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 var fastrandseed uintptr
867 func fastrandinit() {
868 s := (*[unsafe.Sizeof(fastrandseed)]byte)(unsafe.Pointer(&fastrandseed))[:]
872 // Mark gp ready to run.
873 func ready(gp *g, traceskip int, next bool) {
875 traceGoUnpark(gp, traceskip)
878 status := readgstatus(gp)
881 mp := acquirem() // disable preemption because it can be holding p in a local var
882 if status&^_Gscan != _Gwaiting {
884 throw("bad g->status in ready")
887 // status is Gwaiting or Gscanwaiting, make Grunnable and put on runq
888 casgstatus(gp, _Gwaiting, _Grunnable)
889 runqput(mp.p.ptr(), gp, next)
894 // freezeStopWait is a large value that freezetheworld sets
895 // sched.stopwait to in order to request that all Gs permanently stop.
896 const freezeStopWait = 0x7fffffff
898 // freezing is set to non-zero if the runtime is trying to freeze the
900 var freezing atomic.Bool
902 // Similar to stopTheWorld but best-effort and can be called several times.
903 // There is no reverse operation, used during crashing.
904 // This function must not lock any mutexes.
905 func freezetheworld() {
907 // stopwait and preemption requests can be lost
908 // due to races with concurrently executing threads,
909 // so try several times
910 for i := 0; i < 5; i++ {
911 // this should tell the scheduler to not start any new goroutines
912 sched.stopwait = freezeStopWait
913 sched.gcwaiting.Store(true)
914 // this should stop running goroutines
916 break // no running goroutines
926 // All reads and writes of g's status go through readgstatus, casgstatus
927 // castogscanstatus, casfrom_Gscanstatus.
930 func readgstatus(gp *g) uint32 {
931 return gp.atomicstatus.Load()
934 // The Gscanstatuses are acting like locks and this releases them.
935 // If it proves to be a performance hit we should be able to make these
936 // simple atomic stores but for now we are going to throw if
937 // we see an inconsistent state.
938 func casfrom_Gscanstatus(gp *g, oldval, newval uint32) {
941 // Check that transition is valid.
944 print("runtime: casfrom_Gscanstatus bad oldval gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n")
946 throw("casfrom_Gscanstatus:top gp->status is not in scan state")
952 if newval == oldval&^_Gscan {
953 success = gp.atomicstatus.CompareAndSwap(oldval, newval)
957 print("runtime: casfrom_Gscanstatus failed gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n")
959 throw("casfrom_Gscanstatus: gp->status is not in scan state")
961 releaseLockRank(lockRankGscan)
964 // This will return false if the gp is not in the expected status and the cas fails.
965 // This acts like a lock acquire while the casfromgstatus acts like a lock release.
966 func castogscanstatus(gp *g, oldval, newval uint32) bool {
972 if newval == oldval|_Gscan {
973 r := gp.atomicstatus.CompareAndSwap(oldval, newval)
975 acquireLockRank(lockRankGscan)
981 print("runtime: castogscanstatus oldval=", hex(oldval), " newval=", hex(newval), "\n")
982 throw("castogscanstatus")
986 // casgstatusAlwaysTrack is a debug flag that causes casgstatus to always track
987 // various latencies on every transition instead of sampling them.
988 var casgstatusAlwaysTrack = false
990 // If asked to move to or from a Gscanstatus this will throw. Use the castogscanstatus
991 // and casfrom_Gscanstatus instead.
992 // casgstatus will loop if the g->atomicstatus is in a Gscan status until the routine that
993 // put it in the Gscan state is finished.
996 func casgstatus(gp *g, oldval, newval uint32) {
997 if (oldval&_Gscan != 0) || (newval&_Gscan != 0) || oldval == newval {
999 print("runtime: casgstatus: oldval=", hex(oldval), " newval=", hex(newval), "\n")
1000 throw("casgstatus: bad incoming values")
1004 acquireLockRank(lockRankGscan)
1005 releaseLockRank(lockRankGscan)
1007 // See https://golang.org/cl/21503 for justification of the yield delay.
1008 const yieldDelay = 5 * 1000
1011 // loop if gp->atomicstatus is in a scan state giving
1012 // GC time to finish and change the state to oldval.
1013 for i := 0; !gp.atomicstatus.CompareAndSwap(oldval, newval); i++ {
1014 if oldval == _Gwaiting && gp.atomicstatus.Load() == _Grunnable {
1015 throw("casgstatus: waiting for Gwaiting but is Grunnable")
1018 nextYield = nanotime() + yieldDelay
1020 if nanotime() < nextYield {
1021 for x := 0; x < 10 && gp.atomicstatus.Load() != oldval; x++ {
1026 nextYield = nanotime() + yieldDelay/2
1030 if oldval == _Grunning {
1031 // Track every gTrackingPeriod time a goroutine transitions out of running.
1032 if casgstatusAlwaysTrack || gp.trackingSeq%gTrackingPeriod == 0 {
1041 // Handle various kinds of tracking.
1044 // - Time spent in runnable.
1045 // - Time spent blocked on a sync.Mutex or sync.RWMutex.
1048 // We transitioned out of runnable, so measure how much
1049 // time we spent in this state and add it to
1052 gp.runnableTime += now - gp.trackingStamp
1053 gp.trackingStamp = 0
1055 if !gp.waitreason.isMutexWait() {
1056 // Not blocking on a lock.
1059 // Blocking on a lock, measure it. Note that because we're
1060 // sampling, we have to multiply by our sampling period to get
1061 // a more representative estimate of the absolute value.
1062 // gTrackingPeriod also represents an accurate sampling period
1063 // because we can only enter this state from _Grunning.
1065 sched.totalMutexWaitTime.Add((now - gp.trackingStamp) * gTrackingPeriod)
1066 gp.trackingStamp = 0
1070 if !gp.waitreason.isMutexWait() {
1071 // Not blocking on a lock.
1074 // Blocking on a lock. Write down the timestamp.
1076 gp.trackingStamp = now
1078 // We just transitioned into runnable, so record what
1079 // time that happened.
1081 gp.trackingStamp = now
1083 // We're transitioning into running, so turn off
1084 // tracking and record how much time we spent in
1087 sched.timeToRun.record(gp.runnableTime)
1092 // casGToWaiting transitions gp from old to _Gwaiting, and sets the wait reason.
1094 // Use this over casgstatus when possible to ensure that a waitreason is set.
1095 func casGToWaiting(gp *g, old uint32, reason waitReason) {
1096 // Set the wait reason before calling casgstatus, because casgstatus will use it.
1097 gp.waitreason = reason
1098 casgstatus(gp, old, _Gwaiting)
1101 // casgstatus(gp, oldstatus, Gcopystack), assuming oldstatus is Gwaiting or Grunnable.
1102 // Returns old status. Cannot call casgstatus directly, because we are racing with an
1103 // async wakeup that might come in from netpoll. If we see Gwaiting from the readgstatus,
1104 // it might have become Grunnable by the time we get to the cas. If we called casgstatus,
1105 // it would loop waiting for the status to go back to Gwaiting, which it never will.
1108 func casgcopystack(gp *g) uint32 {
1110 oldstatus := readgstatus(gp) &^ _Gscan
1111 if oldstatus != _Gwaiting && oldstatus != _Grunnable {
1112 throw("copystack: bad status, not Gwaiting or Grunnable")
1114 if gp.atomicstatus.CompareAndSwap(oldstatus, _Gcopystack) {
1120 // casGToPreemptScan transitions gp from _Grunning to _Gscan|_Gpreempted.
1122 // TODO(austin): This is the only status operation that both changes
1123 // the status and locks the _Gscan bit. Rethink this.
1124 func casGToPreemptScan(gp *g, old, new uint32) {
1125 if old != _Grunning || new != _Gscan|_Gpreempted {
1126 throw("bad g transition")
1128 acquireLockRank(lockRankGscan)
1129 for !gp.atomicstatus.CompareAndSwap(_Grunning, _Gscan|_Gpreempted) {
1133 // casGFromPreempted attempts to transition gp from _Gpreempted to
1134 // _Gwaiting. If successful, the caller is responsible for
1135 // re-scheduling gp.
1136 func casGFromPreempted(gp *g, old, new uint32) bool {
1137 if old != _Gpreempted || new != _Gwaiting {
1138 throw("bad g transition")
1140 gp.waitreason = waitReasonPreempted
1141 return gp.atomicstatus.CompareAndSwap(_Gpreempted, _Gwaiting)
1144 // stopTheWorld stops all P's from executing goroutines, interrupting
1145 // all goroutines at GC safe points and records reason as the reason
1146 // for the stop. On return, only the current goroutine's P is running.
1147 // stopTheWorld must not be called from a system stack and the caller
1148 // must not hold worldsema. The caller must call startTheWorld when
1149 // other P's should resume execution.
1151 // stopTheWorld is safe for multiple goroutines to call at the
1152 // same time. Each will execute its own stop, and the stops will
1155 // This is also used by routines that do stack dumps. If the system is
1156 // in panic or being exited, this may not reliably stop all
1158 func stopTheWorld(reason string) {
1159 semacquire(&worldsema)
1161 gp.m.preemptoff = reason
1162 systemstack(func() {
1163 // Mark the goroutine which called stopTheWorld preemptible so its
1164 // stack may be scanned.
1165 // This lets a mark worker scan us while we try to stop the world
1166 // since otherwise we could get in a mutual preemption deadlock.
1167 // We must not modify anything on the G stack because a stack shrink
1168 // may occur. A stack shrink is otherwise OK though because in order
1169 // to return from this function (and to leave the system stack) we
1170 // must have preempted all goroutines, including any attempting
1171 // to scan our stack, in which case, any stack shrinking will
1172 // have already completed by the time we exit.
1173 // Don't provide a wait reason because we're still executing.
1174 casGToWaiting(gp, _Grunning, waitReasonStoppingTheWorld)
1175 stopTheWorldWithSema()
1176 casgstatus(gp, _Gwaiting, _Grunning)
1180 // startTheWorld undoes the effects of stopTheWorld.
1181 func startTheWorld() {
1182 systemstack(func() { startTheWorldWithSema(false) })
1184 // worldsema must be held over startTheWorldWithSema to ensure
1185 // gomaxprocs cannot change while worldsema is held.
1187 // Release worldsema with direct handoff to the next waiter, but
1188 // acquirem so that semrelease1 doesn't try to yield our time.
1190 // Otherwise if e.g. ReadMemStats is being called in a loop,
1191 // it might stomp on other attempts to stop the world, such as
1192 // for starting or ending GC. The operation this blocks is
1193 // so heavy-weight that we should just try to be as fair as
1196 // We don't want to just allow us to get preempted between now
1197 // and releasing the semaphore because then we keep everyone
1198 // (including, for example, GCs) waiting longer.
1201 semrelease1(&worldsema, true, 0)
1205 // stopTheWorldGC has the same effect as stopTheWorld, but blocks
1206 // until the GC is not running. It also blocks a GC from starting
1207 // until startTheWorldGC is called.
1208 func stopTheWorldGC(reason string) {
1210 stopTheWorld(reason)
1213 // startTheWorldGC undoes the effects of stopTheWorldGC.
1214 func startTheWorldGC() {
1219 // Holding worldsema grants an M the right to try to stop the world.
1220 var worldsema uint32 = 1
1222 // Holding gcsema grants the M the right to block a GC, and blocks
1223 // until the current GC is done. In particular, it prevents gomaxprocs
1224 // from changing concurrently.
1226 // TODO(mknyszek): Once gomaxprocs and the execution tracer can handle
1227 // being changed/enabled during a GC, remove this.
1228 var gcsema uint32 = 1
1230 // stopTheWorldWithSema is the core implementation of stopTheWorld.
1231 // The caller is responsible for acquiring worldsema and disabling
1232 // preemption first and then should stopTheWorldWithSema on the system
1235 // semacquire(&worldsema, 0)
1236 // m.preemptoff = "reason"
1237 // systemstack(stopTheWorldWithSema)
1239 // When finished, the caller must either call startTheWorld or undo
1240 // these three operations separately:
1242 // m.preemptoff = ""
1243 // systemstack(startTheWorldWithSema)
1244 // semrelease(&worldsema)
1246 // It is allowed to acquire worldsema once and then execute multiple
1247 // startTheWorldWithSema/stopTheWorldWithSema pairs.
1248 // Other P's are able to execute between successive calls to
1249 // startTheWorldWithSema and stopTheWorldWithSema.
1250 // Holding worldsema causes any other goroutines invoking
1251 // stopTheWorld to block.
1252 func stopTheWorldWithSema() {
1255 // If we hold a lock, then we won't be able to stop another M
1256 // that is blocked trying to acquire the lock.
1258 throw("stopTheWorld: holding locks")
1262 sched.stopwait = gomaxprocs
1263 sched.gcwaiting.Store(true)
1266 gp.m.p.ptr().status = _Pgcstop // Pgcstop is only diagnostic.
1268 // try to retake all P's in Psyscall status
1269 for _, pp := range allp {
1271 if s == _Psyscall && atomic.Cas(&pp.status, s, _Pgcstop) {
1283 pp, _ := pidleget(now)
1287 pp.status = _Pgcstop
1290 wait := sched.stopwait > 0
1293 // wait for remaining P's to stop voluntarily
1296 // wait for 100us, then try to re-preempt in case of any races
1297 if notetsleep(&sched.stopnote, 100*1000) {
1298 noteclear(&sched.stopnote)
1307 if sched.stopwait != 0 {
1308 bad = "stopTheWorld: not stopped (stopwait != 0)"
1310 for _, pp := range allp {
1311 if pp.status != _Pgcstop {
1312 bad = "stopTheWorld: not stopped (status != _Pgcstop)"
1316 if freezing.Load() {
1317 // Some other thread is panicking. This can cause the
1318 // sanity checks above to fail if the panic happens in
1319 // the signal handler on a stopped thread. Either way,
1320 // we should halt this thread.
1331 func startTheWorldWithSema(emitTraceEvent bool) int64 {
1332 assertWorldStopped()
1334 mp := acquirem() // disable preemption because it can be holding p in a local var
1335 if netpollinited() {
1336 list := netpoll(0) // non-blocking
1346 p1 := procresize(procs)
1347 sched.gcwaiting.Store(false)
1348 if sched.sysmonwait.Load() {
1349 sched.sysmonwait.Store(false)
1350 notewakeup(&sched.sysmonnote)
1363 throw("startTheWorld: inconsistent mp->nextp")
1366 notewakeup(&mp.park)
1368 // Start M to run P. Do not start another M below.
1373 // Capture start-the-world time before doing clean-up tasks.
1374 startTime := nanotime()
1379 // Wakeup an additional proc in case we have excessive runnable goroutines
1380 // in local queues or in the global queue. If we don't, the proc will park itself.
