在大多数的编程语言中,main函数都是用户程序的入口函数,go中也是如此。那么main.main是整个程序的入口吗, 肯定不是,因为go程序依赖于runtime,在程序的初始阶段需要初始化运行时,之后才会运行到用户的main函数,那么main.main是在哪里被调用的呢?接下来就从go程序的入口,再到go的GMP模型进行一个探究。

注意:本文使用的go sdk的版本为go1.20

1.go程序的入口

1 首先,编写一个简单的go程序,并将其进行编译,在此使用linux系统:

package main

import "fmt"

func main() {
	fmt.Println("hello,world")
}

编译:-N -l 用于阻止编译时进行优化和内联

go build -gcflags "-N -l" main.go

2 然后使用gdb来调试go程序:

首先,使用gdb加载支持调试go语言的脚本文件:

gdbsource /usr/local/go/src/runtime/runtime-gdb.py
➜  RemoteWorking git:(master) ✗ gdb
(gdb) source /usr/local/go/src/runtime/runtime-gdb.py

3 调试程序:

gdb main
info files
0x45c020_rt0_amd64_linuxsrc/runtime/rt0_liunx_amd64.s
TEXT _rt0_amd64_linux(SB),NOSPLIT,$-8
	JMP	_rt0_amd64(SB)
_rt0_amd64asm_amd64.s
TEXT _rt0_amd64(SB),NOSPLIT,$-8
	MOVQ	0(SP), DI	// argc
	LEAQ	8(SP), SI	// argv
	JMP	runtime·rt0_go(SB)
rt0_go
TEXT runtime·rt0_go(SB),NOSPLIT|TOPFRAME,$0
	...
	MOVL	24(SP), AX		// copy argc
	MOVL	AX, 0(SP)
	MOVQ	32(SP), AX		// copy argv
	MOVQ	AX, 8(SP)
	CALL	runtime·args(SB)
	CALL	runtime·osinit(SB)
	CALL	runtime·schedinit(SB)

	// create a new goroutine to start program
	MOVQ	$runtime·mainPC(SB), AX		// entry
	PUSHQ	AX
	CALL	runtime·newproc(SB)
	POPQ	AX

	// start this M
	CALL	runtime·mstart(SB)

	CALL	runtime·abort(SB)	// mstart should never return
	RET
...

// mainPC is a function value for runtime.main, to be passed to newproc.
// The reference to runtime.main is made via ABIInternal, since the
// actual function (not the ABI0 wrapper) is needed by newproc.
DATA	runtime·mainPC+0(SB)/8,$runtime·main<ABIInternal>(SB)
GLOBL	runtime·mainPC(SB),RODATA,$8
rt0_goruntime.osinitruntime.schedinit
runtime.mainPCruntime.newprocruntime.mainnewprocgoroutine
go func()newprocruntime.maingoroutine
runtime.mstartsysmonscheduleschedulefindrunnablegoroutineexecutegoroutineggogomain goroutine

如下图所示:

 

2 GMP模型

GMP模型是go语言goroutine的调度系统,调度是将goroutine调度到线程上执行的过程,而操作系统的调度器则负责将线程调度到CPU上运行。

2.1 GM模型

GMGgoroutineMgoroutinemgoroutinengoroutineglobrunqglobrunqgoroutine

而且一个goroutine创建的goroutine也会被放入全局队列中,同时也需要加锁。这样也会造成程序的局部性较差,因为一个goroutine创建的另一个goroutine大概率不会在同一个线程上运行。

 

2.2 改进的GMP模型

1 所有线程都从全局队列获取goroutine,造成锁争用强度大。2. 程序的局部性较差

go语言引入了GMP模型,G同样代表一个goroutine,M代表machine,也就是worker thread,p代表processor,包含了运行go代码所需的资源。

官方解释:

