本文次要讲 wakep startm
wakep 在 newproc 可能会调用(main 起来之后)会调用 wakep
startm 在 wakep 中调用
mstart 在 rt0_go 中调用,执行 main
零碎线程 m
在 golang 中有三种零碎线程:
主线程:golang 程序启动加载的时候就运行在主线程上,代码中由一个全局的 m0 示意
运行 sysmon 的线程
普通用户线程,用来与 p 绑定,运行 g 中的工作的线程,
主线程和运行 sysmon 都是单实例,独自一个线程。而用户线程会有很多事例,他会依据调度器的需要新建,休眠和唤醒。
wakep startm
// Tries to add one more P to execute G's.
// Called when a G is made runnable (newproc, ready).
// 增加一个闲置的 p 来执行 g
func wakep() {if atomic.Load(&sched.npidle) == 0 { // 没有限度的 g, 返回
return
}
// be conservative about spinning threads
if atomic.Load(&sched.nmspinning) != 0 || !atomic.Cas(&sched.nmspinning, 0, 1) { //
return
}
startm(nil, true)
}
// Schedules some M to run the p (creates an M if necessary).
// If p==nil, tries to get an idle P, if no idle P's does nothing.
// May run with m.p==nil, so write barriers are not allowed.
// If spinning is set, the caller has incremented nmspinning and startm will
// either decrement nmspinning or set m.spinning in the newly started M.
//
// Callers passing a non-nil P must call from a non-preemptible context. See
// comment on acquirem below.
//
// Must not have write barriers because this may be called without a P.
// 调度一些 M 去运行 P
// 如果 p 不存在则从缓存获取 P,没有 P 就返回
// 如果 spinning 是 true,那个相应的 startm 须要减去 nmspinning
// 如果调用者调用含有非空 p,那么 Callers 就不能被抢占
//go:nowritebarrierrec
func startm(_p_ *p, spinning bool) {
// Disable preemption.
//
// Every owned P must have an owner that will eventually stop it in the
// event of a GC stop request. startm takes transient ownership of a P
// (either from argument or pidleget below) and transfers ownership to
// a started M, which will be responsible for performing the stop.
// 每个领有的 P 必须具备一个所有者,该所有者将在 GC 申请终止的状况下最终将其进行。// startm 获得 P 的长期所有权(从上面的参数或 pidleget 中获取),并将所有权转移给已启动的 M,该 M 将负责执行进行。//
// Preemption must be disabled during this transient ownership,
// otherwise the P this is running on may enter GC stop while still
// holding the transient P, leaving that P in limbo and deadlocking the
// STW.
// 在此短暂所有权期间,必须禁用抢占,否则正在运行的 P 可能会在依然放弃该暂态 P 的同时进入 GC 进行,// 从而使该 P 处于混乱状态并使 STW 死锁。// Callers passing a non-nil P must already be in non-preemptible
// context, otherwise such preemption could occur on function entry to
// startm. Callers passing a nil P may be preemptible, so we must
// disable preemption before acquiring a P from pidleget below.
// 传递非 nil P 的调用者必须曾经在不可抢占的上下文中,否则这种抢占可能产生在向进入 startm 的函数时。// 传递 nil P 的调用者可能是可抢占的,因而在从上面的 pidleget 获取 P 之前,咱们必须先禁用抢占。mp := acquirem() // 禁止抢占
lock(&sched.lock)
if _p_ == nil {_p_ = pidleget()
if _p_ == nil { // 没有闲暇的 p 了,只能回去了
unlock(&sched.lock)
if spinning {
// The caller incremented nmspinning, but there are no idle Ps,
// so it's okay to just undo the increment and give up.
if int32(atomic.Xadd(&sched.nmspinning, -1)) < 0 {throw("startm: negative nmspinning")
}
}
releasem(mp) // 能够抢占了
return
}
}
nmp := mget() // 获取一个全局闲暇的 m
if nmp == nil { // 没有闲暇的 m 了
// No M is available, we must drop sched.lock and call newm.
// However, we already own a P to assign to the M.
//
// Once sched.lock is released, another G (e.g., in a syscall),
// could find no idle P while checkdead finds a runnable G but
// no running M's because this new M hasn't started yet, thus
// throwing in an apparent deadlock.