1381 // If we have lots of excessive work, resetspinning will unpark additional procs as necessary.
1389 // usesLibcall indicates whether this runtime performs system calls
1391 func usesLibcall() bool {
1393 case "aix", "darwin", "illumos", "ios", "solaris", "windows":
1396 return GOARCH == "386" || GOARCH == "amd64" || GOARCH == "arm" || GOARCH == "arm64"
1401 // mStackIsSystemAllocated indicates whether this runtime starts on a
1402 // system-allocated stack.
1403 func mStackIsSystemAllocated() bool {
1405 case "aix", "darwin", "plan9", "illumos", "ios", "solaris", "windows":
1409 case "386", "amd64", "arm", "arm64":
1416 // mstart is the entry-point for new Ms.
1417 // It is written in assembly, uses ABI0, is marked TOPFRAME, and calls mstart0.
1420 // mstart0 is the Go entry-point for new Ms.
1421 // This must not split the stack because we may not even have stack
1422 // bounds set up yet.
1424 // May run during STW (because it doesn't have a P yet), so write
1425 // barriers are not allowed.
1428 //go:nowritebarrierrec
1432 osStack := gp.stack.lo == 0
1434 // Initialize stack bounds from system stack.
1435 // Cgo may have left stack size in stack.hi.
1436 // minit may update the stack bounds.
1438 // Note: these bounds may not be very accurate.
1439 // We set hi to &size, but there are things above
1440 // it. The 1024 is supposed to compensate this,
1441 // but is somewhat arbitrary.
1444 size = 8192 * sys.StackGuardMultiplier
1446 gp.stack.hi = uintptr(noescape(unsafe.Pointer(&size)))
1447 gp.stack.lo = gp.stack.hi - size + 1024
1449 // Initialize stack guard so that we can start calling regular
1451 gp.stackguard0 = gp.stack.lo + _StackGuard
1452 // This is the g0, so we can also call go:systemstack
1453 // functions, which check stackguard1.
1454 gp.stackguard1 = gp.stackguard0
1457 // Exit this thread.
1458 if mStackIsSystemAllocated() {
1459 // Windows, Solaris, illumos, Darwin, AIX and Plan 9 always system-allocate
1460 // the stack, but put it in gp.stack before mstart,
1461 // so the logic above hasn't set osStack yet.
1467 // The go:noinline is to guarantee the getcallerpc/getcallersp below are safe,
1468 // so that we can set up g0.sched to return to the call of mstart1 above.
1475 throw("bad runtime·mstart")
1478 // Set up m.g0.sched as a label returning to just
1479 // after the mstart1 call in mstart0 above, for use by goexit0 and mcall.
1480 // We're never coming back to mstart1 after we call schedule,
1481 // so other calls can reuse the current frame.
1482 // And goexit0 does a gogo that needs to return from mstart1
1483 // and let mstart0 exit the thread.
1484 gp.sched.g = guintptr(unsafe.Pointer(gp))
1485 gp.sched.pc = getcallerpc()
1486 gp.sched.sp = getcallersp()
1491 // Install signal handlers; after minit so that minit can
1492 // prepare the thread to be able to handle the signals.
1497 if fn := gp.m.mstartfn; fn != nil {
1502 acquirep(gp.m.nextp.ptr())
1508 // mstartm0 implements part of mstart1 that only runs on the m0.
1510 // Write barriers are allowed here because we know the GC can't be
1511 // running yet, so they'll be no-ops.
1513 //go:yeswritebarrierrec
1515 // Create an extra M for callbacks on threads not created by Go.
1516 // An extra M is also needed on Windows for callbacks created by
1517 // syscall.NewCallback. See issue #6751 for details.
1518 if (iscgo || GOOS == "windows") && !cgoHasExtraM {
1525 // mPark causes a thread to park itself, returning once woken.
1530 notesleep(&gp.m.park)
1531 noteclear(&gp.m.park)
1534 // mexit tears down and exits the current thread.
1536 // Don't call this directly to exit the thread, since it must run at
1537 // the top of the thread stack. Instead, use gogo(&gp.m.g0.sched) to
1538 // unwind the stack to the point that exits the thread.
1540 // It is entered with m.p != nil, so write barriers are allowed. It
1541 // will release the P before exiting.
1543 //go:yeswritebarrierrec
1544 func mexit(osStack bool) {
1548 // This is the main thread. Just wedge it.
1550 // On Linux, exiting the main thread puts the process
1551 // into a non-waitable zombie state. On Plan 9,
1552 // exiting the main thread unblocks wait even though
1553 // other threads are still running. On Solaris we can
1554 // neither exitThread nor return from mstart. Other
1555 // bad things probably happen on other platforms.
1557 // We could try to clean up this M more before wedging
1558 // it, but that complicates signal handling.
1559 handoffp(releasep())
1565 throw("locked m0 woke up")
1571 // Free the gsignal stack.
1572 if mp.gsignal != nil {
1573 stackfree(mp.gsignal.stack)
1574 // On some platforms, when calling into VDSO (e.g. nanotime)
1575 // we store our g on the gsignal stack, if there is one.
1576 // Now the stack is freed, unlink it from the m, so we
1577 // won't write to it when calling VDSO code.
1581 // Remove m from allm.
1583 for pprev := &allm; *pprev != nil; pprev = &(*pprev).alllink {
1589 throw("m not found in allm")
1591 // Delay reaping m until it's done with the stack.
1593 // Put mp on the free list, though it will not be reaped while freeWait
1594 // is freeMWait. mp is no longer reachable via allm, so even if it is
1595 // on an OS stack, we must keep a reference to mp alive so that the GC
1596 // doesn't free mp while we are still using it.
1598 // Note that the free list must not be linked through alllink because
1599 // some functions walk allm without locking, so may be using alllink.
1600 mp.freeWait.Store(freeMWait)
1601 mp.freelink = sched.freem
1605 atomic.Xadd64(&ncgocall, int64(mp.ncgocall))
1608 handoffp(releasep())
1609 // After this point we must not have write barriers.
1611 // Invoke the deadlock detector. This must happen after
1612 // handoffp because it may have started a new M to take our
1619 if GOOS == "darwin" || GOOS == "ios" {
1620 // Make sure pendingPreemptSignals is correct when an M exits.
1622 if mp.signalPending.Load() != 0 {
1623 pendingPreemptSignals.Add(-1)
1627 // Destroy all allocated resources. After this is called, we may no
1628 // longer take any locks.
1632 // No more uses of mp, so it is safe to drop the reference.
1633 mp.freeWait.Store(freeMRef)
1635 // Return from mstart and let the system thread
1636 // library free the g0 stack and terminate the thread.
1640 // mstart is the thread's entry point, so there's nothing to
1641 // return to. Exit the thread directly. exitThread will clear
1642 // m.freeWait when it's done with the stack and the m can be
1644 exitThread(&mp.freeWait)
1647 // forEachP calls fn(p) for every P p when p reaches a GC safe point.
1648 // If a P is currently executing code, this will bring the P to a GC
1649 // safe point and execute fn on that P. If the P is not executing code
1650 // (it is idle or in a syscall), this will call fn(p) directly while
1651 // preventing the P from exiting its state. This does not ensure that
1652 // fn will run on every CPU executing Go code, but it acts as a global
1653 // memory barrier. GC uses this as a "ragged barrier."
1655 // The caller must hold worldsema.
1658 func forEachP(fn func(*p)) {
1660 pp := getg().m.p.ptr()
1663 if sched.safePointWait != 0 {
1664 throw("forEachP: sched.safePointWait != 0")
1666 sched.safePointWait = gomaxprocs - 1
1667 sched.safePointFn = fn
1669 // Ask all Ps to run the safe point function.
1670 for _, p2 := range allp {
1672 atomic.Store(&p2.runSafePointFn, 1)
1677 // Any P entering _Pidle or _Psyscall from now on will observe
1678 // p.runSafePointFn == 1 and will call runSafePointFn when
1679 // changing its status to _Pidle/_Psyscall.
1681 // Run safe point function for all idle Ps. sched.pidle will
1682 // not change because we hold sched.lock.
1683 for p := sched.pidle.ptr(); p != nil; p = p.link.ptr() {
1684 if atomic.Cas(&p.runSafePointFn, 1, 0) {
1686 sched.safePointWait--
1690 wait := sched.safePointWait > 0
1693 // Run fn for the current P.
1696 // Force Ps currently in _Psyscall into _Pidle and hand them
1697 // off to induce safe point function execution.
1698 for _, p2 := range allp {
1700 if s == _Psyscall && p2.runSafePointFn == 1 && atomic.Cas(&p2.status, s, _Pidle) {
1710 // Wait for remaining Ps to run fn.
1713 // Wait for 100us, then try to re-preempt in
1714 // case of any races.
1716 // Requires system stack.
1717 if notetsleep(&sched.safePointNote, 100*1000) {
1718 noteclear(&sched.safePointNote)
1724 if sched.safePointWait != 0 {
1725 throw("forEachP: not done")
1727 for _, p2 := range allp {
1728 if p2.runSafePointFn != 0 {
1729 throw("forEachP: P did not run fn")
1734 sched.safePointFn = nil
1739 // runSafePointFn runs the safe point function, if any, for this P.
1740 // This should be called like
1742 // if getg().m.p.runSafePointFn != 0 {
1746 // runSafePointFn must be checked on any transition in to _Pidle or
1747 // _Psyscall to avoid a race where forEachP sees that the P is running
1748 // just before the P goes into _Pidle/_Psyscall and neither forEachP
1749 // nor the P run the safe-point function.
1750 func runSafePointFn() {
1751 p := getg().m.p.ptr()
1752 // Resolve the race between forEachP running the safe-point
1753 // function on this P's behalf and this P running the
1754 // safe-point function directly.
1755 if !atomic.Cas(&p.runSafePointFn, 1, 0) {
1758 sched.safePointFn(p)
1760 sched.safePointWait--
1761 if sched.safePointWait == 0 {
1762 notewakeup(&sched.safePointNote)
1767 // When running with cgo, we call _cgo_thread_start
1768 // to start threads for us so that we can play nicely with
1770 var cgoThreadStart unsafe.Pointer
1772 type cgothreadstart struct {
1778 // Allocate a new m unassociated with any thread.
1779 // Can use p for allocation context if needed.
1780 // fn is recorded as the new m's m.mstartfn.
1781 // id is optional pre-allocated m ID. Omit by passing -1.
1783 // This function is allowed to have write barriers even if the caller
1784 // isn't because it borrows pp.
1786 //go:yeswritebarrierrec
1787 func allocm(pp *p, fn func(), id int64) *m {
1790 // The caller owns pp, but we may borrow (i.e., acquirep) it. We must
1791 // disable preemption to ensure it is not stolen, which would make the
1792 // caller lose ownership.
1797 acquirep(pp) // temporarily borrow p for mallocs in this function
1800 // Release the free M list. We need to do this somewhere and
1801 // this may free up a stack we can use.
1802 if sched.freem != nil {
1805 for freem := sched.freem; freem != nil; {
1806 wait := freem.freeWait.Load()
1807 if wait == freeMWait {
1808 next := freem.freelink
1809 freem.freelink = newList
1814 // Free the stack if needed. For freeMRef, there is
1815 // nothing to do except drop freem from the sched.freem
1817 if wait == freeMStack {
1818 // stackfree must be on the system stack, but allocm is
1819 // reachable off the system stack transitively from
1821 systemstack(func() {
1822 stackfree(freem.g0.stack)
1825 freem = freem.freelink
1827 sched.freem = newList
1835 // In case of cgo or Solaris or illumos or Darwin, pthread_create will make us a stack.
1836 // Windows and Plan 9 will layout sched stack on OS stack.
1837 if iscgo || mStackIsSystemAllocated() {
1840 mp.g0 = malg(8192 * sys.StackGuardMultiplier)
1844 if pp == gp.m.p.ptr() {
1849 allocmLock.runlock()
1853 // needm is called when a cgo callback happens on a
1854 // thread without an m (a thread not created by Go).
1855 // In this case, needm is expected to find an m to use
1856 // and return with m, g initialized correctly.
1857 // Since m and g are not set now (likely nil, but see below)
1858 // needm is limited in what routines it can call. In particular
1859 // it can only call nosplit functions (textflag 7) and cannot
1860 // do any scheduling that requires an m.
1862 // In order to avoid needing heavy lifting here, we adopt
1863 // the following strategy: there is a stack of available m's
1864 // that can be stolen. Using compare-and-swap
1865 // to pop from the stack has ABA races, so we simulate
1866 // a lock by doing an exchange (via Casuintptr) to steal the stack
1867 // head and replace the top pointer with MLOCKED (1).
1868 // This serves as a simple spin lock that we can use even
1869 // without an m. The thread that locks the stack in this way
1870 // unlocks the stack by storing a valid stack head pointer.
1872 // In order to make sure that there is always an m structure
1873 // available to be stolen, we maintain the invariant that there
1874 // is always one more than needed. At the beginning of the
1875 // program (if cgo is in use) the list is seeded with a single m.
1876 // If needm finds that it has taken the last m off the list, its job
1877 // is - once it has installed its own m so that it can do things like
1878 // allocate memory - to create a spare m and put it on the list.
1880 // Each of these extra m's also has a g0 and a curg that are
1881 // pressed into service as the scheduling stack and current
1882 // goroutine for the duration of the cgo callback.
1884 // It calls dropm to put the m back on the list,
1885 // 1. when the callback is done with the m in non-pthread platforms,
1886 // 2. or when the C thread exiting on pthread platforms.
1890 if (iscgo || GOOS == "windows") && !cgoHasExtraM {
1891 // Can happen if C/C++ code calls Go from a global ctor.
1892 // Can also happen on Windows if a global ctor uses a
1893 // callback created by syscall.NewCallback. See issue #6751
1896 // Can not throw, because scheduler is not initialized yet.
1897 writeErrStr("fatal error: cgo callback before cgo call\n")
1901 // Save and block signals before getting an M.
1902 // The signal handler may call needm itself,
1903 // and we must avoid a deadlock. Also, once g is installed,
1904 // any incoming signals will try to execute,
1905 // but we won't have the sigaltstack settings and other data
1906 // set up appropriately until the end of minit, which will
1907 // unblock the signals. This is the same dance as when
1908 // starting a new m to run Go code via newosproc.
1913 // Lock extra list, take head, unlock popped list.
1914 // nilokay=false is safe here because of the invariant above,
1915 // that the extra list always contains or will soon contain
1917 mp := lockextra(false)
1919 // Set needextram when we've just emptied the list,
1920 // so that the eventual call into cgocallbackg will
1921 // allocate a new m for the extra list. We delay the
1922 // allocation until then so that it can be done
1923 // after exitsyscall makes sure it is okay to be
1924 // running at all (that is, there's no garbage collection
1925 // running right now).