// Goroutine scheduler
// The scheduler's job is to distribute ready-to-run goroutines over worker threads.
//
// The main concepts are:
// G - goroutine.
// M - worker thread, or machine.
// P - processor, a resource that is required to execute Go code.
//     M must have an associated P to execute Go code, however it can be
//     blocked or in a syscall w/o an associated P.
goroutinegoroutinegoroutinepprunqgoroutinerunq256goroutineglobrunqgoroutineprunqrunqgoroutine
p的数量pCPUgoroutineruntime.GOMAXPROCm的数量

 

2.3 相关数据结构

2.3.1 runtime.g

goroutine在runtime中表示为一个g结构体:

type g struct {
	stack       stack   // offset known to runtime/cgo
	stackguard0 uintptr // offset known to liblink

	...
	m         *m      // current m; offset known to arm liblink
	sched     gobuf
	...
	atomicstatus atomic.Uint32
	
	goid         uint64
	
	preempt       bool // preemption signal, duplicates stackguard0 = stackpreempt
}

type stack struct {
	lo uintptr
	hi uintptr
}

type gobuf struct {
	sp   uintptr
	pc   uintptr
	g    guintptr
	...
}

省略了一些不太关心的字段,其中的一些字段的含义如下:

stackstackguard0mschedatomicstatusgoidpreempt
goroutine是一个有栈协程,stack字段用于描述协程的栈,goroutine的初始栈大小为2K,并且是从堆中分配的,是可以动态增长的。
sched用来存储goroutine执行的上下文,它与goroutine切换的底层实现相关,其中sp标识stack pointer,pc为program counter,g用来反向关联到当前g。

 

g的状态:

atomicstatus字段表示goroutine的状态,goroutine有多种状态:

状态含义
_Gidle当前goroutine刚被分配,还没有被初始化
_Grunnable当前goroutine处于待运行状态,他可能处于p的本地runq或者globrunq中,当前并没有在运行用户代码,它的栈也不归自己所有。
_Grunning当前goroutine正在运行用户代码,有关联的M和P。不会处于任何runq中,栈归该goroutine所有。
_Gsyscall当前goroutine正在执行系统调用,并没有在执行用户代码,拥有栈,而且被分配了M。
_Gwaiting当前goroutine处于阻塞状态,即不再runq中,也没有得到运行。它肯定被记录在某个地方,比如chan的阻塞队列、mutex的阻塞队列中。
_Gdead当前goroutine没有在使用,可能存在一个free list中或者刚刚被初始化。
_Gcopystack当前goroutine的栈正在被移动,没有在执行用户代码也不在一个runq中。

 

2.3.2 runtime.m

GMP中的M代表一个工作线程,在runtime中使用m结构体来表示:

type m struct {
	g0      *g     // goroutine with scheduling stack
	gsignal       *g                // signal-handling g	
	curg          *g       // current running goroutine	
	p             puintptr // attached p for executing go code (nil if not executing go code)	
	id            int64	
	preemptoff    string // if != "", keep curg running on this m
	locks         int32
	spinning      bool // m is out of work and is actively looking for work
	mOS
}

省略了其中一些不太关心的字段,其中一些字段的含义如下:

g0gsignalcurgpidpreemptofflocksspiningmOS

 

2.3.3 runtime.p

GMP中的p代表processor,其中包含了一系列用于运行goroutine的资源,比如本地runq、堆内存缓存、栈内存缓存、goroutine id缓存等,在runtime中使用p结构体表示:

type p struct {
	id          int32
	status      uint32 // one of pidle/prunning/...

	schedtick   uint32     // incremented on every scheduler call
	syscalltick uint32     // incremented on every system call
	sysmontick  sysmontick // last tick observed by sysmon
	m           muintptr   // back-link to associated m (nil if idle)
    
	// Cache of goroutine ids, amortizes accesses to runtime·sched.goidgen.
	goidcache    uint64
	goidcacheend uint64

	// Queue of runnable goroutines. Accessed without lock.
	runqhead uint32
	runqtail uint32
	runq     [256]guintptr
	
	runnext guintptr

	// Available G's (status == Gdead)
	gFree struct {
		gList
		n int32
	}

	// preempt is set to indicate that this P should be enter the
	// scheduler ASAP (regardless of what G is running on it).
	preempt bool
}

