1 // Copyright 2014 The Go Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style
3 // license that can be found in the LICENSE file.
11 "runtime/internal/atomic"
12 "runtime/internal/sys"
16 // set using cmd/go/internal/modload.ModInfoProg
19 // Goroutine scheduler
20 // The scheduler's job is to distribute ready-to-run goroutines over worker threads.
22 // The main concepts are:
24 // M - worker thread, or machine.
25 // P - processor, a resource that is required to execute Go code.
26 // M must have an associated P to execute Go code, however it can be
27 // blocked or in a syscall w/o an associated P.
29 // Design doc at https://golang.org/s/go11sched.
31 // Worker thread parking/unparking.
32 // We need to balance between keeping enough running worker threads to utilize
33 // available hardware parallelism and parking excessive running worker threads
34 // to conserve CPU resources and power. This is not simple for two reasons:
35 // (1) scheduler state is intentionally distributed (in particular, per-P work
36 // queues), so it is not possible to compute global predicates on fast paths;
37 // (2) for optimal thread management we would need to know the future (don't park
38 // a worker thread when a new goroutine will be readied in near future).
40 // Three rejected approaches that would work badly:
41 // 1. Centralize all scheduler state (would inhibit scalability).
42 // 2. Direct goroutine handoff. That is, when we ready a new goroutine and there
43 // is a spare P, unpark a thread and handoff it the thread and the goroutine.
44 // This would lead to thread state thrashing, as the thread that readied the
45 // goroutine can be out of work the very next moment, we will need to park it.
46 // Also, it would destroy locality of computation as we want to preserve
47 // dependent goroutines on the same thread; and introduce additional latency.
48 // 3. Unpark an additional thread whenever we ready a goroutine and there is an
49 // idle P, but don't do handoff. This would lead to excessive thread parking/
50 // unparking as the additional threads will instantly park without discovering
53 // The current approach:
55 // This approach applies to three primary sources of potential work: readying a
56 // goroutine, new/modified-earlier timers, and idle-priority GC. See below for
57 // additional details.
59 // We unpark an additional thread when we submit work if (this is wakep()):
60 // 1. There is an idle P, and
61 // 2. There are no "spinning" worker threads.
63 // A worker thread is considered spinning if it is out of local work and did
64 // not find work in the global run queue or netpoller; the spinning state is
65 // denoted in m.spinning and in sched.nmspinning. Threads unparked this way are
66 // also considered spinning; we don't do goroutine handoff so such threads are
67 // out of work initially. Spinning threads spin on looking for work in per-P
68 // run queues and timer heaps or from the GC before parking. If a spinning
69 // thread finds work it takes itself out of the spinning state and proceeds to
70 // execution. If it does not find work it takes itself out of the spinning
71 // state and then parks.
73 // If there is at least one spinning thread (sched.nmspinning>1), we don't
74 // unpark new threads when submitting work. To compensate for that, if the last
75 // spinning thread finds work and stops spinning, it must unpark a new spinning
76 // thread. This approach smooths out unjustified spikes of thread unparking,
77 // but at the same time guarantees eventual maximal CPU parallelism
80 // The main implementation complication is that we need to be very careful
81 // during spinning->non-spinning thread transition. This transition can race
82 // with submission of new work, and either one part or another needs to unpark
83 // another worker thread. If they both fail to do that, we can end up with
84 // semi-persistent CPU underutilization.
86 // The general pattern for submission is:
87 // 1. Submit work to the local run queue, timer heap, or GC state.
88 // 2. #StoreLoad-style memory barrier.
89 // 3. Check sched.nmspinning.
91 // The general pattern for spinning->non-spinning transition is:
92 // 1. Decrement nmspinning.
93 // 2. #StoreLoad-style memory barrier.
94 // 3. Check all per-P work queues and GC for new work.
96 // Note that all this complexity does not apply to global run queue as we are
97 // not sloppy about thread unparking when submitting to global queue. Also see
98 // comments for nmspinning manipulation.
100 // How these different sources of work behave varies, though it doesn't affect
101 // the synchronization approach:
102 // * Ready goroutine: this is an obvious source of work; the goroutine is
103 // immediately ready and must run on some thread eventually.
104 // * New/modified-earlier timer: The current timer implementation (see time.go)
105 // uses netpoll in a thread with no work available to wait for the soonest
106 // timer. If there is no thread waiting, we want a new spinning thread to go
108 // * Idle-priority GC: The GC wakes a stopped idle thread to contribute to
109 // background GC work (note: currently disabled per golang.org/issue/19112).
110 // Also see golang.org/issue/44313, as this should be extended to all GC
120 //go:linkname runtime_inittask runtime..inittask
121 var runtime_inittask initTask
123 //go:linkname main_inittask main..inittask
124 var main_inittask initTask
126 // main_init_done is a signal used by cgocallbackg that initialization
127 // has been completed. It is made before _cgo_notify_runtime_init_done,
128 // so all cgo calls can rely on it existing. When main_init is complete,
129 // it is closed, meaning cgocallbackg can reliably receive from it.
130 var main_init_done chan bool
132 //go:linkname main_main main.main
135 // mainStarted indicates that the main M has started.
138 // runtimeInitTime is the nanotime() at which the runtime started.
139 var runtimeInitTime int64
141 // Value to use for signal mask for newly created M's.
142 var initSigmask sigset
144 // The main goroutine.
148 // Racectx of m0->g0 is used only as the parent of the main goroutine.
149 // It must not be used for anything else.
152 // Max stack size is 1 GB on 64-bit, 250 MB on 32-bit.
153 // Using decimal instead of binary GB and MB because
154 // they look nicer in the stack overflow failure message.
155 if goarch.PtrSize == 8 {
156 maxstacksize = 1000000000
158 maxstacksize = 250000000
161 // An upper limit for max stack size. Used to avoid random crashes
162 // after calling SetMaxStack and trying to allocate a stack that is too big,
163 // since stackalloc works with 32-bit sizes.
164 maxstackceiling = 2 * maxstacksize
166 // Allow newproc to start new Ms.
169 if GOARCH != "wasm" { // no threads on wasm yet, so no sysmon
171 newm(sysmon, nil, -1)
175 // Lock the main goroutine onto this, the main OS thread,
176 // during initialization. Most programs won't care, but a few
177 // do require certain calls to be made by the main thread.
178 // Those can arrange for main.main to run in the main thread
179 // by calling runtime.LockOSThread during initialization
180 // to preserve the lock.
184 throw("runtime.main not on m0")
187 // Record when the world started.
188 // Must be before doInit for tracing init.
189 runtimeInitTime = nanotime()
190 if runtimeInitTime == 0 {
191 throw("nanotime returning zero")
194 if debug.inittrace != 0 {
195 inittrace.id = getg().goid
196 inittrace.active = true
199 doInit(&runtime_inittask) // Must be before defer.
201 // Defer unlock so that runtime.Goexit during init does the unlock too.
211 main_init_done = make(chan bool)
213 if _cgo_pthread_key_created == nil {
214 throw("_cgo_pthread_key_created missing")
217 if _cgo_thread_start == nil {
218 throw("_cgo_thread_start missing")
220 if GOOS != "windows" {
221 if _cgo_setenv == nil {
222 throw("_cgo_setenv missing")
224 if _cgo_unsetenv == nil {
225 throw("_cgo_unsetenv missing")
228 if _cgo_notify_runtime_init_done == nil {
229 throw("_cgo_notify_runtime_init_done missing")
232 // Set the x_crosscall2_ptr C function pointer variable point to crosscall2.
233 if set_crosscall2 == nil {
234 throw("set_crosscall2 missing")
238 // Start the template thread in case we enter Go from
239 // a C-created thread and need to create a new thread.
240 startTemplateThread()
241 cgocall(_cgo_notify_runtime_init_done, nil)
244 doInit(&main_inittask)
246 // Disable init tracing after main init done to avoid overhead
247 // of collecting statistics in malloc and newproc
248 inittrace.active = false
250 close(main_init_done)
255 if isarchive || islibrary {
256 // A program compiled with -buildmode=c-archive or c-shared
257 // has a main, but it is not executed.
260 fn := main_main // make an indirect call, as the linker doesn't know the address of the main package when laying down the runtime
263 runExitHooks(0) // run hooks now, since racefini does not return
267 // Make racy client program work: if panicking on
268 // another goroutine at the same time as main returns,
269 // let the other goroutine finish printing the panic trace.
270 // Once it does, it will exit. See issues 3934 and 20018.
271 if runningPanicDefers.Load() != 0 {
272 // Running deferred functions should not take long.
273 for c := 0; c < 1000; c++ {
274 if runningPanicDefers.Load() == 0 {
280 if panicking.Load() != 0 {
281 gopark(nil, nil, waitReasonPanicWait, traceEvGoStop, 1)
292 // os_beforeExit is called from os.Exit(0).
294 //go:linkname os_beforeExit os.runtime_beforeExit
295 func os_beforeExit(exitCode int) {
296 runExitHooks(exitCode)
297 if exitCode == 0 && raceenabled {
302 // start forcegc helper goroutine
307 func forcegchelper() {
309 lockInit(&forcegc.lock, lockRankForcegc)
312 if forcegc.idle.Load() {
313 throw("forcegc: phase error")
315 forcegc.idle.Store(true)
316 goparkunlock(&forcegc.lock, waitReasonForceGCIdle, traceEvGoBlock, 1)
317 // this goroutine is explicitly resumed by sysmon
318 if debug.gctrace > 0 {
321 // Time-triggered, fully concurrent.
322 gcStart(gcTrigger{kind: gcTriggerTime, now: nanotime()})
326 // Gosched yields the processor, allowing other goroutines to run. It does not
327 // suspend the current goroutine, so execution resumes automatically.
335 // goschedguarded yields the processor like gosched, but also checks
336 // for forbidden states and opts out of the yield in those cases.
339 func goschedguarded() {
340 mcall(goschedguarded_m)
343 // goschedIfBusy yields the processor like gosched, but only does so if
344 // there are no idle Ps or if we're on the only P and there's nothing in
345 // the run queue. In both cases, there is freely available idle time.
348 func goschedIfBusy() {
350 // Call gosched if gp.preempt is set; we may be in a tight loop that
351 // doesn't otherwise yield.
352 if !gp.preempt && sched.npidle.Load() > 0 {
358 // Puts the current goroutine into a waiting state and calls unlockf on the
361 // If unlockf returns false, the goroutine is resumed.
363 // unlockf must not access this G's stack, as it may be moved between
364 // the call to gopark and the call to unlockf.
366 // Note that because unlockf is called after putting the G into a waiting
367 // state, the G may have already been readied by the time unlockf is called
368 // unless there is external synchronization preventing the G from being
369 // readied. If unlockf returns false, it must guarantee that the G cannot be
370 // externally readied.
372 // Reason explains why the goroutine has been parked. It is displayed in stack
373 // traces and heap dumps. Reasons should be unique and descriptive. Do not
374 // re-use reasons, add new ones.
375 func gopark(unlockf func(*g, unsafe.Pointer) bool, lock unsafe.Pointer, reason waitReason, traceEv byte, traceskip int) {
376 if reason != waitReasonSleep {
377 checkTimeouts() // timeouts may expire while two goroutines keep the scheduler busy
381 status := readgstatus(gp)
382 if status != _Grunning && status != _Gscanrunning {
383 throw("gopark: bad g status")
386 mp.waitunlockf = unlockf
387 gp.waitreason = reason
388 mp.waittraceev = traceEv
389 mp.waittraceskip = traceskip
391 // can't do anything that might move the G between Ms here.
395 // Puts the current goroutine into a waiting state and unlocks the lock.
396 // The goroutine can be made runnable again by calling goready(gp).
397 func goparkunlock(lock *mutex, reason waitReason, traceEv byte, traceskip int) {
398 gopark(parkunlock_c, unsafe.Pointer(lock), reason, traceEv, traceskip)
401 func goready(gp *g, traceskip int) {
403 ready(gp, traceskip, true)
408 func acquireSudog() *sudog {
409 // Delicate dance: the semaphore implementation calls
410 // acquireSudog, acquireSudog calls new(sudog),
411 // new calls malloc, malloc can call the garbage collector,
412 // and the garbage collector calls the semaphore implementation
414 // Break the cycle by doing acquirem/releasem around new(sudog).
415 // The acquirem/releasem increments m.locks during new(sudog),
416 // which keeps the garbage collector from being invoked.
419 if len(pp.sudogcache) == 0 {
420 lock(&sched.sudoglock)
421 // First, try to grab a batch from central cache.
422 for len(pp.sudogcache) < cap(pp.sudogcache)/2 && sched.sudogcache != nil {
423 s := sched.sudogcache
424 sched.sudogcache = s.next
426 pp.sudogcache = append(pp.sudogcache, s)
428 unlock(&sched.sudoglock)
429 // If the central cache is empty, allocate a new one.
430 if len(pp.sudogcache) == 0 {
431 pp.sudogcache = append(pp.sudogcache, new(sudog))
434 n := len(pp.sudogcache)
435 s := pp.sudogcache[n-1]
436 pp.sudogcache[n-1] = nil
437 pp.sudogcache = pp.sudogcache[:n-1]
439 throw("acquireSudog: found s.elem != nil in cache")
446 func releaseSudog(s *sudog) {
448 throw("runtime: sudog with non-nil elem")
451 throw("runtime: sudog with non-false isSelect")
454 throw("runtime: sudog with non-nil next")
457 throw("runtime: sudog with non-nil prev")
459 if s.waitlink != nil {
460 throw("runtime: sudog with non-nil waitlink")
463 throw("runtime: sudog with non-nil c")
467 throw("runtime: releaseSudog with non-nil gp.param")
469 mp := acquirem() // avoid rescheduling to another P
471 if len(pp.sudogcache) == cap(pp.sudogcache) {
472 // Transfer half of local cache to the central cache.
473 var first, last *sudog
474 for len(pp.sudogcache) > cap(pp.sudogcache)/2 {
475 n := len(pp.sudogcache)
476 p := pp.sudogcache[n-1]
477 pp.sudogcache[n-1] = nil
478 pp.sudogcache = pp.sudogcache[:n-1]
486 lock(&sched.sudoglock)
487 last.next = sched.sudogcache
488 sched.sudogcache = first
489 unlock(&sched.sudoglock)
491 pp.sudogcache = append(pp.sudogcache, s)
495 // called from assembly.
496 func badmcall(fn func(*g)) {
497 throw("runtime: mcall called on m->g0 stack")
500 func badmcall2(fn func(*g)) {
501 throw("runtime: mcall function returned")
504 func badreflectcall() {
505 panic(plainError("arg size to reflect.call more than 1GB"))
509 //go:nowritebarrierrec
510 func badmorestackg0() {
511 writeErrStr("fatal: morestack on g0\n")
515 //go:nowritebarrierrec
516 func badmorestackgsignal() {
517 writeErrStr("fatal: morestack on gsignal\n")
525 func lockedOSThread() bool {
527 return gp.lockedm != 0 && gp.m.lockedg != 0
531 // allgs contains all Gs ever created (including dead Gs), and thus
534 // Access via the slice is protected by allglock or stop-the-world.
535 // Readers that cannot take the lock may (carefully!) use the atomic
540 // allglen and allgptr are atomic variables that contain len(allgs) and
541 // &allgs[0] respectively. Proper ordering depends on totally-ordered
542 // loads and stores. Writes are protected by allglock.
544 // allgptr is updated before allglen. Readers should read allglen
545 // before allgptr to ensure that allglen is always <= len(allgptr). New
546 // Gs appended during the race can be missed. For a consistent view of
547 // all Gs, allglock must be held.
549 // allgptr copies should always be stored as a concrete type or
550 // unsafe.Pointer, not uintptr, to ensure that GC can still reach it
551 // even if it points to a stale array.
556 func allgadd(gp *g) {
557 if readgstatus(gp) == _Gidle {
558 throw("allgadd: bad status Gidle")
562 allgs = append(allgs, gp)
563 if &allgs[0] != allgptr {
564 atomicstorep(unsafe.Pointer(&allgptr), unsafe.Pointer(&allgs[0]))
566 atomic.Storeuintptr(&allglen, uintptr(len(allgs)))
570 // allGsSnapshot returns a snapshot of the slice of all Gs.
572 // The world must be stopped or allglock must be held.
573 func allGsSnapshot() []*g {
574 assertWorldStoppedOrLockHeld(&allglock)
576 // Because the world is stopped or allglock is held, allgadd
577 // cannot happen concurrently with this. allgs grows
578 // monotonically and existing entries never change, so we can
579 // simply return a copy of the slice header. For added safety,
580 // we trim everything past len because that can still change.
581 return allgs[:len(allgs):len(allgs)]
584 // atomicAllG returns &allgs[0] and len(allgs) for use with atomicAllGIndex.
585 func atomicAllG() (**g, uintptr) {
586 length := atomic.Loaduintptr(&allglen)
587 ptr := (**g)(atomic.Loadp(unsafe.Pointer(&allgptr)))
591 // atomicAllGIndex returns ptr[i] with the allgptr returned from atomicAllG.
592 func atomicAllGIndex(ptr **g, i uintptr) *g {
593 return *(**g)(add(unsafe.Pointer(ptr), i*goarch.PtrSize))
596 // forEachG calls fn on every G from allgs.
598 // forEachG takes a lock to exclude concurrent addition of new Gs.
599 func forEachG(fn func(gp *g)) {
601 for _, gp := range allgs {
607 // forEachGRace calls fn on every G from allgs.
609 // forEachGRace avoids locking, but does not exclude addition of new Gs during
610 // execution, which may be missed.
611 func forEachGRace(fn func(gp *g)) {
612 ptr, length := atomicAllG()
613 for i := uintptr(0); i < length; i++ {
614 gp := atomicAllGIndex(ptr, i)
621 // Number of goroutine ids to grab from sched.goidgen to local per-P cache at once.
622 // 16 seems to provide enough amortization, but other than that it's mostly arbitrary number.
626 // cpuinit sets up CPU feature flags and calls internal/cpu.Initialize. env should be the complete
627 // value of the GODEBUG environment variable.
628 func cpuinit(env string) {
630 case "aix", "darwin", "ios", "dragonfly", "freebsd", "netbsd", "openbsd", "illumos", "solaris", "linux":
631 cpu.DebugOptions = true
635 // Support cpu feature variables are used in code generated by the compiler
636 // to guard execution of instructions that can not be assumed to be always supported.
639 x86HasPOPCNT = cpu.X86.HasPOPCNT
640 x86HasSSE41 = cpu.X86.HasSSE41
641 x86HasFMA = cpu.X86.HasFMA
644 armHasVFPv4 = cpu.ARM.HasVFPv4
647 arm64HasATOMICS = cpu.ARM64.HasATOMICS
651 // getGodebugEarly extracts the environment variable GODEBUG from the environment on
652 // Unix-like operating systems and returns it. This function exists to extract GODEBUG
653 // early before much of the runtime is initialized.
654 func getGodebugEarly() string {
655 const prefix = "GODEBUG="
658 case "aix", "darwin", "ios", "dragonfly", "freebsd", "netbsd", "openbsd", "illumos", "solaris", "linux":
659 // Similar to goenv_unix but extracts the environment value for
661 // TODO(moehrmann): remove when general goenvs() can be called before cpuinit()
663 for argv_index(argv, argc+1+n) != nil {
667 for i := int32(0); i < n; i++ {
668 p := argv_index(argv, argc+1+i)
669 s := unsafe.String(p, findnull(p))
671 if hasPrefix(s, prefix) {
672 env = gostring(p)[len(prefix):]
680 // The bootstrap sequence is:
684 // make & queue new G
685 // call runtime·mstart
687 // The new G calls runtime·main.
689 lockInit(&sched.lock, lockRankSched)
690 lockInit(&sched.sysmonlock, lockRankSysmon)
691 lockInit(&sched.deferlock, lockRankDefer)
692 lockInit(&sched.sudoglock, lockRankSudog)
693 lockInit(&deadlock, lockRankDeadlock)
694 lockInit(&paniclk, lockRankPanic)
695 lockInit(&allglock, lockRankAllg)
696 lockInit(&allpLock, lockRankAllp)
697 lockInit(&reflectOffs.lock, lockRankReflectOffs)
698 lockInit(&finlock, lockRankFin)
699 lockInit(&trace.bufLock, lockRankTraceBuf)
700 lockInit(&trace.stringsLock, lockRankTraceStrings)
701 lockInit(&trace.lock, lockRankTrace)
702 lockInit(&cpuprof.lock, lockRankCpuprof)
703 lockInit(&trace.stackTab.lock, lockRankTraceStackTab)
704 // Enforce that this lock is always a leaf lock.
705 // All of this lock's critical sections should be
707 lockInit(&memstats.heapStats.noPLock, lockRankLeafRank)
709 // raceinit must be the first call to race detector.
710 // In particular, it must be done before mallocinit below calls racemapshadow.
713 gp.racectx, raceprocctx0 = raceinit()
716 sched.maxmcount = 10000
718 // The world starts stopped.
724 godebug := getGodebugEarly()
725 initPageTrace(godebug) // must run after mallocinit but before anything allocates
726 cpuinit(godebug) // must run before alginit
727 alginit() // maps, hash, fastrand must not be used before this call
728 fastrandinit() // must run before mcommoninit
729 mcommoninit(gp.m, -1)
730 modulesinit() // provides activeModules
731 typelinksinit() // uses maps, activeModules
732 itabsinit() // uses activeModules
733 stkobjinit() // must run before GC starts
735 sigsave(&gp.m.sigmask)
736 initSigmask = gp.m.sigmask
743 // if disableMemoryProfiling is set, update MemProfileRate to 0 to turn off memprofile.
744 // Note: parsedebugvars may update MemProfileRate, but when disableMemoryProfiling is
745 // set to true by the linker, it means that nothing is consuming the profile, it is
746 // safe to set MemProfileRate to 0.
747 if disableMemoryProfiling {
752 sched.lastpoll.Store(nanotime())
754 if n, ok := atoi32(gogetenv("GOMAXPROCS")); ok && n > 0 {
757 if procresize(procs) != nil {
758 throw("unknown runnable goroutine during bootstrap")
762 // World is effectively started now, as P's can run.
765 if buildVersion == "" {
766 // Condition should never trigger. This code just serves
767 // to ensure runtime·buildVersion is kept in the resulting binary.
768 buildVersion = "unknown"
770 if len(modinfo) == 1 {
771 // Condition should never trigger. This code just serves
772 // to ensure runtime·modinfo is kept in the resulting binary.