1926 mp.needextram = mp.schedlink == 0
1928 unlockextra(mp.schedlink.ptr())
1930 // Store the original signal mask for use by minit.
1931 mp.sigmask = sigmask
1933 // Install TLS on some platforms (previously setg
1934 // would do this if necessary).
1937 // Install g (= m->g0) and set the stack bounds
1938 // to match the current stack. We don't actually know
1939 // how big the stack is, like we don't know how big any
1940 // scheduling stack is, but we assume there's at least 32 kB,
1941 // which is more than enough for us.
1944 gp.stack.hi = getcallersp() + 1024
1945 gp.stack.lo = getcallersp() - 32*1024
1946 gp.stackguard0 = gp.stack.lo + _StackGuard
1948 // Should mark we are already in Go now.
1949 // Otherwise, we may call needm again when we get a signal, before cgocallbackg1,
1950 // which means the extram list may be empty, that will cause a deadlock.
1951 mp.isExtraInC = false
1953 // Initialize this thread to use the m.
1957 // mp.curg is now a real goroutine.
1958 casgstatus(mp.curg, _Gdead, _Gsyscall)
1962 // Acquire an extra m and bind it to the C thread when a pthread key has been created.
1965 func needAndBindM() {
1968 if _cgo_pthread_key_created != nil && *(*uintptr)(_cgo_pthread_key_created) != 0 {
1973 // newextram allocates m's and puts them on the extra list.
1974 // It is called with a working local m, so that it can do things
1975 // like call schedlock and allocate.
1977 c := extraMWaiters.Swap(0)
1979 for i := uint32(0); i < c; i++ {
1983 // Make sure there is at least one extra M.
1984 mp := lockextra(true)
1992 // oneNewExtraM allocates an m and puts it on the extra list.
1993 func oneNewExtraM() {
1994 // Create extra goroutine locked to extra m.
1995 // The goroutine is the context in which the cgo callback will run.
1996 // The sched.pc will never be returned to, but setting it to
1997 // goexit makes clear to the traceback routines where
1998 // the goroutine stack ends.
1999 mp := allocm(nil, nil, -1)
2001 gp.sched.pc = abi.FuncPCABI0(goexit) + sys.PCQuantum
2002 gp.sched.sp = gp.stack.hi
2003 gp.sched.sp -= 4 * goarch.PtrSize // extra space in case of reads slightly beyond frame
2005 gp.sched.g = guintptr(unsafe.Pointer(gp))
2006 gp.syscallpc = gp.sched.pc
2007 gp.syscallsp = gp.sched.sp
2008 gp.stktopsp = gp.sched.sp
2009 // malg returns status as _Gidle. Change to _Gdead before
2010 // adding to allg where GC can see it. We use _Gdead to hide
2011 // this from tracebacks and stack scans since it isn't a
2012 // "real" goroutine until needm grabs it.
2013 casgstatus(gp, _Gidle, _Gdead)
2017 // mark we are in C by default.
2018 mp.isExtraInC = true
2022 gp.goid = sched.goidgen.Add(1)
2023 gp.sysblocktraced = true
2025 gp.racectx = racegostart(abi.FuncPCABIInternal(newextram) + sys.PCQuantum)
2028 // Trigger two trace events for the locked g in the extra m,
2029 // since the next event of the g will be traceEvGoSysExit in exitsyscall,
2030 // while calling from C thread to Go.
2031 traceGoCreate(gp, 0) // no start pc
2033 traceEvent(traceEvGoInSyscall, -1, gp.goid)
2035 // put on allg for garbage collector
2038 // gp is now on the allg list, but we don't want it to be
2039 // counted by gcount. It would be more "proper" to increment
2040 // sched.ngfree, but that requires locking. Incrementing ngsys
2041 // has the same effect.
2044 // Add m to the extra list.
2045 mnext := lockextra(true)
2046 mp.schedlink.set(mnext)
2051 // dropm puts the current m back onto the extra list.
2053 // 1. On systems without pthreads, like Windows
2054 // dropm is called when a cgo callback has called needm but is now
2055 // done with the callback and returning back into the non-Go thread.
2057 // The main expense here is the call to signalstack to release the
2058 // m's signal stack, and then the call to needm on the next callback
2059 // from this thread. It is tempting to try to save the m for next time,
2060 // which would eliminate both these costs, but there might not be
2061 // a next time: the current thread (which Go does not control) might exit.
2062 // If we saved the m for that thread, there would be an m leak each time
2063 // such a thread exited. Instead, we acquire and release an m on each
2064 // call. These should typically not be scheduling operations, just a few
2065 // atomics, so the cost should be small.
2067 // 2. On systems with pthreads
2068 // dropm is called while a non-Go thread is exiting.
2069 // We allocate a pthread per-thread variable using pthread_key_create,
2070 // to register a thread-exit-time destructor.
2071 // And store the g into a thread-specific value associated with the pthread key,
2072 // when first return back to C.
2073 // So that the destructor would invoke dropm while the non-Go thread is exiting.
2074 // This is much faster since it avoids expensive signal-related syscalls.
2076 // NOTE: this always runs without a P, so, nowritebarrierrec required.
2078 //go:nowritebarrierrec
2080 // Clear m and g, and return m to the extra list.
2081 // After the call to setg we can only call nosplit functions
2082 // with no pointer manipulation.
2085 // Return mp.curg to dead state.
2086 casgstatus(mp.curg, _Gsyscall, _Gdead)
2087 mp.curg.preemptStop = false
2090 // Block signals before unminit.
2091 // Unminit unregisters the signal handling stack (but needs g on some systems).
2092 // Setg(nil) clears g, which is the signal handler's cue not to run Go handlers.
2093 // It's important not to try to handle a signal between those two steps.
2094 sigmask := mp.sigmask
2098 mnext := lockextra(true)
2100 mp.schedlink.set(mnext)
2104 // Commit the release of mp.
2107 msigrestore(sigmask)
2110 // bindm store the g0 of the current m into a thread-specific value.
2112 // We allocate a pthread per-thread variable using pthread_key_create,
2113 // to register a thread-exit-time destructor.
2114 // We are here setting the thread-specific value of the pthread key, to enable the destructor.
2115 // So that the pthread_key_destructor would dropm while the C thread is exiting.
2117 // And the saved g will be used in pthread_key_destructor,
2118 // since the g stored in the TLS by Go might be cleared in some platforms,
2119 // before the destructor invoked, so, we restore g by the stored g, before dropm.
2121 // We store g0 instead of m, to make the assembly code simpler,
2122 // since we need to restore g0 in runtime.cgocallback.
2124 // On systems without pthreads, like Windows, bindm shouldn't be used.
2126 // NOTE: this always runs without a P, so, nowritebarrierrec required.
2129 //go:nowritebarrierrec
2131 if GOOS == "windows" || GOOS == "plan9" {
2132 fatal("bindm in unexpected GOOS")
2136 fatal("the current g is not g0")
2138 if _cgo_bindm != nil {
2139 asmcgocall(_cgo_bindm, unsafe.Pointer(g))
2143 // A helper function for EnsureDropM.
2144 func getm() uintptr {
2145 return uintptr(unsafe.Pointer(getg().m))
2148 var extram atomic.Uintptr
2149 var extraMCount uint32 // Protected by lockextra
2150 var extraMWaiters atomic.Uint32
2152 // lockextra locks the extra list and returns the list head.
2153 // The caller must unlock the list by storing a new list head
2154 // to extram. If nilokay is true, then lockextra will
2155 // return a nil list head if that's what it finds. If nilokay is false,
2156 // lockextra will keep waiting until the list head is no longer nil.
2159 func lockextra(nilokay bool) *m {
2164 old := extram.Load()
2169 if old == 0 && !nilokay {
2171 // Add 1 to the number of threads
2172 // waiting for an M.
2173 // This is cleared by newextram.
2174 extraMWaiters.Add(1)
2180 if extram.CompareAndSwap(old, locked) {
2181 return (*m)(unsafe.Pointer(old))
2189 func unlockextra(mp *m) {
2190 extram.Store(uintptr(unsafe.Pointer(mp)))
2194 // allocmLock is locked for read when creating new Ms in allocm and their
2195 // addition to allm. Thus acquiring this lock for write blocks the
2196 // creation of new Ms.
2199 // execLock serializes exec and clone to avoid bugs or unspecified
2200 // behaviour around exec'ing while creating/destroying threads. See
2205 // These errors are reported (via writeErrStr) by some OS-specific
2206 // versions of newosproc and newosproc0.
2208 failthreadcreate = "runtime: failed to create new OS thread\n"
2209 failallocatestack = "runtime: failed to allocate stack for the new OS thread\n"
2212 // newmHandoff contains a list of m structures that need new OS threads.
2213 // This is used by newm in situations where newm itself can't safely
2214 // start an OS thread.
2215 var newmHandoff struct {
2218 // newm points to a list of M structures that need new OS
2219 // threads. The list is linked through m.schedlink.
2222 // waiting indicates that wake needs to be notified when an m
2223 // is put on the list.
2227 // haveTemplateThread indicates that the templateThread has
2228 // been started. This is not protected by lock. Use cas to set
2230 haveTemplateThread uint32
2233 // Create a new m. It will start off with a call to fn, or else the scheduler.
2234 // fn needs to be static and not a heap allocated closure.
2235 // May run with m.p==nil, so write barriers are not allowed.
2237 // id is optional pre-allocated m ID. Omit by passing -1.
2239 //go:nowritebarrierrec
2240 func newm(fn func(), pp *p, id int64) {
2241 // allocm adds a new M to allm, but they do not start until created by
2242 // the OS in newm1 or the template thread.
2244 // doAllThreadsSyscall requires that every M in allm will eventually
2245 // start and be signal-able, even with a STW.
2247 // Disable preemption here until we start the thread to ensure that
2248 // newm is not preempted between allocm and starting the new thread,
2249 // ensuring that anything added to allm is guaranteed to eventually
2253 mp := allocm(pp, fn, id)
2255 mp.sigmask = initSigmask
2256 if gp := getg(); gp != nil && gp.m != nil && (gp.m.lockedExt != 0 || gp.m.incgo) && GOOS != "plan9" {
2257 // We're on a locked M or a thread that may have been
2258 // started by C. The kernel state of this thread may
2259 // be strange (the user may have locked it for that
2260 // purpose). We don't want to clone that into another
2261 // thread. Instead, ask a known-good thread to create
2262 // the thread for us.
2264 // This is disabled on Plan 9. See golang.org/issue/22227.
2266 // TODO: This may be unnecessary on Windows, which
2267 // doesn't model thread creation off fork.
2268 lock(&newmHandoff.lock)
2269 if newmHandoff.haveTemplateThread == 0 {
2270 throw("on a locked thread with no template thread")
2272 mp.schedlink = newmHandoff.newm
2273 newmHandoff.newm.set(mp)
2274 if newmHandoff.waiting {
2275 newmHandoff.waiting = false
2276 notewakeup(&newmHandoff.wake)
2278 unlock(&newmHandoff.lock)
2279 // The M has not started yet, but the template thread does not
2280 // participate in STW, so it will always process queued Ms and
2281 // it is safe to releasem.
2291 var ts cgothreadstart
2292 if _cgo_thread_start == nil {
2293 throw("_cgo_thread_start missing")
2296 ts.tls = (*uint64)(unsafe.Pointer(&mp.tls[0]))
2297 ts.fn = unsafe.Pointer(abi.FuncPCABI0(mstart))
2299 msanwrite(unsafe.Pointer(&ts), unsafe.Sizeof(ts))
2302 asanwrite(unsafe.Pointer(&ts), unsafe.Sizeof(ts))
2304 execLock.rlock() // Prevent process clone.
2305 asmcgocall(_cgo_thread_start, unsafe.Pointer(&ts))
2309 execLock.rlock() // Prevent process clone.
2314 // startTemplateThread starts the template thread if it is not already
2317 // The calling thread must itself be in a known-good state.
2318 func startTemplateThread() {
2319 if GOARCH == "wasm" { // no threads on wasm yet
2323 // Disable preemption to guarantee that the template thread will be
2324 // created before a park once haveTemplateThread is set.
2326 if !atomic.Cas(&newmHandoff.haveTemplateThread, 0, 1) {
2330 newm(templateThread, nil, -1)
2334 // templateThread is a thread in a known-good state that exists solely
2335 // to start new threads in known-good states when the calling thread
2336 // may not be in a good state.
2338 // Many programs never need this, so templateThread is started lazily
2339 // when we first enter a state that might lead to running on a thread
2340 // in an unknown state.
2342 // templateThread runs on an M without a P, so it must not have write
2345 //go:nowritebarrierrec
2346 func templateThread() {
2353 lock(&newmHandoff.lock)
2354 for newmHandoff.newm != 0 {
2355 newm := newmHandoff.newm.ptr()
2356 newmHandoff.newm = 0
2357 unlock(&newmHandoff.lock)
2359 next := newm.schedlink.ptr()
2364 lock(&newmHandoff.lock)
2366 newmHandoff.waiting = true
2367 noteclear(&newmHandoff.wake)
2368 unlock(&newmHandoff.lock)
2369 notesleep(&newmHandoff.wake)
2373 // Stops execution of the current m until new work is available.
2374 // Returns with acquired P.
2378 if gp.m.locks != 0 {
2379 throw("stopm holding locks")
2382 throw("stopm holding p")
2385 throw("stopm spinning")
2392 acquirep(gp.m.nextp.ptr())
2397 // startm's caller incremented nmspinning. Set the new M's spinning.
2398 getg().m.spinning = true
2401 // Schedules some M to run the p (creates an M if necessary).
2402 // If p==nil, tries to get an idle P, if no idle P's does nothing.
2403 // May run with m.p==nil, so write barriers are not allowed.
2404 // If spinning is set, the caller has incremented nmspinning and must provide a
2405 // P. startm will set m.spinning in the newly started M.
2407 // Callers passing a non-nil P must call from a non-preemptible context. See
2408 // comment on acquirem below.
2410 // Must not have write barriers because this may be called without a P.
2412 //go:nowritebarrierrec
2413 func startm(pp *p, spinning bool) {
2414 // Disable preemption.
2416 // Every owned P must have an owner that will eventually stop it in the
2417 // event of a GC stop request. startm takes transient ownership of a P
2418 // (either from argument or pidleget below) and transfers ownership to
2419 // a started M, which will be responsible for performing the stop.
2421 // Preemption must be disabled during this transient ownership,
2422 // otherwise the P this is running on may enter GC stop while still
2423 // holding the transient P, leaving that P in limbo and deadlocking the
2426 // Callers passing a non-nil P must already be in non-preemptible
2427 // context, otherwise such preemption could occur on function entry to
2428 // startm. Callers passing a nil P may be preemptible, so we must
2429 // disable preemption before acquiring a P from pidleget below.