其中省略了一些不太关心的字段,一些字段的含义如下:

idstatusschedticksyscallticksysmontickmgoidcache、goidcacheendrunqhead、runqtail、runqrunnextgFreepreempt
p的状态:
状态含义
_Pidle当前p处于空闲状态,没有被用于执行用户代码或调度。p处于idle list中,它的本地runq是空的
_Prunning当前p与一个m进行关联并且被用于执行用户代码或者调度
_Psyscall当前p没有在运行用于代码,它与系统调用中的M有亲和关系,但不属于它,并且可能被另一个M窃取。这类似于_Pidle,但使用轻量级转换并维护M亲和关系。
_Pgcstop当前p因为STW而停止
_Pdead停用状态,因为GOMAXPROC可用收缩,会造成多余的p被停用。一旦GOMAXPROC重新增长,那么停用的p会被重新启用。

 

2.3.4 runtime.schedt

还有另一个和调度相关的数据结构需要关注,就是runtime.schedt,其中包含了调度的一些全局数据,schedt类型的实例只会存在一个:

var (
	allm       *m        // 所有m组成一个链表
	gomaxprocs int32     // 对应与GOMAXPROC
	ncpu       int32     // CPU核心数
	
	sched      schedt   // 调度器相关的数据结构
	


	allpLock mutex     // 保护allp的锁
	allp []*p          // 所有的p
)

schedt结构如下:

其中全局runq就存在与schedt结构中
type schedt struct {
	goidgen   atomic.Uint64    
	
	midle        muintptr // idle m's waiting for work
	nmidle       int32    // number of idle m's waiting for work
	mnext        int64    // number of m's that have been created and next M ID
	maxmcount    int32    // maximum number of m's allowed (or die)
	nmsys        int32    // number of system m's not counted for deadlock
	nmfreed      int64    // cumulative number of freed m's

	ngsys atomic.Int32 // number of system goroutines

	pidle        puintptr // idle p's
	npidle       atomic.Int32
	nmspinning   atomic.Int32  // See "Worker thread parking/unparking" comment in proc.go.

	// Global runnable queue.
	runq     gQueue
	runqsize int32

	// Global cache of dead G's.
	gFree struct {
		lock    mutex
		stack   gList // Gs with stacks
		noStack gList // Gs without stacks
		n       int32
	}
}
goidgenmidlenmidlemnextmaxmcountnmsysnmfreedngsyspidlenpidlenmspiningqunq、runqsizegFree

 

2.4 g0、m0

在每个工作线程M中都存在一个g0,g0的主要功能就是执行调度程序,当需要执行调度程序时会将运行栈切换的g0栈,然后运行调度程序来寻找一个就绪的goroutine并切换运行。

有两个函数可用切换到g0栈来运行:

func mcall(fn func(*g))

func systemstack(fn func())
  • mcall:将调用mcall的协程栈切换的g0栈并且在g0栈上运行fn,mcall仅可以被除了g0、gsingal之外的g调用。
  • systemstack:在系统栈上运行fn,然后再切换回来。

 

m0为进程的第一个线程,也就是运行main goroutine的线程

 

3 G的创建与退出

我们再程序中通常使用下面的方式来创建一个goroutine:

go func()
runtime.newprocnewprocgoroutine

我们可用通过将代码编译为汇编来查看go func()是怎么执行的:

将下面的示例代码编译为汇编程序:

go build -gcflags -S main.go
package main

import (
	"fmt"
	"time"
)

func print() {
	fmt.Println("hello, GMP")
	time.Sleep(time.Second)
}

func main() {
	go print()

	select {}
}

汇编代码如下:

