="token function">proc_dointvec_minmax</span><span class="token punctuation">(</span>table<span class="token punctuation">,</span> write<span class="token punctuation">,</span> buffer<span class="token punctuation">,</span> length<span class="token punctuation">,</span> ppos<span class="token punctuation">)</span><span class="token punctuation">;</span>

<span class="token keyword">if</span> <span class="token punctuation">(</span>ret<span class="token punctuation">)</span>

	<span class="token keyword">return</span> ret<span class="token punctuation">;</span>

<span class="token keyword">if</span> <span class="token punctuation">(</span>write<span class="token punctuation">)</span> <span class="token punctuation">{<!-- --></span>

	<span class="token keyword">static</span> <span class="token keyword">int</span> stfu<span class="token punctuation">;</span>

    <span class="token comment">// echo 1 &gt; drop_caches</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span>sysctl_drop_caches <span class="token operator">&amp;</span> <span class="token number">1</span><span class="token punctuation">)</span> <span class="token punctuation">{<!-- --></span>

		<span class="token function">iterate_supers</span><span class="token punctuation">(</span>drop_pagecache_sb<span class="token punctuation">,</span> <span class="token constant">NULL</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

		<span class="token function">count_vm_event</span><span class="token punctuation">(</span>DROP_PAGECACHE<span class="token punctuation">)</span><span class="token punctuation">;</span>

	<span class="token punctuation">}</span>

    <span class="token comment">// echo 2 &gt; drop_caches</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span>sysctl_drop_caches <span class="token operator">&amp;</span> <span class="token number">2</span><span class="token punctuation">)</span> <span class="token punctuation">{<!-- --></span>

		<span class="token function">drop_slab</span><span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

		<span class="token function">count_vm_event</span><span class="token punctuation">(</span>DROP_SLAB<span class="token punctuation">)</span><span class="token punctuation">;</span>

	<span class="token punctuation">}</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span><span class="token operator">!</span>stfu<span class="token punctuation">)</span> <span class="token punctuation">{<!-- --></span>

		<span class="token function">pr_info</span><span class="token punctuation">(</span><span class="token string">"%s (%d): drop_caches: %d\n"</span><span class="token punctuation">,</span>

			current<span class="token operator">-&gt;</span>comm<span class="token punctuation">,</span> <span class="token function">task_pid_nr</span><span class="token punctuation">(</span>current<span class="token punctuation">)</span><span class="token punctuation">,</span>

			sysctl_drop_caches<span class="token punctuation">)</span><span class="token punctuation">;</span>

	<span class="token punctuation">}</span>

    <span class="token comment">//置位，否则就一直在回收了</span>

	stfu <span class="token operator">|</span><span class="token operator">=</span> sysctl_drop_caches <span class="token operator">&amp;</span> <span class="token number">4</span><span class="token punctuation">;</span>

<span class="token punctuation">}</span>

<span class="token keyword">return</span> <span class="token number">0</span><span class="token punctuation">;</span>


}

可见，echo 1时，会调用drop_pagecache_sb去释放pagecache，我们继续往下查，

drop_pagecache_sb ->

	iput ->

		iput_final->

			write_inode_now -> #提交writeback_control，立即回写

				writeback_single_inode ->

					__writeback_single_inode ->

						do_writepages #调用对应文件系统的writepage写回磁盘

在BDI回写里，一开始提交的是wb_writeback_work，等到实际要执行回写操作时，都会转换为writeback_control，再去执行回写。

因此，echo 1的操作就是，遍历每个超级块，调用drop_pagecache_sb，drop_pagecache_sb中会遍历该超级块所有的inode，对其关联的pagecache进行回写。与BDI不同的是，该操作是立马执行，不需要等待周期执行或者inode过期。

3、释放slab cache

而对于echo 2的情况，就比较复杂一点，

static void drop_slab(void)

{

	int nr_objects;

	struct shrink_control shrink = {

		.gfp_mask = GFP_KERNEL,

	};

    //上次回收缓存数量高于10，就再进行一次回收

    //这个条件其实挺苛刻的，回收后整个系统空闲slab不会超过10

	do {

		nr_objects = shrink_slab(&shrink, 1000, 1000);

	} while (nr_objects > 10);

}

unsigned long shrink_slab(struct shrink_control shrink,


unsigned long nr_pages_scanned,


unsigned long lru_pages)


{


struct shrinker shrinker;


unsigned long ret = 0;


...