2434 // TODO(prattmic): All remaining calls to this function
2435 // with _p_ == nil could be cleaned up to find a P
2436 // before calling startm.
2437 throw("startm: P required for spinning=true")
2448 // No M is available, we must drop sched.lock and call newm.
2449 // However, we already own a P to assign to the M.
2451 // Once sched.lock is released, another G (e.g., in a syscall),
2452 // could find no idle P while checkdead finds a runnable G but
2453 // no running M's because this new M hasn't started yet, thus
2454 // throwing in an apparent deadlock.
2456 // Avoid this situation by pre-allocating the ID for the new M,
2457 // thus marking it as 'running' before we drop sched.lock. This
2458 // new M will eventually run the scheduler to execute any
2465 // The caller incremented nmspinning, so set m.spinning in the new M.
2469 // Ownership transfer of pp committed by start in newm.
2470 // Preemption is now safe.
2476 throw("startm: m is spinning")
2479 throw("startm: m has p")
2481 if spinning && !runqempty(pp) {
2482 throw("startm: p has runnable gs")
2484 // The caller incremented nmspinning, so set m.spinning in the new M.
2485 nmp.spinning = spinning
2487 notewakeup(&nmp.park)
2488 // Ownership transfer of pp committed by wakeup. Preemption is now
2493 // Hands off P from syscall or locked M.
2494 // Always runs without a P, so write barriers are not allowed.
2496 //go:nowritebarrierrec
2497 func handoffp(pp *p) {
2498 // handoffp must start an M in any situation where
2499 // findrunnable would return a G to run on pp.
2501 // if it has local work, start it straight away
2502 if !runqempty(pp) || sched.runqsize != 0 {
2506 // if there's trace work to do, start it straight away
2507 if (trace.enabled || trace.shutdown) && traceReaderAvailable() != nil {
2511 // if it has GC work, start it straight away
2512 if gcBlackenEnabled != 0 && gcMarkWorkAvailable(pp) {
2516 // no local work, check that there are no spinning/idle M's,
2517 // otherwise our help is not required
2518 if sched.nmspinning.Load()+sched.npidle.Load() == 0 && sched.nmspinning.CompareAndSwap(0, 1) { // TODO: fast atomic
2519 sched.needspinning.Store(0)
2524 if sched.gcwaiting.Load() {
2525 pp.status = _Pgcstop
2527 if sched.stopwait == 0 {
2528 notewakeup(&sched.stopnote)
2533 if pp.runSafePointFn != 0 && atomic.Cas(&pp.runSafePointFn, 1, 0) {
2534 sched.safePointFn(pp)
2535 sched.safePointWait--
2536 if sched.safePointWait == 0 {
2537 notewakeup(&sched.safePointNote)
2540 if sched.runqsize != 0 {
2545 // If this is the last running P and nobody is polling network,
2546 // need to wakeup another M to poll network.
2547 if sched.npidle.Load() == gomaxprocs-1 && sched.lastpoll.Load() != 0 {
2553 // The scheduler lock cannot be held when calling wakeNetPoller below
2554 // because wakeNetPoller may call wakep which may call startm.
2555 when := nobarrierWakeTime(pp)
2564 // Tries to add one more P to execute G's.
2565 // Called when a G is made runnable (newproc, ready).
2566 // Must be called with a P.
2568 // Be conservative about spinning threads, only start one if none exist
2570 if sched.nmspinning.Load() != 0 || !sched.nmspinning.CompareAndSwap(0, 1) {
2574 // Disable preemption until ownership of pp transfers to the next M in
2575 // startm. Otherwise preemption here would leave pp stuck waiting to
2578 // See preemption comment on acquirem in startm for more details.
2583 pp, _ = pidlegetSpinning(0)
2585 if sched.nmspinning.Add(-1) < 0 {
2586 throw("wakep: negative nmspinning")
2592 // Since we always have a P, the race in the "No M is available"
2593 // comment in startm doesn't apply during the small window between the
2594 // unlock here and lock in startm. A checkdead in between will always
2595 // see at least one running M (ours).
2603 // Stops execution of the current m that is locked to a g until the g is runnable again.
2604 // Returns with acquired P.
2605 func stoplockedm() {
2608 if gp.m.lockedg == 0 || gp.m.lockedg.ptr().lockedm.ptr() != gp.m {
2609 throw("stoplockedm: inconsistent locking")
2612 // Schedule another M to run this p.
2617 // Wait until another thread schedules lockedg again.
2619 status := readgstatus(gp.m.lockedg.ptr())
2620 if status&^_Gscan != _Grunnable {
2621 print("runtime:stoplockedm: lockedg (atomicstatus=", status, ") is not Grunnable or Gscanrunnable\n")
2622 dumpgstatus(gp.m.lockedg.ptr())
2623 throw("stoplockedm: not runnable")
2625 acquirep(gp.m.nextp.ptr())
2629 // Schedules the locked m to run the locked gp.
2630 // May run during STW, so write barriers are not allowed.
2632 //go:nowritebarrierrec
2633 func startlockedm(gp *g) {
2634 mp := gp.lockedm.ptr()
2636 throw("startlockedm: locked to me")
2639 throw("startlockedm: m has p")
2641 // directly handoff current P to the locked m
2645 notewakeup(&mp.park)
2649 // Stops the current m for stopTheWorld.
2650 // Returns when the world is restarted.
2654 if !sched.gcwaiting.Load() {
2655 throw("gcstopm: not waiting for gc")
2658 gp.m.spinning = false
2659 // OK to just drop nmspinning here,
2660 // startTheWorld will unpark threads as necessary.
2661 if sched.nmspinning.Add(-1) < 0 {
2662 throw("gcstopm: negative nmspinning")
2667 pp.status = _Pgcstop
2669 if sched.stopwait == 0 {
2670 notewakeup(&sched.stopnote)
2676 // Schedules gp to run on the current M.
2677 // If inheritTime is true, gp inherits the remaining time in the
2678 // current time slice. Otherwise, it starts a new time slice.
2681 // Write barriers are allowed because this is called immediately after
2682 // acquiring a P in several places.
2684 //go:yeswritebarrierrec
2685 func execute(gp *g, inheritTime bool) {
2688 if goroutineProfile.active {
2689 // Make sure that gp has had its stack written out to the goroutine
2690 // profile, exactly as it was when the goroutine profiler first stopped
2692 tryRecordGoroutineProfile(gp, osyield)
2695 // Assign gp.m before entering _Grunning so running Gs have an
2699 casgstatus(gp, _Grunnable, _Grunning)
2702 gp.stackguard0 = gp.stack.lo + _StackGuard
2704 mp.p.ptr().schedtick++
2707 // Check whether the profiler needs to be turned on or off.
2708 hz := sched.profilehz
2709 if mp.profilehz != hz {
2710 setThreadCPUProfiler(hz)
2714 // GoSysExit has to happen when we have a P, but before GoStart.
2715 // So we emit it here.
2716 if gp.syscallsp != 0 && gp.sysblocktraced {
2717 traceGoSysExit(gp.sysexitticks)
2725 // Finds a runnable goroutine to execute.
2726 // Tries to steal from other P's, get g from local or global queue, poll network.
2727 // tryWakeP indicates that the returned goroutine is not normal (GC worker, trace
2728 // reader) so the caller should try to wake a P.
2729 func findRunnable() (gp *g, inheritTime, tryWakeP bool) {
2732 // The conditions here and in handoffp must agree: if
2733 // findrunnable would return a G to run, handoffp must start
2738 if sched.gcwaiting.Load() {
2742 if pp.runSafePointFn != 0 {
2746 // now and pollUntil are saved for work stealing later,
2747 // which may steal timers. It's important that between now
2748 // and then, nothing blocks, so these numbers remain mostly
2750 now, pollUntil, _ := checkTimers(pp, 0)
2752 // Try to schedule the trace reader.
2753 if trace.enabled || trace.shutdown {
2756 casgstatus(gp, _Gwaiting, _Grunnable)
2757 traceGoUnpark(gp, 0)
2758 return gp, false, true
2762 // Try to schedule a GC worker.
2763 if gcBlackenEnabled != 0 {
2764 gp, tnow := gcController.findRunnableGCWorker(pp, now)
2766 return gp, false, true
2771 // Check the global runnable queue once in a while to ensure fairness.
2772 // Otherwise two goroutines can completely occupy the local runqueue
2773 // by constantly respawning each other.
2774 if pp.schedtick%61 == 0 && sched.runqsize > 0 {
2776 gp := globrunqget(pp, 1)
2779 return gp, false, false
2783 // Wake up the finalizer G.
2784 if fingStatus.Load()&(fingWait|fingWake) == fingWait|fingWake {
2785 if gp := wakefing(); gp != nil {
2789 if *cgo_yield != nil {
2790 asmcgocall(*cgo_yield, nil)
2794 if gp, inheritTime := runqget(pp); gp != nil {
2795 return gp, inheritTime, false
2799 if sched.runqsize != 0 {
2801 gp := globrunqget(pp, 0)
2804 return gp, false, false
2809 // This netpoll is only an optimization before we resort to stealing.
2810 // We can safely skip it if there are no waiters or a thread is blocked
2811 // in netpoll already. If there is any kind of logical race with that
2812 // blocked thread (e.g. it has already returned from netpoll, but does
2813 // not set lastpoll yet), this thread will do blocking netpoll below
2815 if netpollinited() && netpollWaiters.Load() > 0 && sched.lastpoll.Load() != 0 {
2816 if list := netpoll(0); !list.empty() { // non-blocking
2819 casgstatus(gp, _Gwaiting, _Grunnable)
2821 traceGoUnpark(gp, 0)
2823 return gp, false, false
2827 // Spinning Ms: steal work from other Ps.
2829 // Limit the number of spinning Ms to half the number of busy Ps.
2830 // This is necessary to prevent excessive CPU consumption when
2831 // GOMAXPROCS>>1 but the program parallelism is low.
2832 if mp.spinning || 2*sched.nmspinning.Load() < gomaxprocs-sched.npidle.Load() {
2837 gp, inheritTime, tnow, w, newWork := stealWork(now)
2839 // Successfully stole.
2840 return gp, inheritTime, false
2843 // There may be new timer or GC work; restart to
2849 if w != 0 && (pollUntil == 0 || w < pollUntil) {
2850 // Earlier timer to wait for.
2855 // We have nothing to do.
2857 // If we're in the GC mark phase, can safely scan and blacken objects,
2858 // and have work to do, run idle-time marking rather than give up the P.
2859 if gcBlackenEnabled != 0 && gcMarkWorkAvailable(pp) && gcController.addIdleMarkWorker() {
2860 node := (*gcBgMarkWorkerNode)(gcBgMarkWorkerPool.pop())
2862 pp.gcMarkWorkerMode = gcMarkWorkerIdleMode
2864 casgstatus(gp, _Gwaiting, _Grunnable)
2866 traceGoUnpark(gp, 0)
2868 return gp, false, false
2870 gcController.removeIdleMarkWorker()
2874 // If a callback returned and no other goroutine is awake,
2875 // then wake event handler goroutine which pauses execution
2876 // until a callback was triggered.
2877 gp, otherReady := beforeIdle(now, pollUntil)
2879 casgstatus(gp, _Gwaiting, _Grunnable)
2881 traceGoUnpark(gp, 0)
2883 return gp, false, false
2889 // Before we drop our P, make a snapshot of the allp slice,
2890 // which can change underfoot once we no longer block
2891 // safe-points. We don't need to snapshot the contents because
2892 // everything up to cap(allp) is immutable.
2893 allpSnapshot := allp
2894 // Also snapshot masks. Value changes are OK, but we can't allow
2895 // len to change out from under us.
2896 idlepMaskSnapshot := idlepMask
2897 timerpMaskSnapshot := timerpMask
2899 // return P and block
2901 if sched.gcwaiting.Load() || pp.runSafePointFn != 0 {
2905 if sched.runqsize != 0 {
2906 gp := globrunqget(pp, 0)
2908 return gp, false, false
2910 if !mp.spinning && sched.needspinning.Load() == 1 {
2911 // See "Delicate dance" comment below.
2916 if releasep() != pp {
2917 throw("findrunnable: wrong p")
2919 now = pidleput(pp, now)
2922 // Delicate dance: thread transitions from spinning to non-spinning
2923 // state, potentially concurrently with submission of new work. We must
2924 // drop nmspinning first and then check all sources again (with
2925 // #StoreLoad memory barrier in between). If we do it the other way
2926 // around, another thread can submit work after we've checked all
2927 // sources but before we drop nmspinning; as a result nobody will
2928 // unpark a thread to run the work.
2930 // This applies to the following sources of work:
2932 // * Goroutines added to a per-P run queue.
2933 // * New/modified-earlier timers on a per-P timer heap.
2934 // * Idle-priority GC work (barring golang.org/issue/19112).
2936 // If we discover new work below, we need to restore m.spinning as a
2937 // signal for resetspinning to unpark a new worker thread (because
2938 // there can be more than one starving goroutine).
2940 // However, if after discovering new work we also observe no idle Ps
2941 // (either here or in resetspinning), we have a problem. We may be
2942 // racing with a non-spinning M in the block above, having found no
2943 // work and preparing to release its P and park. Allowing that P to go
2944 // idle will result in loss of work conservation (idle P while there is
2945 // runnable work). This could result in complete deadlock in the
2946 // unlikely event that we discover new work (from netpoll) right as we
2947 // are racing with _all_ other Ps going idle.
2949 // We use sched.needspinning to synchronize with non-spinning Ms going
2950 // idle. If needspinning is set when they are about to drop their P,
2951 // they abort the drop and instead become a new spinning M on our
2952 // behalf. If we are not racing and the system is truly fully loaded
2953 // then no spinning threads are required, and the next thread to
2954 // naturally become spinning will clear the flag.
2956 // Also see "Worker thread parking/unparking" comment at the top of the
2958 wasSpinning := mp.spinning
2961 if sched.nmspinning.Add(-1) < 0 {
2962 throw("findrunnable: negative nmspinning")
2965 // Note the for correctness, only the last M transitioning from
2966 // spinning to non-spinning must perform these rechecks to
2967 // ensure no missed work. However, the runtime has some cases
2968 // of transient increments of nmspinning that are decremented
2969 // without going through this path, so we must be conservative
2970 // and perform the check on all spinning Ms.
2972 // See https://go.dev/issue/43997.
2974 // Check all runqueues once again.
2975 pp := checkRunqsNoP(allpSnapshot, idlepMaskSnapshot)
2982 // Check for idle-priority GC work again.