"".main STEXT size=50 args=0x0 locals=0x10 funcid=0x0 align=0x0
        0x0000 00000 (/root/RemoteWorking/main.go:13)   TEXT    "".main(SB), ABIInternal, $16-0
        0x0000 00000 (/root/RemoteWorking/main.go:13)   CMPQ    SP, 16(R14)
        0x0004 00004 (/root/RemoteWorking/main.go:13)   PCDATA  $0, $-2
        0x0004 00004 (/root/RemoteWorking/main.go:13)   JLS     43
        0x0006 00006 (/root/RemoteWorking/main.go:13)   PCDATA  $0, $-1
        0x0006 00006 (/root/RemoteWorking/main.go:13)   SUBQ    $16, SP
        0x000a 00010 (/root/RemoteWorking/main.go:13)   MOVQ    BP, 8(SP)
        0x000f 00015 (/root/RemoteWorking/main.go:13)   LEAQ    8(SP), BP
        0x0014 00020 (/root/RemoteWorking/main.go:13)   FUNCDATA        $0, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0014 00020 (/root/RemoteWorking/main.go:13)   FUNCDATA        $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0014 00020 (/root/RemoteWorking/main.go:14)   LEAQ    "".print·f(SB), AX
        0x001b 00027 (/root/RemoteWorking/main.go:14)   PCDATA  $1, $0
        0x001b 00027 (/root/RemoteWorking/main.go:14)   NOP
        0x0020 00032 (/root/RemoteWorking/main.go:14)   CALL    runtime.newproc(SB)
AXruntime.newproc
runtime.newproc代码如下
func newproc(fn *funcval) {
	gp := getg()          // 获取当前g
	pc := getcallerpc() 
	systemstack(func() {
		newg := newproc1(fn, gp, pc)     // 创建一个新的goroutine
 
		pp := getg().m.p.ptr()           // 获取当前g运行的m关联的p
		runqput(pp, newg, true)          // 将新的goroutine加入到就绪队列中

		if mainStarted {
			wakep()                     // 唤醒新的p
		}
	})
}
runqputgoroutineprunnextrunnextgoroutinegoroutinegoroutinegoroutine
newproc的主要逻辑就是创建了一个新的g,并将其放入当前g运行的m关联的p的本地runq中。

newproc1的代码如下:

func newproc1(fn *funcval, callergp *g, callerpc uintptr) *g {
	...
	mp := acquirem() // 禁止抢占
	pp := mp.p.ptr()    // 获取当前m关联的p
	newg := gfget(pp)   // 从p的缓存中获取一个g
	if newg == nil {
		newg = malg(_StackMin)    // 如果从缓存中获取不到,则新创建一个,_StackMin的值为2048,也就是2K
		casgstatus(newg, _Gidle, _Gdead)
		allgadd(newg) // publishes with a g->status of Gdead so GC scanner doesn't look at uninitialized stack.
	}
	...

	// goexit函数被放在了pc上,gostartcallfn会对其进行特殊处理
	newg.sched.pc = abi.FuncPCABI0(goexit) + sys.PCQuantum // +PCQuantum so that previous instruction is in same function
	newg.sched.g = guintptr(unsafe.Pointer(newg))
	gostartcallfn(&newg.sched, fn)
	newg.gopc = callerpc
	newg.ancestors = saveAncestors(callergp)
	newg.startpc = fn.fn
	...
	casgstatus(newg, _Gdead, _Grunnable)    // 改变g的状态
	
	newg.goid = pp.goidcache  // 分配goid
	pp.goidcache++
	
	releasem(mp)

	return newg
}

func gostartcallfn(gobuf *gobuf, fv *funcval) {
	var fn unsafe.Pointer
	if fv != nil {
		fn = unsafe.Pointer(fv.fn)
	} else {
		fn = unsafe.Pointer(abi.FuncPCABIInternal(nilfunc))
	}
	gostartcall(gobuf, fn, unsafe.Pointer(fv))
}