777 func dumpgstatus(gp *g) {
779 print("runtime: gp: gp=", gp, ", goid=", gp.goid, ", gp->atomicstatus=", readgstatus(gp), "\n")
780 print("runtime: getg: g=", thisg, ", goid=", thisg.goid, ", g->atomicstatus=", readgstatus(thisg), "\n")
783 // sched.lock must be held.
785 assertLockHeld(&sched.lock)
787 if mcount() > sched.maxmcount {
788 print("runtime: program exceeds ", sched.maxmcount, "-thread limit\n")
789 throw("thread exhaustion")
793 // mReserveID returns the next ID to use for a new m. This new m is immediately
794 // considered 'running' by checkdead.
796 // sched.lock must be held.
797 func mReserveID() int64 {
798 assertLockHeld(&sched.lock)
800 if sched.mnext+1 < sched.mnext {
801 throw("runtime: thread ID overflow")
809 // Pre-allocated ID may be passed as 'id', or omitted by passing -1.
810 func mcommoninit(mp *m, id int64) {
813 // g0 stack won't make sense for user (and is not necessary unwindable).
815 callers(1, mp.createstack[:])
826 lo := uint32(int64Hash(uint64(mp.id), fastrandseed))
827 hi := uint32(int64Hash(uint64(cputicks()), ^fastrandseed))
831 // Same behavior as for 1.17.
832 // TODO: Simplify this.
833 if goarch.BigEndian {
834 mp.fastrand = uint64(lo)<<32 | uint64(hi)
836 mp.fastrand = uint64(hi)<<32 | uint64(lo)
840 if mp.gsignal != nil {
841 mp.gsignal.stackguard1 = mp.gsignal.stack.lo + _StackGuard
844 // Add to allm so garbage collector doesn't free g->m
845 // when it is just in a register or thread-local storage.
848 // NumCgoCall() iterates over allm w/o schedlock,
849 // so we need to publish it safely.
850 atomicstorep(unsafe.Pointer(&allm), unsafe.Pointer(mp))
853 // Allocate memory to hold a cgo traceback if the cgo call crashes.
854 if iscgo || GOOS == "solaris" || GOOS == "illumos" || GOOS == "windows" {
855 mp.cgoCallers = new(cgoCallers)
859 func (mp *m) becomeSpinning() {
861 sched.nmspinning.Add(1)
862 sched.needspinning.Store(0)
865 func (mp *m) incgocallback() bool {
866 return (!mp.incgo && mp.ncgo > 0) || mp.isextra
869 var fastrandseed uintptr
871 func fastrandinit() {
872 s := (*[unsafe.Sizeof(fastrandseed)]byte)(unsafe.Pointer(&fastrandseed))[:]
876 // Mark gp ready to run.
877 func ready(gp *g, traceskip int, next bool) {
879 traceGoUnpark(gp, traceskip)
882 status := readgstatus(gp)
885 mp := acquirem() // disable preemption because it can be holding p in a local var
886 if status&^_Gscan != _Gwaiting {
888 throw("bad g->status in ready")
891 // status is Gwaiting or Gscanwaiting, make Grunnable and put on runq
892 casgstatus(gp, _Gwaiting, _Grunnable)
893 runqput(mp.p.ptr(), gp, next)
898 // freezeStopWait is a large value that freezetheworld sets
899 // sched.stopwait to in order to request that all Gs permanently stop.
900 const freezeStopWait = 0x7fffffff
902 // freezing is set to non-zero if the runtime is trying to freeze the
904 var freezing atomic.Bool
906 // Similar to stopTheWorld but best-effort and can be called several times.
907 // There is no reverse operation, used during crashing.
908 // This function must not lock any mutexes.
909 func freezetheworld() {
911 // stopwait and preemption requests can be lost
912 // due to races with concurrently executing threads,
913 // so try several times
914 for i := 0; i < 5; i++ {
915 // this should tell the scheduler to not start any new goroutines
916 sched.stopwait = freezeStopWait
917 sched.gcwaiting.Store(true)
918 // this should stop running goroutines
920 break // no running goroutines
930 // All reads and writes of g's status go through readgstatus, casgstatus
931 // castogscanstatus, casfrom_Gscanstatus.
934 func readgstatus(gp *g) uint32 {
935 return gp.atomicstatus.Load()
938 // The Gscanstatuses are acting like locks and this releases them.
939 // If it proves to be a performance hit we should be able to make these
940 // simple atomic stores but for now we are going to throw if
941 // we see an inconsistent state.
942 func casfrom_Gscanstatus(gp *g, oldval, newval uint32) {
945 // Check that transition is valid.
948 print("runtime: casfrom_Gscanstatus bad oldval gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n")
950 throw("casfrom_Gscanstatus:top gp->status is not in scan state")
956 if newval == oldval&^_Gscan {
957 success = gp.atomicstatus.CompareAndSwap(oldval, newval)
961 print("runtime: casfrom_Gscanstatus failed gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n")
963 throw("casfrom_Gscanstatus: gp->status is not in scan state")
965 releaseLockRank(lockRankGscan)
968 // This will return false if the gp is not in the expected status and the cas fails.
969 // This acts like a lock acquire while the casfromgstatus acts like a lock release.
970 func castogscanstatus(gp *g, oldval, newval uint32) bool {
976 if newval == oldval|_Gscan {
977 r := gp.atomicstatus.CompareAndSwap(oldval, newval)
979 acquireLockRank(lockRankGscan)
985 print("runtime: castogscanstatus oldval=", hex(oldval), " newval=", hex(newval), "\n")
986 throw("castogscanstatus")
990 // casgstatusAlwaysTrack is a debug flag that causes casgstatus to always track
991 // various latencies on every transition instead of sampling them.
992 var casgstatusAlwaysTrack = false
994 // If asked to move to or from a Gscanstatus this will throw. Use the castogscanstatus
995 // and casfrom_Gscanstatus instead.
996 // casgstatus will loop if the g->atomicstatus is in a Gscan status until the routine that
997 // put it in the Gscan state is finished.
1000 func casgstatus(gp *g, oldval, newval uint32) {
1001 if (oldval&_Gscan != 0) || (newval&_Gscan != 0) || oldval == newval {
1002 systemstack(func() {
1003 print("runtime: casgstatus: oldval=", hex(oldval), " newval=", hex(newval), "\n")
1004 throw("casgstatus: bad incoming values")
1008 acquireLockRank(lockRankGscan)
1009 releaseLockRank(lockRankGscan)
1011 // See https://golang.org/cl/21503 for justification of the yield delay.
1012 const yieldDelay = 5 * 1000
1015 // loop if gp->atomicstatus is in a scan state giving
1016 // GC time to finish and change the state to oldval.
1017 for i := 0; !gp.atomicstatus.CompareAndSwap(oldval, newval); i++ {
1018 if oldval == _Gwaiting && gp.atomicstatus.Load() == _Grunnable {
1019 throw("casgstatus: waiting for Gwaiting but is Grunnable")
1022 nextYield = nanotime() + yieldDelay
1024 if nanotime() < nextYield {
1025 for x := 0; x < 10 && gp.atomicstatus.Load() != oldval; x++ {
1030 nextYield = nanotime() + yieldDelay/2
1034 if oldval == _Grunning {
1035 // Track every gTrackingPeriod time a goroutine transitions out of running.
1036 if casgstatusAlwaysTrack || gp.trackingSeq%gTrackingPeriod == 0 {
1045 // Handle various kinds of tracking.
1048 // - Time spent in runnable.
1049 // - Time spent blocked on a sync.Mutex or sync.RWMutex.
1052 // We transitioned out of runnable, so measure how much
1053 // time we spent in this state and add it to
1056 gp.runnableTime += now - gp.trackingStamp
1057 gp.trackingStamp = 0
1059 if !gp.waitreason.isMutexWait() {
1060 // Not blocking on a lock.
1063 // Blocking on a lock, measure it. Note that because we're
1064 // sampling, we have to multiply by our sampling period to get
1065 // a more representative estimate of the absolute value.
1066 // gTrackingPeriod also represents an accurate sampling period
1067 // because we can only enter this state from _Grunning.
1069 sched.totalMutexWaitTime.Add((now - gp.trackingStamp) * gTrackingPeriod)
1070 gp.trackingStamp = 0
1074 if !gp.waitreason.isMutexWait() {
1075 // Not blocking on a lock.
1078 // Blocking on a lock. Write down the timestamp.
1080 gp.trackingStamp = now
1082 // We just transitioned into runnable, so record what
1083 // time that happened.
1085 gp.trackingStamp = now
1087 // We're transitioning into running, so turn off
1088 // tracking and record how much time we spent in
1091 sched.timeToRun.record(gp.runnableTime)
1096 // casGToWaiting transitions gp from old to _Gwaiting, and sets the wait reason.
1098 // Use this over casgstatus when possible to ensure that a waitreason is set.
1099 func casGToWaiting(gp *g, old uint32, reason waitReason) {
1100 // Set the wait reason before calling casgstatus, because casgstatus will use it.
1101 gp.waitreason = reason
1102 casgstatus(gp, old, _Gwaiting)
1105 // casgstatus(gp, oldstatus, Gcopystack), assuming oldstatus is Gwaiting or Grunnable.
1106 // Returns old status. Cannot call casgstatus directly, because we are racing with an
1107 // async wakeup that might come in from netpoll. If we see Gwaiting from the readgstatus,
1108 // it might have become Grunnable by the time we get to the cas. If we called casgstatus,
1109 // it would loop waiting for the status to go back to Gwaiting, which it never will.
1112 func casgcopystack(gp *g) uint32 {
1114 oldstatus := readgstatus(gp) &^ _Gscan
1115 if oldstatus != _Gwaiting && oldstatus != _Grunnable {
1116 throw("copystack: bad status, not Gwaiting or Grunnable")
1118 if gp.atomicstatus.CompareAndSwap(oldstatus, _Gcopystack) {
1124 // casGToPreemptScan transitions gp from _Grunning to _Gscan|_Gpreempted.
1126 // TODO(austin): This is the only status operation that both changes
1127 // the status and locks the _Gscan bit. Rethink this.
1128 func casGToPreemptScan(gp *g, old, new uint32) {
1129 if old != _Grunning || new != _Gscan|_Gpreempted {
1130 throw("bad g transition")
1132 acquireLockRank(lockRankGscan)
1133 for !gp.atomicstatus.CompareAndSwap(_Grunning, _Gscan|_Gpreempted) {
1137 // casGFromPreempted attempts to transition gp from _Gpreempted to
1138 // _Gwaiting. If successful, the caller is responsible for
1139 // re-scheduling gp.
1140 func casGFromPreempted(gp *g, old, new uint32) bool {
1141 if old != _Gpreempted || new != _Gwaiting {
1142 throw("bad g transition")
1144 gp.waitreason = waitReasonPreempted
1145 return gp.atomicstatus.CompareAndSwap(_Gpreempted, _Gwaiting)
1148 // stopTheWorld stops all P's from executing goroutines, interrupting
1149 // all goroutines at GC safe points and records reason as the reason
1150 // for the stop. On return, only the current goroutine's P is running.
1151 // stopTheWorld must not be called from a system stack and the caller
1152 // must not hold worldsema. The caller must call startTheWorld when
1153 // other P's should resume execution.
1155 // stopTheWorld is safe for multiple goroutines to call at the
1156 // same time. Each will execute its own stop, and the stops will
1159 // This is also used by routines that do stack dumps. If the system is
1160 // in panic or being exited, this may not reliably stop all
1162 func stopTheWorld(reason string) {
1163 semacquire(&worldsema)
1165 gp.m.preemptoff = reason
1166 systemstack(func() {
1167 // Mark the goroutine which called stopTheWorld preemptible so its
1168 // stack may be scanned.
1169 // This lets a mark worker scan us while we try to stop the world
1170 // since otherwise we could get in a mutual preemption deadlock.
1171 // We must not modify anything on the G stack because a stack shrink
1172 // may occur. A stack shrink is otherwise OK though because in order
1173 // to return from this function (and to leave the system stack) we
1174 // must have preempted all goroutines, including any attempting
1175 // to scan our stack, in which case, any stack shrinking will
1176 // have already completed by the time we exit.
1177 // Don't provide a wait reason because we're still executing.
1178 casGToWaiting(gp, _Grunning, waitReasonStoppingTheWorld)
1179 stopTheWorldWithSema()
1180 casgstatus(gp, _Gwaiting, _Grunning)
1184 // startTheWorld undoes the effects of stopTheWorld.
1185 func startTheWorld() {
1186 systemstack(func() { startTheWorldWithSema(false) })
1188 // worldsema must be held over startTheWorldWithSema to ensure
1189 // gomaxprocs cannot change while worldsema is held.
1191 // Release worldsema with direct handoff to the next waiter, but
1192 // acquirem so that semrelease1 doesn't try to yield our time.
1194 // Otherwise if e.g. ReadMemStats is being called in a loop,
1195 // it might stomp on other attempts to stop the world, such as
1196 // for starting or ending GC. The operation this blocks is
1197 // so heavy-weight that we should just try to be as fair as
1200 // We don't want to just allow us to get preempted between now
1201 // and releasing the semaphore because then we keep everyone
1202 // (including, for example, GCs) waiting longer.
1205 semrelease1(&worldsema, true, 0)
1209 // stopTheWorldGC has the same effect as stopTheWorld, but blocks
1210 // until the GC is not running. It also blocks a GC from starting
1211 // until startTheWorldGC is called.
1212 func stopTheWorldGC(reason string) {
1214 stopTheWorld(reason)
1217 // startTheWorldGC undoes the effects of stopTheWorldGC.
1218 func startTheWorldGC() {
1223 // Holding worldsema grants an M the right to try to stop the world.
1224 var worldsema uint32 = 1
1226 // Holding gcsema grants the M the right to block a GC, and blocks
1227 // until the current GC is done. In particular, it prevents gomaxprocs
1228 // from changing concurrently.
1230 // TODO(mknyszek): Once gomaxprocs and the execution tracer can handle
1231 // being changed/enabled during a GC, remove this.
1232 var gcsema uint32 = 1
1234 // stopTheWorldWithSema is the core implementation of stopTheWorld.
1235 // The caller is responsible for acquiring worldsema and disabling
1236 // preemption first and then should stopTheWorldWithSema on the system
1239 // semacquire(&worldsema, 0)
1240 // m.preemptoff = "reason"
1241 // systemstack(stopTheWorldWithSema)
1243 // When finished, the caller must either call startTheWorld or undo
1244 // these three operations separately:
1246 // m.preemptoff = ""
1247 // systemstack(startTheWorldWithSema)
1248 // semrelease(&worldsema)
1250 // It is allowed to acquire worldsema once and then execute multiple
1251 // startTheWorldWithSema/stopTheWorldWithSema pairs.
1252 // Other P's are able to execute between successive calls to
1253 // startTheWorldWithSema and stopTheWorldWithSema.
1254 // Holding worldsema causes any other goroutines invoking
1255 // stopTheWorld to block.
1256 func stopTheWorldWithSema() {
1259 // If we hold a lock, then we won't be able to stop another M
1260 // that is blocked trying to acquire the lock.
1262 throw("stopTheWorld: holding locks")
1266 sched.stopwait = gomaxprocs
1267 sched.gcwaiting.Store(true)
1270 gp.m.p.ptr().status = _Pgcstop // Pgcstop is only diagnostic.
1272 // try to retake all P's in Psyscall status
1273 for _, pp := range allp {
1275 if s == _Psyscall && atomic.Cas(&pp.status, s, _Pgcstop) {
1287 pp, _ := pidleget(now)
1291 pp.status = _Pgcstop
1294 wait := sched.stopwait > 0
1297 // wait for remaining P's to stop voluntarily
1300 // wait for 100us, then try to re-preempt in case of any races
1301 if notetsleep(&sched.stopnote, 100*1000) {
1302 noteclear(&sched.stopnote)
1311 if sched.stopwait != 0 {
1312 bad = "stopTheWorld: not stopped (stopwait != 0)"
1314 for _, pp := range allp {
1315 if pp.status != _Pgcstop {
1316 bad = "stopTheWorld: not stopped (status != _Pgcstop)"
1320 if freezing.Load() {
1321 // Some other thread is panicking. This can cause the
1322 // sanity checks above to fail if the panic happens in
1323 // the signal handler on a stopped thread. Either way,
1324 // we should halt this thread.
1335 func startTheWorldWithSema(emitTraceEvent bool) int64 {
1336 assertWorldStopped()
1338 mp := acquirem() // disable preemption because it can be holding p in a local var
1339 if netpollinited() {
1340 list := netpoll(0) // non-blocking
1350 p1 := procresize(procs)
1351 sched.gcwaiting.Store(false)
1352 if sched.sysmonwait.Load() {
1353 sched.sysmonwait.Store(false)
1354 notewakeup(&sched.sysmonnote)
1367 throw("startTheWorld: inconsistent mp->nextp")
1370 notewakeup(&mp.park)
1372 // Start M to run P. Do not start another M below.
1377 // Capture start-the-world time before doing clean-up tasks.
1378 startTime := nanotime()
1383 // Wakeup an additional proc in case we have excessive runnable goroutines
1384 // in local queues or in the global queue. If we don't, the proc will park itself.
1385 // If we have lots of excessive work, resetspinning will unpark additional procs as necessary.
1393 // usesLibcall indicates whether this runtime performs system calls
1395 func usesLibcall() bool {
1397 case "aix", "darwin", "illumos", "ios", "solaris", "windows":
1400 return GOARCH == "386" || GOARCH == "amd64" || GOARCH == "arm" || GOARCH == "arm64"
1405 // mStackIsSystemAllocated indicates whether this runtime starts on a
1406 // system-allocated stack.
1407 func mStackIsSystemAllocated() bool {
1409 case "aix", "darwin", "plan9", "illumos", "ios", "solaris", "windows":
1413 case "386", "amd64", "arm", "arm64":
1420 // mstart is the entry-point for new Ms.
1421 // It is written in assembly, uses ABI0, is marked TOPFRAME, and calls mstart0.
1424 // mstart0 is the Go entry-point for new Ms.
1425 // This must not split the stack because we may not even have stack
1426 // bounds set up yet.
1428 // May run during STW (because it doesn't have a P yet), so write
1429 // barriers are not allowed.
1432 //go:nowritebarrierrec
1436 osStack := gp.stack.lo == 0
1438 // Initialize stack bounds from system stack.
1439 // Cgo may have left stack size in stack.hi.
1440 // minit may update the stack bounds.
1442 // Note: these bounds may not be very accurate.
1443 // We set hi to &size, but there are things above
1444 // it. The 1024 is supposed to compensate this,
1445 // but is somewhat arbitrary.
1448 size = 8192 * sys.StackGuardMultiplier
1450 gp.stack.hi = uintptr(noescape(unsafe.Pointer(&size)))
1451 gp.stack.lo = gp.stack.hi - size + 1024
1453 // Initialize stack guard so that we can start calling regular
1455 gp.stackguard0 = gp.stack.lo + _StackGuard
1456 // This is the g0, so we can also call go:systemstack
1457 // functions, which check stackguard1.
1458 gp.stackguard1 = gp.stackguard0
1461 // Exit this thread.
1462 if mStackIsSystemAllocated() {
1463 // Windows, Solaris, illumos, Darwin, AIX and Plan 9 always system-allocate
1464 // the stack, but put it in gp.stack before mstart,
1465 // so the logic above hasn't set osStack yet.
1471 // The go:noinline is to guarantee the getcallerpc/getcallersp below are safe,
1472 // so that we can set up g0.sched to return to the call of mstart1 above.
1479 throw("bad runtime·mstart")
1482 // Set up m.g0.sched as a label returning to just
1483 // after the mstart1 call in mstart0 above, for use by goexit0 and mcall.
1484 // We're never coming back to mstart1 after we call schedule,
1485 // so other calls can reuse the current frame.
1486 // And goexit0 does a gogo that needs to return from mstart1
1487 // and let mstart0 exit the thread.
1488 gp.sched.g = guintptr(unsafe.Pointer(gp))
1489 gp.sched.pc = getcallerpc()
1490 gp.sched.sp = getcallersp()
1495 // Install signal handlers; after minit so that minit can
1496 // prepare the thread to be able to handle the signals.
1501 if fn := gp.m.mstartfn; fn != nil {
1506 acquirep(gp.m.nextp.ptr())
1512 // mstartm0 implements part of mstart1 that only runs on the m0.
1514 // Write barriers are allowed here because we know the GC can't be
1515 // running yet, so they'll be no-ops.
1517 //go:yeswritebarrierrec
1519 // Create an extra M for callbacks on threads not created by Go.
1520 // An extra M is also needed on Windows for callbacks created by
1521 // syscall.NewCallback. See issue #6751 for details.
1522 if (iscgo || GOOS == "windows") && !cgoHasExtraM {
1529 // mPark causes a thread to park itself, returning once woken.
1534 notesleep(&gp.m.park)
1535 noteclear(&gp.m.park)
1538 // mexit tears down and exits the current thread.
1540 // Don't call this directly to exit the thread, since it must run at
1541 // the top of the thread stack. Instead, use gogo(&gp.m.g0.sched) to
1542 // unwind the stack to the point that exits the thread.
1544 // It is entered with m.p != nil, so write barriers are allowed. It
1545 // will release the P before exiting.
1547 //go:yeswritebarrierrec
1548 func mexit(osStack bool) {
1552 // This is the main thread. Just wedge it.
1554 // On Linux, exiting the main thread puts the process
1555 // into a non-waitable zombie state. On Plan 9,
1556 // exiting the main thread unblocks wait even though
1557 // other threads are still running. On Solaris we can
1558 // neither exitThread nor return from mstart. Other
1559 // bad things probably happen on other platforms.
1561 // We could try to clean up this M more before wedging
1562 // it, but that complicates signal handling.
1563 handoffp(releasep())
1569 throw("locked m0 woke up")
1575 // Free the gsignal stack.
1576 if mp.gsignal != nil {
1577 stackfree(mp.gsignal.stack)
1578 // On some platforms, when calling into VDSO (e.g. nanotime)
1579 // we store our g on the gsignal stack, if there is one.
1580 // Now the stack is freed, unlink it from the m, so we
1581 // won't write to it when calling VDSO code.
1585 // Remove m from allm.
1587 for pprev := &allm; *pprev != nil; pprev = &(*pprev).alllink {
1593 throw("m not found in allm")
1595 // Delay reaping m until it's done with the stack.