2983 pp, gp := checkIdleGCNoP()
2988 // Run the idle worker.
2989 pp.gcMarkWorkerMode = gcMarkWorkerIdleMode
2990 casgstatus(gp, _Gwaiting, _Grunnable)
2992 traceGoUnpark(gp, 0)
2994 return gp, false, false
2997 // Finally, check for timer creation or expiry concurrently with
2998 // transitioning from spinning to non-spinning.
3000 // Note that we cannot use checkTimers here because it calls
3001 // adjusttimers which may need to allocate memory, and that isn't
3002 // allowed when we don't have an active P.
3003 pollUntil = checkTimersNoP(allpSnapshot, timerpMaskSnapshot, pollUntil)
3006 // Poll network until next timer.
3007 if netpollinited() && (netpollWaiters.Load() > 0 || pollUntil != 0) && sched.lastpoll.Swap(0) != 0 {
3008 sched.pollUntil.Store(pollUntil)
3010 throw("findrunnable: netpoll with p")
3013 throw("findrunnable: netpoll with spinning")
3019 delay = pollUntil - now
3025 // When using fake time, just poll.
3028 list := netpoll(delay) // block until new work is available
3029 sched.pollUntil.Store(0)
3030 sched.lastpoll.Store(now)
3031 if faketime != 0 && list.empty() {
3032 // Using fake time and nothing is ready; stop M.
3033 // When all M's stop, checkdead will call timejump.
3038 pp, _ := pidleget(now)
3047 casgstatus(gp, _Gwaiting, _Grunnable)
3049 traceGoUnpark(gp, 0)
3051 return gp, false, false
3058 } else if pollUntil != 0 && netpollinited() {
3059 pollerPollUntil := sched.pollUntil.Load()
3060 if pollerPollUntil == 0 || pollerPollUntil > pollUntil {
3068 // pollWork reports whether there is non-background work this P could
3069 // be doing. This is a fairly lightweight check to be used for
3070 // background work loops, like idle GC. It checks a subset of the
3071 // conditions checked by the actual scheduler.
3072 func pollWork() bool {
3073 if sched.runqsize != 0 {
3076 p := getg().m.p.ptr()
3080 if netpollinited() && netpollWaiters.Load() > 0 && sched.lastpoll.Load() != 0 {
3081 if list := netpoll(0); !list.empty() {
3089 // stealWork attempts to steal a runnable goroutine or timer from any P.
3091 // If newWork is true, new work may have been readied.
3093 // If now is not 0 it is the current time. stealWork returns the passed time or
3094 // the current time if now was passed as 0.
3095 func stealWork(now int64) (gp *g, inheritTime bool, rnow, pollUntil int64, newWork bool) {
3096 pp := getg().m.p.ptr()
3100 const stealTries = 4
3101 for i := 0; i < stealTries; i++ {
3102 stealTimersOrRunNextG := i == stealTries-1
3104 for enum := stealOrder.start(fastrand()); !enum.done(); enum.next() {
3105 if sched.gcwaiting.Load() {
3106 // GC work may be available.
3107 return nil, false, now, pollUntil, true
3109 p2 := allp[enum.position()]
3114 // Steal timers from p2. This call to checkTimers is the only place
3115 // where we might hold a lock on a different P's timers. We do this
3116 // once on the last pass before checking runnext because stealing
3117 // from the other P's runnext should be the last resort, so if there
3118 // are timers to steal do that first.
3120 // We only check timers on one of the stealing iterations because
3121 // the time stored in now doesn't change in this loop and checking
3122 // the timers for each P more than once with the same value of now
3123 // is probably a waste of time.
3125 // timerpMask tells us whether the P may have timers at all. If it
3126 // can't, no need to check at all.
3127 if stealTimersOrRunNextG && timerpMask.read(enum.position()) {
3128 tnow, w, ran := checkTimers(p2, now)
3130 if w != 0 && (pollUntil == 0 || w < pollUntil) {
3134 // Running the timers may have
3135 // made an arbitrary number of G's
3136 // ready and added them to this P's
3137 // local run queue. That invalidates
3138 // the assumption of runqsteal
3139 // that it always has room to add
3140 // stolen G's. So check now if there
3141 // is a local G to run.
3142 if gp, inheritTime := runqget(pp); gp != nil {
3143 return gp, inheritTime, now, pollUntil, ranTimer
3149 // Don't bother to attempt to steal if p2 is idle.
3150 if !idlepMask.read(enum.position()) {
3151 if gp := runqsteal(pp, p2, stealTimersOrRunNextG); gp != nil {
3152 return gp, false, now, pollUntil, ranTimer
3158 // No goroutines found to steal. Regardless, running a timer may have
3159 // made some goroutine ready that we missed. Indicate the next timer to
3161 return nil, false, now, pollUntil, ranTimer
3164 // Check all Ps for a runnable G to steal.
3166 // On entry we have no P. If a G is available to steal and a P is available,
3167 // the P is returned which the caller should acquire and attempt to steal the
3169 func checkRunqsNoP(allpSnapshot []*p, idlepMaskSnapshot pMask) *p {
3170 for id, p2 := range allpSnapshot {
3171 if !idlepMaskSnapshot.read(uint32(id)) && !runqempty(p2) {
3173 pp, _ := pidlegetSpinning(0)
3175 // Can't get a P, don't bother checking remaining Ps.
3184 // No work available.
3188 // Check all Ps for a timer expiring sooner than pollUntil.
3190 // Returns updated pollUntil value.
3191 func checkTimersNoP(allpSnapshot []*p, timerpMaskSnapshot pMask, pollUntil int64) int64 {
3192 for id, p2 := range allpSnapshot {
3193 if timerpMaskSnapshot.read(uint32(id)) {
3194 w := nobarrierWakeTime(p2)
3195 if w != 0 && (pollUntil == 0 || w < pollUntil) {
3204 // Check for idle-priority GC, without a P on entry.
3206 // If some GC work, a P, and a worker G are all available, the P and G will be
3207 // returned. The returned P has not been wired yet.
3208 func checkIdleGCNoP() (*p, *g) {
3209 // N.B. Since we have no P, gcBlackenEnabled may change at any time; we
3210 // must check again after acquiring a P. As an optimization, we also check
3211 // if an idle mark worker is needed at all. This is OK here, because if we
3212 // observe that one isn't needed, at least one is currently running. Even if
3213 // it stops running, its own journey into the scheduler should schedule it
3214 // again, if need be (at which point, this check will pass, if relevant).
3215 if atomic.Load(&gcBlackenEnabled) == 0 || !gcController.needIdleMarkWorker() {
3218 if !gcMarkWorkAvailable(nil) {
3222 // Work is available; we can start an idle GC worker only if there is
3223 // an available P and available worker G.
3225 // We can attempt to acquire these in either order, though both have
3226 // synchronization concerns (see below). Workers are almost always
3227 // available (see comment in findRunnableGCWorker for the one case
3228 // there may be none). Since we're slightly less likely to find a P,
3229 // check for that first.
3231 // Synchronization: note that we must hold sched.lock until we are
3232 // committed to keeping it. Otherwise we cannot put the unnecessary P
3233 // back in sched.pidle without performing the full set of idle
3234 // transition checks.
3236 // If we were to check gcBgMarkWorkerPool first, we must somehow handle
3237 // the assumption in gcControllerState.findRunnableGCWorker that an
3238 // empty gcBgMarkWorkerPool is only possible if gcMarkDone is running.
3240 pp, now := pidlegetSpinning(0)
3246 // Now that we own a P, gcBlackenEnabled can't change (as it requires STW).
3247 if gcBlackenEnabled == 0 || !gcController.addIdleMarkWorker() {
3253 node := (*gcBgMarkWorkerNode)(gcBgMarkWorkerPool.pop())
3257 gcController.removeIdleMarkWorker()
3263 return pp, node.gp.ptr()
3266 // wakeNetPoller wakes up the thread sleeping in the network poller if it isn't
3267 // going to wake up before the when argument; or it wakes an idle P to service
3268 // timers and the network poller if there isn't one already.
3269 func wakeNetPoller(when int64) {
3270 if sched.lastpoll.Load() == 0 {
3271 // In findrunnable we ensure that when polling the pollUntil
3272 // field is either zero or the time to which the current
3273 // poll is expected to run. This can have a spurious wakeup
3274 // but should never miss a wakeup.
3275 pollerPollUntil := sched.pollUntil.Load()
3276 if pollerPollUntil == 0 || pollerPollUntil > when {
3280 // There are no threads in the network poller, try to get
3281 // one there so it can handle new timers.
3282 if GOOS != "plan9" { // Temporary workaround - see issue #42303.
3288 func resetspinning() {
3291 throw("resetspinning: not a spinning m")
3293 gp.m.spinning = false
3294 nmspinning := sched.nmspinning.Add(-1)
3296 throw("findrunnable: negative nmspinning")
3298 // M wakeup policy is deliberately somewhat conservative, so check if we
3299 // need to wakeup another P here. See "Worker thread parking/unparking"
3300 // comment at the top of the file for details.
3304 // injectglist adds each runnable G on the list to some run queue,
3305 // and clears glist. If there is no current P, they are added to the
3306 // global queue, and up to npidle M's are started to run them.
3307 // Otherwise, for each idle P, this adds a G to the global queue
3308 // and starts an M. Any remaining G's are added to the current P's
3310 // This may temporarily acquire sched.lock.
3311 // Can run concurrently with GC.
3312 func injectglist(glist *gList) {
3317 for gp := glist.head.ptr(); gp != nil; gp = gp.schedlink.ptr() {
3318 traceGoUnpark(gp, 0)
3322 // Mark all the goroutines as runnable before we put them
3323 // on the run queues.
3324 head := glist.head.ptr()
3327 for gp := head; gp != nil; gp = gp.schedlink.ptr() {
3330 casgstatus(gp, _Gwaiting, _Grunnable)
3333 // Turn the gList into a gQueue.
3339 startIdle := func(n int) {
3340 for i := 0; i < n; i++ {
3341 mp := acquirem() // See comment in startm.
3344 pp, _ := pidlegetSpinning(0)
3357 pp := getg().m.p.ptr()
3360 globrunqputbatch(&q, int32(qsize))
3366 npidle := int(sched.npidle.Load())
3369 for n = 0; n < npidle && !q.empty(); n++ {
3375 globrunqputbatch(&globq, int32(n))
3382 runqputbatch(pp, &q, qsize)
3386 // One round of scheduler: find a runnable goroutine and execute it.
3392 throw("schedule: holding locks")
3395 if mp.lockedg != 0 {
3397 execute(mp.lockedg.ptr(), false) // Never returns.
3400 // We should not schedule away from a g that is executing a cgo call,
3401 // since the cgo call is using the m's g0 stack.
3403 throw("schedule: in cgo")
3410 // Safety check: if we are spinning, the run queue should be empty.
3411 // Check this before calling checkTimers, as that might call
3412 // goready to put a ready goroutine on the local run queue.
3413 if mp.spinning && (pp.runnext != 0 || pp.runqhead != pp.runqtail) {
3414 throw("schedule: spinning with local work")
3417 gp, inheritTime, tryWakeP := findRunnable() // blocks until work is available
3419 // This thread is going to run a goroutine and is not spinning anymore,
3420 // so if it was marked as spinning we need to reset it now and potentially
3421 // start a new spinning M.
3426 if sched.disable.user && !schedEnabled(gp) {
3427 // Scheduling of this goroutine is disabled. Put it on
3428 // the list of pending runnable goroutines for when we
3429 // re-enable user scheduling and look again.
3431 if schedEnabled(gp) {
3432 // Something re-enabled scheduling while we
3433 // were acquiring the lock.
3436 sched.disable.runnable.pushBack(gp)
3443 // If about to schedule a not-normal goroutine (a GCworker or tracereader),
3444 // wake a P if there is one.
3448 if gp.lockedm != 0 {
3449 // Hands off own p to the locked m,
3450 // then blocks waiting for a new p.
3455 execute(gp, inheritTime)
3458 // dropg removes the association between m and the current goroutine m->curg (gp for short).
3459 // Typically a caller sets gp's status away from Grunning and then
3460 // immediately calls dropg to finish the job. The caller is also responsible
3461 // for arranging that gp will be restarted using ready at an
3462 // appropriate time. After calling dropg and arranging for gp to be
3463 // readied later, the caller can do other work but eventually should
3464 // call schedule to restart the scheduling of goroutines on this m.
3468 setMNoWB(&gp.m.curg.m, nil)
3469 setGNoWB(&gp.m.curg, nil)
3472 // checkTimers runs any timers for the P that are ready.
3473 // If now is not 0 it is the current time.
3474 // It returns the passed time or the current time if now was passed as 0.
3475 // and the time when the next timer should run or 0 if there is no next timer,
3476 // and reports whether it ran any timers.
3477 // If the time when the next timer should run is not 0,
3478 // it is always larger than the returned time.
3479 // We pass now in and out to avoid extra calls of nanotime.
3481 //go:yeswritebarrierrec
3482 func checkTimers(pp *p, now int64) (rnow, pollUntil int64, ran bool) {
3483 // If it's not yet time for the first timer, or the first adjusted
3484 // timer, then there is nothing to do.
3485 next := pp.timer0When.Load()
3486 nextAdj := pp.timerModifiedEarliest.Load()
3487 if next == 0 || (nextAdj != 0 && nextAdj < next) {
3492 // No timers to run or adjust.
3493 return now, 0, false
3500 // Next timer is not ready to run, but keep going
3501 // if we would clear deleted timers.
3502 // This corresponds to the condition below where
3503 // we decide whether to call clearDeletedTimers.
3504 if pp != getg().m.p.ptr() || int(pp.deletedTimers.Load()) <= int(pp.numTimers.Load()/4) {
3505 return now, next, false
3509 lock(&pp.timersLock)
3511 if len(pp.timers) > 0 {
3512 adjusttimers(pp, now)
3513 for len(pp.timers) > 0 {
3514 // Note that runtimer may temporarily unlock
3516 if tw := runtimer(pp, now); tw != 0 {
3526 // If this is the local P, and there are a lot of deleted timers,
3527 // clear them out. We only do this for the local P to reduce
3528 // lock contention on timersLock.
3529 if pp == getg().m.p.ptr() && int(pp.deletedTimers.Load()) > len(pp.timers)/4 {
3530 clearDeletedTimers(pp)
3533 unlock(&pp.timersLock)
3535 return now, pollUntil, ran
3538 func parkunlock_c(gp *g, lock unsafe.Pointer) bool {
3539 unlock((*mutex)(lock))
3543 // park continuation on g0.
3544 func park_m(gp *g) {
3548 traceGoPark(mp.waittraceev, mp.waittraceskip)
3551 // N.B. Not using casGToWaiting here because the waitreason is
3552 // set by park_m's caller.