//遍历系统中所有的shrinker，回收各个slab管理区的空闲缓存


list_for_each_entry(shrinker, &shrinker_list, list) {


unsigned long long delta;


long total_scan;


long max_pass;


int shrink_ret = 0;


long nr;


long new_nr;


//获取批处理数量，默认每次回收128，对于超级块而言是1024


long batch_size = shrinker->batch ? shrinker->batch


: SHRINK_BATCH;


//获取该slab管理区可回收的缓存数量


max_pass = do_shrinker_shrink(shrinker, shrink, 0);


if (max_pass <= 0)


continue;
	nr <span class="token operator">=</span> <span class="token function">atomic_long_xchg</span><span class="token punctuation">(</span><span class="token operator">&amp;</span>shrinker<span class="token operator">-&gt;</span>nr_in_batch<span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

	total_scan <span class="token operator">=</span> nr<span class="token punctuation">;</span>

    <span class="token comment">//计算该slab管理区此次缓存回收额度，一堆操作</span>

    <span class="token comment">//针对手动释放缓存的场景，基本上是两倍的max_pass，也就是尽可能去释放</span>

    <span class="token comment">//对于kswap或其他路径上，不会超过一倍的max_pass</span>

	delta <span class="token operator">=</span> <span class="token punctuation">(</span><span class="token number">4</span> <span class="token operator">*</span> nr_pages_scanned<span class="token punctuation">)</span> <span class="token operator">/</span> shrinker<span class="token operator">-&gt;</span>seeks<span class="token punctuation">;</span>

	delta <span class="token operator">*</span><span class="token operator">=</span> max_pass<span class="token punctuation">;</span>

	<span class="token function">do_div</span><span class="token punctuation">(</span>delta<span class="token punctuation">,</span> lru_pages <span class="token operator">+</span> <span class="token number">1</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

	total_scan <span class="token operator">+</span><span class="token operator">=</span> delta<span class="token punctuation">;</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span>total_scan <span class="token operator">&lt;</span> <span class="token number">0</span><span class="token punctuation">)</span> <span class="token punctuation">{<!-- --></span>

		<span class="token function">printk</span><span class="token punctuation">(</span>KERN_ERR <span class="token string">"shrink_slab: %pF negative objects to "</span>

		       <span class="token string">"delete nr=%ld\n"</span><span class="token punctuation">,</span>

		       shrinker<span class="token operator">-&gt;</span>shrink<span class="token punctuation">,</span> total_scan<span class="token punctuation">)</span><span class="token punctuation">;</span>

		total_scan <span class="token operator">=</span> max_pass<span class="token punctuation">;</span>

	<span class="token punctuation">}</span>

    <span class="token comment">//如果delta偏小，意味着系统中inactive的缓存偏少，我们回收的额度也不能设置太大</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span>delta <span class="token operator">&lt;</span> max_pass <span class="token operator">/</span> <span class="token number">4</span><span class="token punctuation">)</span>

		total_scan <span class="token operator">=</span> <span class="token function">min</span><span class="token punctuation">(</span>total_scan<span class="token punctuation">,</span> max_pass <span class="token operator">/</span> <span class="token number">2</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

    <span class="token comment">//控制回收总额上限，避免死循环</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span>total_scan <span class="token operator">&gt;</span> max_pass <span class="token operator">*</span> <span class="token number">2</span><span class="token punctuation">)</span>

		total_scan <span class="token operator">=</span> max_pass <span class="token operator">*</span> <span class="token number">2</span><span class="token punctuation">;</span>

	<span class="token function">trace_mm_shrink_slab_start</span><span class="token punctuation">(</span>shrinker<span class="token punctuation">,</span> shrink<span class="token punctuation">,</span> nr<span class="token punctuation">,</span>

				nr_pages_scanned<span class="token punctuation">,</span> lru_pages<span class="token punctuation">,</span>

				max_pass<span class="token punctuation">,</span> delta<span class="token punctuation">,</span> total_scan<span class="token punctuation">)</span><span class="token punctuation">;</span>

    <span class="token comment">//循环回收缓存</span>

	<span class="token keyword">while</span> <span class="token punctuation">(</span>total_scan <span class="token operator">&gt;=</span> batch_size<span class="token punctuation">)</span> <span class="token punctuation">{<!-- --></span>

		<span class="token keyword">int</span> nr_before<span class="token punctuation">;</span>

        <span class="token comment">//记录处理前缓存数量</span>

		nr_before <span class="token operator">=</span> <span class="token function">do_shrinker_shrink</span><span class="token punctuation">(</span>shrinker<span class="token punctuation">,</span> shrink<span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

        <span class="token comment">//回收后缓存数量</span>

		shrink_ret <span class="token operator">=</span> <span class="token function">do_shrinker_shrink</span><span class="token punctuation">(</span>shrinker<span class="token punctuation">,</span> shrink<span class="token punctuation">,</span>

						batch_size<span class="token punctuation">)</span><span class="token punctuation">;</span>

