整體資源消耗：CPU、內(nèi)存使用增高，IO調(diào)度頻率增高

內(nèi)核源碼

kernel/sysctl.c文件

static struct ctl_table vm_table[] = {
	...
	{
		.procname	= "dirty_background_ratio",
		.data		= &dirty_background_ratio,
		.maxlen		= sizeof(dirty_background_ratio),
		.mode		= 0644,
		.proc_handler	= dirty_background_ratio_handler,
		.extra1		= &zero,
		.extra2		= &one_hundred,
	},
	{
		.procname	= "dirty_ratio",
		.data		= &vm_dirty_ratio,
		.maxlen		= sizeof(vm_dirty_ratio),
		.mode		= 0644,
		.proc_handler	= dirty_ratio_handler,
		.extra1		= &zero,
		.extra2		= &one_hundred,
	},
	{
		.procname	= "dirty_writeback_centisecs",
		.data		= &dirty_writeback_interval,
		.maxlen		= sizeof(dirty_writeback_interval),
		.mode		= 0644,
		.proc_handler	= dirty_writeback_centisecs_handler,
	},
}

修改/proc/sys/vm配置項(xiàng)的信息，實(shí)際上修改了對應(yīng)的某個(gè)全局變量的值，如dirty_background_ratio對應(yīng)的變量為&dirty_background_ratio。

每個(gè)全局變量都有默認(rèn)值，追溯這些全局變量的定義

<mm/page-writeback.c>

int dirty_background_ratio = 10;
unsigned long dirty_background_bytes;
int vm_dirty_ratio = 20;
unsigned long vm_dirty_bytes;
unsigned int dirty_writeback_interval = 5 * 100; /* centiseconds */
unsigned int dirty_expire_interval = 30 * 100; /* centiseconds */

默認(rèn)值：

回刷進(jìn)程

通過ps aux，我們總能看到writeback的內(nèi)核進(jìn)程。

$ ps aux | grep "writeback" 
root 21 0.0 0.0 0 0 ? S< Mar19 0:00 [writeback]

實(shí)際上是一個(gè)工作隊(duì)列對應(yīng)的進(jìn)程，在default_bdi_init()中創(chuàng)建。

/* bdi_wq serves all asynchronous writeback tasks */
 struct workqueue_struct *bdi_wq;
 
static int __init default_bdi_init(void)
{
	...
	bdi_wq = alloc_workqueue("writeback", WQ_MEM_RECLAIM | WQ_FREEZABLE |
			WQ_UNBOUND | WQ_SYSFS, 0);
	...
}

回刷進(jìn)程的核心是函數(shù)wb_workfn()，通過函數(shù)wb_init()綁定。

static int wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi
		int blkcg_id, gfp_t gfp)
{
	...
	INIT_DELAYED_WORK(&wb->dwork, wb_workfn);
	...
}

喚醒回刷進(jìn)程的操作如下：

static void wb_wakeup(struct bdi_writeback *wb)
{
	spin_lock_bh(&wb->work_lock);
	if (test_bit(WB_registered, &wb->state))
		mod_delayed_work(bdi_wq, &wb->dwork, 0);
	spin_unlock_bh(&wb->work_lock);
}

表示喚醒的回刷任務(wù)在工作隊(duì)列writeback中執(zhí)行，這樣，就把工作隊(duì)列和回刷工作綁定了

在wb_workfn()的最后，有如下代碼：

void wb_workfn(struct work_struct *work)
{
	...
	/* 如果還有需要回收的內(nèi)存，再次喚醒 */
	if (!list_empty(&wb->work_list))
		wb_wakeup(wb);
	/* 如果還有臟數(shù)據(jù)，延遲喚醒 */
	else if (wb_has_dirty_io(wb) && dirty_writeback_interval)
		wb_wakeup_delayed(wb);
}

static void wb_wakeup(struct bdi_writeback *wb)
{
	spin_lock_bh(&wb->work_lock);
	if (test_bit(WB_registered, &wb->state))
		mod_delayed_work(bdi_wq, &wb->dwork, 0);
	spin_unlock_bh(&wb->work_lock);
}

void wb_wakeup_delayed(struct bdi_writeback *wb)
{
	unsigned long timeout;

	/* 在這里使用dirty_writeback_interval，設(shè)置下次喚醒時(shí)間 */
	timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
	spin_lock_bh(&wb->work_lock);
	if (test_bit(WB_registered, &wb->state))
		queue_delayed_work(bdi_wq, &wb->dwork, timeout);
	spin_unlock_bh(&wb->work_lock);
}

根據(jù)kernel/sysctl.c的內(nèi)容，我們知道dirty_writeback_centisecs配置項(xiàng)對應(yīng)的全局變量是dirty_writeback_interval。

可以看出dirty_writeback_interval在wb_wakeup_delayed()中起作用，在wb_workfn()的最后根據(jù)dirty_writeback_interval設(shè)置下一次喚醒時(shí)間。

我們還發(fā)現(xiàn)通過msecs_to_jiffies(XXX * 10)來換算單位，表示dirty_writeback_interval乘以10之后的計(jì)量單位才是毫秒msecs。怪不得說dirty_writeback_centisecs的單位是1/100秒。

臟數(shù)據(jù)量

臟數(shù)據(jù)量通過dirty_background_XXX和dirty_XXX表示，根據(jù)kernel/sysctl.c的內(nèi)容，我們知道dirty_background_XXX配置項(xiàng)對應(yīng)的全局變量是dirty_background_XXX，dirty_XXX對于的全局變量是vm_dirty_XXX。