3553 casgstatus(gp, _Grunning, _Gwaiting)
3556 if fn := mp.waitunlockf; fn != nil {
3557 ok := fn(gp, mp.waitlock)
3558 mp.waitunlockf = nil
3562 traceGoUnpark(gp, 2)
3564 casgstatus(gp, _Gwaiting, _Grunnable)
3565 execute(gp, true) // Schedule it back, never returns.
3571 func goschedImpl(gp *g) {
3572 status := readgstatus(gp)
3573 if status&^_Gscan != _Grunning {
3575 throw("bad g status")
3577 casgstatus(gp, _Grunning, _Grunnable)
3586 // Gosched continuation on g0.
3587 func gosched_m(gp *g) {
3594 // goschedguarded is a forbidden-states-avoided version of gosched_m.
3595 func goschedguarded_m(gp *g) {
3597 if !canPreemptM(gp.m) {
3598 gogo(&gp.sched) // never return
3607 func gopreempt_m(gp *g) {
3614 // preemptPark parks gp and puts it in _Gpreempted.
3617 func preemptPark(gp *g) {
3619 traceGoPark(traceEvGoBlock, 0)
3621 status := readgstatus(gp)
3622 if status&^_Gscan != _Grunning {
3624 throw("bad g status")
3627 if gp.asyncSafePoint {
3628 // Double-check that async preemption does not
3629 // happen in SPWRITE assembly functions.
3630 // isAsyncSafePoint must exclude this case.
3631 f := findfunc(gp.sched.pc)
3633 throw("preempt at unknown pc")
3635 if f.flag&funcFlag_SPWRITE != 0 {
3636 println("runtime: unexpected SPWRITE function", funcname(f), "in async preempt")
3637 throw("preempt SPWRITE")
3641 // Transition from _Grunning to _Gscan|_Gpreempted. We can't
3642 // be in _Grunning when we dropg because then we'd be running
3643 // without an M, but the moment we're in _Gpreempted,
3644 // something could claim this G before we've fully cleaned it
3645 // up. Hence, we set the scan bit to lock down further
3646 // transitions until we can dropg.
3647 casGToPreemptScan(gp, _Grunning, _Gscan|_Gpreempted)
3649 casfrom_Gscanstatus(gp, _Gscan|_Gpreempted, _Gpreempted)
3653 // goyield is like Gosched, but it:
3654 // - emits a GoPreempt trace event instead of a GoSched trace event
3655 // - puts the current G on the runq of the current P instead of the globrunq
3661 func goyield_m(gp *g) {
3666 casgstatus(gp, _Grunning, _Grunnable)
3668 runqput(pp, gp, false)
3672 // Finishes execution of the current goroutine.
3683 // goexit continuation on g0.
3684 func goexit0(gp *g) {
3688 casgstatus(gp, _Grunning, _Gdead)
3689 gcController.addScannableStack(pp, -int64(gp.stack.hi-gp.stack.lo))
3690 if isSystemGoroutine(gp, false) {
3694 locked := gp.lockedm != 0
3697 gp.preemptStop = false
3698 gp.paniconfault = false
3699 gp._defer = nil // should be true already but just in case.
3700 gp._panic = nil // non-nil for Goexit during panic. points at stack-allocated data.
3702 gp.waitreason = waitReasonZero
3707 if gcBlackenEnabled != 0 && gp.gcAssistBytes > 0 {
3708 // Flush assist credit to the global pool. This gives
3709 // better information to pacing if the application is
3710 // rapidly creating an exiting goroutines.
3711 assistWorkPerByte := gcController.assistWorkPerByte.Load()
3712 scanCredit := int64(assistWorkPerByte * float64(gp.gcAssistBytes))
3713 gcController.bgScanCredit.Add(scanCredit)
3714 gp.gcAssistBytes = 0
3719 if GOARCH == "wasm" { // no threads yet on wasm
3721 schedule() // never returns
3724 if mp.lockedInt != 0 {
3725 print("invalid m->lockedInt = ", mp.lockedInt, "\n")
3726 throw("internal lockOSThread error")
3730 // The goroutine may have locked this thread because
3731 // it put it in an unusual kernel state. Kill it
3732 // rather than returning it to the thread pool.
3734 // Return to mstart, which will release the P and exit
3736 if GOOS != "plan9" { // See golang.org/issue/22227.
3739 // Clear lockedExt on plan9 since we may end up re-using
3747 // save updates getg().sched to refer to pc and sp so that a following
3748 // gogo will restore pc and sp.
3750 // save must not have write barriers because invoking a write barrier
3751 // can clobber getg().sched.
3754 //go:nowritebarrierrec
3755 func save(pc, sp uintptr) {
3758 if gp == gp.m.g0 || gp == gp.m.gsignal {
3759 // m.g0.sched is special and must describe the context
3760 // for exiting the thread. mstart1 writes to it directly.
3761 // m.gsignal.sched should not be used at all.
3762 // This check makes sure save calls do not accidentally
3763 // run in contexts where they'd write to system g's.
3764 throw("save on system g not allowed")
3771 // We need to ensure ctxt is zero, but can't have a write
3772 // barrier here. However, it should always already be zero.
3774 if gp.sched.ctxt != nil {
3779 // The goroutine g is about to enter a system call.
3780 // Record that it's not using the cpu anymore.
3781 // This is called only from the go syscall library and cgocall,
3782 // not from the low-level system calls used by the runtime.
3784 // Entersyscall cannot split the stack: the save must
3785 // make g->sched refer to the caller's stack segment, because
3786 // entersyscall is going to return immediately after.
3788 // Nothing entersyscall calls can split the stack either.
3789 // We cannot safely move the stack during an active call to syscall,
3790 // because we do not know which of the uintptr arguments are
3791 // really pointers (back into the stack).
3792 // In practice, this means that we make the fast path run through
3793 // entersyscall doing no-split things, and the slow path has to use systemstack
3794 // to run bigger things on the system stack.
3796 // reentersyscall is the entry point used by cgo callbacks, where explicitly
3797 // saved SP and PC are restored. This is needed when exitsyscall will be called
3798 // from a function further up in the call stack than the parent, as g->syscallsp
3799 // must always point to a valid stack frame. entersyscall below is the normal
3800 // entry point for syscalls, which obtains the SP and PC from the caller.
3803 // At the start of a syscall we emit traceGoSysCall to capture the stack trace.
3804 // If the syscall does not block, that is it, we do not emit any other events.
3805 // If the syscall blocks (that is, P is retaken), retaker emits traceGoSysBlock;
3806 // when syscall returns we emit traceGoSysExit and when the goroutine starts running
3807 // (potentially instantly, if exitsyscallfast returns true) we emit traceGoStart.
3808 // To ensure that traceGoSysExit is emitted strictly after traceGoSysBlock,
3809 // we remember current value of syscalltick in m (gp.m.syscalltick = gp.m.p.ptr().syscalltick),
3810 // whoever emits traceGoSysBlock increments p.syscalltick afterwards;
3811 // and we wait for the increment before emitting traceGoSysExit.
3812 // Note that the increment is done even if tracing is not enabled,
3813 // because tracing can be enabled in the middle of syscall. We don't want the wait to hang.
3816 func reentersyscall(pc, sp uintptr) {
3819 // Disable preemption because during this function g is in Gsyscall status,
3820 // but can have inconsistent g->sched, do not let GC observe it.
3823 // Entersyscall must not call any function that might split/grow the stack.
3824 // (See details in comment above.)
3825 // Catch calls that might, by replacing the stack guard with something that
3826 // will trip any stack check and leaving a flag to tell newstack to die.
3827 gp.stackguard0 = stackPreempt
3828 gp.throwsplit = true
3830 // Leave SP around for GC and traceback.
3834 casgstatus(gp, _Grunning, _Gsyscall)
3835 if gp.syscallsp < gp.stack.lo || gp.stack.hi < gp.syscallsp {
3836 systemstack(func() {
3837 print("entersyscall inconsistent ", hex(gp.syscallsp), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n")
3838 throw("entersyscall")
3843 systemstack(traceGoSysCall)
3844 // systemstack itself clobbers g.sched.{pc,sp} and we might
3845 // need them later when the G is genuinely blocked in a
3850 if sched.sysmonwait.Load() {
3851 systemstack(entersyscall_sysmon)
3855 if gp.m.p.ptr().runSafePointFn != 0 {
3856 // runSafePointFn may stack split if run on this stack
3857 systemstack(runSafePointFn)
3861 gp.m.syscalltick = gp.m.p.ptr().syscalltick
3862 gp.sysblocktraced = true
3867 atomic.Store(&pp.status, _Psyscall)
3868 if sched.gcwaiting.Load() {
3869 systemstack(entersyscall_gcwait)
3876 // Standard syscall entry used by the go syscall library and normal cgo calls.
3878 // This is exported via linkname to assembly in the syscall package and x/sys.
3881 //go:linkname entersyscall
3882 func entersyscall() {
3883 reentersyscall(getcallerpc(), getcallersp())
3886 func entersyscall_sysmon() {
3888 if sched.sysmonwait.Load() {
3889 sched.sysmonwait.Store(false)
3890 notewakeup(&sched.sysmonnote)
3895 func entersyscall_gcwait() {
3897 pp := gp.m.oldp.ptr()
3900 if sched.stopwait > 0 && atomic.Cas(&pp.status, _Psyscall, _Pgcstop) {
3906 if sched.stopwait--; sched.stopwait == 0 {
3907 notewakeup(&sched.stopnote)
3913 // The same as entersyscall(), but with a hint that the syscall is blocking.
3916 func entersyscallblock() {
3919 gp.m.locks++ // see comment in entersyscall
3920 gp.throwsplit = true
3921 gp.stackguard0 = stackPreempt // see comment in entersyscall
3922 gp.m.syscalltick = gp.m.p.ptr().syscalltick
3923 gp.sysblocktraced = true
3924 gp.m.p.ptr().syscalltick++
3926 // Leave SP around for GC and traceback.
3930 gp.syscallsp = gp.sched.sp
3931 gp.syscallpc = gp.sched.pc
3932 if gp.syscallsp < gp.stack.lo || gp.stack.hi < gp.syscallsp {
3936 systemstack(func() {
3937 print("entersyscallblock inconsistent ", hex(sp1), " ", hex(sp2), " ", hex(sp3), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n")
3938 throw("entersyscallblock")
3941 casgstatus(gp, _Grunning, _Gsyscall)
3942 if gp.syscallsp < gp.stack.lo || gp.stack.hi < gp.syscallsp {
3943 systemstack(func() {
3944 print("entersyscallblock inconsistent ", hex(sp), " ", hex(gp.sched.sp), " ", hex(gp.syscallsp), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n")
3945 throw("entersyscallblock")
3949 systemstack(entersyscallblock_handoff)
3951 // Resave for traceback during blocked call.
3952 save(getcallerpc(), getcallersp())
3957 func entersyscallblock_handoff() {
3960 traceGoSysBlock(getg().m.p.ptr())
3962 handoffp(releasep())
3965 // The goroutine g exited its system call.
3966 // Arrange for it to run on a cpu again.
3967 // This is called only from the go syscall library, not
3968 // from the low-level system calls used by the runtime.
3970 // Write barriers are not allowed because our P may have been stolen.
3972 // This is exported via linkname to assembly in the syscall package.
3975 //go:nowritebarrierrec
3976 //go:linkname exitsyscall
3977 func exitsyscall() {
3980 gp.m.locks++ // see comment in entersyscall
3981 if getcallersp() > gp.syscallsp {
3982 throw("exitsyscall: syscall frame is no longer valid")
3986 oldp := gp.m.oldp.ptr()
3988 if exitsyscallfast(oldp) {
3989 // When exitsyscallfast returns success, we have a P so can now use
3991 if goroutineProfile.active {
3992 // Make sure that gp has had its stack written out to the goroutine
3993 // profile, exactly as it was when the goroutine profiler first
3994 // stopped the world.
3995 systemstack(func() {
3996 tryRecordGoroutineProfileWB(gp)
4000 if oldp != gp.m.p.ptr() || gp.m.syscalltick != gp.m.p.ptr().syscalltick {
4001 systemstack(traceGoStart)
4004 // There's a cpu for us, so we can run.
4005 gp.m.p.ptr().syscalltick++
4006 // We need to cas the status and scan before resuming...
4007 casgstatus(gp, _Gsyscall, _Grunning)
4009 // Garbage collector isn't running (since we are),
4010 // so okay to clear syscallsp.
4014 // restore the preemption request in case we've cleared it in newstack
4015 gp.stackguard0 = stackPreempt
4017 // otherwise restore the real _StackGuard, we've spoiled it in entersyscall/entersyscallblock
4018 gp.stackguard0 = gp.stack.lo + _StackGuard
4020 gp.throwsplit = false
4022 if sched.disable.user && !schedEnabled(gp) {
4023 // Scheduling of this goroutine is disabled.
4032 // Wait till traceGoSysBlock event is emitted.
4033 // This ensures consistency of the trace (the goroutine is started after it is blocked).
4034 for oldp != nil && oldp.syscalltick == gp.m.syscalltick {
4037 // We can't trace syscall exit right now because we don't have a P.
4038 // Tracing code can invoke write barriers that cannot run without a P.
4039 // So instead we remember the syscall exit time and emit the event
4040 // in execute when we have a P.
4041 gp.sysexitticks = cputicks()
4046 // Call the scheduler.
4049 // Scheduler returned, so we're allowed to run now.
4050 // Delete the syscallsp information that we left for
4051 // the garbage collector during the system call.
4052 // Must wait until now because until gosched returns
4053 // we don't know for sure that the garbage collector
4056 gp.m.p.ptr().syscalltick++
4057 gp.throwsplit = false
4061 func exitsyscallfast(oldp *p) bool {
4064 // Freezetheworld sets stopwait but does not retake P's.
4065 if sched.stopwait == freezeStopWait {
4069 // Try to re-acquire the last P.
4070 if oldp != nil && oldp.status == _Psyscall && atomic.Cas(&oldp.status, _Psyscall, _Pidle) {
4071 // There's a cpu for us, so we can run.
4073 exitsyscallfast_reacquired()
4077 // Try to get any other idle P.
4078 if sched.pidle != 0 {
4080 systemstack(func() {
4081 ok = exitsyscallfast_pidle()
4082 if ok && trace.enabled {
4084 // Wait till traceGoSysBlock event is emitted.
4085 // This ensures consistency of the trace (the goroutine is started after it is blocked).
4086 for oldp.syscalltick == gp.m.syscalltick {
4100 // exitsyscallfast_reacquired is the exitsyscall path on which this G
4101 // has successfully reacquired the P it was running on before the
4105 func exitsyscallfast_reacquired() {
4107 if gp.m.syscalltick != gp.m.p.ptr().syscalltick {
4109 // The p was retaken and then enter into syscall again (since gp.m.syscalltick has changed).