func gostartcall(buf *gobuf, fn, ctxt unsafe.Pointer) {
	sp := buf.sp
	sp -= goarch.PtrSize
	*(*uintptr)(unsafe.Pointer(sp)) = buf.pc   // 在goroutine的栈帧中插入了goexit函数
	buf.sp = sp
	buf.pc = uintptr(fn)
	buf.ctxt = ctxt
}
在newproc1中获取了一个g实例,对其中的字段进行了设置,为其分配id,并修改状态为*_Grunnable*。特别需要注意的时,在gostartcall函数中,往goroutine的栈帧中插入了一个goexit函数,因此当goroutine从运行的函数退出时,就会返回到goexit函数中。
使用goland调试go程序时,可以从调用栈中查看到runtime.goexit函数,仿佛是runtime.goexit函数调用了runtime.main,而runtime.main又调用了main.main函数

而runtime.goexit是一段使用汇编实现的代码:

TEXT runtime·goexit(SB),NOSPLIT|TOPFRAME,$0-0
	BYTE	$0x90	// NOP
	CALL	runtime·goexit1(SB)	// does not return
	// traceback from goexit1 must hit code range of goexit
	BYTE	$0x90	// NOP
runtime.goexit1
func goexit1() {
	mcall(goexit0)
}
func goexit0(gp *g) {
    // 重置g的状态
	mp := getg().m
	pp := mp.p.ptr()

	casgstatus(gp, _Grunning, _Gdead)
	gcController.addScannableStack(pp, -int64(gp.stack.hi-gp.stack.lo))
	if isSystemGoroutine(gp, false) {
		sched.ngsys.Add(-1)
	}
	gp.m = nil
	locked := gp.lockedm != 0
	gp.lockedm = 0
	mp.lockedg = 0
	gp.preemptStop = false
	gp.paniconfault = false
	gp._defer = nil // should be true already but just in case.
	gp._panic = nil // non-nil for Goexit during panic. points at stack-allocated data.
	gp.writebuf = nil
	gp.waitreason = waitReasonZero
	gp.param = nil
	gp.labels = nil
	gp.timer = nil

	dropg()    // 将当前g从m移除
  
	gfput(pp, gp)  //将g放入p的gFreelist中
	
	schedule()   // 触发新一轮的调度
}
从runtime.goexit到runtime.goexit1,最终到runtime.goexit0函数中,对g的状态进行了重置,然后将g从m中移除,放入p的gFree List中,,以便后续重用。然后调用了scheduler函数,scheduler函数正是调度的入口,如此一来便形成了一个闭环。

 

总结

 

4 调度循环

go的调度器会不断调度goroutine到线程上运行,当一个goroutine结束运行、发生阻塞、主动让出、或者时间片用尽时就会触发新一轮的调度,重新选择一个goroutine来运行。整个流程如下:

mstartscheduleexecuteuser code

 

4.1 runtime.schedule

schedulescheduleg0

代码如下:

func schedule() {
	mp := getg().m
	
    // 线程持有锁时,不能进行调度,以免造成runtime内部错误
	if mp.locks != 0 {       
		throw("schedule: holding locks")
	}
	
    // 判断当前M有没有和G绑定,如果有,这个M就不能用来执行其它的G
	if mp.lockedg != 0 {
		stoplockedm()
		execute(mp.lockedg.ptr(), false) // Never returns.
	}

	// 判断是否在进行cgo调用,如果在就不能进行调度,因为g0栈正在被cgo使用
	if mp.incgo {
		throw("schedule: in cgo")
	}

top:
	pp := mp.p.ptr()        // 获取当前m关联的p
	pp.preempt = false

	if mp.spinning && (pp.runnext != 0 || pp.runqhead != pp.runqtail) {
		throw("schedule: spinning with local work")
	}
	
    // 寻找一个可运行的g,阻塞直到找到
	gp, inheritTime, tryWakeP := findRunnable() // blocks until work is available

	// 如果当前线程正在自旋寻找新的工作,因为已经找到工作了,重置自旋状态
	if mp.spinning {
		resetspinning()
	}

	...