1597 // Put mp on the free list, though it will not be reaped while freeWait
1598 // is freeMWait. mp is no longer reachable via allm, so even if it is
1599 // on an OS stack, we must keep a reference to mp alive so that the GC
1600 // doesn't free mp while we are still using it.
1602 // Note that the free list must not be linked through alllink because
1603 // some functions walk allm without locking, so may be using alllink.
1604 mp.freeWait.Store(freeMWait)
1605 mp.freelink = sched.freem
1609 atomic.Xadd64(&ncgocall, int64(mp.ncgocall))
1612 handoffp(releasep())
1613 // After this point we must not have write barriers.
1615 // Invoke the deadlock detector. This must happen after
1616 // handoffp because it may have started a new M to take our
1623 if GOOS == "darwin" || GOOS == "ios" {
1624 // Make sure pendingPreemptSignals is correct when an M exits.
1626 if mp.signalPending.Load() != 0 {
1627 pendingPreemptSignals.Add(-1)
1631 // Destroy all allocated resources. After this is called, we may no
1632 // longer take any locks.
1636 // No more uses of mp, so it is safe to drop the reference.
1637 mp.freeWait.Store(freeMRef)
1639 // Return from mstart and let the system thread
1640 // library free the g0 stack and terminate the thread.
1644 // mstart is the thread's entry point, so there's nothing to
1645 // return to. Exit the thread directly. exitThread will clear
1646 // m.freeWait when it's done with the stack and the m can be
1648 exitThread(&mp.freeWait)
1651 // forEachP calls fn(p) for every P p when p reaches a GC safe point.
1652 // If a P is currently executing code, this will bring the P to a GC
1653 // safe point and execute fn on that P. If the P is not executing code
1654 // (it is idle or in a syscall), this will call fn(p) directly while
1655 // preventing the P from exiting its state. This does not ensure that
1656 // fn will run on every CPU executing Go code, but it acts as a global
1657 // memory barrier. GC uses this as a "ragged barrier."
1659 // The caller must hold worldsema.
1662 func forEachP(fn func(*p)) {
1664 pp := getg().m.p.ptr()
1667 if sched.safePointWait != 0 {
1668 throw("forEachP: sched.safePointWait != 0")
1670 sched.safePointWait = gomaxprocs - 1
1671 sched.safePointFn = fn
1673 // Ask all Ps to run the safe point function.
1674 for _, p2 := range allp {
1676 atomic.Store(&p2.runSafePointFn, 1)
1681 // Any P entering _Pidle or _Psyscall from now on will observe
1682 // p.runSafePointFn == 1 and will call runSafePointFn when
1683 // changing its status to _Pidle/_Psyscall.
1685 // Run safe point function for all idle Ps. sched.pidle will
1686 // not change because we hold sched.lock.
1687 for p := sched.pidle.ptr(); p != nil; p = p.link.ptr() {
1688 if atomic.Cas(&p.runSafePointFn, 1, 0) {
1690 sched.safePointWait--
1694 wait := sched.safePointWait > 0
1697 // Run fn for the current P.
1700 // Force Ps currently in _Psyscall into _Pidle and hand them
1701 // off to induce safe point function execution.
1702 for _, p2 := range allp {
1704 if s == _Psyscall && p2.runSafePointFn == 1 && atomic.Cas(&p2.status, s, _Pidle) {
1714 // Wait for remaining Ps to run fn.
1717 // Wait for 100us, then try to re-preempt in
1718 // case of any races.
1720 // Requires system stack.
1721 if notetsleep(&sched.safePointNote, 100*1000) {
1722 noteclear(&sched.safePointNote)
1728 if sched.safePointWait != 0 {
1729 throw("forEachP: not done")
1731 for _, p2 := range allp {
1732 if p2.runSafePointFn != 0 {
1733 throw("forEachP: P did not run fn")
1738 sched.safePointFn = nil
1743 // runSafePointFn runs the safe point function, if any, for this P.
1744 // This should be called like
1746 // if getg().m.p.runSafePointFn != 0 {
1750 // runSafePointFn must be checked on any transition in to _Pidle or
1751 // _Psyscall to avoid a race where forEachP sees that the P is running
1752 // just before the P goes into _Pidle/_Psyscall and neither forEachP
1753 // nor the P run the safe-point function.
1754 func runSafePointFn() {
1755 p := getg().m.p.ptr()
1756 // Resolve the race between forEachP running the safe-point
1757 // function on this P's behalf and this P running the
1758 // safe-point function directly.
1759 if !atomic.Cas(&p.runSafePointFn, 1, 0) {
1762 sched.safePointFn(p)
1764 sched.safePointWait--
1765 if sched.safePointWait == 0 {
1766 notewakeup(&sched.safePointNote)
1771 // When running with cgo, we call _cgo_thread_start
1772 // to start threads for us so that we can play nicely with
1774 var cgoThreadStart unsafe.Pointer
1776 type cgothreadstart struct {
1782 // Allocate a new m unassociated with any thread.
1783 // Can use p for allocation context if needed.
1784 // fn is recorded as the new m's m.mstartfn.
1785 // id is optional pre-allocated m ID. Omit by passing -1.
1787 // This function is allowed to have write barriers even if the caller
1788 // isn't because it borrows pp.
1790 //go:yeswritebarrierrec
1791 func allocm(pp *p, fn func(), id int64) *m {
1794 // The caller owns pp, but we may borrow (i.e., acquirep) it. We must
1795 // disable preemption to ensure it is not stolen, which would make the
1796 // caller lose ownership.
1801 acquirep(pp) // temporarily borrow p for mallocs in this function
1804 // Release the free M list. We need to do this somewhere and
1805 // this may free up a stack we can use.
1806 if sched.freem != nil {
1809 for freem := sched.freem; freem != nil; {
1810 wait := freem.freeWait.Load()
1811 if wait == freeMWait {
1812 next := freem.freelink
1813 freem.freelink = newList
1818 // Free the stack if needed. For freeMRef, there is
1819 // nothing to do except drop freem from the sched.freem
1821 if wait == freeMStack {
1822 // stackfree must be on the system stack, but allocm is
1823 // reachable off the system stack transitively from
1825 systemstack(func() {
1826 stackfree(freem.g0.stack)
1829 freem = freem.freelink
1831 sched.freem = newList
1839 // In case of cgo or Solaris or illumos or Darwin, pthread_create will make us a stack.
1840 // Windows and Plan 9 will layout sched stack on OS stack.
1841 if iscgo || mStackIsSystemAllocated() {
1844 mp.g0 = malg(8192 * sys.StackGuardMultiplier)
1848 if pp == gp.m.p.ptr() {
1853 allocmLock.runlock()
1857 // needm is called when a cgo callback happens on a
1858 // thread without an m (a thread not created by Go).
1859 // In this case, needm is expected to find an m to use
1860 // and return with m, g initialized correctly.
1861 // Since m and g are not set now (likely nil, but see below)
1862 // needm is limited in what routines it can call. In particular
1863 // it can only call nosplit functions (textflag 7) and cannot
1864 // do any scheduling that requires an m.
1866 // In order to avoid needing heavy lifting here, we adopt
1867 // the following strategy: there is a stack of available m's
1868 // that can be stolen. Using compare-and-swap
1869 // to pop from the stack has ABA races, so we simulate
1870 // a lock by doing an exchange (via Casuintptr) to steal the stack
1871 // head and replace the top pointer with MLOCKED (1).
1872 // This serves as a simple spin lock that we can use even
1873 // without an m. The thread that locks the stack in this way
1874 // unlocks the stack by storing a valid stack head pointer.
1876 // In order to make sure that there is always an m structure
1877 // available to be stolen, we maintain the invariant that there
1878 // is always one more than needed. At the beginning of the
1879 // program (if cgo is in use) the list is seeded with a single m.
1880 // If needm finds that it has taken the last m off the list, its job
1881 // is - once it has installed its own m so that it can do things like
1882 // allocate memory - to create a spare m and put it on the list.
1884 // Each of these extra m's also has a g0 and a curg that are
1885 // pressed into service as the scheduling stack and current
1886 // goroutine for the duration of the cgo callback.
1888 // It calls dropm to put the m back on the list,
1889 // 1. when the callback is done with the m in non-pthread platforms,
1890 // 2. or when the C thread exiting on pthread platforms.
1894 if (iscgo || GOOS == "windows") && !cgoHasExtraM {
1895 // Can happen if C/C++ code calls Go from a global ctor.
1896 // Can also happen on Windows if a global ctor uses a
1897 // callback created by syscall.NewCallback. See issue #6751
1900 // Can not throw, because scheduler is not initialized yet.
1901 writeErrStr("fatal error: cgo callback before cgo call\n")
1905 // Save and block signals before getting an M.
1906 // The signal handler may call needm itself,
1907 // and we must avoid a deadlock. Also, once g is installed,
1908 // any incoming signals will try to execute,
1909 // but we won't have the sigaltstack settings and other data
1910 // set up appropriately until the end of minit, which will
1911 // unblock the signals. This is the same dance as when
1912 // starting a new m to run Go code via newosproc.
1917 // Lock extra list, take head, unlock popped list.
1918 // nilokay=false is safe here because of the invariant above,
1919 // that the extra list always contains or will soon contain
1921 mp := lockextra(false)
1923 // Set needextram when we've just emptied the list,
1924 // so that the eventual call into cgocallbackg will
1925 // allocate a new m for the extra list. We delay the
1926 // allocation until then so that it can be done
1927 // after exitsyscall makes sure it is okay to be
1928 // running at all (that is, there's no garbage collection
1929 // running right now).
1930 mp.needextram = mp.schedlink == 0
1932 unlockextra(mp.schedlink.ptr())
1934 // Store the original signal mask for use by minit.
1935 mp.sigmask = sigmask
1937 // Install TLS on some platforms (previously setg
1938 // would do this if necessary).
1941 // Install g (= m->g0) and set the stack bounds
1942 // to match the current stack. We don't actually know
1943 // how big the stack is, like we don't know how big any
1944 // scheduling stack is, but we assume there's at least 32 kB,
1945 // which is more than enough for us.
1948 gp.stack.hi = getcallersp() + 1024
1949 gp.stack.lo = getcallersp() - 32*1024
1950 gp.stackguard0 = gp.stack.lo + _StackGuard
1952 // Should mark we are already in Go now.
1953 // Otherwise, we may call needm again when we get a signal, before cgocallbackg1,
1954 // which means the extram list may be empty, that will cause a deadlock.
1955 mp.isExtraInC = false
1957 // Initialize this thread to use the m.
1961 // mp.curg is now a real goroutine.
1962 casgstatus(mp.curg, _Gdead, _Gsyscall)
1966 // Acquire an extra m and bind it to the C thread when a pthread key has been created.
1969 func needAndBindM() {
1972 if _cgo_pthread_key_created != nil && *(*uintptr)(_cgo_pthread_key_created) != 0 {
1977 // newextram allocates m's and puts them on the extra list.
1978 // It is called with a working local m, so that it can do things
1979 // like call schedlock and allocate.
1981 c := extraMWaiters.Swap(0)
1983 for i := uint32(0); i < c; i++ {
1987 // Make sure there is at least one extra M.
1988 mp := lockextra(true)
1996 // oneNewExtraM allocates an m and puts it on the extra list.
1997 func oneNewExtraM() {
1998 // Create extra goroutine locked to extra m.
1999 // The goroutine is the context in which the cgo callback will run.
2000 // The sched.pc will never be returned to, but setting it to
2001 // goexit makes clear to the traceback routines where
2002 // the goroutine stack ends.
2003 mp := allocm(nil, nil, -1)
2005 gp.sched.pc = abi.FuncPCABI0(goexit) + sys.PCQuantum
2006 gp.sched.sp = gp.stack.hi
2007 gp.sched.sp -= 4 * goarch.PtrSize // extra space in case of reads slightly beyond frame
2009 gp.sched.g = guintptr(unsafe.Pointer(gp))
2010 gp.syscallpc = gp.sched.pc
2011 gp.syscallsp = gp.sched.sp
2012 gp.stktopsp = gp.sched.sp
2013 // malg returns status as _Gidle. Change to _Gdead before
2014 // adding to allg where GC can see it. We use _Gdead to hide
2015 // this from tracebacks and stack scans since it isn't a
2016 // "real" goroutine until needm grabs it.
2017 casgstatus(gp, _Gidle, _Gdead)
2021 // mark we are in C by default.
2022 mp.isExtraInC = true
2026 gp.goid = sched.goidgen.Add(1)
2027 gp.sysblocktraced = true
2029 gp.racectx = racegostart(abi.FuncPCABIInternal(newextram) + sys.PCQuantum)
2032 // Trigger two trace events for the locked g in the extra m,
2033 // since the next event of the g will be traceEvGoSysExit in exitsyscall,
2034 // while calling from C thread to Go.
2035 traceGoCreate(gp, 0) // no start pc
2037 traceEvent(traceEvGoInSyscall, -1, gp.goid)
2039 // put on allg for garbage collector
2042 // gp is now on the allg list, but we don't want it to be
2043 // counted by gcount. It would be more "proper" to increment
2044 // sched.ngfree, but that requires locking. Incrementing ngsys
2045 // has the same effect.
2048 // Add m to the extra list.
2049 mnext := lockextra(true)
2050 mp.schedlink.set(mnext)
2055 // dropm puts the current m back onto the extra list.
2057 // 1. On systems without pthreads, like Windows
2058 // dropm is called when a cgo callback has called needm but is now
2059 // done with the callback and returning back into the non-Go thread.
2061 // The main expense here is the call to signalstack to release the
2062 // m's signal stack, and then the call to needm on the next callback
2063 // from this thread. It is tempting to try to save the m for next time,
2064 // which would eliminate both these costs, but there might not be
2065 // a next time: the current thread (which Go does not control) might exit.
2066 // If we saved the m for that thread, there would be an m leak each time
2067 // such a thread exited. Instead, we acquire and release an m on each
2068 // call. These should typically not be scheduling operations, just a few
2069 // atomics, so the cost should be small.
2071 // 2. On systems with pthreads
2072 // dropm is called while a non-Go thread is exiting.
2073 // We allocate a pthread per-thread variable using pthread_key_create,
2074 // to register a thread-exit-time destructor.
2075 // And store the g into a thread-specific value associated with the pthread key,
2076 // when first return back to C.
2077 // So that the destructor would invoke dropm while the non-Go thread is exiting.
2078 // This is much faster since it avoids expensive signal-related syscalls.
2080 // NOTE: this always runs without a P, so, nowritebarrierrec required.
2082 //go:nowritebarrierrec
2084 // Clear m and g, and return m to the extra list.
2085 // After the call to setg we can only call nosplit functions
2086 // with no pointer manipulation.
2089 // Return mp.curg to dead state.
2090 casgstatus(mp.curg, _Gsyscall, _Gdead)
2091 mp.curg.preemptStop = false
2094 // Block signals before unminit.
2095 // Unminit unregisters the signal handling stack (but needs g on some systems).
2096 // Setg(nil) clears g, which is the signal handler's cue not to run Go handlers.
2097 // It's important not to try to handle a signal between those two steps.
2098 sigmask := mp.sigmask
2102 mnext := lockextra(true)
2104 mp.schedlink.set(mnext)
2108 // Commit the release of mp.
2111 msigrestore(sigmask)
2114 // bindm store the g0 of the current m into a thread-specific value.
2116 // We allocate a pthread per-thread variable using pthread_key_create,
2117 // to register a thread-exit-time destructor.
2118 // We are here setting the thread-specific value of the pthread key, to enable the destructor.
2119 // So that the pthread_key_destructor would dropm while the C thread is exiting.
2121 // And the saved g will be used in pthread_key_destructor,
2122 // since the g stored in the TLS by Go might be cleared in some platforms,
2123 // before the destructor invoked, so, we restore g by the stored g, before dropm.
2125 // We store g0 instead of m, to make the assembly code simpler,
2126 // since we need to restore g0 in runtime.cgocallback.
2128 // On systems without pthreads, like Windows, bindm shouldn't be used.
2130 // NOTE: this always runs without a P, so, nowritebarrierrec required.
2133 //go:nowritebarrierrec
2135 if GOOS == "windows" || GOOS == "plan9" {
2136 fatal("bindm in unexpected GOOS")
2140 fatal("the current g is not g0")
2142 if _cgo_bindm != nil {
2143 asmcgocall(_cgo_bindm, unsafe.Pointer(g))
2147 // A helper function for EnsureDropM.
2148 func getm() uintptr {
2149 return uintptr(unsafe.Pointer(getg().m))
2152 var extram atomic.Uintptr
2153 var extraMCount uint32 // Protected by lockextra
2154 var extraMWaiters atomic.Uint32
2156 // lockextra locks the extra list and returns the list head.
2157 // The caller must unlock the list by storing a new list head
2158 // to extram. If nilokay is true, then lockextra will
2159 // return a nil list head if that's what it finds. If nilokay is false,
2160 // lockextra will keep waiting until the list head is no longer nil.
2163 func lockextra(nilokay bool) *m {
2168 old := extram.Load()
2173 if old == 0 && !nilokay {
2175 // Add 1 to the number of threads
2176 // waiting for an M.
2177 // This is cleared by newextram.
2178 extraMWaiters.Add(1)
2184 if extram.CompareAndSwap(old, locked) {
2185 return (*m)(unsafe.Pointer(old))
2193 func unlockextra(mp *m) {
2194 extram.Store(uintptr(unsafe.Pointer(mp)))
2198 // allocmLock is locked for read when creating new Ms in allocm and their
2199 // addition to allm. Thus acquiring this lock for write blocks the
2200 // creation of new Ms.
2203 // execLock serializes exec and clone to avoid bugs or unspecified
2204 // behaviour around exec'ing while creating/destroying threads. See
2209 // These errors are reported (via writeErrStr) by some OS-specific
2210 // versions of newosproc and newosproc0.
2212 failthreadcreate = "runtime: failed to create new OS thread\n"
2213 failallocatestack = "runtime: failed to allocate stack for the new OS thread\n"
2216 // newmHandoff contains a list of m structures that need new OS threads.
2217 // This is used by newm in situations where newm itself can't safely
2218 // start an OS thread.
2219 var newmHandoff struct {
2222 // newm points to a list of M structures that need new OS
2223 // threads. The list is linked through m.schedlink.
2226 // waiting indicates that wake needs to be notified when an m
2227 // is put on the list.
2231 // haveTemplateThread indicates that the templateThread has
2232 // been started. This is not protected by lock. Use cas to set
2234 haveTemplateThread uint32
2237 // Create a new m. It will start off with a call to fn, or else the scheduler.
2238 // fn needs to be static and not a heap allocated closure.
2239 // May run with m.p==nil, so write barriers are not allowed.
2241 // id is optional pre-allocated m ID. Omit by passing -1.
2243 //go:nowritebarrierrec
2244 func newm(fn func(), pp *p, id int64) {
2245 // allocm adds a new M to allm, but they do not start until created by
2246 // the OS in newm1 or the template thread.
2248 // doAllThreadsSyscall requires that every M in allm will eventually
2249 // start and be signal-able, even with a STW.
2251 // Disable preemption here until we start the thread to ensure that
2252 // newm is not preempted between allocm and starting the new thread,
2253 // ensuring that anything added to allm is guaranteed to eventually
2257 mp := allocm(pp, fn, id)
2259 mp.sigmask = initSigmask
2260 if gp := getg(); gp != nil && gp.m != nil && (gp.m.lockedExt != 0 || gp.m.incgo) && GOOS != "plan9" {
2261 // We're on a locked M or a thread that may have been
2262 // started by C. The kernel state of this thread may
2263 // be strange (the user may have locked it for that
2264 // purpose). We don't want to clone that into another
2265 // thread. Instead, ask a known-good thread to create
2266 // the thread for us.
2268 // This is disabled on Plan 9. See golang.org/issue/22227.
2270 // TODO: This may be unnecessary on Windows, which
2271 // doesn't model thread creation off fork.
2272 lock(&newmHandoff.lock)
2273 if newmHandoff.haveTemplateThread == 0 {
2274 throw("on a locked thread with no template thread")
2276 mp.schedlink = newmHandoff.newm
2277 newmHandoff.newm.set(mp)
2278 if newmHandoff.waiting {
2279 newmHandoff.waiting = false
2280 notewakeup(&newmHandoff.wake)
2282 unlock(&newmHandoff.lock)
2283 // The M has not started yet, but the template thread does not
2284 // participate in STW, so it will always process queued Ms and
2285 // it is safe to releasem.
2295 var ts cgothreadstart
2296 if _cgo_thread_start == nil {
2297 throw("_cgo_thread_start missing")
2300 ts.tls = (*uint64)(unsafe.Pointer(&mp.tls[0]))
2301 ts.fn = unsafe.Pointer(abi.FuncPCABI0(mstart))
2303 msanwrite(unsafe.Pointer(&ts), unsafe.Sizeof(ts))
2306 asanwrite(unsafe.Pointer(&ts), unsafe.Sizeof(ts))
2308 execLock.rlock() // Prevent process clone.
2309 asmcgocall(_cgo_thread_start, unsafe.Pointer(&ts))
2313 execLock.rlock() // Prevent process clone.
2318 // startTemplateThread starts the template thread if it is not already
2321 // The calling thread must itself be in a known-good state.
2322 func startTemplateThread() {
2323 if GOARCH == "wasm" { // no threads on wasm yet
2327 // Disable preemption to guarantee that the template thread will be
2328 // created before a park once haveTemplateThread is set.
2330 if !atomic.Cas(&newmHandoff.haveTemplateThread, 0, 1) {
2334 newm(templateThread, nil, -1)
2338 // templateThread is a thread in a known-good state that exists solely
2339 // to start new threads in known-good states when the calling thread
2340 // may not be in a good state.
2342 // Many programs never need this, so templateThread is started lazily
2343 // when we first enter a state that might lead to running on a thread
2344 // in an unknown state.
2346 // templateThread runs on an M without a P, so it must not have write
2349 //go:nowritebarrierrec
2350 func templateThread() {
2357 lock(&newmHandoff.lock)
2358 for newmHandoff.newm != 0 {
2359 newm := newmHandoff.newm.ptr()
2360 newmHandoff.newm = 0
2361 unlock(&newmHandoff.lock)
2363 next := newm.schedlink.ptr()
2368 lock(&newmHandoff.lock)
2370 newmHandoff.waiting = true
2371 noteclear(&newmHandoff.wake)
2372 unlock(&newmHandoff.lock)
2373 notesleep(&newmHandoff.wake)
2377 // Stops execution of the current m until new work is available.