4110 // traceGoSysBlock for this syscall was already emitted,
4111 // but here we effectively retake the p from the new syscall running on the same p.
4112 systemstack(func() {
4113 // Denote blocking of the new syscall.
4114 traceGoSysBlock(gp.m.p.ptr())
4115 // Denote completion of the current syscall.
4119 gp.m.p.ptr().syscalltick++
4123 func exitsyscallfast_pidle() bool {
4125 pp, _ := pidleget(0)
4126 if pp != nil && sched.sysmonwait.Load() {
4127 sched.sysmonwait.Store(false)
4128 notewakeup(&sched.sysmonnote)
4138 // exitsyscall slow path on g0.
4139 // Failed to acquire P, enqueue gp as runnable.
4141 // Called via mcall, so gp is the calling g from this M.
4143 //go:nowritebarrierrec
4144 func exitsyscall0(gp *g) {
4145 casgstatus(gp, _Gsyscall, _Grunnable)
4149 if schedEnabled(gp) {
4156 // Below, we stoplockedm if gp is locked. globrunqput releases
4157 // ownership of gp, so we must check if gp is locked prior to
4158 // committing the release by unlocking sched.lock, otherwise we
4159 // could race with another M transitioning gp from unlocked to
4161 locked = gp.lockedm != 0
4162 } else if sched.sysmonwait.Load() {
4163 sched.sysmonwait.Store(false)
4164 notewakeup(&sched.sysmonnote)
4169 execute(gp, false) // Never returns.
4172 // Wait until another thread schedules gp and so m again.
4174 // N.B. lockedm must be this M, as this g was running on this M
4175 // before entersyscall.
4177 execute(gp, false) // Never returns.
4180 schedule() // Never returns.
4183 // Called from syscall package before fork.
4185 //go:linkname syscall_runtime_BeforeFork syscall.runtime_BeforeFork
4187 func syscall_runtime_BeforeFork() {
4190 // Block signals during a fork, so that the child does not run
4191 // a signal handler before exec if a signal is sent to the process
4192 // group. See issue #18600.
4194 sigsave(&gp.m.sigmask)
4197 // This function is called before fork in syscall package.
4198 // Code between fork and exec must not allocate memory nor even try to grow stack.
4199 // Here we spoil g->_StackGuard to reliably detect any attempts to grow stack.
4200 // runtime_AfterFork will undo this in parent process, but not in child.
4201 gp.stackguard0 = stackFork
4204 // Called from syscall package after fork in parent.
4206 //go:linkname syscall_runtime_AfterFork syscall.runtime_AfterFork
4208 func syscall_runtime_AfterFork() {
4211 // See the comments in beforefork.
4212 gp.stackguard0 = gp.stack.lo + _StackGuard
4214 msigrestore(gp.m.sigmask)
4219 // inForkedChild is true while manipulating signals in the child process.
4220 // This is used to avoid calling libc functions in case we are using vfork.
4221 var inForkedChild bool
4223 // Called from syscall package after fork in child.
4224 // It resets non-sigignored signals to the default handler, and
4225 // restores the signal mask in preparation for the exec.
4227 // Because this might be called during a vfork, and therefore may be
4228 // temporarily sharing address space with the parent process, this must
4229 // not change any global variables or calling into C code that may do so.
4231 //go:linkname syscall_runtime_AfterForkInChild syscall.runtime_AfterForkInChild
4233 //go:nowritebarrierrec
4234 func syscall_runtime_AfterForkInChild() {
4235 // It's OK to change the global variable inForkedChild here
4236 // because we are going to change it back. There is no race here,
4237 // because if we are sharing address space with the parent process,
4238 // then the parent process can not be running concurrently.
4239 inForkedChild = true
4241 clearSignalHandlers()
4243 // When we are the child we are the only thread running,
4244 // so we know that nothing else has changed gp.m.sigmask.
4245 msigrestore(getg().m.sigmask)
4247 inForkedChild = false
4250 // pendingPreemptSignals is the number of preemption signals
4251 // that have been sent but not received. This is only used on Darwin.
4253 var pendingPreemptSignals atomic.Int32
4255 // Called from syscall package before Exec.
4257 //go:linkname syscall_runtime_BeforeExec syscall.runtime_BeforeExec
4258 func syscall_runtime_BeforeExec() {
4259 // Prevent thread creation during exec.
4262 // On Darwin, wait for all pending preemption signals to
4263 // be received. See issue #41702.
4264 if GOOS == "darwin" || GOOS == "ios" {
4265 for pendingPreemptSignals.Load() > 0 {
4271 // Called from syscall package after Exec.
4273 //go:linkname syscall_runtime_AfterExec syscall.runtime_AfterExec
4274 func syscall_runtime_AfterExec() {
4278 // Allocate a new g, with a stack big enough for stacksize bytes.
4279 func malg(stacksize int32) *g {
4282 stacksize = round2(_StackSystem + stacksize)
4283 systemstack(func() {
4284 newg.stack = stackalloc(uint32(stacksize))
4286 newg.stackguard0 = newg.stack.lo + _StackGuard
4287 newg.stackguard1 = ^uintptr(0)
4288 // Clear the bottom word of the stack. We record g
4289 // there on gsignal stack during VDSO on ARM and ARM64.
4290 *(*uintptr)(unsafe.Pointer(newg.stack.lo)) = 0
4295 // Create a new g running fn.
4296 // Put it on the queue of g's waiting to run.
4297 // The compiler turns a go statement into a call to this.
4298 func newproc(fn *funcval) {
4301 systemstack(func() {
4302 newg := newproc1(fn, gp, pc)
4304 pp := getg().m.p.ptr()
4305 runqput(pp, newg, true)
4313 // Create a new g in state _Grunnable, starting at fn. callerpc is the
4314 // address of the go statement that created this. The caller is responsible
4315 // for adding the new g to the scheduler.
4316 func newproc1(fn *funcval, callergp *g, callerpc uintptr) *g {
4318 fatal("go of nil func value")
4321 mp := acquirem() // disable preemption because we hold M and P in local vars.
4325 newg = malg(_StackMin)
4326 casgstatus(newg, _Gidle, _Gdead)
4327 allgadd(newg) // publishes with a g->status of Gdead so GC scanner doesn't look at uninitialized stack.
4329 if newg.stack.hi == 0 {
4330 throw("newproc1: newg missing stack")
4333 if readgstatus(newg) != _Gdead {
4334 throw("newproc1: new g is not Gdead")
4337 totalSize := uintptr(4*goarch.PtrSize + sys.MinFrameSize) // extra space in case of reads slightly beyond frame
4338 totalSize = alignUp(totalSize, sys.StackAlign)
4339 sp := newg.stack.hi - totalSize
4343 *(*uintptr)(unsafe.Pointer(sp)) = 0
4345 spArg += sys.MinFrameSize
4348 memclrNoHeapPointers(unsafe.Pointer(&newg.sched), unsafe.Sizeof(newg.sched))
4351 newg.sched.pc = abi.FuncPCABI0(goexit) + sys.PCQuantum // +PCQuantum so that previous instruction is in same function
4352 newg.sched.g = guintptr(unsafe.Pointer(newg))
4353 gostartcallfn(&newg.sched, fn)
4354 newg.parentGoid = callergp.goid
4355 newg.gopc = callerpc
4356 newg.ancestors = saveAncestors(callergp)
4357 newg.startpc = fn.fn
4358 if isSystemGoroutine(newg, false) {
4361 // Only user goroutines inherit pprof labels.
4363 newg.labels = mp.curg.labels
4365 if goroutineProfile.active {
4366 // A concurrent goroutine profile is running. It should include
4367 // exactly the set of goroutines that were alive when the goroutine
4368 // profiler first stopped the world. That does not include newg, so
4369 // mark it as not needing a profile before transitioning it from
4371 newg.goroutineProfiled.Store(goroutineProfileSatisfied)
4374 // Track initial transition?
4375 newg.trackingSeq = uint8(fastrand())
4376 if newg.trackingSeq%gTrackingPeriod == 0 {
4377 newg.tracking = true
4379 casgstatus(newg, _Gdead, _Grunnable)
4380 gcController.addScannableStack(pp, int64(newg.stack.hi-newg.stack.lo))
4382 if pp.goidcache == pp.goidcacheend {
4383 // Sched.goidgen is the last allocated id,
4384 // this batch must be [sched.goidgen+1, sched.goidgen+GoidCacheBatch].
4385 // At startup sched.goidgen=0, so main goroutine receives goid=1.
4386 pp.goidcache = sched.goidgen.Add(_GoidCacheBatch)
4387 pp.goidcache -= _GoidCacheBatch - 1
4388 pp.goidcacheend = pp.goidcache + _GoidCacheBatch
4390 newg.goid = pp.goidcache
4393 newg.racectx = racegostart(callerpc)
4394 if newg.labels != nil {
4395 // See note in proflabel.go on labelSync's role in synchronizing
4396 // with the reads in the signal handler.
4397 racereleasemergeg(newg, unsafe.Pointer(&labelSync))
4401 traceGoCreate(newg, newg.startpc)
4408 // saveAncestors copies previous ancestors of the given caller g and
4409 // includes info for the current caller into a new set of tracebacks for
4410 // a g being created.
4411 func saveAncestors(callergp *g) *[]ancestorInfo {
4412 // Copy all prior info, except for the root goroutine (goid 0).
4413 if debug.tracebackancestors <= 0 || callergp.goid == 0 {
4416 var callerAncestors []ancestorInfo
4417 if callergp.ancestors != nil {
4418 callerAncestors = *callergp.ancestors
4420 n := int32(len(callerAncestors)) + 1
4421 if n > debug.tracebackancestors {
4422 n = debug.tracebackancestors
4424 ancestors := make([]ancestorInfo, n)
4425 copy(ancestors[1:], callerAncestors)
4427 var pcs [tracebackInnerFrames]uintptr
4428 npcs := gcallers(callergp, 0, pcs[:])
4429 ipcs := make([]uintptr, npcs)
4431 ancestors[0] = ancestorInfo{
4433 goid: callergp.goid,
4434 gopc: callergp.gopc,
4437 ancestorsp := new([]ancestorInfo)
4438 *ancestorsp = ancestors
4442 // Put on gfree list.
4443 // If local list is too long, transfer a batch to the global list.
4444 func gfput(pp *p, gp *g) {
4445 if readgstatus(gp) != _Gdead {
4446 throw("gfput: bad status (not Gdead)")
4449 stksize := gp.stack.hi - gp.stack.lo
4451 if stksize != uintptr(startingStackSize) {
4452 // non-standard stack size - free it.
4461 if pp.gFree.n >= 64 {
4467 for pp.gFree.n >= 32 {
4468 gp := pp.gFree.pop()
4470 if gp.stack.lo == 0 {
4477 lock(&sched.gFree.lock)
4478 sched.gFree.noStack.pushAll(noStackQ)
4479 sched.gFree.stack.pushAll(stackQ)
4480 sched.gFree.n += inc
4481 unlock(&sched.gFree.lock)
4485 // Get from gfree list.
4486 // If local list is empty, grab a batch from global list.
4487 func gfget(pp *p) *g {
4489 if pp.gFree.empty() && (!sched.gFree.stack.empty() || !sched.gFree.noStack.empty()) {
4490 lock(&sched.gFree.lock)
4491 // Move a batch of free Gs to the P.
4492 for pp.gFree.n < 32 {
4493 // Prefer Gs with stacks.
4494 gp := sched.gFree.stack.pop()
4496 gp = sched.gFree.noStack.pop()
4505 unlock(&sched.gFree.lock)
4508 gp := pp.gFree.pop()
4513 if gp.stack.lo != 0 && gp.stack.hi-gp.stack.lo != uintptr(startingStackSize) {
4514 // Deallocate old stack. We kept it in gfput because it was the
4515 // right size when the goroutine was put on the free list, but
4516 // the right size has changed since then.
4517 systemstack(func() {
4524 if gp.stack.lo == 0 {
4525 // Stack was deallocated in gfput or just above. Allocate a new one.
4526 systemstack(func() {
4527 gp.stack = stackalloc(startingStackSize)
4529 gp.stackguard0 = gp.stack.lo + _StackGuard
4532 racemalloc(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo)
4535 msanmalloc(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo)
4538 asanunpoison(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo)
4544 // Purge all cached G's from gfree list to the global list.
4545 func gfpurge(pp *p) {
4551 for !pp.gFree.empty() {
4552 gp := pp.gFree.pop()
4554 if gp.stack.lo == 0 {
4561 lock(&sched.gFree.lock)
4562 sched.gFree.noStack.pushAll(noStackQ)
4563 sched.gFree.stack.pushAll(stackQ)
4564 sched.gFree.n += inc
4565 unlock(&sched.gFree.lock)
4568 // Breakpoint executes a breakpoint trap.
4573 // dolockOSThread is called by LockOSThread and lockOSThread below
4574 // after they modify m.locked. Do not allow preemption during this call,
4575 // or else the m might be different in this function than in the caller.
4578 func dolockOSThread() {
4579 if GOARCH == "wasm" {
4580 return // no threads on wasm yet
4583 gp.m.lockedg.set(gp)
4584 gp.lockedm.set(gp.m)
4587 // LockOSThread wires the calling goroutine to its current operating system thread.
4588 // The calling goroutine will always execute in that thread,
4589 // and no other goroutine will execute in it,
4590 // until the calling goroutine has made as many calls to
4591 // UnlockOSThread as to LockOSThread.
4592 // If the calling goroutine exits without unlocking the thread,
4593 // the thread will be terminated.
4595 // All init functions are run on the startup thread. Calling LockOSThread
4596 // from an init function will cause the main function to be invoked on
4599 // A goroutine should call LockOSThread before calling OS services or
4600 // non-Go library functions that depend on per-thread state.
4603 func LockOSThread() {
4604 if atomic.Load(&newmHandoff.haveTemplateThread) == 0 && GOOS != "plan9" {
4605 // If we need to start a new thread from the locked
4606 // thread, we need the template thread. Start it now
4607 // while we're in a known-good state.
4608 startTemplateThread()
4612 if gp.m.lockedExt == 0 {
4614 panic("LockOSThread nesting overflow")
4620 func lockOSThread() {
4621 getg().m.lockedInt++
4625 // dounlockOSThread is called by UnlockOSThread and unlockOSThread below
4626 // after they update m->locked. Do not allow preemption during this call,
4627 // or else the m might be in different in this function than in the caller.
4630 func dounlockOSThread() {
4631 if GOARCH == "wasm" {
4632 return // no threads on wasm yet
4635 if gp.m.lockedInt != 0 || gp.m.lockedExt != 0 {
4642 // UnlockOSThread undoes an earlier call to LockOSThread.