		<span class="token keyword">if</span> <span class="token punctuation">(</span>shrink_ret <span class="token operator">==</span> <span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>

			<span class="token keyword">break</span><span class="token punctuation">;</span>

        <span class="token comment">//统计此次回收的缓存数量</span>

		<span class="token keyword">if</span> <span class="token punctuation">(</span>shrink_ret <span class="token operator">&lt;</span> nr_before<span class="token punctuation">)</span>

			ret <span class="token operator">+</span><span class="token operator">=</span> nr_before <span class="token operator">-</span> shrink_ret<span class="token punctuation">;</span>

		<span class="token function">count_vm_events</span><span class="token punctuation">(</span>SLABS_SCANNED<span class="token punctuation">,</span> batch_size<span class="token punctuation">)</span><span class="token punctuation">;</span>

        <span class="token comment">//减少扫描总额</span>

		total_scan <span class="token operator">-</span><span class="token operator">=</span> batch_size<span class="token punctuation">;</span>

		<span class="token function">cond_resched</span><span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">;</span>

	<span class="token punctuation">}</span>

    <span class="token comment">//如果剩下的额度不够一个batch_size，留着下次使用，记录在nr_in_batch</span>

	<span class="token keyword">if</span> <span class="token punctuation">(</span>total_scan <span class="token operator">&gt;</span> <span class="token number">0</span><span class="token punctuation">)</span>

		new_nr <span class="token operator">=</span> <span class="token function">atomic_long_add_return</span><span class="token punctuation">(</span>total_scan<span class="token punctuation">,</span>

				<span class="token operator">&amp;</span>shrinker<span class="token operator">-&gt;</span>nr_in_batch<span class="token punctuation">)</span><span class="token punctuation">;</span>

	<span class="token keyword">else</span>

		new_nr <span class="token operator">=</span> <span class="token function">atomic_long_read</span><span class="token punctuation">(</span><span class="token operator">&amp;</span>shrinker<span class="token operator">-&gt;</span>nr_in_batch<span class="token punctuation">)</span><span class="token punctuation">;</span>

	<span class="token function">trace_mm_shrink_slab_end</span><span class="token punctuation">(</span>shrinker<span class="token punctuation">,</span> shrink_ret<span class="token punctuation">,</span> nr<span class="token punctuation">,</span> new_nr<span class="token punctuation">)</span><span class="token punctuation">;</span>

<span class="token punctuation">}</span>

<span class="token function">up_read</span><span class="token punctuation">(</span><span class="token operator">&amp;</span>shrinker_rwsem<span class="token punctuation">)</span><span class="token punctuation">;</span>


out:


cond_resched();


return ret;


}

空闲slab缓存计算和回收都是在do_shrinker_shrink完成，它其实调用的是一个函数指针，不同slab管理区有自己定义的shrink函数，第三个入参nr_to_scan为0时，是计算空闲slab缓存；不为空时，表示扫描和回收缓存的数量。

static inline int do_shrinker_shrink(struct shrinker *shrinker,

				     struct shrink_control *sc,

				     unsigned long nr_to_scan)

{

	int objects;

	sc->nr_to_scan = nr_to_scan;

	objects = (*shrinker->shrink)(shrinker, sc);
<span class="token keyword">if</span> <span class="token punctuation">(</span>objects <span class="token operator">&lt;</span> <span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>

	<span class="token keyword">return</span> INT_MAX<span class="token punctuation">;</span>

<span class="token keyword">return</span> objects<span class="token punctuation">;</span>


}

总的来说，drop_slab就是调用每个slab管理区定义的shrink函数，先计算出可回收的slab缓存数量，然后确定扫描数量，最后调用shrink函数执行缓存扫描和回收。