2378 // Returns with acquired P.
2382 if gp.m.locks != 0 {
2383 throw("stopm holding locks")
2386 throw("stopm holding p")
2389 throw("stopm spinning")
2396 acquirep(gp.m.nextp.ptr())
2401 // startm's caller incremented nmspinning. Set the new M's spinning.
2402 getg().m.spinning = true
2405 // Schedules some M to run the p (creates an M if necessary).
2406 // If p==nil, tries to get an idle P, if no idle P's does nothing.
2407 // May run with m.p==nil, so write barriers are not allowed.
2408 // If spinning is set, the caller has incremented nmspinning and must provide a
2409 // P. startm will set m.spinning in the newly started M.
2411 // Callers passing a non-nil P must call from a non-preemptible context. See
2412 // comment on acquirem below.
2414 // Must not have write barriers because this may be called without a P.
2416 //go:nowritebarrierrec
2417 func startm(pp *p, spinning bool) {
2418 // Disable preemption.
2420 // Every owned P must have an owner that will eventually stop it in the
2421 // event of a GC stop request. startm takes transient ownership of a P
2422 // (either from argument or pidleget below) and transfers ownership to
2423 // a started M, which will be responsible for performing the stop.
2425 // Preemption must be disabled during this transient ownership,
2426 // otherwise the P this is running on may enter GC stop while still
2427 // holding the transient P, leaving that P in limbo and deadlocking the
2430 // Callers passing a non-nil P must already be in non-preemptible
2431 // context, otherwise such preemption could occur on function entry to
2432 // startm. Callers passing a nil P may be preemptible, so we must
2433 // disable preemption before acquiring a P from pidleget below.
2438 // TODO(prattmic): All remaining calls to this function
2439 // with _p_ == nil could be cleaned up to find a P
2440 // before calling startm.
2441 throw("startm: P required for spinning=true")
2452 // No M is available, we must drop sched.lock and call newm.
2453 // However, we already own a P to assign to the M.
2455 // Once sched.lock is released, another G (e.g., in a syscall),
2456 // could find no idle P while checkdead finds a runnable G but
2457 // no running M's because this new M hasn't started yet, thus
2458 // throwing in an apparent deadlock.
2460 // Avoid this situation by pre-allocating the ID for the new M,
2461 // thus marking it as 'running' before we drop sched.lock. This
2462 // new M will eventually run the scheduler to execute any
2469 // The caller incremented nmspinning, so set m.spinning in the new M.
2473 // Ownership transfer of pp committed by start in newm.
2474 // Preemption is now safe.
2480 throw("startm: m is spinning")
2483 throw("startm: m has p")
2485 if spinning && !runqempty(pp) {
2486 throw("startm: p has runnable gs")
2488 // The caller incremented nmspinning, so set m.spinning in the new M.
2489 nmp.spinning = spinning
2491 notewakeup(&nmp.park)
2492 // Ownership transfer of pp committed by wakeup. Preemption is now
2497 // Hands off P from syscall or locked M.
2498 // Always runs without a P, so write barriers are not allowed.
2500 //go:nowritebarrierrec
2501 func handoffp(pp *p) {
2502 // handoffp must start an M in any situation where
2503 // findrunnable would return a G to run on pp.
2505 // if it has local work, start it straight away
2506 if !runqempty(pp) || sched.runqsize != 0 {
2510 // if there's trace work to do, start it straight away
2511 if (trace.enabled || trace.shutdown) && traceReaderAvailable() != nil {
2515 // if it has GC work, start it straight away
2516 if gcBlackenEnabled != 0 && gcMarkWorkAvailable(pp) {
2520 // no local work, check that there are no spinning/idle M's,
2521 // otherwise our help is not required
2522 if sched.nmspinning.Load()+sched.npidle.Load() == 0 && sched.nmspinning.CompareAndSwap(0, 1) { // TODO: fast atomic
2523 sched.needspinning.Store(0)
2528 if sched.gcwaiting.Load() {
2529 pp.status = _Pgcstop
2531 if sched.stopwait == 0 {
2532 notewakeup(&sched.stopnote)
2537 if pp.runSafePointFn != 0 && atomic.Cas(&pp.runSafePointFn, 1, 0) {
2538 sched.safePointFn(pp)
2539 sched.safePointWait--
2540 if sched.safePointWait == 0 {
2541 notewakeup(&sched.safePointNote)
2544 if sched.runqsize != 0 {
2549 // If this is the last running P and nobody is polling network,
2550 // need to wakeup another M to poll network.
2551 if sched.npidle.Load() == gomaxprocs-1 && sched.lastpoll.Load() != 0 {
2557 // The scheduler lock cannot be held when calling wakeNetPoller below
2558 // because wakeNetPoller may call wakep which may call startm.
2559 when := nobarrierWakeTime(pp)
2568 // Tries to add one more P to execute G's.
2569 // Called when a G is made runnable (newproc, ready).
2570 // Must be called with a P.
2572 // Be conservative about spinning threads, only start one if none exist
2574 if sched.nmspinning.Load() != 0 || !sched.nmspinning.CompareAndSwap(0, 1) {
2578 // Disable preemption until ownership of pp transfers to the next M in
2579 // startm. Otherwise preemption here would leave pp stuck waiting to
2582 // See preemption comment on acquirem in startm for more details.
2587 pp, _ = pidlegetSpinning(0)
2589 if sched.nmspinning.Add(-1) < 0 {
2590 throw("wakep: negative nmspinning")
2596 // Since we always have a P, the race in the "No M is available"
2597 // comment in startm doesn't apply during the small window between the
2598 // unlock here and lock in startm. A checkdead in between will always
2599 // see at least one running M (ours).
2607 // Stops execution of the current m that is locked to a g until the g is runnable again.
2608 // Returns with acquired P.
2609 func stoplockedm() {
2612 if gp.m.lockedg == 0 || gp.m.lockedg.ptr().lockedm.ptr() != gp.m {
2613 throw("stoplockedm: inconsistent locking")
2616 // Schedule another M to run this p.
2621 // Wait until another thread schedules lockedg again.
2623 status := readgstatus(gp.m.lockedg.ptr())
2624 if status&^_Gscan != _Grunnable {
2625 print("runtime:stoplockedm: lockedg (atomicstatus=", status, ") is not Grunnable or Gscanrunnable\n")
2626 dumpgstatus(gp.m.lockedg.ptr())
2627 throw("stoplockedm: not runnable")
2629 acquirep(gp.m.nextp.ptr())
2633 // Schedules the locked m to run the locked gp.
2634 // May run during STW, so write barriers are not allowed.
2636 //go:nowritebarrierrec
2637 func startlockedm(gp *g) {
2638 mp := gp.lockedm.ptr()
2640 throw("startlockedm: locked to me")
2643 throw("startlockedm: m has p")
2645 // directly handoff current P to the locked m
2649 notewakeup(&mp.park)
2653 // Stops the current m for stopTheWorld.
2654 // Returns when the world is restarted.
2658 if !sched.gcwaiting.Load() {
2659 throw("gcstopm: not waiting for gc")
2662 gp.m.spinning = false
2663 // OK to just drop nmspinning here,
2664 // startTheWorld will unpark threads as necessary.
2665 if sched.nmspinning.Add(-1) < 0 {
2666 throw("gcstopm: negative nmspinning")
2671 pp.status = _Pgcstop
2673 if sched.stopwait == 0 {
2674 notewakeup(&sched.stopnote)
2680 // Schedules gp to run on the current M.
2681 // If inheritTime is true, gp inherits the remaining time in the
2682 // current time slice. Otherwise, it starts a new time slice.
2685 // Write barriers are allowed because this is called immediately after
2686 // acquiring a P in several places.
2688 //go:yeswritebarrierrec
2689 func execute(gp *g, inheritTime bool) {
2692 if goroutineProfile.active {
2693 // Make sure that gp has had its stack written out to the goroutine
2694 // profile, exactly as it was when the goroutine profiler first stopped
2696 tryRecordGoroutineProfile(gp, osyield)
2699 // Assign gp.m before entering _Grunning so running Gs have an
2703 casgstatus(gp, _Grunnable, _Grunning)
2706 gp.stackguard0 = gp.stack.lo + _StackGuard
2708 mp.p.ptr().schedtick++
2711 // Check whether the profiler needs to be turned on or off.
2712 hz := sched.profilehz
2713 if mp.profilehz != hz {
2714 setThreadCPUProfiler(hz)
2718 // GoSysExit has to happen when we have a P, but before GoStart.
2719 // So we emit it here.
2720 if gp.syscallsp != 0 && gp.sysblocktraced {
2721 traceGoSysExit(gp.sysexitticks)
2729 // Finds a runnable goroutine to execute.
2730 // Tries to steal from other P's, get g from local or global queue, poll network.
2731 // tryWakeP indicates that the returned goroutine is not normal (GC worker, trace
2732 // reader) so the caller should try to wake a P.
2733 func findRunnable() (gp *g, inheritTime, tryWakeP bool) {
2736 // The conditions here and in handoffp must agree: if
2737 // findrunnable would return a G to run, handoffp must start
2742 if sched.gcwaiting.Load() {
2746 if pp.runSafePointFn != 0 {
2750 // now and pollUntil are saved for work stealing later,
2751 // which may steal timers. It's important that between now
2752 // and then, nothing blocks, so these numbers remain mostly
2754 now, pollUntil, _ := checkTimers(pp, 0)
2756 // Try to schedule the trace reader.
2757 if trace.enabled || trace.shutdown {
2760 casgstatus(gp, _Gwaiting, _Grunnable)
2761 traceGoUnpark(gp, 0)
2762 return gp, false, true
2766 // Try to schedule a GC worker.
2767 if gcBlackenEnabled != 0 {
2768 gp, tnow := gcController.findRunnableGCWorker(pp, now)
2770 return gp, false, true
2775 // Check the global runnable queue once in a while to ensure fairness.
2776 // Otherwise two goroutines can completely occupy the local runqueue
2777 // by constantly respawning each other.
2778 if pp.schedtick%61 == 0 && sched.runqsize > 0 {
2780 gp := globrunqget(pp, 1)
2783 return gp, false, false
2787 // Wake up the finalizer G.
2788 if fingStatus.Load()&(fingWait|fingWake) == fingWait|fingWake {
2789 if gp := wakefing(); gp != nil {
2793 if *cgo_yield != nil {
2794 asmcgocall(*cgo_yield, nil)
2798 if gp, inheritTime := runqget(pp); gp != nil {
2799 return gp, inheritTime, false
2803 if sched.runqsize != 0 {
2805 gp := globrunqget(pp, 0)
2808 return gp, false, false
2813 // This netpoll is only an optimization before we resort to stealing.
2814 // We can safely skip it if there are no waiters or a thread is blocked
2815 // in netpoll already. If there is any kind of logical race with that
2816 // blocked thread (e.g. it has already returned from netpoll, but does
2817 // not set lastpoll yet), this thread will do blocking netpoll below
2819 if netpollinited() && netpollWaiters.Load() > 0 && sched.lastpoll.Load() != 0 {
2820 if list := netpoll(0); !list.empty() { // non-blocking
2823 casgstatus(gp, _Gwaiting, _Grunnable)
2825 traceGoUnpark(gp, 0)
2827 return gp, false, false
2831 // Spinning Ms: steal work from other Ps.
2833 // Limit the number of spinning Ms to half the number of busy Ps.
2834 // This is necessary to prevent excessive CPU consumption when
2835 // GOMAXPROCS>>1 but the program parallelism is low.
2836 if mp.spinning || 2*sched.nmspinning.Load() < gomaxprocs-sched.npidle.Load() {
2841 gp, inheritTime, tnow, w, newWork := stealWork(now)
2843 // Successfully stole.
2844 return gp, inheritTime, false
2847 // There may be new timer or GC work; restart to
2853 if w != 0 && (pollUntil == 0 || w < pollUntil) {
2854 // Earlier timer to wait for.
2859 // We have nothing to do.
2861 // If we're in the GC mark phase, can safely scan and blacken objects,
2862 // and have work to do, run idle-time marking rather than give up the P.
2863 if gcBlackenEnabled != 0 && gcMarkWorkAvailable(pp) && gcController.addIdleMarkWorker() {
2864 node := (*gcBgMarkWorkerNode)(gcBgMarkWorkerPool.pop())
2866 pp.gcMarkWorkerMode = gcMarkWorkerIdleMode
2868 casgstatus(gp, _Gwaiting, _Grunnable)
2870 traceGoUnpark(gp, 0)
2872 return gp, false, false
2874 gcController.removeIdleMarkWorker()
2878 // If a callback returned and no other goroutine is awake,
2879 // then wake event handler goroutine which pauses execution
2880 // until a callback was triggered.
2881 gp, otherReady := beforeIdle(now, pollUntil)
2883 casgstatus(gp, _Gwaiting, _Grunnable)
2885 traceGoUnpark(gp, 0)
2887 return gp, false, false
2893 // Before we drop our P, make a snapshot of the allp slice,
2894 // which can change underfoot once we no longer block
2895 // safe-points. We don't need to snapshot the contents because
2896 // everything up to cap(allp) is immutable.
2897 allpSnapshot := allp
2898 // Also snapshot masks. Value changes are OK, but we can't allow
2899 // len to change out from under us.
2900 idlepMaskSnapshot := idlepMask
2901 timerpMaskSnapshot := timerpMask
2903 // return P and block
2905 if sched.gcwaiting.Load() || pp.runSafePointFn != 0 {
2909 if sched.runqsize != 0 {
2910 gp := globrunqget(pp, 0)
2912 return gp, false, false
2914 if !mp.spinning && sched.needspinning.Load() == 1 {
2915 // See "Delicate dance" comment below.
2920 if releasep() != pp {
2921 throw("findrunnable: wrong p")
2923 now = pidleput(pp, now)
2926 // Delicate dance: thread transitions from spinning to non-spinning
2927 // state, potentially concurrently with submission of new work. We must
2928 // drop nmspinning first and then check all sources again (with
2929 // #StoreLoad memory barrier in between). If we do it the other way
2930 // around, another thread can submit work after we've checked all
2931 // sources but before we drop nmspinning; as a result nobody will
2932 // unpark a thread to run the work.
2934 // This applies to the following sources of work:
2936 // * Goroutines added to a per-P run queue.
2937 // * New/modified-earlier timers on a per-P timer heap.
2938 // * Idle-priority GC work (barring golang.org/issue/19112).
2940 // If we discover new work below, we need to restore m.spinning as a
2941 // signal for resetspinning to unpark a new worker thread (because
2942 // there can be more than one starving goroutine).
2944 // However, if after discovering new work we also observe no idle Ps
2945 // (either here or in resetspinning), we have a problem. We may be
2946 // racing with a non-spinning M in the block above, having found no
2947 // work and preparing to release its P and park. Allowing that P to go
2948 // idle will result in loss of work conservation (idle P while there is
2949 // runnable work). This could result in complete deadlock in the
2950 // unlikely event that we discover new work (from netpoll) right as we
2951 // are racing with _all_ other Ps going idle.
2953 // We use sched.needspinning to synchronize with non-spinning Ms going
2954 // idle. If needspinning is set when they are about to drop their P,
2955 // they abort the drop and instead become a new spinning M on our
2956 // behalf. If we are not racing and the system is truly fully loaded
2957 // then no spinning threads are required, and the next thread to
2958 // naturally become spinning will clear the flag.
2960 // Also see "Worker thread parking/unparking" comment at the top of the
2962 wasSpinning := mp.spinning
2965 if sched.nmspinning.Add(-1) < 0 {
2966 throw("findrunnable: negative nmspinning")
2969 // Note the for correctness, only the last M transitioning from
2970 // spinning to non-spinning must perform these rechecks to
2971 // ensure no missed work. However, the runtime has some cases
2972 // of transient increments of nmspinning that are decremented
2973 // without going through this path, so we must be conservative
2974 // and perform the check on all spinning Ms.
2976 // See https://go.dev/issue/43997.
2978 // Check all runqueues once again.
2979 pp := checkRunqsNoP(allpSnapshot, idlepMaskSnapshot)
2986 // Check for idle-priority GC work again.
2987 pp, gp := checkIdleGCNoP()
2992 // Run the idle worker.
2993 pp.gcMarkWorkerMode = gcMarkWorkerIdleMode
2994 casgstatus(gp, _Gwaiting, _Grunnable)
2996 traceGoUnpark(gp, 0)
2998 return gp, false, false
3001 // Finally, check for timer creation or expiry concurrently with
3002 // transitioning from spinning to non-spinning.
3004 // Note that we cannot use checkTimers here because it calls
3005 // adjusttimers which may need to allocate memory, and that isn't
3006 // allowed when we don't have an active P.
3007 pollUntil = checkTimersNoP(allpSnapshot, timerpMaskSnapshot, pollUntil)
3010 // Poll network until next timer.
3011 if netpollinited() && (netpollWaiters.Load() > 0 || pollUntil != 0) && sched.lastpoll.Swap(0) != 0 {
3012 sched.pollUntil.Store(pollUntil)
3014 throw("findrunnable: netpoll with p")
3017 throw("findrunnable: netpoll with spinning")
3023 delay = pollUntil - now
3029 // When using fake time, just poll.
3032 list := netpoll(delay) // block until new work is available
3033 sched.pollUntil.Store(0)
3034 sched.lastpoll.Store(now)
3035 if faketime != 0 && list.empty() {
3036 // Using fake time and nothing is ready; stop M.
3037 // When all M's stop, checkdead will call timejump.
3042 pp, _ := pidleget(now)
3051 casgstatus(gp, _Gwaiting, _Grunnable)
3053 traceGoUnpark(gp, 0)
3055 return gp, false, false
3062 } else if pollUntil != 0 && netpollinited() {
3063 pollerPollUntil := sched.pollUntil.Load()
3064 if pollerPollUntil == 0 || pollerPollUntil > pollUntil {
3072 // pollWork reports whether there is non-background work this P could
3073 // be doing. This is a fairly lightweight check to be used for
3074 // background work loops, like idle GC. It checks a subset of the
3075 // conditions checked by the actual scheduler.
3076 func pollWork() bool {
3077 if sched.runqsize != 0 {
3080 p := getg().m.p.ptr()
3084 if netpollinited() && netpollWaiters.Load() > 0 && sched.lastpoll.Load() != 0 {
3085 if list := netpoll(0); !list.empty() {
3093 // stealWork attempts to steal a runnable goroutine or timer from any P.
3095 // If newWork is true, new work may have been readied.
3097 // If now is not 0 it is the current time. stealWork returns the passed time or
3098 // the current time if now was passed as 0.
3099 func stealWork(now int64) (gp *g, inheritTime bool, rnow, pollUntil int64, newWork bool) {
3100 pp := getg().m.p.ptr()
3104 const stealTries = 4
3105 for i := 0; i < stealTries; i++ {
3106 stealTimersOrRunNextG := i == stealTries-1
3108 for enum := stealOrder.start(fastrand()); !enum.done(); enum.next() {
3109 if sched.gcwaiting.Load() {
3110 // GC work may be available.
3111 return nil, false, now, pollUntil, true
3113 p2 := allp[enum.position()]
3118 // Steal timers from p2. This call to checkTimers is the only place
3119 // where we might hold a lock on a different P's timers. We do this
3120 // once on the last pass before checking runnext because stealing
3121 // from the other P's runnext should be the last resort, so if there
3122 // are timers to steal do that first.
3124 // We only check timers on one of the stealing iterations because
3125 // the time stored in now doesn't change in this loop and checking
3126 // the timers for each P more than once with the same value of now
3127 // is probably a waste of time.
3129 // timerpMask tells us whether the P may have timers at all. If it
3130 // can't, no need to check at all.
3131 if stealTimersOrRunNextG && timerpMask.read(enum.position()) {
3132 tnow, w, ran := checkTimers(p2, now)
3134 if w != 0 && (pollUntil == 0 || w < pollUntil) {
3138 // Running the timers may have
3139 // made an arbitrary number of G's
3140 // ready and added them to this P's
3141 // local run queue. That invalidates
3142 // the assumption of runqsteal
3143 // that it always has room to add
3144 // stolen G's. So check now if there
3145 // is a local G to run.
3146 if gp, inheritTime := runqget(pp); gp != nil {
3147 return gp, inheritTime, now, pollUntil, ranTimer
3153 // Don't bother to attempt to steal if p2 is idle.
3154 if !idlepMask.read(enum.position()) {
3155 if gp := runqsteal(pp, p2, stealTimersOrRunNextG); gp != nil {
3156 return gp, false, now, pollUntil, ranTimer
3162 // No goroutines found to steal. Regardless, running a timer may have
3163 // made some goroutine ready that we missed. Indicate the next timer to
3165 return nil, false, now, pollUntil, ranTimer
3168 // Check all Ps for a runnable G to steal.
3170 // On entry we have no P. If a G is available to steal and a P is available,
3171 // the P is returned which the caller should acquire and attempt to steal the
3173 func checkRunqsNoP(allpSnapshot []*p, idlepMaskSnapshot pMask) *p {
3174 for id, p2 := range allpSnapshot {
3175 if !idlepMaskSnapshot.read(uint32(id)) && !runqempty(p2) {
3177 pp, _ := pidlegetSpinning(0)
3179 // Can't get a P, don't bother checking remaining Ps.
3188 // No work available.
3192 // Check all Ps for a timer expiring sooner than pollUntil.
3194 // Returns updated pollUntil value.
3195 func checkTimersNoP(allpSnapshot []*p, timerpMaskSnapshot pMask, pollUntil int64) int64 {
3196 for id, p2 := range allpSnapshot {
3197 if timerpMaskSnapshot.read(uint32(id)) {
3198 w := nobarrierWakeTime(p2)
3199 if w != 0 && (pollUntil == 0 || w < pollUntil) {
3208 // Check for idle-priority GC, without a P on entry.
3210 // If some GC work, a P, and a worker G are all available, the P and G will be
3211 // returned. The returned P has not been wired yet.
3212 func checkIdleGCNoP() (*p, *g) {
3213 // N.B. Since we have no P, gcBlackenEnabled may change at any time; we
3214 // must check again after acquiring a P. As an optimization, we also check
3215 // if an idle mark worker is needed at all. This is OK here, because if we
3216 // observe that one isn't needed, at least one is currently running. Even if
3217 // it stops running, its own journey into the scheduler should schedule it
3218 // again, if need be (at which point, this check will pass, if relevant).