我們把目光聚焦到函數(shù)domain_dirty_limits()，通過這個(gè)函數(shù)換算臟數(shù)據(jù)閾值。

static void domain_dirty_limits(struct dirty_throttle_control *dtc)
{
	...
	unsigned long bytes = vm_dirty_bytes;
	unsigned long bg_bytes = dirty_background_bytes;
	/* convert ratios to per-PAGE_SIZE for higher precision */
	unsigned long ratio = (vm_dirty_ratio * PAGE_SIZE) / 100;
	unsigned long bg_ratio = (dirty_background_ratio * PAGE_SIZE) / 100;
	...
	if (bytes)
		thresh = DIV_ROUND_UP(bytes, PAGE_SIZE);
	else
		thresh = (ratio * available_memory) / PAGE_SIZE;

	if (bg_bytes)
		bg_thresh = DIV_ROUND_UP(bg_bytes, PAGE_SIZE);
	else
		bg_thresh = (bg_ratio * available_memory) / PAGE_SIZE;

	if (bg_thresh >= thresh)
		bg_thresh = thresh / 2;

	dtc->thresh = thresh;
	dtc->bg_thresh = bg_thresh;

• dirty_background_bytes/dirty_bytes的優(yōu)先級高于dirty_background_ratio/dirty_ratio
• dirty_background_bytes/ratio和dirty_bytes/ratio最終會統(tǒng)一換算成頁做計(jì)量單位
• dirty_background_bytes/dirty_bytes做進(jìn)一除法，表示如果值為4097Bytes，換算后是2頁
• dirty_background_ratio/dirty_ratio相乘的基數(shù)是available_memory，表示可用內(nèi)存
• 如果dirty_background_XXX大于dirty_XXX，則取dirty_XXX的一半

內(nèi)存計(jì)算方法：

static unsigned long global_dirtyable_memory(void)
{
	unsigned long x;
	
	x = global_zone_page_state(NR_FREE_PAGES);
	/*
	 * Pages reserved for the kernel should not be considered
	 * dirtyable, to prevent a situation where reclaim has to
	 * clean pages in order to balance the zones.
	 */
	 
	 x += global_node_page_state(NR_INACTIVE_FILE);
	 x += global_node_page_state(NR_ACTIVE_FILE); 
	 
	 if (!vm_highmem_is_dirtyable)
	 	x -= highmem_dirtyable_memory(x);
	 
	 return x + 1; /* Ensure that we never return 0 */
}

因此，

# cat /proc/meminfo
MemTotal:        3880184 kB
MemFree:          562764 kB
MemAvailable:    3263916 kB
Buffers:          372732 kB
Cached:          2330740 kB
SwapCached:            0 kB
Active:          2064100 kB
Inactive:         849240 kB
Active(anon):     216636 kB
Inactive(anon):      240 kB
Active(file):    1847464 kB
Inactive(file):   849000 kB
Unevictable:        7936 kB
Mlocked:            7936 kB
SwapTotal:             0 kB
SwapFree:              0 kB
Dirty:                 0 kB
Writeback:            12 kB
AnonPages:        217868 kB
Mapped:           140412 kB
Shmem:               584 kB
Slab:             319784 kB
SReclaimable:     293908 kB
SUnreclaim:        25876 kB
KernelStack:        3920 kB
PageTables:         6392 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:     1940092 kB
Committed_AS:    2266836 kB
VmallocTotal:   34359738367 kB
VmallocUsed:       13392 kB
VmallocChunk:   34359718524 kB
Percpu:              352 kB
HardwareCorrupted:     0 kB
AnonHugePages:     40960 kB
CmaTotal:              0 kB
CmaFree:               0 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
DirectMap4k:      100208 kB
DirectMap2M:     4093952 kB
DirectMap1G:     2097152 kB

可用內(nèi)存 = 空閑頁 - 內(nèi)核預(yù)留頁 + 活動文件頁 + 非活動文件頁 ( - 高端內(nèi)存)
即：
$available = $memfree - $water_low_total;

$pagecache = $Active_file + $Inactive_file;
$pagecache -=min($pagecache/2, $water_low_total);
$available += $pagecache;

$available += $SReclaimable - min($SReclaimable/2, $water_low_total);

簡化為：
可用內(nèi)存 = 空閑頁 +  可回收頁
可回收頁 = Active_file + Inactive_file + SReclaimable

臟數(shù)據(jù)達(dá)到閾值后是怎么觸發(fā)回刷的？文章來源地址http://www.zghlxwxcb.cn/news/detail-499602.html

static void balance_dirty_pages(struct bdi_writeback *wb,
				unsigned long pages_dirtied)
{
	unsigned long nr_reclaimable;   /* = file_dirty + unstable_nfs */
	...
	/*
	 * Unstable writes are a feature of certain networked
	 * filesystems (i.e. NFS) in which data may have been
	 * written to the server's write cache, but has not yet
	 * been flushed to permanent storage.
	 */
	nr_reclaimable = global_node_page_state(NR_FILE_DIRTY) +
					global_node_page_state(NR_UNSTABLE_NFS);
	...
	if (nr_reclaimable > gdtc->bg_thresh)
		wb_start_background_writeback(wb);
}

void wb_start_background_writeback(struct bdi_writeback *wb)
{
	wb_wakeup(wb);
}

總結(jié)

• 可回收內(nèi)存 = 文件臟頁 + 文件系統(tǒng)不穩(wěn)定頁(NFS)
• 可回收內(nèi)存達(dá)到dirty_background_XXX計(jì)算的閾值，只是喚醒臟數(shù)據(jù)回刷工作后直接返回，并不會等待回收完成，最終回收工作還是看writeback進(jìn)程

到了這里，關(guān)于linux性能優(yōu)化-IO調(diào)度優(yōu)化的文章就介紹完了。如果您還想了解更多內(nèi)容，請?jiān)谟疑辖撬阉鱐OY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！