[转帖]vm内核参数之缓存回收drop_caches的更多相关文章

vm内核参数优化设置
http://www.cnblogs.com/wjoyxt/archive/2014/06/08/3777042.html (1)vm.overcommit_memory 执行grep -i com ...
linux 内核参数VM调优之参数调节和场景分析
1. pdflush刷新脏数据条件 (linux IO 内核参数调优之原理和参数介绍)上一章节讲述了IO内核调优介个重要参数参数. 总结可知cached中的脏数据满足如下几个条件中一个或者多个的时 ...
谨慎调整内核参数:vm.min_free_kbytes
内核参数:内存相关内存管理从三个层次管理内存,分别是node, zone ,page; 64位的x86物理机内存从高地址到低地址分为: Normal DMA32 DMA.随着地址降低. [root@ ...
Centos内核参数
内核参数 abi.vsyscall32 = 1 在2.6.25版本以后的x86-64内核中,默认启用了VDSO32. 虚拟动态共享对象 http://man7.org/linux/man-pages ...
oracle内核参数详解
一.前言在生产中,我们安装oracle数据库时,为达到最优我们需要对操作系统的内核参数进行一定的调整.主要从内存.cpu.io以及网络等方面,根据实际情况进行调整.以下参数可供大家参考,如有不当之处 ...
LINUX内核参数网络相关
有助于提高网络性能和吞吐量的参数 net.core.somaxconn = 128 已完成连接队列(completed connection queue) (1)三次握手已经完成,但还未被应用层接收( ...
linux下TCP/IP及内核参数优化调优(转）
Linux下TCP/IP及内核参数优化有多种方式,参数配置得当可以大大提高系统的性能,也可以根据特定场景进行专门的优化,如TIME_WAIT过高,DDOS攻击等等. 如下配置是写在sysctl.con ...
Linux内核参数配置
Linux在系统运行时修改内核参数(/proc/sys与/etc/sysctl.conf),而不需要重新引导系统,这个功能是通过/proc虚拟文件系统实现的. 在/proc/sys目录下存放着大多数的 ...
/proc/sys/vm/ 内存参数
linux下proc里关于磁盘性能的参数 http://blog.csdn.net/eroswang/article/details/6126646 我们在磁盘写操作持续繁忙的服务器上曾经碰到一个特 ...
Nginx的10万并发内核参数优化
关于内核参数的优化: net.ipv4.tcp_max_tw_buckets = 6000timewait的数量,默认是180000.net.ipv4.ip_local_port_range = 10 ...

随机推荐

华为云MetaStudio全新升级，盘古数字人大模型助力数字人自由
摘要:基于盘古大模型能力,华为云MetaStudio数字内容生产线全新升级,推出数字人模型生成服务和模型驱动服务. 近日,华为开发者大会2023 ( Cloud ) 在东莞拉开帷幕.基于盘古大模型能力 ...
AI推理实践丨多路极致性能目标检测最佳实践设计解密
摘要:基于CANN的多路极致性能目标检测最佳实践设计解密. 本文分享自华为云社区<基于CANN的AI推理最佳实践丨多路极致性能目标检测应用设计解密>,作者: 昇腾CANN . 当前人工智能 ...
GaussDB(DWS)查询过滤器原理与应用
摘要:GaussDB(DWS)查询过滤器(黑名单)提供查询过滤功能,支持自动隔离反复被终止的查询,防止烂SQL再次执行. 本文分享自华为云社区<GaussDB(DWS)查询过滤器原理与应用> ...
聊聊数仓中TPCD-DS&TPC-H与查询性能的那些事儿
摘要:详细讲述使用GaussDB(DWS)时,如何使用TPC-DS/TPC-H等标准数据模型,获取DWS的查询性能数据. 本文分享自华为云社区<GaussDB(DWS) <DWS之TPCD ...
NDPQ(NDP+PQ)，定义分布式数据库新方向
摘要:云服务提供商构建新的云原生关系数据库系统,专门为云基础架构设计,通常采用将计算和存储分离到独立扩展的分布式层的设计. 本文分享自华为云社区<性能提升100倍!GaussDB(for MyS ...
xcode打包导出ipa
xcode打包导出ipa 众所周知,在开发苹果应用时需要使用签名(证书)才能进行打包安装苹果IPA,作为刚接触ios开发的同学,只是学习ios app开发内测,并没有上架appstore需求,对于苹果 ...
如何快速从 ETL 到 ELT？火山引擎 ByteHouse 做了这三件事
更多技术交流.求职机会,欢迎关注字节跳动数据平台微信公众号,回复[1]进入官方交流群前言当涉及到企业分析场景时,所使用的数据通常源自多样的业务数据,这些数据系统大多采用以行为主的存储结构,比如支付 ...
Python MatplotlibDeprecationWarning Matplotlib 3.6 and will be removed two minor releases later
百度飞桨(PaddlePaddle)-数字识别在Pycharm中使用Matplotlib中的pyplot时,运行代码报错: MatplotlibDeprecationWarning: Support ...
C-Shopping基于Next.js，开源电商平台全新亮相
嗨,大家好!欢迎来到C-Shopping,这是一场揭开科技面纱的电商之旅.我是C-Shopping开源作者"继小鹏",今天将为你介绍一款基于最新技术的开源电商平台.让我们一同探索吧 ...
Django rest_framework使用自定义异常
完整代码 https://gitee.com/mom925/django-system 在settings.py中配置 REST_FRAMEWORK = { "EXCEPTION_HANDL ...

[转帖]vm内核参数之缓存回收drop_caches

1、关于drop_caches

2、释放pagecache

3、释放slab cache

[转帖]vm内核参数之缓存回收drop_caches的更多相关文章

随机推荐

热门专题