	// If about to schedule a not-normal goroutine (a GCworker or tracereader),
	// wake a P if there is one.
	if tryWakeP {
		wakep()
	}
	if gp.lockedm != 0 {
		// Hands off own p to the locked m,
		// then blocks waiting for a new p.
		startlockedm(gp)
		goto top
	}
	
    // 调用execute来运行g
	execute(gp, inheritTime)
}
scheduleGGcgocgog0
findrunnablegexecuteg

 

4.2 runtime.findrunnable

_Runnablegoroutine
tracetrace reader
func findRunnable() (gp *g, inheritTime, tryWakeP bool) {
	mp := getg().m

	// The conditions here and in handoffp must agree: if
	// findrunnable would return a G to run, handoffp must start
	// an M.
	...
	// Try to schedule the trace reader.
	if trace.enabled || trace.shutdown {
		gp := traceReader()
		if gp != nil {
			casgstatus(gp, _Gwaiting, _Grunnable)
			traceGoUnpark(gp, 0)
			return gp, false, true
		}
	}
	...
}
trace readergcGC Worker
// Try to schedule a GC worker.
	if gcBlackenEnabled != 0 {
		gp, tnow := gcController.findRunnableGCWorker(pp, now)    // 尝试获取一个GC Worker
		if gp != nil {
			return gp, false, true
		}
		now = tnow
	}
goroutinegoroutine61goroutinep.schedtick
	// Check the global runnable queue once in a while to ensure fairness.
	// Otherwise two goroutines can completely occupy the local runqueue
	// by constantly respawning each other.
	if pp.schedtick%61 == 0 && sched.runqsize > 0 {
		lock(&sched.lock)
		gp := globrunqget(pp, 1)   // 每执行61次调度,就从全局队列中获取一个goroutine来运行    
		unlock(&sched.lock)
		if gp != nil {
			return gp, false, false
		}
	}
prunqgoroutine
// local runq
	if gp, inheritTime := runqget(pp); gp != nil {       // 从p的本地runq中获取
		return gp, inheritTime, false
	}
prunqgoroutinegoroutinerunqgoroutineprunq
// global runq
	if sched.runqsize != 0 {
		lock(&sched.lock)
		gp := globrunqget(pp, 0)
		unlock(&sched.lock)
		if gp != nil {
			return gp, false, false
		}
	}
goroutinenetpollergoroutine
// 如果netpoller启动了,并且其中管理的fd数量大于0,调用netpoll来轮询网络,以此来获取在网络中就绪的goroutine
if netpollinited() && netpollWaiters.Load() > 0 && sched.lastpoll.Load() != 0 {
		if list := netpoll(0); !list.empty() { // non-blocking
			gp := list.pop()        // 获取到一批goroutine,组成一个链表,获取链表头第一个
			injectglist(&list)      // 将其它goroutine放入本地runq中
			casgstatus(gp, _Gwaiting, _Grunnable)
			if trace.enabled {
				traceGoUnpark(gp, 0)
			}
			return gp, false, false
		}
	}
goroutinerunqgoroutinecpuPPallpP
if mp.spinning || 2*sched.nmspinning.Load() < gomaxprocs-sched.npidle.Load() {
		if !mp.spinning {
			mp.becomeSpinning()
		}

		gp, inheritTime, tnow, w, newWork := stealWork(now)      // 从其它P那偷取工作
		if gp != nil {
			// Successfully stole.
			return gp, inheritTime, false
		}
		if newWork {
			// There may be new timer or GC work; restart to
			// discover.
			goto top
		}

		now = tnow
		if w != 0 && (pollUntil == 0 || w < pollUntil) {
			// Earlier timer to wait for.
			pollUntil = w
		}
	}

 

4.3 runtime.execute、runtime.gogo

runtime.executegggogogoroutinegogogoroutine
func execute(gp *g, inheritTime bool) {
	mp := getg().m

	...