3219 if atomic.Load(&gcBlackenEnabled) == 0 || !gcController.needIdleMarkWorker() {
3222 if !gcMarkWorkAvailable(nil) {
3226 // Work is available; we can start an idle GC worker only if there is
3227 // an available P and available worker G.
3229 // We can attempt to acquire these in either order, though both have
3230 // synchronization concerns (see below). Workers are almost always
3231 // available (see comment in findRunnableGCWorker for the one case
3232 // there may be none). Since we're slightly less likely to find a P,
3233 // check for that first.
3235 // Synchronization: note that we must hold sched.lock until we are
3236 // committed to keeping it. Otherwise we cannot put the unnecessary P
3237 // back in sched.pidle without performing the full set of idle
3238 // transition checks.
3240 // If we were to check gcBgMarkWorkerPool first, we must somehow handle
3241 // the assumption in gcControllerState.findRunnableGCWorker that an
3242 // empty gcBgMarkWorkerPool is only possible if gcMarkDone is running.
3244 pp, now := pidlegetSpinning(0)
3250 // Now that we own a P, gcBlackenEnabled can't change (as it requires STW).
3251 if gcBlackenEnabled == 0 || !gcController.addIdleMarkWorker() {
3257 node := (*gcBgMarkWorkerNode)(gcBgMarkWorkerPool.pop())
3261 gcController.removeIdleMarkWorker()
3267 return pp, node.gp.ptr()
3270 // wakeNetPoller wakes up the thread sleeping in the network poller if it isn't
3271 // going to wake up before the when argument; or it wakes an idle P to service
3272 // timers and the network poller if there isn't one already.
3273 func wakeNetPoller(when int64) {
3274 if sched.lastpoll.Load() == 0 {
3275 // In findrunnable we ensure that when polling the pollUntil
3276 // field is either zero or the time to which the current
3277 // poll is expected to run. This can have a spurious wakeup
3278 // but should never miss a wakeup.
3279 pollerPollUntil := sched.pollUntil.Load()
3280 if pollerPollUntil == 0 || pollerPollUntil > when {
3284 // There are no threads in the network poller, try to get
3285 // one there so it can handle new timers.
3286 if GOOS != "plan9" { // Temporary workaround - see issue #42303.
3292 func resetspinning() {
3295 throw("resetspinning: not a spinning m")
3297 gp.m.spinning = false
3298 nmspinning := sched.nmspinning.Add(-1)
3300 throw("findrunnable: negative nmspinning")
3302 // M wakeup policy is deliberately somewhat conservative, so check if we
3303 // need to wakeup another P here. See "Worker thread parking/unparking"
3304 // comment at the top of the file for details.
3308 // injectglist adds each runnable G on the list to some run queue,
3309 // and clears glist. If there is no current P, they are added to the
3310 // global queue, and up to npidle M's are started to run them.
3311 // Otherwise, for each idle P, this adds a G to the global queue
3312 // and starts an M. Any remaining G's are added to the current P's
3314 // This may temporarily acquire sched.lock.
3315 // Can run concurrently with GC.
3316 func injectglist(glist *gList) {
3321 for gp := glist.head.ptr(); gp != nil; gp = gp.schedlink.ptr() {
3322 traceGoUnpark(gp, 0)
3326 // Mark all the goroutines as runnable before we put them
3327 // on the run queues.
3328 head := glist.head.ptr()
3331 for gp := head; gp != nil; gp = gp.schedlink.ptr() {
3334 casgstatus(gp, _Gwaiting, _Grunnable)
3337 // Turn the gList into a gQueue.
3343 startIdle := func(n int) {
3344 for i := 0; i < n; i++ {
3345 mp := acquirem() // See comment in startm.
3348 pp, _ := pidlegetSpinning(0)
3361 pp := getg().m.p.ptr()
3364 globrunqputbatch(&q, int32(qsize))
3370 npidle := int(sched.npidle.Load())
3373 for n = 0; n < npidle && !q.empty(); n++ {
3379 globrunqputbatch(&globq, int32(n))
3386 runqputbatch(pp, &q, qsize)
3390 // One round of scheduler: find a runnable goroutine and execute it.
3396 throw("schedule: holding locks")
3399 if mp.lockedg != 0 {
3401 execute(mp.lockedg.ptr(), false) // Never returns.
3404 // We should not schedule away from a g that is executing a cgo call,
3405 // since the cgo call is using the m's g0 stack.
3407 throw("schedule: in cgo")
3414 // Safety check: if we are spinning, the run queue should be empty.
3415 // Check this before calling checkTimers, as that might call
3416 // goready to put a ready goroutine on the local run queue.
3417 if mp.spinning && (pp.runnext != 0 || pp.runqhead != pp.runqtail) {
3418 throw("schedule: spinning with local work")
3421 gp, inheritTime, tryWakeP := findRunnable() // blocks until work is available
3423 // This thread is going to run a goroutine and is not spinning anymore,
3424 // so if it was marked as spinning we need to reset it now and potentially
3425 // start a new spinning M.
3430 if sched.disable.user && !schedEnabled(gp) {
3431 // Scheduling of this goroutine is disabled. Put it on
3432 // the list of pending runnable goroutines for when we
3433 // re-enable user scheduling and look again.
3435 if schedEnabled(gp) {
3436 // Something re-enabled scheduling while we
3437 // were acquiring the lock.
3440 sched.disable.runnable.pushBack(gp)
3447 // If about to schedule a not-normal goroutine (a GCworker or tracereader),
3448 // wake a P if there is one.
3452 if gp.lockedm != 0 {
3453 // Hands off own p to the locked m,
3454 // then blocks waiting for a new p.
3459 execute(gp, inheritTime)
3462 // dropg removes the association between m and the current goroutine m->curg (gp for short).
3463 // Typically a caller sets gp's status away from Grunning and then
3464 // immediately calls dropg to finish the job. The caller is also responsible
3465 // for arranging that gp will be restarted using ready at an
3466 // appropriate time. After calling dropg and arranging for gp to be
3467 // readied later, the caller can do other work but eventually should
3468 // call schedule to restart the scheduling of goroutines on this m.
3472 setMNoWB(&gp.m.curg.m, nil)
3473 setGNoWB(&gp.m.curg, nil)
3476 // checkTimers runs any timers for the P that are ready.
3477 // If now is not 0 it is the current time.
3478 // It returns the passed time or the current time if now was passed as 0.
3479 // and the time when the next timer should run or 0 if there is no next timer,
3480 // and reports whether it ran any timers.
3481 // If the time when the next timer should run is not 0,
3482 // it is always larger than the returned time.
3483 // We pass now in and out to avoid extra calls of nanotime.
3485 //go:yeswritebarrierrec
3486 func checkTimers(pp *p, now int64) (rnow, pollUntil int64, ran bool) {
3487 // If it's not yet time for the first timer, or the first adjusted
3488 // timer, then there is nothing to do.
3489 next := pp.timer0When.Load()
3490 nextAdj := pp.timerModifiedEarliest.Load()
3491 if next == 0 || (nextAdj != 0 && nextAdj < next) {
3496 // No timers to run or adjust.
3497 return now, 0, false
3504 // Next timer is not ready to run, but keep going
3505 // if we would clear deleted timers.
3506 // This corresponds to the condition below where
3507 // we decide whether to call clearDeletedTimers.
3508 if pp != getg().m.p.ptr() || int(pp.deletedTimers.Load()) <= int(pp.numTimers.Load()/4) {
3509 return now, next, false
3513 lock(&pp.timersLock)
3515 if len(pp.timers) > 0 {
3516 adjusttimers(pp, now)
3517 for len(pp.timers) > 0 {
3518 // Note that runtimer may temporarily unlock
3520 if tw := runtimer(pp, now); tw != 0 {
3530 // If this is the local P, and there are a lot of deleted timers,
3531 // clear them out. We only do this for the local P to reduce
3532 // lock contention on timersLock.
3533 if pp == getg().m.p.ptr() && int(pp.deletedTimers.Load()) > len(pp.timers)/4 {
3534 clearDeletedTimers(pp)
3537 unlock(&pp.timersLock)
3539 return now, pollUntil, ran
3542 func parkunlock_c(gp *g, lock unsafe.Pointer) bool {
3543 unlock((*mutex)(lock))
3547 // park continuation on g0.
3548 func park_m(gp *g) {
3552 traceGoPark(mp.waittraceev, mp.waittraceskip)
3555 // N.B. Not using casGToWaiting here because the waitreason is
3556 // set by park_m's caller.
3557 casgstatus(gp, _Grunning, _Gwaiting)
3560 if fn := mp.waitunlockf; fn != nil {
3561 ok := fn(gp, mp.waitlock)
3562 mp.waitunlockf = nil
3566 traceGoUnpark(gp, 2)
3568 casgstatus(gp, _Gwaiting, _Grunnable)
3569 execute(gp, true) // Schedule it back, never returns.
3575 func goschedImpl(gp *g) {
3576 status := readgstatus(gp)
3577 if status&^_Gscan != _Grunning {
3579 throw("bad g status")
3581 casgstatus(gp, _Grunning, _Grunnable)
3590 // Gosched continuation on g0.
3591 func gosched_m(gp *g) {
3598 // goschedguarded is a forbidden-states-avoided version of gosched_m.
3599 func goschedguarded_m(gp *g) {
3601 if !canPreemptM(gp.m) {
3602 gogo(&gp.sched) // never return
3611 func gopreempt_m(gp *g) {
3618 // preemptPark parks gp and puts it in _Gpreempted.
3621 func preemptPark(gp *g) {
3623 traceGoPark(traceEvGoBlock, 0)
3625 status := readgstatus(gp)
3626 if status&^_Gscan != _Grunning {
3628 throw("bad g status")
3631 if gp.asyncSafePoint {
3632 // Double-check that async preemption does not
3633 // happen in SPWRITE assembly functions.
3634 // isAsyncSafePoint must exclude this case.
3635 f := findfunc(gp.sched.pc)
3637 throw("preempt at unknown pc")
3639 if f.flag&funcFlag_SPWRITE != 0 {
3640 println("runtime: unexpected SPWRITE function", funcname(f), "in async preempt")
3641 throw("preempt SPWRITE")
3645 // Transition from _Grunning to _Gscan|_Gpreempted. We can't
3646 // be in _Grunning when we dropg because then we'd be running
3647 // without an M, but the moment we're in _Gpreempted,
3648 // something could claim this G before we've fully cleaned it
3649 // up. Hence, we set the scan bit to lock down further
3650 // transitions until we can dropg.
3651 casGToPreemptScan(gp, _Grunning, _Gscan|_Gpreempted)
3653 casfrom_Gscanstatus(gp, _Gscan|_Gpreempted, _Gpreempted)
3657 // goyield is like Gosched, but it:
3658 // - emits a GoPreempt trace event instead of a GoSched trace event
3659 // - puts the current G on the runq of the current P instead of the globrunq
3665 func goyield_m(gp *g) {
3670 casgstatus(gp, _Grunning, _Grunnable)
3672 runqput(pp, gp, false)
3676 // Finishes execution of the current goroutine.
3687 // goexit continuation on g0.
3688 func goexit0(gp *g) {
3692 casgstatus(gp, _Grunning, _Gdead)
3693 gcController.addScannableStack(pp, -int64(gp.stack.hi-gp.stack.lo))
3694 if isSystemGoroutine(gp, false) {
3698 locked := gp.lockedm != 0
3701 gp.preemptStop = false
3702 gp.paniconfault = false
3703 gp._defer = nil // should be true already but just in case.
3704 gp._panic = nil // non-nil for Goexit during panic. points at stack-allocated data.
3706 gp.waitreason = waitReasonZero
3711 if gcBlackenEnabled != 0 && gp.gcAssistBytes > 0 {
3712 // Flush assist credit to the global pool. This gives
3713 // better information to pacing if the application is
3714 // rapidly creating an exiting goroutines.
3715 assistWorkPerByte := gcController.assistWorkPerByte.Load()
3716 scanCredit := int64(assistWorkPerByte * float64(gp.gcAssistBytes))
3717 gcController.bgScanCredit.Add(scanCredit)
3718 gp.gcAssistBytes = 0
3723 if GOARCH == "wasm" { // no threads yet on wasm
3725 schedule() // never returns
3728 if mp.lockedInt != 0 {
3729 print("invalid m->lockedInt = ", mp.lockedInt, "\n")
3730 throw("internal lockOSThread error")
3734 // The goroutine may have locked this thread because
3735 // it put it in an unusual kernel state. Kill it
3736 // rather than returning it to the thread pool.
3738 // Return to mstart, which will release the P and exit
3740 if GOOS != "plan9" { // See golang.org/issue/22227.
3743 // Clear lockedExt on plan9 since we may end up re-using
3751 // save updates getg().sched to refer to pc and sp so that a following
3752 // gogo will restore pc and sp.
3754 // save must not have write barriers because invoking a write barrier
3755 // can clobber getg().sched.
3758 //go:nowritebarrierrec
3759 func save(pc, sp uintptr) {
3762 if gp == gp.m.g0 || gp == gp.m.gsignal {
3763 // m.g0.sched is special and must describe the context
3764 // for exiting the thread. mstart1 writes to it directly.
3765 // m.gsignal.sched should not be used at all.
3766 // This check makes sure save calls do not accidentally
3767 // run in contexts where they'd write to system g's.
3768 throw("save on system g not allowed")
3775 // We need to ensure ctxt is zero, but can't have a write
3776 // barrier here. However, it should always already be zero.
3778 if gp.sched.ctxt != nil {
3783 // The goroutine g is about to enter a system call.
3784 // Record that it's not using the cpu anymore.
3785 // This is called only from the go syscall library and cgocall,
3786 // not from the low-level system calls used by the runtime.
3788 // Entersyscall cannot split the stack: the save must
3789 // make g->sched refer to the caller's stack segment, because
3790 // entersyscall is going to return immediately after.
3792 // Nothing entersyscall calls can split the stack either.
3793 // We cannot safely move the stack during an active call to syscall,
3794 // because we do not know which of the uintptr arguments are
3795 // really pointers (back into the stack).
3796 // In practice, this means that we make the fast path run through
3797 // entersyscall doing no-split things, and the slow path has to use systemstack
3798 // to run bigger things on the system stack.
3800 // reentersyscall is the entry point used by cgo callbacks, where explicitly
3801 // saved SP and PC are restored. This is needed when exitsyscall will be called
3802 // from a function further up in the call stack than the parent, as g->syscallsp
3803 // must always point to a valid stack frame. entersyscall below is the normal
3804 // entry point for syscalls, which obtains the SP and PC from the caller.
3807 // At the start of a syscall we emit traceGoSysCall to capture the stack trace.
3808 // If the syscall does not block, that is it, we do not emit any other events.
3809 // If the syscall blocks (that is, P is retaken), retaker emits traceGoSysBlock;
3810 // when syscall returns we emit traceGoSysExit and when the goroutine starts running
3811 // (potentially instantly, if exitsyscallfast returns true) we emit traceGoStart.
3812 // To ensure that traceGoSysExit is emitted strictly after traceGoSysBlock,
3813 // we remember current value of syscalltick in m (gp.m.syscalltick = gp.m.p.ptr().syscalltick),
3814 // whoever emits traceGoSysBlock increments p.syscalltick afterwards;
3815 // and we wait for the increment before emitting traceGoSysExit.
3816 // Note that the increment is done even if tracing is not enabled,
3817 // because tracing can be enabled in the middle of syscall. We don't want the wait to hang.
3820 func reentersyscall(pc, sp uintptr) {
3823 // Disable preemption because during this function g is in Gsyscall status,
3824 // but can have inconsistent g->sched, do not let GC observe it.
3827 // Entersyscall must not call any function that might split/grow the stack.
3828 // (See details in comment above.)
3829 // Catch calls that might, by replacing the stack guard with something that
3830 // will trip any stack check and leaving a flag to tell newstack to die.
3831 gp.stackguard0 = stackPreempt
3832 gp.throwsplit = true
3834 // Leave SP around for GC and traceback.
3838 casgstatus(gp, _Grunning, _Gsyscall)
3839 if gp.syscallsp < gp.stack.lo || gp.stack.hi < gp.syscallsp {
3840 systemstack(func() {
3841 print("entersyscall inconsistent ", hex(gp.syscallsp), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n")
3842 throw("entersyscall")
3847 systemstack(traceGoSysCall)
3848 // systemstack itself clobbers g.sched.{pc,sp} and we might
3849 // need them later when the G is genuinely blocked in a
3854 if sched.sysmonwait.Load() {
3855 systemstack(entersyscall_sysmon)
3859 if gp.m.p.ptr().runSafePointFn != 0 {
3860 // runSafePointFn may stack split if run on this stack
3861 systemstack(runSafePointFn)
3865 gp.m.syscalltick = gp.m.p.ptr().syscalltick
3866 gp.sysblocktraced = true
3871 atomic.Store(&pp.status, _Psyscall)
3872 if sched.gcwaiting.Load() {
3873 systemstack(entersyscall_gcwait)
3880 // Standard syscall entry used by the go syscall library and normal cgo calls.
3882 // This is exported via linkname to assembly in the syscall package and x/sys.
3885 //go:linkname entersyscall
3886 func entersyscall() {
3887 reentersyscall(getcallerpc(), getcallersp())
3890 func entersyscall_sysmon() {
3892 if sched.sysmonwait.Load() {
3893 sched.sysmonwait.Store(false)
3894 notewakeup(&sched.sysmonnote)
3899 func entersyscall_gcwait() {
3901 pp := gp.m.oldp.ptr()
3904 if sched.stopwait > 0 && atomic.Cas(&pp.status, _Psyscall, _Pgcstop) {
3910 if sched.stopwait--; sched.stopwait == 0 {
3911 notewakeup(&sched.stopnote)
3917 // The same as entersyscall(), but with a hint that the syscall is blocking.
3920 func entersyscallblock() {
3923 gp.m.locks++ // see comment in entersyscall
3924 gp.throwsplit = true
3925 gp.stackguard0 = stackPreempt // see comment in entersyscall
3926 gp.m.syscalltick = gp.m.p.ptr().syscalltick
3927 gp.sysblocktraced = true
3928 gp.m.p.ptr().syscalltick++
3930 // Leave SP around for GC and traceback.
3934 gp.syscallsp = gp.sched.sp
3935 gp.syscallpc = gp.sched.pc
3936 if gp.syscallsp < gp.stack.lo || gp.stack.hi < gp.syscallsp {
3940 systemstack(func() {
3941 print("entersyscallblock inconsistent ", hex(sp1), " ", hex(sp2), " ", hex(sp3), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n")
3942 throw("entersyscallblock")
3945 casgstatus(gp, _Grunning, _Gsyscall)
3946 if gp.syscallsp < gp.stack.lo || gp.stack.hi < gp.syscallsp {
3947 systemstack(func() {
3948 print("entersyscallblock inconsistent ", hex(sp), " ", hex(gp.sched.sp), " ", hex(gp.syscallsp), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n")
3949 throw("entersyscallblock")
3953 systemstack(entersyscallblock_handoff)
3955 // Resave for traceback during blocked call.
3956 save(getcallerpc(), getcallersp())
3961 func entersyscallblock_handoff() {
3964 traceGoSysBlock(getg().m.p.ptr())
3966 handoffp(releasep())
3969 // The goroutine g exited its system call.
3970 // Arrange for it to run on a cpu again.
3971 // This is called only from the go syscall library, not
3972 // from the low-level system calls used by the runtime.
3974 // Write barriers are not allowed because our P may have been stolen.
3976 // This is exported via linkname to assembly in the syscall package.
3979 //go:nowritebarrierrec
3980 //go:linkname exitsyscall
3981 func exitsyscall() {
3984 gp.m.locks++ // see comment in entersyscall
3985 if getcallersp() > gp.syscallsp {
3986 throw("exitsyscall: syscall frame is no longer valid")
3990 oldp := gp.m.oldp.ptr()
3992 if exitsyscallfast(oldp) {
3993 // When exitsyscallfast returns success, we have a P so can now use
3995 if goroutineProfile.active {
3996 // Make sure that gp has had its stack written out to the goroutine
3997 // profile, exactly as it was when the goroutine profiler first
3998 // stopped the world.
3999 systemstack(func() {
4000 tryRecordGoroutineProfileWB(gp)
4004 if oldp != gp.m.p.ptr() || gp.m.syscalltick != gp.m.p.ptr().syscalltick {
4005 systemstack(traceGoStart)
4008 // There's a cpu for us, so we can run.
4009 gp.m.p.ptr().syscalltick++
4010 // We need to cas the status and scan before resuming...
4011 casgstatus(gp, _Gsyscall, _Grunning)
4013 // Garbage collector isn't running (since we are),
4014 // so okay to clear syscallsp.
4018 // restore the preemption request in case we've cleared it in newstack
4019 gp.stackguard0 = stackPreempt
4021 // otherwise restore the real _StackGuard, we've spoiled it in entersyscall/entersyscallblock
4022 gp.stackguard0 = gp.stack.lo + _StackGuard
4024 gp.throwsplit = false
4026 if sched.disable.user && !schedEnabled(gp) {
4027 // Scheduling of this goroutine is disabled.
4036 // Wait till traceGoSysBlock event is emitted.
4037 // This ensures consistency of the trace (the goroutine is started after it is blocked).
4038 for oldp != nil && oldp.syscalltick == gp.m.syscalltick {
4041 // We can't trace syscall exit right now because we don't have a P.
4042 // Tracing code can invoke write barriers that cannot run without a P.
4043 // So instead we remember the syscall exit time and emit the event
4044 // in execute when we have a P.
4045 gp.sysexitticks = cputicks()
4050 // Call the scheduler.
4053 // Scheduler returned, so we're allowed to run now.
4054 // Delete the syscallsp information that we left for
4055 // the garbage collector during the system call.
4056 // Must wait until now because until gosched returns
4057 // we don't know for sure that the garbage collector
4060 gp.m.p.ptr().syscalltick++
4061 gp.throwsplit = false
4065 func exitsyscallfast(oldp *p) bool {
4068 // Freezetheworld sets stopwait but does not retake P's.
4069 if sched.stopwait == freezeStopWait {
4073 // Try to re-acquire the last P.
4074 if oldp != nil && oldp.status == _Psyscall && atomic.Cas(&oldp.status, _Psyscall, _Pidle) {
4075 // There's a cpu for us, so we can run.