//
// Avoid this situation by pre-allocating the ID for the new M,
// thus marking it as 'running' before we drop sched.lock. This
// new M will eventually run the scheduler to execute any
// queued G's.
id := mReserveID()
unlock(&sched.lock)
var fn func()
if spinning {
// The caller incremented nmspinning, so set m.spinning in the new M.
fn = mspinning
}
newm(fn, _p_, id) // 简略新建一个 m,就能够回去了
// Ownership transfer of _p_ committed by start in newm.
// Preemption is now safe.
releasem(mp) // 开释以后 g 的 m,能够被抢占了
return
}
unlock(&sched.lock)
if nmp.spinning {throw("startm: m is spinning")
}
if nmp.nextp != 0 {throw("startm: m has p")
}
if spinning && !runqempty(_p_) {throw("startm: p has runnable gs")
}
// The caller incremented nmspinning, so set m.spinning in the new M.
nmp.spinning = spinning // 标记该 M 是否在自旋
nmp.nextp.set(_p_) // 暂存 P
notewakeup(&nmp.park) // 唤醒 M
// Ownership transfer of _p_ committed by wakeup. Preemption is now
// safe.
releasem(mp)
}
startm 次要实现工作:
- 如果_p_为空就获取缓存的_p_
- 如果没有闲暇的 m, new 一个 m 并且初始化 m, 包含创立 go 和 gsignal, 新建零碎线程,并且在下面执行 mstart
-
如果有闲暇的 m, 唤醒 m
newm
// Create a new m. It will start off with a call to fn, or else the scheduler. // fn needs to be static and not a heap allocated closure. // May run with m.p==nil, so write barriers are not allowed. // // id is optional pre-allocated m ID. Omit by passing -1. //go:nowritebarrierrec func newm(fn func(), _p_ *p, id int64) {mp := allocm(_p_, fn, id) // new 一个 m 并且初始化 m, 包含创立 go 和 gsignal mp.doesPark = (_p_ != nil) // m 是否应该挂起,P!=nil 就能够间接用 p 执行了 就不必挂起了 mp.nextp.set(_p_) mp.sigmask = initSigmask if gp := getg(); gp != nil && gp.m != nil && (gp.m.lockedExt != 0 || gp.m.incgo) && GOOS != "plan9" { // We're on a locked M or a thread that may have been // started by C. The kernel state of this thread may // be strange (the user may have locked it for that // purpose). We don't want to clone that into another // thread. Instead, ask a known-good thread to create // the thread for us. // // This is disabled on Plan 9. See golang.org/issue/22227. // // TODO: This may be unnecessary on Windows, which // doesn't model thread creation off fork. lock(&newmHandoff.lock) if newmHandoff.haveTemplateThread == 0 {throw("on a locked thread with no template thread") } mp.schedlink = newmHandoff.newm newmHandoff.newm.set(mp) if newmHandoff.waiting { newmHandoff.waiting = false notewakeup(&newmHandoff.wake) } unlock(&newmHandoff.lock) return } // 关联真正的调配 os thread // 调配一个零碎线程,且实现 g0 上的栈调配 // 传入 mstart 函数,让线程执行 mstart newm1(mp) }
newm 的次要工作:
- new 一个 m 并且初始化 m, 包含创立 go 和 gsignal
- 初始化一些参数
-
新建一个零碎线程并且执行 mstart
func newm1(mp *m) { if iscgo { var ts cgothreadstart if _cgo_thread_start == nil {throw("_cgo_thread_start missing") } ts.g.set(mp.g0) ts.tls = (*uint64)(unsafe.Pointer(&mp.tls[0])) ts.fn = unsafe.Pointer(funcPC(mstart)) if msanenabled {msanwrite(unsafe.Pointer(&ts), unsafe.Sizeof(ts)) } execLock.rlock() // Prevent process clone. asmcgocall(_cgo_thread_start, unsafe.Pointer(&ts)) execLock.runlock() return } execLock.rlock() // Prevent process clone. newosproc(mp) execLock.runlock()}
Linux newosproc