	mp.curg = gp  // 设置当前运行的g
	gp.m = mp     // 关联当前的m 
	casgstatus(gp, _Grunnable, _Grunning)   // 将当前g的状态切换为_Grunning
	gp.waitsince = 0
	gp.preempt = false
	gp.stackguard0 = gp.stack.lo + _StackGuard
	if !inheritTime {
		mp.p.ptr().schedtick++
	}

	...
    
	gogo(&gp.sched)       // 调用gogo来切换协程
}

 

4.4 runtime.gopark、runtime.goready

goparkgoroutine_Gwaiting
func gopark(unlockf func(*g, unsafe.Pointer) bool, lock unsafe.Pointer, reason waitReason, traceEv byte, traceskip int) {
	...
	// can't do anything that might move the G between Ms here.
	mcall(park_m)
}

func park_m(gp *g) {
	mp := getg().m

	if trace.enabled {
		traceGoPark(mp.waittraceev, mp.waittraceskip)
	}

	// 修改g的状态为Gwaiting
	casgstatus(gp, _Grunning, _Gwaiting)
	dropg()

	if fn := mp.waitunlockf; fn != nil {
		ok := fn(gp, mp.waitlock)
		mp.waitunlockf = nil
		mp.waitlock = nil
		if !ok {
			if trace.enabled {
				traceGoUnpark(gp, 2)
			}
			casgstatus(gp, _Gwaiting, _Grunnable)
			execute(gp, true) // Schedule it back, never returns.
		}
	}
    // 触发新一轮的调度
	schedule()
}
goroutinegoparkchangoroutinechangopark
goreadychangoreadysystemstackg0readyreadygoroutine_Grunnablerunq中goroutine
func goready(gp *g, traceskip int) {
	systemstack(func() {
		ready(gp, traceskip, true)
	})
}

func ready(gp *g, traceskip int, next bool) {
	status := readgstatus(gp)

	// Mark runnable.
	mp := acquirem() // disable preemption because it can be holding p in a local var
	if status&^_Gscan != _Gwaiting {
		dumpgstatus(gp)
		throw("bad g->status in ready")
	}

	// 将g的状态切换为_Grunnable
	casgstatus(gp, _Gwaiting, _Grunnable)
    // 将g添加到runq中
	runqput(mp.p.ptr(), gp, next)
    // 唤醒新的p
	wakep()
	releasem(mp)
}

 

4.5 work stealing和handoff机制

work stealing

当一个线程没有可用的工作并且从全局队列中也找不到时,该线程并不会立马陷入休眠或者被销毁,而且尝试从其它P中窃取一部分的工作来运行。

handoff

当一个goroutine处于系统调用时,可能会导致整个线程发生阻塞。为了充分利用多核CPU,当前P会与M进行解绑,并且寻找或创建一个新的M来运行工作。

 

5 抢占式调度

gogoroutine抢占式时间片时钟中断上下文上下文
CPUgoroutinegoroutine 10msgo用户态时钟中断go
goroutineGoschedgo调度器的抢占式调度实际上是通过hook的方式来实现的goroutinego插入栈检测的相关代码抢占调度
runtime.morestackruntime.morestack_noctxt
runtime.newstack

抢占相关的代码如下:

func newstack() {
    ...
    // 加载 stackguard0
    stackguard0 := atomic.Loaduintptr(&gp.stackguard0)
	
    // 判断stackguard0是否被标记为了抢占
	preempt := stackguard0 == stackPreempt
    
    if preempt {
		...

		// 触发抢占
		gopreempt_m(gp) // never return
	}
}

func gopreempt_m(gp *g) {
	...
	goschedImpl(gp)
}

func goschedImpl(gp *g) {
	status := readgstatus(gp)
	if status&^_Gscan != _Grunning {
		dumpgstatus(gp)
		throw("bad g status")
	}
	casgstatus(gp, _Grunning, _Grunnable)
	dropg()
	lock(&sched.lock)
	globrunqput(gp)
	unlock(&sched.lock)
	
    // 触发新的一轮调度
	schedule()
}
newstackstackguard0stackPreempt0xfffffade
goroutinegoroutinegoroutinegoroutinemorestackgoroutineschedule
goroutineCPU