4077 exitsyscallfast_reacquired()
4081 // Try to get any other idle P.
4082 if sched.pidle != 0 {
4084 systemstack(func() {
4085 ok = exitsyscallfast_pidle()
4086 if ok && trace.enabled {
4088 // Wait till traceGoSysBlock event is emitted.
4089 // This ensures consistency of the trace (the goroutine is started after it is blocked).
4090 for oldp.syscalltick == gp.m.syscalltick {
4104 // exitsyscallfast_reacquired is the exitsyscall path on which this G
4105 // has successfully reacquired the P it was running on before the
4109 func exitsyscallfast_reacquired() {
4111 if gp.m.syscalltick != gp.m.p.ptr().syscalltick {
4113 // The p was retaken and then enter into syscall again (since gp.m.syscalltick has changed).
4114 // traceGoSysBlock for this syscall was already emitted,
4115 // but here we effectively retake the p from the new syscall running on the same p.
4116 systemstack(func() {
4117 // Denote blocking of the new syscall.
4118 traceGoSysBlock(gp.m.p.ptr())
4119 // Denote completion of the current syscall.
4123 gp.m.p.ptr().syscalltick++
4127 func exitsyscallfast_pidle() bool {
4129 pp, _ := pidleget(0)
4130 if pp != nil && sched.sysmonwait.Load() {
4131 sched.sysmonwait.Store(false)
4132 notewakeup(&sched.sysmonnote)
4142 // exitsyscall slow path on g0.
4143 // Failed to acquire P, enqueue gp as runnable.
4145 // Called via mcall, so gp is the calling g from this M.
4147 //go:nowritebarrierrec
4148 func exitsyscall0(gp *g) {
4149 casgstatus(gp, _Gsyscall, _Grunnable)
4153 if schedEnabled(gp) {
4160 // Below, we stoplockedm if gp is locked. globrunqput releases
4161 // ownership of gp, so we must check if gp is locked prior to
4162 // committing the release by unlocking sched.lock, otherwise we
4163 // could race with another M transitioning gp from unlocked to
4165 locked = gp.lockedm != 0
4166 } else if sched.sysmonwait.Load() {
4167 sched.sysmonwait.Store(false)
4168 notewakeup(&sched.sysmonnote)
4173 execute(gp, false) // Never returns.
4176 // Wait until another thread schedules gp and so m again.
4178 // N.B. lockedm must be this M, as this g was running on this M
4179 // before entersyscall.
4181 execute(gp, false) // Never returns.
4184 schedule() // Never returns.
4187 // Called from syscall package before fork.
4189 //go:linkname syscall_runtime_BeforeFork syscall.runtime_BeforeFork
4191 func syscall_runtime_BeforeFork() {
4194 // Block signals during a fork, so that the child does not run
4195 // a signal handler before exec if a signal is sent to the process
4196 // group. See issue #18600.
4198 sigsave(&gp.m.sigmask)
4201 // This function is called before fork in syscall package.
4202 // Code between fork and exec must not allocate memory nor even try to grow stack.
4203 // Here we spoil g->_StackGuard to reliably detect any attempts to grow stack.
4204 // runtime_AfterFork will undo this in parent process, but not in child.
4205 gp.stackguard0 = stackFork
4208 // Called from syscall package after fork in parent.
4210 //go:linkname syscall_runtime_AfterFork syscall.runtime_AfterFork
4212 func syscall_runtime_AfterFork() {
4215 // See the comments in beforefork.
4216 gp.stackguard0 = gp.stack.lo + _StackGuard
4218 msigrestore(gp.m.sigmask)
4223 // inForkedChild is true while manipulating signals in the child process.
4224 // This is used to avoid calling libc functions in case we are using vfork.
4225 var inForkedChild bool
4227 // Called from syscall package after fork in child.
4228 // It resets non-sigignored signals to the default handler, and
4229 // restores the signal mask in preparation for the exec.
4231 // Because this might be called during a vfork, and therefore may be
4232 // temporarily sharing address space with the parent process, this must
4233 // not change any global variables or calling into C code that may do so.
4235 //go:linkname syscall_runtime_AfterForkInChild syscall.runtime_AfterForkInChild
4237 //go:nowritebarrierrec
4238 func syscall_runtime_AfterForkInChild() {
4239 // It's OK to change the global variable inForkedChild here
4240 // because we are going to change it back. There is no race here,
4241 // because if we are sharing address space with the parent process,
4242 // then the parent process can not be running concurrently.
4243 inForkedChild = true
4245 clearSignalHandlers()
4247 // When we are the child we are the only thread running,
4248 // so we know that nothing else has changed gp.m.sigmask.
4249 msigrestore(getg().m.sigmask)
4251 inForkedChild = false
4254 // pendingPreemptSignals is the number of preemption signals
4255 // that have been sent but not received. This is only used on Darwin.
4257 var pendingPreemptSignals atomic.Int32
4259 // Called from syscall package before Exec.
4261 //go:linkname syscall_runtime_BeforeExec syscall.runtime_BeforeExec
4262 func syscall_runtime_BeforeExec() {
4263 // Prevent thread creation during exec.
4266 // On Darwin, wait for all pending preemption signals to
4267 // be received. See issue #41702.
4268 if GOOS == "darwin" || GOOS == "ios" {
4269 for pendingPreemptSignals.Load() > 0 {
4275 // Called from syscall package after Exec.
4277 //go:linkname syscall_runtime_AfterExec syscall.runtime_AfterExec
4278 func syscall_runtime_AfterExec() {
4282 // Allocate a new g, with a stack big enough for stacksize bytes.
4283 func malg(stacksize int32) *g {
4286 stacksize = round2(_StackSystem + stacksize)
4287 systemstack(func() {
4288 newg.stack = stackalloc(uint32(stacksize))
4290 newg.stackguard0 = newg.stack.lo + _StackGuard
4291 newg.stackguard1 = ^uintptr(0)
4292 // Clear the bottom word of the stack. We record g
4293 // there on gsignal stack during VDSO on ARM and ARM64.
4294 *(*uintptr)(unsafe.Pointer(newg.stack.lo)) = 0
4299 // Create a new g running fn.
4300 // Put it on the queue of g's waiting to run.
4301 // The compiler turns a go statement into a call to this.
4302 func newproc(fn *funcval) {
4305 systemstack(func() {
4306 newg := newproc1(fn, gp, pc)
4308 pp := getg().m.p.ptr()
4309 runqput(pp, newg, true)
4317 // Create a new g in state _Grunnable, starting at fn. callerpc is the
4318 // address of the go statement that created this. The caller is responsible
4319 // for adding the new g to the scheduler.
4320 func newproc1(fn *funcval, callergp *g, callerpc uintptr) *g {
4322 fatal("go of nil func value")
4325 mp := acquirem() // disable preemption because we hold M and P in local vars.
4329 newg = malg(_StackMin)
4330 casgstatus(newg, _Gidle, _Gdead)
4331 allgadd(newg) // publishes with a g->status of Gdead so GC scanner doesn't look at uninitialized stack.
4333 if newg.stack.hi == 0 {
4334 throw("newproc1: newg missing stack")
4337 if readgstatus(newg) != _Gdead {
4338 throw("newproc1: new g is not Gdead")
4341 totalSize := uintptr(4*goarch.PtrSize + sys.MinFrameSize) // extra space in case of reads slightly beyond frame
4342 totalSize = alignUp(totalSize, sys.StackAlign)
4343 sp := newg.stack.hi - totalSize
4347 *(*uintptr)(unsafe.Pointer(sp)) = 0
4349 spArg += sys.MinFrameSize
4352 memclrNoHeapPointers(unsafe.Pointer(&newg.sched), unsafe.Sizeof(newg.sched))
4355 newg.sched.pc = abi.FuncPCABI0(goexit) + sys.PCQuantum // +PCQuantum so that previous instruction is in same function
4356 newg.sched.g = guintptr(unsafe.Pointer(newg))
4357 gostartcallfn(&newg.sched, fn)
4358 newg.parentGoid = callergp.goid
4359 newg.gopc = callerpc
4360 newg.ancestors = saveAncestors(callergp)
4361 newg.startpc = fn.fn
4362 if isSystemGoroutine(newg, false) {
4365 // Only user goroutines inherit pprof labels.
4367 newg.labels = mp.curg.labels
4369 if goroutineProfile.active {
4370 // A concurrent goroutine profile is running. It should include
4371 // exactly the set of goroutines that were alive when the goroutine
4372 // profiler first stopped the world. That does not include newg, so
4373 // mark it as not needing a profile before transitioning it from
4375 newg.goroutineProfiled.Store(goroutineProfileSatisfied)
4378 // Track initial transition?
4379 newg.trackingSeq = uint8(fastrand())
4380 if newg.trackingSeq%gTrackingPeriod == 0 {
4381 newg.tracking = true
4383 casgstatus(newg, _Gdead, _Grunnable)
4384 gcController.addScannableStack(pp, int64(newg.stack.hi-newg.stack.lo))
4386 if pp.goidcache == pp.goidcacheend {
4387 // Sched.goidgen is the last allocated id,
4388 // this batch must be [sched.goidgen+1, sched.goidgen+GoidCacheBatch].
4389 // At startup sched.goidgen=0, so main goroutine receives goid=1.
4390 pp.goidcache = sched.goidgen.Add(_GoidCacheBatch)
4391 pp.goidcache -= _GoidCacheBatch - 1
4392 pp.goidcacheend = pp.goidcache + _GoidCacheBatch
4394 newg.goid = pp.goidcache
4397 newg.racectx = racegostart(callerpc)
4398 if newg.labels != nil {
4399 // See note in proflabel.go on labelSync's role in synchronizing
4400 // with the reads in the signal handler.
4401 racereleasemergeg(newg, unsafe.Pointer(&labelSync))
4405 traceGoCreate(newg, newg.startpc)
4412 // saveAncestors copies previous ancestors of the given caller g and
4413 // includes info for the current caller into a new set of tracebacks for
4414 // a g being created.
4415 func saveAncestors(callergp *g) *[]ancestorInfo {
4416 // Copy all prior info, except for the root goroutine (goid 0).
4417 if debug.tracebackancestors <= 0 || callergp.goid == 0 {
4420 var callerAncestors []ancestorInfo
4421 if callergp.ancestors != nil {
4422 callerAncestors = *callergp.ancestors
4424 n := int32(len(callerAncestors)) + 1
4425 if n > debug.tracebackancestors {
4426 n = debug.tracebackancestors
4428 ancestors := make([]ancestorInfo, n)
4429 copy(ancestors[1:], callerAncestors)
4431 var pcs [tracebackInnerFrames]uintptr
4432 npcs := gcallers(callergp, 0, pcs[:])
4433 ipcs := make([]uintptr, npcs)
4435 ancestors[0] = ancestorInfo{
4437 goid: callergp.goid,
4438 gopc: callergp.gopc,
4441 ancestorsp := new([]ancestorInfo)
4442 *ancestorsp = ancestors
4446 // Put on gfree list.
4447 // If local list is too long, transfer a batch to the global list.
4448 func gfput(pp *p, gp *g) {
4449 if readgstatus(gp) != _Gdead {
4450 throw("gfput: bad status (not Gdead)")
4453 stksize := gp.stack.hi - gp.stack.lo
4455 if stksize != uintptr(startingStackSize) {
4456 // non-standard stack size - free it.
4465 if pp.gFree.n >= 64 {
4471 for pp.gFree.n >= 32 {
4472 gp := pp.gFree.pop()
4474 if gp.stack.lo == 0 {
4481 lock(&sched.gFree.lock)
4482 sched.gFree.noStack.pushAll(noStackQ)
4483 sched.gFree.stack.pushAll(stackQ)
4484 sched.gFree.n += inc
4485 unlock(&sched.gFree.lock)
4489 // Get from gfree list.
4490 // If local list is empty, grab a batch from global list.
4491 func gfget(pp *p) *g {
4493 if pp.gFree.empty() && (!sched.gFree.stack.empty() || !sched.gFree.noStack.empty()) {
4494 lock(&sched.gFree.lock)
4495 // Move a batch of free Gs to the P.
4496 for pp.gFree.n < 32 {
4497 // Prefer Gs with stacks.
4498 gp := sched.gFree.stack.pop()
4500 gp = sched.gFree.noStack.pop()
4509 unlock(&sched.gFree.lock)
4512 gp := pp.gFree.pop()
4517 if gp.stack.lo != 0 && gp.stack.hi-gp.stack.lo != uintptr(startingStackSize) {
4518 // Deallocate old stack. We kept it in gfput because it was the
4519 // right size when the goroutine was put on the free list, but
4520 // the right size has changed since then.
4521 systemstack(func() {
4528 if gp.stack.lo == 0 {
4529 // Stack was deallocated in gfput or just above. Allocate a new one.
4530 systemstack(func() {
4531 gp.stack = stackalloc(startingStackSize)
4533 gp.stackguard0 = gp.stack.lo + _StackGuard
4536 racemalloc(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo)
4539 msanmalloc(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo)
4542 asanunpoison(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo)
4548 // Purge all cached G's from gfree list to the global list.
4549 func gfpurge(pp *p) {
4555 for !pp.gFree.empty() {
4556 gp := pp.gFree.pop()
4558 if gp.stack.lo == 0 {
4565 lock(&sched.gFree.lock)
4566 sched.gFree.noStack.pushAll(noStackQ)
4567 sched.gFree.stack.pushAll(stackQ)
4568 sched.gFree.n += inc
4569 unlock(&sched.gFree.lock)
4572 // Breakpoint executes a breakpoint trap.
4577 // dolockOSThread is called by LockOSThread and lockOSThread below
4578 // after they modify m.locked. Do not allow preemption during this call,
4579 // or else the m might be different in this function than in the caller.
4582 func dolockOSThread() {
4583 if GOARCH == "wasm" {
4584 return // no threads on wasm yet
4587 gp.m.lockedg.set(gp)
4588 gp.lockedm.set(gp.m)
4591 // LockOSThread wires the calling goroutine to its current operating system thread.
4592 // The calling goroutine will always execute in that thread,
4593 // and no other goroutine will execute in it,
4594 // until the calling goroutine has made as many calls to
4595 // UnlockOSThread as to LockOSThread.
4596 // If the calling goroutine exits without unlocking the thread,
4597 // the thread will be terminated.
4599 // All init functions are run on the startup thread. Calling LockOSThread
4600 // from an init function will cause the main function to be invoked on
4603 // A goroutine should call LockOSThread before calling OS services or
4604 // non-Go library functions that depend on per-thread state.
4607 func LockOSThread() {
4608 if atomic.Load(&newmHandoff.haveTemplateThread) == 0 && GOOS != "plan9" {
4609 // If we need to start a new thread from the locked
4610 // thread, we need the template thread. Start it now
4611 // while we're in a known-good state.
4612 startTemplateThread()
4616 if gp.m.lockedExt == 0 {
4618 panic("LockOSThread nesting overflow")
4624 func lockOSThread() {
4625 getg().m.lockedInt++
4629 // dounlockOSThread is called by UnlockOSThread and unlockOSThread below
4630 // after they update m->locked. Do not allow preemption during this call,
4631 // or else the m might be in different in this function than in the caller.
4634 func dounlockOSThread() {
4635 if GOARCH == "wasm" {
4636 return // no threads on wasm yet
4639 if gp.m.lockedInt != 0 || gp.m.lockedExt != 0 {
4646 // UnlockOSThread undoes an earlier call to LockOSThread.
4647 // If this drops the number of active LockOSThread calls on the
4648 // calling goroutine to zero, it unwires the calling goroutine from
4649 // its fixed operating system thread.
4650 // If there are no active LockOSThread calls, this is a no-op.
4652 // Before calling UnlockOSThread, the caller must ensure that the OS
4653 // thread is suitable for running other goroutines. If the caller made
4654 // any permanent changes to the state of the thread that would affect
4655 // other goroutines, it should not call this function and thus leave
4656 // the goroutine locked to the OS thread until the goroutine (and
4657 // hence the thread) exits.
4660 func UnlockOSThread() {
4662 if gp.m.lockedExt == 0 {
4670 func unlockOSThread() {
4672 if gp.m.lockedInt == 0 {
4673 systemstack(badunlockosthread)
4679 func badunlockosthread() {
4680 throw("runtime: internal error: misuse of lockOSThread/unlockOSThread")
4683 func gcount() int32 {
4684 n := int32(atomic.Loaduintptr(&allglen)) - sched.gFree.n - sched.ngsys.Load()
4685 for _, pp := range allp {
4689 // All these variables can be changed concurrently, so the result can be inconsistent.
4690 // But at least the current goroutine is running.
4697 func mcount() int32 {
4698 return int32(sched.mnext - sched.nmfreed)
4702 signalLock atomic.Uint32
4704 // Must hold signalLock to write. Reads may be lock-free, but
4705 // signalLock should be taken to synchronize with changes.
4709 func _System() { _System() }
4710 func _ExternalCode() { _ExternalCode() }
4711 func _LostExternalCode() { _LostExternalCode() }
4712 func _GC() { _GC() }
4713 func _LostSIGPROFDuringAtomic64() { _LostSIGPROFDuringAtomic64() }
4714 func _VDSO() { _VDSO() }
4716 // Called if we receive a SIGPROF signal.
4717 // Called by the signal handler, may run during STW.
4719 //go:nowritebarrierrec
4720 func sigprof(pc, sp, lr uintptr, gp *g, mp *m) {
4721 if prof.hz.Load() == 0 {
4725 // If mp.profilehz is 0, then profiling is not enabled for this thread.
4726 // We must check this to avoid a deadlock between setcpuprofilerate
4727 // and the call to cpuprof.add, below.
4728 if mp != nil && mp.profilehz == 0 {
4732 // On mips{,le}/arm, 64bit atomics are emulated with spinlocks, in
4733 // runtime/internal/atomic. If SIGPROF arrives while the program is inside
4734 // the critical section, it creates a deadlock (when writing the sample).
4735 // As a workaround, create a counter of SIGPROFs while in critical section
4736 // to store the count, and pass it to sigprof.add() later when SIGPROF is
4737 // received from somewhere else (with _LostSIGPROFDuringAtomic64 as pc).
4738 if GOARCH == "mips" || GOARCH == "mipsle" || GOARCH == "arm" {
4739 if f := findfunc(pc); f.valid() {
4740 if hasPrefix(funcname(f), "runtime/internal/atomic") {
4741 cpuprof.lostAtomic++
4745 if GOARCH == "arm" && goarm < 7 && GOOS == "linux" && pc&0xffff0000 == 0xffff0000 {
4746 // runtime/internal/atomic functions call into kernel
4747 // helpers on arm < 7. See
4748 // runtime/internal/atomic/sys_linux_arm.s.
4749 cpuprof.lostAtomic++
4754 // Profiling runs concurrently with GC, so it must not allocate.
4755 // Set a trap in case the code does allocate.
4756 // Note that on windows, one thread takes profiles of all the
4757 // other threads, so mp is usually not getg().m.
4758 // In fact mp may not even be stopped.
4759 // See golang.org/issue/17165.
4760 getg().m.mallocing++
4763 var stk [maxCPUProfStack]uintptr
4765 if mp.ncgo > 0 && mp.curg != nil && mp.curg.syscallpc != 0 && mp.curg.syscallsp != 0 {
4767 // Check cgoCallersUse to make sure that we are not
4768 // interrupting other code that is fiddling with
4769 // cgoCallers. We are running in a signal handler
4770 // with all signals blocked, so we don't have to worry
4771 // about any other code interrupting us.
4772 if mp.cgoCallersUse.Load() == 0 && mp.cgoCallers != nil && mp.cgoCallers[0] != 0 {
4773 for cgoOff < len(mp.cgoCallers) && mp.cgoCallers[cgoOff] != 0 {
4776 n += copy(stk[:], mp.cgoCallers[:cgoOff])
4777 mp.cgoCallers[0] = 0
4780 // Collect Go stack that leads to the cgo call.
4781 u.initAt(mp.curg.syscallpc, mp.curg.syscallsp, 0, mp.curg, unwindSilentErrors)
4782 } else if usesLibcall() && mp.libcallg != 0 && mp.libcallpc != 0 && mp.libcallsp != 0 {
4783 // Libcall, i.e. runtime syscall on windows.
4784 // Collect Go stack that leads to the call.
4785 u.initAt(mp.libcallpc, mp.libcallsp, 0, mp.libcallg.ptr(), unwindSilentErrors)
4786 } else if mp != nil && mp.vdsoSP != 0 {
4787 // VDSO call, e.g. nanotime1 on Linux.
4788 // Collect Go stack that leads to the call.
4789 u.initAt(mp.vdsoPC, mp.vdsoSP, 0, gp, unwindSilentErrors|unwindJumpStack)
4791 u.initAt(pc, sp, lr, gp, unwindSilentErrors|unwindTrap|unwindJumpStack)
4793 n += tracebackPCs(&u, 0, stk[n:])
4796 // Normal traceback is impossible or has failed.
4797 // Account it against abstract "System" or "GC".
4800 pc = abi.FuncPCABIInternal(_VDSO) + sys.PCQuantum
4801 } else if pc > firstmoduledata.etext {
4802 // "ExternalCode" is better than "etext".
4803 pc = abi.FuncPCABIInternal(_ExternalCode) + sys.PCQuantum
4806 if mp.preemptoff != "" {
4807 stk[1] = abi.FuncPCABIInternal(_GC) + sys.PCQuantum
4809 stk[1] = abi.FuncPCABIInternal(_System) + sys.PCQuantum
4813 if prof.hz.Load() != 0 {
4814 // Note: it can happen on Windows that we interrupted a system thread
4815 // with no g, so gp could nil. The other nil checks are done out of
4816 // caution, but not expected to be nil in practice.
4817 var tagPtr *unsafe.Pointer
4818 if gp != nil && gp.m != nil && gp.m.curg != nil {
4819 tagPtr = &gp.m.curg.labels
4821 cpuprof.add(tagPtr, stk[:n])
4825 if gp != nil && gp.m != nil {
4826 if gp.m.curg != nil {
4831 traceCPUSample(gprof, pp, stk[:n])
4833 getg().m.mallocing--
4836 // setcpuprofilerate sets the CPU profiling rate to hz times per second.
4837 // If hz <= 0, setcpuprofilerate turns off CPU profiling.
4838 func setcpuprofilerate(hz int32) {
4839 // Force sane arguments.