// May run with m.p==nil, so write barriers are not allowed. //go:nowritebarrier func newosproc(mp *m) { // 调配一个零碎线程,且实现 g0 上的栈调配 // 传入 mstart 函数,让线程执行 mstart stk := unsafe.Pointer(mp.g0.stack.hi) /* * note: strace gets confused if we use CLONE_PTRACE here. */ if false {print("newosproc stk=", stk, "m=", mp, "g=", mp.g0, "clone=", funcPC(clone), "id=", mp.id, "ostk=", &mp, "\n") } // Disable signals during clone, so that the new thread starts // with signals disabled. It will enable them in minit. var oset sigset sigprocmask(_SIG_SETMASK, &sigset_all, &oset) ret := clone(cloneFlags, stk, unsafe.Pointer(mp), unsafe.Pointer(mp.g0), unsafe.Pointer(funcPC(mstart))) sigprocmask(_SIG_SETMASK, &oset, nil) if ret < 0 {print("runtime: failed to create new OS thread (have", mcount(), "already; errno=", -ret, ")\n") if ret == -_EAGAIN {println("runtime: may need to increase max user processes (ulimit -u)") } throw("newosproc") } }
Windows newosproc
// May run with m.p==nil, so write barriers are not allowed. This // function is called by newosproc0, so it is also required to // operate without stack guards. //go:nowritebarrierrec //go:nosplit func newosproc(mp *m) { // 调配一个零碎线程,且实现 g0 和 g0 上的栈调配 // 传入 mstart 函数,让线程执行 mstart // We pass 0 for the stack size to use the default for this binary. thandle := stdcall6(_CreateThread, 0, 0, funcPC(tstart_stdcall), uintptr(unsafe.Pointer(mp)), 0, 0) if thandle == 0 {if atomic.Load(&exiting) != 0 { // CreateThread may fail if called // concurrently with ExitProcess. If this // happens, just freeze this thread and let // the process exit. See issue #18253. lock(&deadlock) lock(&deadlock) } print("runtime: failed to create new OS thread (have", mcount(), "already; errno=", getlasterror(), ")\n") throw("runtime.newosproc") } // Close thandle to avoid leaking the thread object if it exits. stdcall1(_CloseHandle, thandle) }
allocm
调配一个 m,且不关联任何一个 os thread
// Allocate a new m unassociated with any thread. // Can use p for allocation context if needed. // fn is recorded as the new m's m.mstartfn. // id is optional pre-allocated m ID. Omit by passing -1. // // This function is allowed to have write barriers even if the caller // isn't because it borrows _p_. // //go:yeswritebarrierrec func allocm(_p_ *p, fn func(), id int64) *m {_g_ := getg() acquirem() // disable GC because it can be called from sysmon if _g_.m.p == 0 { // 为什么会可能没有绑定 p 呢 // 把__p__和 g.m 互相绑定,并且把_p_.status 从_Pidle 转为_Prunning acquirep(_p_) // temporarily borrow p for mallocs in this function } // Release the free M list. We need to do this somewhere and // this may free up a stack we can use. // mexit 的时候会加到 freem, m.gsignal 会在那时候开释,这个构造 // 因为 m 是又 new 创立的,能够由 gc 开释 if sched.freem != nil {lock(&sched.lock) var newList *m for freem := sched.freem; freem != nil; { if freem.freeWait != 0 { next := freem.freelink freem.freelink = newList newList = freem freem = next continue } // stackfree must be on the system stack, but allocm is // reachable off the system stack transitively from // startm. systemstack(func() {stackfree(freem.g0.stack) }) freem = freem.freelink } sched.freem = newList unlock(&sched.lock) } mp := new(m) mp.mstartfn = fn mcommoninit(mp, id) // 在文章(一)有介绍,次要是创立 gsignal,并且把 m 退出 allm // In case of cgo or Solaris or illumos or Darwin, pthread_create will make us a stack. // Windows and Plan 9 will layout sched stack on OS stack. // 创立 g0 if iscgo || mStackIsSystemAllocated() {mp.g0 = malg(-1) } else {mp.g0 = malg(8192 * sys.StackGuardMultiplier) } mp.g0.m = mp if _p_ == _g_.m.p.ptr() {releasep() // 互相_g_.m 和__p__互相解绑 } releasem(_g_.m) // 能够抢占 return mp }
_P_的作用是_g_.m 为空的时候借用来申请堆内存的, 借完_p_.status 设置成_Pidle 并且还回去
allocm 次要实现一下工作: - new 一个 m 并且初始化 m, 包含创立 go 和 gsignal
援用文章
[1] Go 语言底细(6):启动和内存调配初始化
https://studygolang.com/artic…