 

5.1 异步抢占

go1.13go1.13
package main

import "fmt"

func fn(i int) {
	for {
		i++
		fmt.Println(i)
	}
}

func main() {
	go fn(0)

	for {
	}
}

程序启动后,会一直打印数字,但是在我机器上打印到15万多的时候就会停止,整个程序卡死了。

fmt.PrintlnGCGCSTW(Stop The Worldmain goroutien没有发生函数调用,也就无法对其抢占,因此造成了死锁
go1.14异步抢占机制
goroutine
操作系统信goroutinegoroutine信号sigHandlegoroutine
SIGURG
sigHandler
func sighandler(sig uint32, info *siginfo, ctxt unsafe.Pointer, gp *g) {
	...
	// 如果当前信号是sigPreempt 而且 开启了异步抢占
	if sig == sigPreempt && debug.asyncpreemptoff == 0 && !delayedSignal {
		//进行抢占
		doSigPreempt(gp, c)
		
	}
	...
}
doSigPreemptgoroutineasyncPreemptgoroutineasyncPreempt
func doSigPreempt(gp *g, ctxt *sigctxt) {
	...
	if wantAsyncPreempt(gp) {
		if ok, newpc := isAsyncSafePoint(gp, ctxt.sigpc(), ctxt.sigsp(), ctxt.siglr()); ok {
			// Adjust the PC and inject a call to asyncPreempt.
			ctxt.pushCall(abi.FuncPCABI0(asyncPreempt), newpc)
		}
	}

	...
}
asyncPreemptasyncPreempt2schedule
func asyncPreempt2() {
	...
		mcall(gopreempt_m)
	...
}

func gopreempt_m(gp *g) {
	...
	goschedImpl(gp)
}

func goschedImpl(gp *g) {
	...

	schedule()
}

 

6 系统监控线程sysmon

sysmon(system mointor)
runtime.main
sysmon
sysmon
死锁检测轮询网络夺取处于系统调用中的P抢占长时间运行的G周期性触发GC
func sysmon() {
	lock(&sched.lock)
	sched.nmsys++
    // 1.死锁检测
	checkdead()
	unlock(&sched.lock)

	...

	for {
        // 休眠一定时间
		if idle == 0 { // start with 20us sleep...
			delay = 20
		} else if idle > 50 { // start doubling the sleep after 1ms...
			delay *= 2
		}
		if delay > 10*1000 { // up to 10ms
			delay = 10 * 1000
		}
		usleep(delay)

		...
		
		// poll network if not polled for more than 10ms
        // 2.轮询网络如果截至上次已经超过10ms
		lastpoll := sched.lastpoll.Load()
		if netpollinited() && lastpoll != 0 && lastpoll+10*1000*1000 < now {
			sched.lastpoll.CompareAndSwap(lastpoll, now)
			list := netpoll(0) // non-blocking - returns list of goroutines
			if !list.empty() {
				incidlelocked(-1)
				injectglist(&list)
				incidlelocked(1)
			}
		}
		...
        
		// retake P's blocked in syscalls
		// and preempt long running G's
        // 3.夺回阻塞与系统调用中的P
        // 4.抢占长时间运行的G
		if retake(now) != 0 {
			idle = 0
		} else {
			idle++
		}
        
		// check if we need to force a GC
        // 5.周期性触发GC
		if t := (gcTrigger{kind: gcTriggerTime, now: now}); t.test() && forcegc.idle.Load() {
			lock(&forcegc.lock)
			forcegc.idle.Store(false)
			var list gList
			list.push(forcegc.g)
			injectglist(&list)
			unlock(&forcegc.lock)
		}
		if debug.schedtrace > 0 && lasttrace+int64(debug.schedtrace)*1000000 <= now {
			lasttrace = now
			schedtrace(debug.scheddetail > 0)
		}
		unlock(&sched.sysmonlock)
	}
}