4844 // Disable preemption, otherwise we can be rescheduled to another thread
4845 // that has profiling enabled.
4849 // Stop profiler on this thread so that it is safe to lock prof.
4850 // if a profiling signal came in while we had prof locked,
4851 // it would deadlock.
4852 setThreadCPUProfiler(0)
4854 for !prof.signalLock.CompareAndSwap(0, 1) {
4857 if prof.hz.Load() != hz {
4858 setProcessCPUProfiler(hz)
4861 prof.signalLock.Store(0)
4864 sched.profilehz = hz
4868 setThreadCPUProfiler(hz)
4874 // init initializes pp, which may be a freshly allocated p or a
4875 // previously destroyed p, and transitions it to status _Pgcstop.
4876 func (pp *p) init(id int32) {
4878 pp.status = _Pgcstop
4879 pp.sudogcache = pp.sudogbuf[:0]
4880 pp.deferpool = pp.deferpoolbuf[:0]
4882 if pp.mcache == nil {
4885 throw("missing mcache?")
4887 // Use the bootstrap mcache0. Only one P will get
4888 // mcache0: the one with ID 0.
4891 pp.mcache = allocmcache()
4894 if raceenabled && pp.raceprocctx == 0 {
4896 pp.raceprocctx = raceprocctx0
4897 raceprocctx0 = 0 // bootstrap
4899 pp.raceprocctx = raceproccreate()
4902 lockInit(&pp.timersLock, lockRankTimers)
4904 // This P may get timers when it starts running. Set the mask here
4905 // since the P may not go through pidleget (notably P 0 on startup).
4907 // Similarly, we may not go through pidleget before this P starts
4908 // running if it is P 0 on startup.
4912 // destroy releases all of the resources associated with pp and
4913 // transitions it to status _Pdead.
4915 // sched.lock must be held and the world must be stopped.
4916 func (pp *p) destroy() {
4917 assertLockHeld(&sched.lock)
4918 assertWorldStopped()
4920 // Move all runnable goroutines to the global queue
4921 for pp.runqhead != pp.runqtail {
4922 // Pop from tail of local queue
4924 gp := pp.runq[pp.runqtail%uint32(len(pp.runq))].ptr()
4925 // Push onto head of global queue
4928 if pp.runnext != 0 {
4929 globrunqputhead(pp.runnext.ptr())
4932 if len(pp.timers) > 0 {
4933 plocal := getg().m.p.ptr()
4934 // The world is stopped, but we acquire timersLock to
4935 // protect against sysmon calling timeSleepUntil.
4936 // This is the only case where we hold the timersLock of
4937 // more than one P, so there are no deadlock concerns.
4938 lock(&plocal.timersLock)
4939 lock(&pp.timersLock)
4940 moveTimers(plocal, pp.timers)
4942 pp.numTimers.Store(0)
4943 pp.deletedTimers.Store(0)
4944 pp.timer0When.Store(0)
4945 unlock(&pp.timersLock)
4946 unlock(&plocal.timersLock)
4948 // Flush p's write barrier buffer.
4949 if gcphase != _GCoff {
4953 for i := range pp.sudogbuf {
4954 pp.sudogbuf[i] = nil
4956 pp.sudogcache = pp.sudogbuf[:0]
4957 for j := range pp.deferpoolbuf {
4958 pp.deferpoolbuf[j] = nil
4960 pp.deferpool = pp.deferpoolbuf[:0]
4961 systemstack(func() {
4962 for i := 0; i < pp.mspancache.len; i++ {
4963 // Safe to call since the world is stopped.
4964 mheap_.spanalloc.free(unsafe.Pointer(pp.mspancache.buf[i]))
4966 pp.mspancache.len = 0
4968 pp.pcache.flush(&mheap_.pages)
4969 unlock(&mheap_.lock)
4971 freemcache(pp.mcache)
4976 if pp.timerRaceCtx != 0 {
4977 // The race detector code uses a callback to fetch
4978 // the proc context, so arrange for that callback
4979 // to see the right thing.
4980 // This hack only works because we are the only
4986 racectxend(pp.timerRaceCtx)
4991 raceprocdestroy(pp.raceprocctx)
4998 // Change number of processors.
5000 // sched.lock must be held, and the world must be stopped.
5002 // gcworkbufs must not be being modified by either the GC or the write barrier
5003 // code, so the GC must not be running if the number of Ps actually changes.
5005 // Returns list of Ps with local work, they need to be scheduled by the caller.
5006 func procresize(nprocs int32) *p {
5007 assertLockHeld(&sched.lock)
5008 assertWorldStopped()
5011 if old < 0 || nprocs <= 0 {
5012 throw("procresize: invalid arg")
5015 traceGomaxprocs(nprocs)
5018 // update statistics
5020 if sched.procresizetime != 0 {
5021 sched.totaltime += int64(old) * (now - sched.procresizetime)
5023 sched.procresizetime = now
5025 maskWords := (nprocs + 31) / 32
5027 // Grow allp if necessary.
5028 if nprocs > int32(len(allp)) {
5029 // Synchronize with retake, which could be running
5030 // concurrently since it doesn't run on a P.
5032 if nprocs <= int32(cap(allp)) {
5033 allp = allp[:nprocs]
5035 nallp := make([]*p, nprocs)
5036 // Copy everything up to allp's cap so we
5037 // never lose old allocated Ps.
5038 copy(nallp, allp[:cap(allp)])
5042 if maskWords <= int32(cap(idlepMask)) {
5043 idlepMask = idlepMask[:maskWords]
5044 timerpMask = timerpMask[:maskWords]
5046 nidlepMask := make([]uint32, maskWords)
5047 // No need to copy beyond len, old Ps are irrelevant.
5048 copy(nidlepMask, idlepMask)
5049 idlepMask = nidlepMask
5051 ntimerpMask := make([]uint32, maskWords)
5052 copy(ntimerpMask, timerpMask)
5053 timerpMask = ntimerpMask
5058 // initialize new P's
5059 for i := old; i < nprocs; i++ {
5065 atomicstorep(unsafe.Pointer(&allp[i]), unsafe.Pointer(pp))
5069 if gp.m.p != 0 && gp.m.p.ptr().id < nprocs {
5070 // continue to use the current P
5071 gp.m.p.ptr().status = _Prunning
5072 gp.m.p.ptr().mcache.prepareForSweep()
5074 // release the current P and acquire allp[0].
5076 // We must do this before destroying our current P
5077 // because p.destroy itself has write barriers, so we
5078 // need to do that from a valid P.
5081 // Pretend that we were descheduled
5082 // and then scheduled again to keep
5085 traceProcStop(gp.m.p.ptr())
5099 // g.m.p is now set, so we no longer need mcache0 for bootstrapping.
5102 // release resources from unused P's
5103 for i := nprocs; i < old; i++ {
5106 // can't free P itself because it can be referenced by an M in syscall
5110 if int32(len(allp)) != nprocs {
5112 allp = allp[:nprocs]
5113 idlepMask = idlepMask[:maskWords]
5114 timerpMask = timerpMask[:maskWords]
5119 for i := nprocs - 1; i >= 0; i-- {
5121 if gp.m.p.ptr() == pp {
5129 pp.link.set(runnablePs)
5133 stealOrder.reset(uint32(nprocs))
5134 var int32p *int32 = &gomaxprocs // make compiler check that gomaxprocs is an int32
5135 atomic.Store((*uint32)(unsafe.Pointer(int32p)), uint32(nprocs))
5137 // Notify the limiter that the amount of procs has changed.
5138 gcCPULimiter.resetCapacity(now, nprocs)
5143 // Associate p and the current m.
5145 // This function is allowed to have write barriers even if the caller
5146 // isn't because it immediately acquires pp.
5148 //go:yeswritebarrierrec
5149 func acquirep(pp *p) {
5150 // Do the part that isn't allowed to have write barriers.
5153 // Have p; write barriers now allowed.
5155 // Perform deferred mcache flush before this P can allocate
5156 // from a potentially stale mcache.
5157 pp.mcache.prepareForSweep()
5164 // wirep is the first step of acquirep, which actually associates the
5165 // current M to pp. This is broken out so we can disallow write
5166 // barriers for this part, since we don't yet have a P.
5168 //go:nowritebarrierrec
5174 throw("wirep: already in go")
5176 if pp.m != 0 || pp.status != _Pidle {
5181 print("wirep: p->m=", pp.m, "(", id, ") p->status=", pp.status, "\n")
5182 throw("wirep: invalid p state")
5186 pp.status = _Prunning
5189 // Disassociate p and the current m.
5190 func releasep() *p {
5194 throw("releasep: invalid arg")
5197 if pp.m.ptr() != gp.m || pp.status != _Prunning {
5198 print("releasep: m=", gp.m, " m->p=", gp.m.p.ptr(), " p->m=", hex(pp.m), " p->status=", pp.status, "\n")
5199 throw("releasep: invalid p state")
5202 traceProcStop(gp.m.p.ptr())
5210 func incidlelocked(v int32) {
5212 sched.nmidlelocked += v
5219 // Check for deadlock situation.
5220 // The check is based on number of running M's, if 0 -> deadlock.
5221 // sched.lock must be held.
5223 assertLockHeld(&sched.lock)
5225 // For -buildmode=c-shared or -buildmode=c-archive it's OK if
5226 // there are no running goroutines. The calling program is
5227 // assumed to be running.
5228 if islibrary || isarchive {
5232 // If we are dying because of a signal caught on an already idle thread,
5233 // freezetheworld will cause all running threads to block.
5234 // And runtime will essentially enter into deadlock state,
5235 // except that there is a thread that will call exit soon.
5236 if panicking.Load() > 0 {
5240 // If we are not running under cgo, but we have an extra M then account
5241 // for it. (It is possible to have an extra M on Windows without cgo to
5242 // accommodate callbacks created by syscall.NewCallback. See issue #6751
5245 if !iscgo && cgoHasExtraM {
5246 mp := lockextra(true)
5247 haveExtraM := extraMCount > 0
5254 run := mcount() - sched.nmidle - sched.nmidlelocked - sched.nmsys
5259 print("runtime: checkdead: nmidle=", sched.nmidle, " nmidlelocked=", sched.nmidlelocked, " mcount=", mcount(), " nmsys=", sched.nmsys, "\n")
5260 throw("checkdead: inconsistent counts")
5264 forEachG(func(gp *g) {
5265 if isSystemGoroutine(gp, false) {
5268 s := readgstatus(gp)
5269 switch s &^ _Gscan {
5276 print("runtime: checkdead: find g ", gp.goid, " in status ", s, "\n")
5277 throw("checkdead: runnable g")
5280 if grunning == 0 { // possible if main goroutine calls runtime·Goexit()
5281 unlock(&sched.lock) // unlock so that GODEBUG=scheddetail=1 doesn't hang
5282 fatal("no goroutines (main called runtime.Goexit) - deadlock!")
5285 // Maybe jump time forward for playground.
5287 if when := timeSleepUntil(); when < maxWhen {
5290 // Start an M to steal the timer.
5291 pp, _ := pidleget(faketime)
5293 // There should always be a free P since
5294 // nothing is running.
5295 throw("checkdead: no p for timer")
5299 // There should always be a free M since
5300 // nothing is running.
5301 throw("checkdead: no m for timer")
5303 // M must be spinning to steal. We set this to be
5304 // explicit, but since this is the only M it would
5305 // become spinning on its own anyways.
5306 sched.nmspinning.Add(1)
5309 notewakeup(&mp.park)
5314 // There are no goroutines running, so we can look at the P's.
5315 for _, pp := range allp {
5316 if len(pp.timers) > 0 {
5321 unlock(&sched.lock) // unlock so that GODEBUG=scheddetail=1 doesn't hang
5322 fatal("all goroutines are asleep - deadlock!")
5325 // forcegcperiod is the maximum time in nanoseconds between garbage
5326 // collections. If we go this long without a garbage collection, one
5327 // is forced to run.
5329 // This is a variable for testing purposes. It normally doesn't change.
5330 var forcegcperiod int64 = 2 * 60 * 1e9
5332 // needSysmonWorkaround is true if the workaround for
5333 // golang.org/issue/42515 is needed on NetBSD.
5334 var needSysmonWorkaround bool = false
5336 // Always runs without a P, so write barriers are not allowed.
5338 //go:nowritebarrierrec
5345 lasttrace := int64(0)
5346 idle := 0 // how many cycles in succession we had not wokeup somebody
5350 if idle == 0 { // start with 20us sleep...
5352 } else if idle > 50 { // start doubling the sleep after 1ms...
5355 if delay > 10*1000 { // up to 10ms
5360 // sysmon should not enter deep sleep if schedtrace is enabled so that
5361 // it can print that information at the right time.
5363 // It should also not enter deep sleep if there are any active P's so
5364 // that it can retake P's from syscalls, preempt long running G's, and
5365 // poll the network if all P's are busy for long stretches.
5367 // It should wakeup from deep sleep if any P's become active either due
5368 // to exiting a syscall or waking up due to a timer expiring so that it
5369 // can resume performing those duties. If it wakes from a syscall it
5370 // resets idle and delay as a bet that since it had retaken a P from a
5371 // syscall before, it may need to do it again shortly after the
5372 // application starts work again. It does not reset idle when waking
5373 // from a timer to avoid adding system load to applications that spend
5374 // most of their time sleeping.
5376 if debug.schedtrace <= 0 && (sched.gcwaiting.Load() || sched.npidle.Load() == gomaxprocs) {
5378 if sched.gcwaiting.Load() || sched.npidle.Load() == gomaxprocs {
5379 syscallWake := false
5380 next := timeSleepUntil()
5382 sched.sysmonwait.Store(true)
5384 // Make wake-up period small enough
5385 // for the sampling to be correct.
5386 sleep := forcegcperiod / 2
5387 if next-now < sleep {
5390 shouldRelax := sleep >= osRelaxMinNS
5394 syscallWake = notetsleep(&sched.sysmonnote, sleep)
5399 sched.sysmonwait.Store(false)
5400 noteclear(&sched.sysmonnote)
5410 lock(&sched.sysmonlock)
5411 // Update now in case we blocked on sysmonnote or spent a long time
5412 // blocked on schedlock or sysmonlock above.
5415 // trigger libc interceptors if needed
5416 if *cgo_yield != nil {
5417 asmcgocall(*cgo_yield, nil)
5419 // poll network if not polled for more than 10ms
5420 lastpoll := sched.lastpoll.Load()
5421 if netpollinited() && lastpoll != 0 && lastpoll+10*1000*1000 < now {
5422 sched.lastpoll.CompareAndSwap(lastpoll, now)
5423 list := netpoll(0) // non-blocking - returns list of goroutines
5425 // Need to decrement number of idle locked M's
5426 // (pretending that one more is running) before injectglist.
5427 // Otherwise it can lead to the following situation:
5428 // injectglist grabs all P's but before it starts M's to run the P's,
5429 // another M returns from syscall, finishes running its G,
5430 // observes that there is no work to do and no other running M's
5431 // and reports deadlock.
5437 if GOOS == "netbsd" && needSysmonWorkaround {
5438 // netpoll is responsible for waiting for timer
5439 // expiration, so we typically don't have to worry
5440 // about starting an M to service timers. (Note that
5441 // sleep for timeSleepUntil above simply ensures sysmon
5442 // starts running again when that timer expiration may
5443 // cause Go code to run again).
5445 // However, netbsd has a kernel bug that sometimes
5446 // misses netpollBreak wake-ups, which can lead to
5447 // unbounded delays servicing timers. If we detect this
5448 // overrun, then startm to get something to handle the
5451 // See issue 42515 and
5452 // https://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=50094.
5453 if next := timeSleepUntil(); next < now {
5457 if scavenger.sysmonWake.Load() != 0 {
5458 // Kick the scavenger awake if someone requested it.
5461 // retake P's blocked in syscalls
5462 // and preempt long running G's
5463 if retake(now) != 0 {
5468 // check if we need to force a GC
5469 if t := (gcTrigger{kind: gcTriggerTime, now: now}); t.test() && forcegc.idle.Load() {
5471 forcegc.idle.Store(false)
5473 list.push(forcegc.g)
5475 unlock(&forcegc.lock)
5477 if debug.schedtrace > 0 && lasttrace+int64(debug.schedtrace)*1000000 <= now {
5479 schedtrace(debug.scheddetail > 0)
5481 unlock(&sched.sysmonlock)
5485 type sysmontick struct {
5492 // forcePreemptNS is the time slice given to a G before it is
5494 const forcePreemptNS = 10 * 1000 * 1000 // 10ms
5496 func retake(now int64) uint32 {
5498 // Prevent allp slice changes. This lock will be completely
5499 // uncontended unless we're already stopping the world.
5501 // We can't use a range loop over allp because we may
5502 // temporarily drop the allpLock. Hence, we need to re-fetch
5503 // allp each time around the loop.
5504 for i := 0; i < len(allp); i++ {
5507 // This can happen if procresize has grown
5508 // allp but not yet created new Ps.
5511 pd := &pp.sysmontick
5514 if s == _Prunning || s == _Psyscall {
5515 // Preempt G if it's running for too long.
5516 t := int64(pp.schedtick)
5517 if int64(pd.schedtick) != t {
5518 pd.schedtick = uint32(t)
5520 } else if pd.schedwhen+forcePreemptNS <= now {
5522 // In case of syscall, preemptone() doesn't
5523 // work, because there is no M wired to P.
5528 // Retake P from syscall if it's there for more than 1 sysmon tick (at least 20us).
5529 t := int64(pp.syscalltick)
5530 if !sysretake && int64(pd.syscalltick) != t {
5531 pd.syscalltick = uint32(t)
5532 pd.syscallwhen = now
5535 // On the one hand we don't want to retake Ps if there is no other work to do,
5536 // but on the other hand we want to retake them eventually
5537 // because they can prevent the sysmon thread from deep sleep.
5538 if runqempty(pp) && sched.nmspinning.Load()+sched.npidle.Load() > 0 && pd.syscallwhen+10*1000*1000 > now {
5541 // Drop allpLock so we can take sched.lock.
5543 // Need to decrement number of idle locked M's
5544 // (pretending that one more is running) before the CAS.
5545 // Otherwise the M from which we retake can exit the syscall,
5546 // increment nmidle and report deadlock.
5548 if atomic.Cas(&pp.status, s, _Pidle) {
5565 // Tell all goroutines that they have been preempted and they should stop.
5566 // This function is purely best-effort. It can fail to inform a goroutine if a
5567 // processor just started running it.
5568 // No locks need to be held.
5569 // Returns true if preemption request was issued to at least one goroutine.
5570 func preemptall() bool {
5572 for _, pp := range allp {
5573 if pp.status != _Prunning {
5583 // Tell the goroutine running on processor P to stop.
5584 // This function is purely best-effort. It can incorrectly fail to inform the
5585 // goroutine. It can inform the wrong goroutine. Even if it informs the
5586 // correct goroutine, that goroutine might ignore the request if it is
5587 // simultaneously executing newstack.
5588 // No lock needs to be held.
5589 // Returns true if preemption request was issued.
5590 // The actual preemption will happen at some point in the future
5591 // and will be indicated by the gp->status no longer being
5593 func preemptone(pp *p) bool {
5595 if mp == nil || mp == getg().m {
5599 if gp == nil || gp == mp.g0 {
5605 // Every call in a goroutine checks for stack overflow by
5606 // comparing the current stack pointer to gp->stackguard0.
5607 // Setting gp->stackguard0 to StackPreempt folds
5608 // preemption into the normal stack overflow check.
5609 gp.stackguard0 = stackPreempt
5611 // Request an async preemption of this P.
5612 if preemptMSupported && debug.asyncpreemptoff == 0 {
5622 func schedtrace(detailed bool) {
5629 print("SCHED ", (now-starttime)/1e6, "ms: gomaxprocs=", gomaxprocs, " idleprocs=", sched.npidle.Load(), " threads=", mcount(), " spinningthreads=", sched.nmspinning.Load(), " needspinning=", sched.needspinning.Load(), " idlethreads=", sched.nmidle, " runqueue=", sched.runqsize)
5631 print(" gcwaiting=", sched.gcwaiting.Load(), " nmidlelocked=", sched.nmidlelocked, " stopwait=", sched.stopwait, " sysmonwait=", sched.sysmonwait.Load(), "\n")
5633 // We must be careful while reading data from P's, M's and G's.
5634 // Even if we hold schedlock, most data can be changed concurrently.
5635 // E.g. (p->m ? p->m->id : -1) can crash if p->m changes from non-nil to nil.
5636 for i, pp := range allp {
5638 h := atomic.Load(&pp.runqhead)
5639 t := atomic.Load(&pp.runqtail)
5641 print(" P", i, ": status=", pp.status, " schedtick=", pp.schedtick, " syscalltick=", pp.syscalltick, " m=")
5647 print(" runqsize=", t-h, " gfreecnt=", pp.gFree.n, " timerslen=", len(pp.timers), "\n")
5649 // In non-detailed mode format lengths of per-P run queues as:
5650 // [len1 len2 len3 len4]
5656 if i == len(allp)-1 {
5667 for mp := allm; mp != nil; mp = mp.alllink {
5669 print(" M", mp.id, ": p=")
5681 print(" mallocing=", mp.mallocing, " throwing=", mp.throwing, " preemptoff=", mp.preemptoff, " locks=", mp.locks, " dying=", mp.dying, " spinning=", mp.spinning, " blocked=", mp.blocked, " lockedg=")
5682 if lockedg := mp.lockedg.ptr(); lockedg != nil {
5690 forEachG(func(gp *g) {
5691 print(" G", gp.goid, ": status=", readgstatus(gp), "(", gp.waitreason.String(), ") m=")
5698 if lockedm := gp.lockedm.ptr(); lockedm != nil {
5708 // schedEnableUser enables or disables the scheduling of user
5711 // This does not stop already running user goroutines, so the caller
5712 // should first stop the world when disabling user goroutines.
5713 func schedEnableUser(enable bool) {
5715 if sched.disable.user == !enable {
5719 sched.disable.user = !enable
5721 n := sched.disable.n
5723 globrunqputbatch(&sched.disable.runnable, n)
5725 for ; n != 0 && sched.npidle.Load() != 0; n-- {
5733 // schedEnabled reports whether gp should be scheduled. It returns
5734 // false is scheduling of gp is disabled.
5736 // sched.lock must be held.
5737 func schedEnabled(gp *g) bool {
5738 assertLockHeld(&sched.lock)
5740 if sched.disable.user {
5741 return isSystemGoroutine(gp, true)
5746 // Put mp on midle list.
5747 // sched.lock must be held.
5748 // May run during STW, so write barriers are not allowed.
5750 //go:nowritebarrierrec
5752 assertLockHeld(&sched.lock)
5754 mp.schedlink = sched.midle
5760 // Try to get an m from midle list.
5761 // sched.lock must be held.
5762 // May run during STW, so write barriers are not allowed.
5764 //go:nowritebarrierrec
5766 assertLockHeld(&sched.lock)
5768 mp := sched.midle.ptr()
5770 sched.midle = mp.schedlink
5776 // Put gp on the global runnable queue.
5777 // sched.lock must be held.
5778 // May run during STW, so write barriers are not allowed.
5780 //go:nowritebarrierrec
5781 func globrunqput(gp *g) {
5782 assertLockHeld(&sched.lock)
5784 sched.runq.pushBack(gp)
5788 // Put gp at the head of the global runnable queue.
5789 // sched.lock must be held.
5790 // May run during STW, so write barriers are not allowed.
5792 //go:nowritebarrierrec
5793 func globrunqputhead(gp *g) {
5794 assertLockHeld(&sched.lock)
5800 // Put a batch of runnable goroutines on the global runnable queue.
5801 // This clears *batch.
5802 // sched.lock must be held.
5803 // May run during STW, so write barriers are not allowed.
5805 //go:nowritebarrierrec
5806 func globrunqputbatch(batch *gQueue, n int32) {
5807 assertLockHeld(&sched.lock)
5809 sched.runq.pushBackAll(*batch)
5814 // Try get a batch of G's from the global runnable queue.
5815 // sched.lock must be held.
5816 func globrunqget(pp *p, max int32) *g {
5817 assertLockHeld(&sched.lock)
5819 if sched.runqsize == 0 {
5823 n := sched.runqsize/gomaxprocs + 1
5824 if n > sched.runqsize {
5827 if max > 0 && n > max {
5830 if n > int32(len(pp.runq))/2 {
5831 n = int32(len(pp.runq)) / 2
5836 gp := sched.runq.pop()
5839 gp1 := sched.runq.pop()
5840 runqput(pp, gp1, false)
5845 // pMask is an atomic bitstring with one bit per P.
5848 // read returns true if P id's bit is set.
5849 func (p pMask) read(id uint32) bool {
5851 mask := uint32(1) << (id % 32)
5852 return (atomic.Load(&p[word]) & mask) != 0
5855 // set sets P id's bit.
5856 func (p pMask) set(id int32) {
5858 mask := uint32(1) << (id % 32)
5859 atomic.Or(&p[word], mask)
5862 // clear clears P id's bit.
5863 func (p pMask) clear(id int32) {
5865 mask := uint32(1) << (id % 32)
5866 atomic.And(&p[word], ^mask)
5869 // updateTimerPMask clears pp's timer mask if it has no timers on its heap.
5871 // Ideally, the timer mask would be kept immediately consistent on any timer
5872 // operations. Unfortunately, updating a shared global data structure in the
5873 // timer hot path adds too much overhead in applications frequently switching
5874 // between no timers and some timers.
5876 // As a compromise, the timer mask is updated only on pidleget / pidleput. A
5877 // running P (returned by pidleget) may add a timer at any time, so its mask
5878 // must be set. An idle P (passed to pidleput) cannot add new timers while
5879 // idle, so if it has no timers at that time, its mask may be cleared.
5881 // Thus, we get the following effects on timer-stealing in findrunnable:
5883 // - Idle Ps with no timers when they go idle are never checked in findrunnable
5884 // (for work- or timer-stealing; this is the ideal case).
5885 // - Running Ps must always be checked.
5886 // - Idle Ps whose timers are stolen must continue to be checked until they run
5887 // again, even after timer expiration.
5889 // When the P starts running again, the mask should be set, as a timer may be
5890 // added at any time.
5892 // TODO(prattmic): Additional targeted updates may improve the above cases.
5893 // e.g., updating the mask when stealing a timer.
5894 func updateTimerPMask(pp *p) {
5895 if pp.numTimers.Load() > 0 {
5899 // Looks like there are no timers, however another P may transiently
5900 // decrement numTimers when handling a timerModified timer in
5901 // checkTimers. We must take timersLock to serialize with these changes.
5902 lock(&pp.timersLock)
5903 if pp.numTimers.Load() == 0 {
5904 timerpMask.clear(pp.id)
5906 unlock(&pp.timersLock)
5909 // pidleput puts p on the _Pidle list. now must be a relatively recent call
5910 // to nanotime or zero. Returns now or the current time if now was zero.
5912 // This releases ownership of p. Once sched.lock is released it is no longer
5915 // sched.lock must be held.
5917 // May run during STW, so write barriers are not allowed.
5919 //go:nowritebarrierrec
5920 func pidleput(pp *p, now int64) int64 {
5921 assertLockHeld(&sched.lock)
5924 throw("pidleput: P has non-empty run queue")
5929 updateTimerPMask(pp) // clear if there are no timers.
5930 idlepMask.set(pp.id)
5931 pp.link = sched.pidle
5934 if !pp.limiterEvent.start(limiterEventIdle, now) {
5935 throw("must be able to track idle limiter event")
5940 // pidleget tries to get a p from the _Pidle list, acquiring ownership.
5942 // sched.lock must be held.
5944 // May run during STW, so write barriers are not allowed.
5946 //go:nowritebarrierrec
5947 func pidleget(now int64) (*p, int64) {
5948 assertLockHeld(&sched.lock)
5950 pp := sched.pidle.ptr()
5952 // Timer may get added at any time now.
5956 timerpMask.set(pp.id)
5957 idlepMask.clear(pp.id)
5958 sched.pidle = pp.link
5959 sched.npidle.Add(-1)
5960 pp.limiterEvent.stop(limiterEventIdle, now)
5965 // pidlegetSpinning tries to get a p from the _Pidle list, acquiring ownership.
5966 // This is called by spinning Ms (or callers than need a spinning M) that have
5967 // found work. If no P is available, this must synchronized with non-spinning
5968 // Ms that may be preparing to drop their P without discovering this work.
5970 // sched.lock must be held.
5972 // May run during STW, so write barriers are not allowed.
5974 //go:nowritebarrierrec
5975 func pidlegetSpinning(now int64) (*p, int64) {
5976 assertLockHeld(&sched.lock)
5978 pp, now := pidleget(now)
5980 // See "Delicate dance" comment in findrunnable. We found work
5981 // that we cannot take, we must synchronize with non-spinning
5982 // Ms that may be preparing to drop their P.
5983 sched.needspinning.Store(1)
5990 // runqempty reports whether pp has no Gs on its local run queue.
5991 // It never returns true spuriously.
5992 func runqempty(pp *p) bool {
5993 // Defend against a race where 1) pp has G1 in runqnext but runqhead == runqtail,
5994 // 2) runqput on pp kicks G1 to the runq, 3) runqget on pp empties runqnext.
5995 // Simply observing that runqhead == runqtail and then observing that runqnext == nil
5996 // does not mean the queue is empty.
5998 head := atomic.Load(&pp.runqhead)
5999 tail := atomic.Load(&pp.runqtail)
6000 runnext := atomic.Loaduintptr((*uintptr)(unsafe.Pointer(&pp.runnext)))
6001 if tail == atomic.Load(&pp.runqtail) {
6002 return head == tail && runnext == 0
6007 // To shake out latent assumptions about scheduling order,
6008 // we introduce some randomness into scheduling decisions
6009 // when running with the race detector.
6010 // The need for this was made obvious by changing the
6011 // (deterministic) scheduling order in Go 1.5 and breaking
6012 // many poorly-written tests.
6013 // With the randomness here, as long as the tests pass
6014 // consistently with -race, they shouldn't have latent scheduling
6016 const randomizeScheduler = raceenabled
6018 // runqput tries to put g on the local runnable queue.
6019 // If next is false, runqput adds g to the tail of the runnable queue.
6020 // If next is true, runqput puts g in the pp.runnext slot.
6021 // If the run queue is full, runnext puts g on the global queue.
6022 // Executed only by the owner P.
6023 func runqput(pp *p, gp *g, next bool) {
6024 if randomizeScheduler && next && fastrandn(2) == 0 {
6030 oldnext := pp.runnext
6031 if !pp.runnext.cas(oldnext, guintptr(unsafe.Pointer(gp))) {
6037 // Kick the old runnext out to the regular run queue.
6042 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with consumers
6044 if t-h < uint32(len(pp.runq)) {
6045 pp.runq[t%uint32(len(pp.runq))].set(gp)
6046 atomic.StoreRel(&pp.runqtail, t+1) // store-release, makes the item available for consumption
6049 if runqputslow(pp, gp, h, t) {
6052 // the queue is not full, now the put above must succeed
6056 // Put g and a batch of work from local runnable queue on global queue.
6057 // Executed only by the owner P.
6058 func runqputslow(pp *p, gp *g, h, t uint32) bool {
6059 var batch [len(pp.runq)/2 + 1]*g
6061 // First, grab a batch from local queue.
6064 if n != uint32(len(pp.runq)/2) {
6065 throw("runqputslow: queue is not full")
6067 for i := uint32(0); i < n; i++ {
6068 batch[i] = pp.runq[(h+i)%uint32(len(pp.runq))].ptr()
6070 if !atomic.CasRel(&pp.runqhead, h, h+n) { // cas-release, commits consume
6075 if randomizeScheduler {
6076 for i := uint32(1); i <= n; i++ {
6077 j := fastrandn(i + 1)
6078 batch[i], batch[j] = batch[j], batch[i]
6082 // Link the goroutines.
6083 for i := uint32(0); i < n; i++ {
6084 batch[i].schedlink.set(batch[i+1])
6087 q.head.set(batch[0])
6088 q.tail.set(batch[n])
6090 // Now put the batch on global queue.
6092 globrunqputbatch(&q, int32(n+1))
6097 // runqputbatch tries to put all the G's on q on the local runnable queue.
6098 // If the queue is full, they are put on the global queue; in that case
6099 // this will temporarily acquire the scheduler lock.
6100 // Executed only by the owner P.
6101 func runqputbatch(pp *p, q *gQueue, qsize int) {
6102 h := atomic.LoadAcq(&pp.runqhead)
6105 for !q.empty() && t-h < uint32(len(pp.runq)) {
6107 pp.runq[t%uint32(len(pp.runq))].set(gp)
6113 if randomizeScheduler {
6114 off := func(o uint32) uint32 {
6115 return (pp.runqtail + o) % uint32(len(pp.runq))
6117 for i := uint32(1); i < n; i++ {
6118 j := fastrandn(i + 1)
6119 pp.runq[off(i)], pp.runq[off(j)] = pp.runq[off(j)], pp.runq[off(i)]
6123 atomic.StoreRel(&pp.runqtail, t)
6126 globrunqputbatch(q, int32(qsize))
6131 // Get g from local runnable queue.
6132 // If inheritTime is true, gp should inherit the remaining time in the
6133 // current time slice. Otherwise, it should start a new time slice.
6134 // Executed only by the owner P.
6135 func runqget(pp *p) (gp *g, inheritTime bool) {
6136 // If there's a runnext, it's the next G to run.
6138 // If the runnext is non-0 and the CAS fails, it could only have been stolen by another P,
6139 // because other Ps can race to set runnext to 0, but only the current P can set it to non-0.
6140 // Hence, there's no need to retry this CAS if it fails.
6141 if next != 0 && pp.runnext.cas(next, 0) {
6142 return next.ptr(), true
6146 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with other consumers
6151 gp := pp.runq[h%uint32(len(pp.runq))].ptr()
6152 if atomic.CasRel(&pp.runqhead, h, h+1) { // cas-release, commits consume
6158 // runqdrain drains the local runnable queue of pp and returns all goroutines in it.
6159 // Executed only by the owner P.
6160 func runqdrain(pp *p) (drainQ gQueue, n uint32) {
6161 oldNext := pp.runnext
6162 if oldNext != 0 && pp.runnext.cas(oldNext, 0) {
6163 drainQ.pushBack(oldNext.ptr())
6168 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with other consumers
6174 if qn > uint32(len(pp.runq)) { // read inconsistent h and t
6178 if !atomic.CasRel(&pp.runqhead, h, h+qn) { // cas-release, commits consume
6182 // We've inverted the order in which it gets G's from the local P's runnable queue
6183 // and then advances the head pointer because we don't want to mess up the statuses of G's
6184 // while runqdrain() and runqsteal() are running in parallel.
6185 // Thus we should advance the head pointer before draining the local P into a gQueue,
6186 // so that we can update any gp.schedlink only after we take the full ownership of G,
6187 // meanwhile, other P's can't access to all G's in local P's runnable queue and steal them.
6188 // See https://groups.google.com/g/golang-dev/c/0pTKxEKhHSc/m/6Q85QjdVBQAJ for more details.
6189 for i := uint32(0); i < qn; i++ {
6190 gp := pp.runq[(h+i)%uint32(len(pp.runq))].ptr()
6197 // Grabs a batch of goroutines from pp's runnable queue into batch.
6198 // Batch is a ring buffer starting at batchHead.
6199 // Returns number of grabbed goroutines.
6200 // Can be executed by any P.
6201 func runqgrab(pp *p, batch *[256]guintptr, batchHead uint32, stealRunNextG bool) uint32 {
6203 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with other consumers
6204 t := atomic.LoadAcq(&pp.runqtail) // load-acquire, synchronize with the producer
6209 // Try to steal from pp.runnext.
6210 if next := pp.runnext; next != 0 {
6211 if pp.status == _Prunning {
6212 // Sleep to ensure that pp isn't about to run the g
6213 // we are about to steal.
6214 // The important use case here is when the g running
6215 // on pp ready()s another g and then almost
6216 // immediately blocks. Instead of stealing runnext
6217 // in this window, back off to give pp a chance to
6218 // schedule runnext. This will avoid thrashing gs
6219 // between different Ps.
6220 // A sync chan send/recv takes ~50ns as of time of
6221 // writing, so 3us gives ~50x overshoot.
6222 if GOOS != "windows" && GOOS != "openbsd" && GOOS != "netbsd" {
6225 // On some platforms system timer granularity is
6226 // 1-15ms, which is way too much for this
6227 // optimization. So just yield.
6231 if !pp.runnext.cas(next, 0) {
6234 batch[batchHead%uint32(len(batch))] = next
6240 if n > uint32(len(pp.runq)/2) { // read inconsistent h and t
6243 for i := uint32(0); i < n; i++ {
6244 g := pp.runq[(h+i)%uint32(len(pp.runq))]
6245 batch[(batchHead+i)%uint32(len(batch))] = g
6247 if atomic.CasRel(&pp.runqhead, h, h+n) { // cas-release, commits consume
6253 // Steal half of elements from local runnable queue of p2
6254 // and put onto local runnable queue of p.
6255 // Returns one of the stolen elements (or nil if failed).
6256 func runqsteal(pp, p2 *p, stealRunNextG bool) *g {
6258 n := runqgrab(p2, &pp.runq, t, stealRunNextG)
6263 gp := pp.runq[(t+n)%uint32(len(pp.runq))].ptr()
6267 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with consumers
6268 if t-h+n >= uint32(len(pp.runq)) {
6269 throw("runqsteal: runq overflow")
6271 atomic.StoreRel(&pp.runqtail, t+n) // store-release, makes the item available for consumption
6275 // A gQueue is a dequeue of Gs linked through g.schedlink. A G can only
6276 // be on one gQueue or gList at a time.
6277 type gQueue struct {
6282 // empty reports whether q is empty.
6283 func (q *gQueue) empty() bool {
6287 // push adds gp to the head of q.
6288 func (q *gQueue) push(gp *g) {
6289 gp.schedlink = q.head
6296 // pushBack adds gp to the tail of q.
6297 func (q *gQueue) pushBack(gp *g) {
6300 q.tail.ptr().schedlink.set(gp)
6307 // pushBackAll adds all Gs in q2 to the tail of q. After this q2 must
6309 func (q *gQueue) pushBackAll(q2 gQueue) {
6313 q2.tail.ptr().schedlink = 0
6315 q.tail.ptr().schedlink = q2.head
6322 // pop removes and returns the head of queue q. It returns nil if
6324 func (q *gQueue) pop() *g {
6327 q.head = gp.schedlink
6335 // popList takes all Gs in q and returns them as a gList.
6336 func (q *gQueue) popList() gList {
6337 stack := gList{q.head}
6342 // A gList is a list of Gs linked through g.schedlink. A G can only be
6343 // on one gQueue or gList at a time.
6348 // empty reports whether l is empty.
6349 func (l *gList) empty() bool {
6353 // push adds gp to the head of l.
6354 func (l *gList) push(gp *g) {
6355 gp.schedlink = l.head
6359 // pushAll prepends all Gs in q to l.
6360 func (l *gList) pushAll(q gQueue) {
6362 q.tail.ptr().schedlink = l.head
6367 // pop removes and returns the head of l. If l is empty, it returns nil.
6368 func (l *gList) pop() *g {
6371 l.head = gp.schedlink
6376 //go:linkname setMaxThreads runtime/debug.setMaxThreads
6377 func setMaxThreads(in int) (out int) {
6379 out = int(sched.maxmcount)
6380 if in > 0x7fffffff { // MaxInt32
6381 sched.maxmcount = 0x7fffffff
6383 sched.maxmcount = int32(in)
6391 func procPin() int {
6396 return int(mp.p.ptr().id)
6405 //go:linkname sync_runtime_procPin sync.runtime_procPin
6407 func sync_runtime_procPin() int {
6411 //go:linkname sync_runtime_procUnpin sync.runtime_procUnpin
6413 func sync_runtime_procUnpin() {
6417 //go:linkname sync_atomic_runtime_procPin sync/atomic.runtime_procPin
6419 func sync_atomic_runtime_procPin() int {
6423 //go:linkname sync_atomic_runtime_procUnpin sync/atomic.runtime_procUnpin
6425 func sync_atomic_runtime_procUnpin() {
6429 // Active spinning for sync.Mutex.
6431 //go:linkname sync_runtime_canSpin sync.runtime_canSpin
6433 func sync_runtime_canSpin(i int) bool {
6434 // sync.Mutex is cooperative, so we are conservative with spinning.
6435 // Spin only few times and only if running on a multicore machine and
6436 // GOMAXPROCS>1 and there is at least one other running P and local runq is empty.
6437 // As opposed to runtime mutex we don't do passive spinning here,
6438 // because there can be work on global runq or on other Ps.
6439 if i >= active_spin || ncpu <= 1 || gomaxprocs <= sched.npidle.Load()+sched.nmspinning.Load()+1 {
6442 if p := getg().m.p.ptr(); !runqempty(p) {
6448 //go:linkname sync_runtime_doSpin sync.runtime_doSpin
6450 func sync_runtime_doSpin() {
6451 procyield(active_spin_cnt)
6454 var stealOrder randomOrder
6456 // randomOrder/randomEnum are helper types for randomized work stealing.
6457 // They allow to enumerate all Ps in different pseudo-random orders without repetitions.
6458 // The algorithm is based on the fact that if we have X such that X and GOMAXPROCS
6459 // are coprime, then a sequences of (i + X) % GOMAXPROCS gives the required enumeration.
6460 type randomOrder struct {
6465 type randomEnum struct {
6472 func (ord *randomOrder) reset(count uint32) {
6474 ord.coprimes = ord.coprimes[:0]
6475 for i := uint32(1); i <= count; i++ {
6476 if gcd(i, count) == 1 {
6477 ord.coprimes = append(ord.coprimes, i)
6482 func (ord *randomOrder) start(i uint32) randomEnum {
6486 inc: ord.coprimes[i/ord.count%uint32(len(ord.coprimes))],
6490 func (enum *randomEnum) done() bool {
6491 return enum.i == enum.count
6494 func (enum *randomEnum) next() {
6496 enum.pos = (enum.pos + enum.inc) % enum.count
6499 func (enum *randomEnum) position() uint32 {
6503 func gcd(a, b uint32) uint32 {
6510 // An initTask represents the set of initializations that need to be done for a package.
6511 // Keep in sync with ../../test/initempty.go:initTask
6512 type initTask struct {
6513 // TODO: pack the first 3 fields more tightly?
6514 state uintptr // 0 = uninitialized, 1 = in progress, 2 = done
6517 // followed by ndeps instances of an *initTask, one per package depended on
6518 // followed by nfns pcs, one per init function to run
6521 // inittrace stores statistics for init functions which are
6522 // updated by malloc and newproc when active is true.
6523 var inittrace tracestat
6525 type tracestat struct {
6526 active bool // init tracing activation status
6527 id uint64 // init goroutine id
6528 allocs uint64 // heap allocations
6529 bytes uint64 // heap allocated bytes
6532 func doInit(t *initTask) {
6534 case 2: // fully initialized
6536 case 1: // initialization in progress
6537 throw("recursive call during initialization - linker skew")
6538 default: // not initialized yet
6539 t.state = 1 // initialization in progress
6541 for i := uintptr(0); i < t.ndeps; i++ {
6542 p := add(unsafe.Pointer(t), (3+i)*goarch.PtrSize)
6543 t2 := *(**initTask)(p)
6548 t.state = 2 // initialization done
6557 if inittrace.active {
6559 // Load stats non-atomically since tracinit is updated only by this init goroutine.
6563 firstFunc := add(unsafe.Pointer(t), (3+t.ndeps)*goarch.PtrSize)
6564 for i := uintptr(0); i < t.nfns; i++ {
6565 p := add(firstFunc, i*goarch.PtrSize)
6566 f := *(*func())(unsafe.Pointer(&p))
6570 if inittrace.active {
6572 // Load stats non-atomically since tracinit is updated only by this init goroutine.
6575 f := *(*func())(unsafe.Pointer(&firstFunc))
6576 pkg := funcpkgpath(findfunc(abi.FuncPCABIInternal(f)))
6579 print("init ", pkg, " @")
6580 print(string(fmtNSAsMS(sbuf[:], uint64(start-runtimeInitTime))), " ms, ")
6581 print(string(fmtNSAsMS(sbuf[:], uint64(end-start))), " ms clock, ")
6582 print(string(itoa(sbuf[:], after.bytes-before.bytes)), " bytes, ")
6583 print(string(itoa(sbuf[:], after.allocs-before.allocs)), " allocs")
6587 t.state = 2 // initialization done