4643 // If this drops the number of active LockOSThread calls on the
4644 // calling goroutine to zero, it unwires the calling goroutine from
4645 // its fixed operating system thread.
4646 // If there are no active LockOSThread calls, this is a no-op.
4648 // Before calling UnlockOSThread, the caller must ensure that the OS
4649 // thread is suitable for running other goroutines. If the caller made
4650 // any permanent changes to the state of the thread that would affect
4651 // other goroutines, it should not call this function and thus leave
4652 // the goroutine locked to the OS thread until the goroutine (and
4653 // hence the thread) exits.
4656 func UnlockOSThread() {
4658 if gp.m.lockedExt == 0 {
4666 func unlockOSThread() {
4668 if gp.m.lockedInt == 0 {
4669 systemstack(badunlockosthread)
4675 func badunlockosthread() {
4676 throw("runtime: internal error: misuse of lockOSThread/unlockOSThread")
4679 func gcount() int32 {
4680 n := int32(atomic.Loaduintptr(&allglen)) - sched.gFree.n - sched.ngsys.Load()
4681 for _, pp := range allp {
4685 // All these variables can be changed concurrently, so the result can be inconsistent.
4686 // But at least the current goroutine is running.
4693 func mcount() int32 {
4694 return int32(sched.mnext - sched.nmfreed)
4698 signalLock atomic.Uint32
4700 // Must hold signalLock to write. Reads may be lock-free, but
4701 // signalLock should be taken to synchronize with changes.
4705 func _System() { _System() }
4706 func _ExternalCode() { _ExternalCode() }
4707 func _LostExternalCode() { _LostExternalCode() }
4708 func _GC() { _GC() }
4709 func _LostSIGPROFDuringAtomic64() { _LostSIGPROFDuringAtomic64() }
4710 func _VDSO() { _VDSO() }
4712 // Called if we receive a SIGPROF signal.
4713 // Called by the signal handler, may run during STW.
4715 //go:nowritebarrierrec
4716 func sigprof(pc, sp, lr uintptr, gp *g, mp *m) {
4717 if prof.hz.Load() == 0 {
4721 // If mp.profilehz is 0, then profiling is not enabled for this thread.
4722 // We must check this to avoid a deadlock between setcpuprofilerate
4723 // and the call to cpuprof.add, below.
4724 if mp != nil && mp.profilehz == 0 {
4728 // On mips{,le}/arm, 64bit atomics are emulated with spinlocks, in
4729 // runtime/internal/atomic. If SIGPROF arrives while the program is inside
4730 // the critical section, it creates a deadlock (when writing the sample).
4731 // As a workaround, create a counter of SIGPROFs while in critical section
4732 // to store the count, and pass it to sigprof.add() later when SIGPROF is
4733 // received from somewhere else (with _LostSIGPROFDuringAtomic64 as pc).
4734 if GOARCH == "mips" || GOARCH == "mipsle" || GOARCH == "arm" {
4735 if f := findfunc(pc); f.valid() {
4736 if hasPrefix(funcname(f), "runtime/internal/atomic") {
4737 cpuprof.lostAtomic++
4741 if GOARCH == "arm" && goarm < 7 && GOOS == "linux" && pc&0xffff0000 == 0xffff0000 {
4742 // runtime/internal/atomic functions call into kernel
4743 // helpers on arm < 7. See
4744 // runtime/internal/atomic/sys_linux_arm.s.
4745 cpuprof.lostAtomic++
4750 // Profiling runs concurrently with GC, so it must not allocate.
4751 // Set a trap in case the code does allocate.
4752 // Note that on windows, one thread takes profiles of all the
4753 // other threads, so mp is usually not getg().m.
4754 // In fact mp may not even be stopped.
4755 // See golang.org/issue/17165.
4756 getg().m.mallocing++
4759 var stk [maxCPUProfStack]uintptr
4761 if mp.ncgo > 0 && mp.curg != nil && mp.curg.syscallpc != 0 && mp.curg.syscallsp != 0 {
4763 // Check cgoCallersUse to make sure that we are not
4764 // interrupting other code that is fiddling with
4765 // cgoCallers. We are running in a signal handler
4766 // with all signals blocked, so we don't have to worry
4767 // about any other code interrupting us.
4768 if mp.cgoCallersUse.Load() == 0 && mp.cgoCallers != nil && mp.cgoCallers[0] != 0 {
4769 for cgoOff < len(mp.cgoCallers) && mp.cgoCallers[cgoOff] != 0 {
4772 n += copy(stk[:], mp.cgoCallers[:cgoOff])
4773 mp.cgoCallers[0] = 0
4776 // Collect Go stack that leads to the cgo call.
4777 u.initAt(mp.curg.syscallpc, mp.curg.syscallsp, 0, mp.curg, unwindSilentErrors)
4778 } else if usesLibcall() && mp.libcallg != 0 && mp.libcallpc != 0 && mp.libcallsp != 0 {
4779 // Libcall, i.e. runtime syscall on windows.
4780 // Collect Go stack that leads to the call.
4781 u.initAt(mp.libcallpc, mp.libcallsp, 0, mp.libcallg.ptr(), unwindSilentErrors)
4782 } else if mp != nil && mp.vdsoSP != 0 {
4783 // VDSO call, e.g. nanotime1 on Linux.
4784 // Collect Go stack that leads to the call.
4785 u.initAt(mp.vdsoPC, mp.vdsoSP, 0, gp, unwindSilentErrors|unwindJumpStack)
4787 u.initAt(pc, sp, lr, gp, unwindSilentErrors|unwindTrap|unwindJumpStack)
4789 n += tracebackPCs(&u, 0, stk[n:])
4792 // Normal traceback is impossible or has failed.
4793 // Account it against abstract "System" or "GC".
4796 pc = abi.FuncPCABIInternal(_VDSO) + sys.PCQuantum
4797 } else if pc > firstmoduledata.etext {
4798 // "ExternalCode" is better than "etext".
4799 pc = abi.FuncPCABIInternal(_ExternalCode) + sys.PCQuantum
4802 if mp.preemptoff != "" {
4803 stk[1] = abi.FuncPCABIInternal(_GC) + sys.PCQuantum
4805 stk[1] = abi.FuncPCABIInternal(_System) + sys.PCQuantum
4809 if prof.hz.Load() != 0 {
4810 // Note: it can happen on Windows that we interrupted a system thread
4811 // with no g, so gp could nil. The other nil checks are done out of
4812 // caution, but not expected to be nil in practice.
4813 var tagPtr *unsafe.Pointer
4814 if gp != nil && gp.m != nil && gp.m.curg != nil {
4815 tagPtr = &gp.m.curg.labels
4817 cpuprof.add(tagPtr, stk[:n])
4821 if gp != nil && gp.m != nil {
4822 if gp.m.curg != nil {
4827 traceCPUSample(gprof, pp, stk[:n])
4829 getg().m.mallocing--
4832 // setcpuprofilerate sets the CPU profiling rate to hz times per second.
4833 // If hz <= 0, setcpuprofilerate turns off CPU profiling.
4834 func setcpuprofilerate(hz int32) {
4835 // Force sane arguments.
4840 // Disable preemption, otherwise we can be rescheduled to another thread
4841 // that has profiling enabled.
4845 // Stop profiler on this thread so that it is safe to lock prof.
4846 // if a profiling signal came in while we had prof locked,
4847 // it would deadlock.
4848 setThreadCPUProfiler(0)
4850 for !prof.signalLock.CompareAndSwap(0, 1) {
4853 if prof.hz.Load() != hz {
4854 setProcessCPUProfiler(hz)
4857 prof.signalLock.Store(0)
4860 sched.profilehz = hz
4864 setThreadCPUProfiler(hz)
4870 // init initializes pp, which may be a freshly allocated p or a
4871 // previously destroyed p, and transitions it to status _Pgcstop.
4872 func (pp *p) init(id int32) {
4874 pp.status = _Pgcstop
4875 pp.sudogcache = pp.sudogbuf[:0]
4876 pp.deferpool = pp.deferpoolbuf[:0]
4878 if pp.mcache == nil {
4881 throw("missing mcache?")
4883 // Use the bootstrap mcache0. Only one P will get
4884 // mcache0: the one with ID 0.
4887 pp.mcache = allocmcache()
4890 if raceenabled && pp.raceprocctx == 0 {
4892 pp.raceprocctx = raceprocctx0
4893 raceprocctx0 = 0 // bootstrap
4895 pp.raceprocctx = raceproccreate()
4898 lockInit(&pp.timersLock, lockRankTimers)
4900 // This P may get timers when it starts running. Set the mask here
4901 // since the P may not go through pidleget (notably P 0 on startup).
4903 // Similarly, we may not go through pidleget before this P starts
4904 // running if it is P 0 on startup.
4908 // destroy releases all of the resources associated with pp and
4909 // transitions it to status _Pdead.
4911 // sched.lock must be held and the world must be stopped.
4912 func (pp *p) destroy() {
4913 assertLockHeld(&sched.lock)
4914 assertWorldStopped()
4916 // Move all runnable goroutines to the global queue
4917 for pp.runqhead != pp.runqtail {
4918 // Pop from tail of local queue
4920 gp := pp.runq[pp.runqtail%uint32(len(pp.runq))].ptr()
4921 // Push onto head of global queue
4924 if pp.runnext != 0 {
4925 globrunqputhead(pp.runnext.ptr())
4928 if len(pp.timers) > 0 {
4929 plocal := getg().m.p.ptr()
4930 // The world is stopped, but we acquire timersLock to
4931 // protect against sysmon calling timeSleepUntil.
4932 // This is the only case where we hold the timersLock of
4933 // more than one P, so there are no deadlock concerns.
4934 lock(&plocal.timersLock)
4935 lock(&pp.timersLock)
4936 moveTimers(plocal, pp.timers)
4938 pp.numTimers.Store(0)
4939 pp.deletedTimers.Store(0)
4940 pp.timer0When.Store(0)
4941 unlock(&pp.timersLock)
4942 unlock(&plocal.timersLock)
4944 // Flush p's write barrier buffer.
4945 if gcphase != _GCoff {
4949 for i := range pp.sudogbuf {
4950 pp.sudogbuf[i] = nil
4952 pp.sudogcache = pp.sudogbuf[:0]
4953 for j := range pp.deferpoolbuf {
4954 pp.deferpoolbuf[j] = nil
4956 pp.deferpool = pp.deferpoolbuf[:0]
4957 systemstack(func() {
4958 for i := 0; i < pp.mspancache.len; i++ {
4959 // Safe to call since the world is stopped.
4960 mheap_.spanalloc.free(unsafe.Pointer(pp.mspancache.buf[i]))
4962 pp.mspancache.len = 0
4964 pp.pcache.flush(&mheap_.pages)
4965 unlock(&mheap_.lock)
4967 freemcache(pp.mcache)
4972 if pp.timerRaceCtx != 0 {
4973 // The race detector code uses a callback to fetch
4974 // the proc context, so arrange for that callback
4975 // to see the right thing.
4976 // This hack only works because we are the only
4982 racectxend(pp.timerRaceCtx)
4987 raceprocdestroy(pp.raceprocctx)
4994 // Change number of processors.
4996 // sched.lock must be held, and the world must be stopped.
4998 // gcworkbufs must not be being modified by either the GC or the write barrier
4999 // code, so the GC must not be running if the number of Ps actually changes.
5001 // Returns list of Ps with local work, they need to be scheduled by the caller.
5002 func procresize(nprocs int32) *p {
5003 assertLockHeld(&sched.lock)
5004 assertWorldStopped()
5007 if old < 0 || nprocs <= 0 {
5008 throw("procresize: invalid arg")
5011 traceGomaxprocs(nprocs)
5014 // update statistics
5016 if sched.procresizetime != 0 {
5017 sched.totaltime += int64(old) * (now - sched.procresizetime)
5019 sched.procresizetime = now
5021 maskWords := (nprocs + 31) / 32
5023 // Grow allp if necessary.
5024 if nprocs > int32(len(allp)) {
5025 // Synchronize with retake, which could be running
5026 // concurrently since it doesn't run on a P.
5028 if nprocs <= int32(cap(allp)) {
5029 allp = allp[:nprocs]
5031 nallp := make([]*p, nprocs)
5032 // Copy everything up to allp's cap so we
5033 // never lose old allocated Ps.
5034 copy(nallp, allp[:cap(allp)])
5038 if maskWords <= int32(cap(idlepMask)) {
5039 idlepMask = idlepMask[:maskWords]
5040 timerpMask = timerpMask[:maskWords]
5042 nidlepMask := make([]uint32, maskWords)
5043 // No need to copy beyond len, old Ps are irrelevant.
5044 copy(nidlepMask, idlepMask)
5045 idlepMask = nidlepMask
5047 ntimerpMask := make([]uint32, maskWords)
5048 copy(ntimerpMask, timerpMask)
5049 timerpMask = ntimerpMask
5054 // initialize new P's
5055 for i := old; i < nprocs; i++ {
5061 atomicstorep(unsafe.Pointer(&allp[i]), unsafe.Pointer(pp))
5065 if gp.m.p != 0 && gp.m.p.ptr().id < nprocs {
5066 // continue to use the current P
5067 gp.m.p.ptr().status = _Prunning
5068 gp.m.p.ptr().mcache.prepareForSweep()
5070 // release the current P and acquire allp[0].
5072 // We must do this before destroying our current P
5073 // because p.destroy itself has write barriers, so we
5074 // need to do that from a valid P.
5077 // Pretend that we were descheduled
5078 // and then scheduled again to keep
5081 traceProcStop(gp.m.p.ptr())
5095 // g.m.p is now set, so we no longer need mcache0 for bootstrapping.
5098 // release resources from unused P's
5099 for i := nprocs; i < old; i++ {
5102 // can't free P itself because it can be referenced by an M in syscall
5106 if int32(len(allp)) != nprocs {
5108 allp = allp[:nprocs]
5109 idlepMask = idlepMask[:maskWords]
5110 timerpMask = timerpMask[:maskWords]
5115 for i := nprocs - 1; i >= 0; i-- {
5117 if gp.m.p.ptr() == pp {
5125 pp.link.set(runnablePs)
5129 stealOrder.reset(uint32(nprocs))
5130 var int32p *int32 = &gomaxprocs // make compiler check that gomaxprocs is an int32
5131 atomic.Store((*uint32)(unsafe.Pointer(int32p)), uint32(nprocs))
5133 // Notify the limiter that the amount of procs has changed.
5134 gcCPULimiter.resetCapacity(now, nprocs)
5139 // Associate p and the current m.
5141 // This function is allowed to have write barriers even if the caller
5142 // isn't because it immediately acquires pp.
5144 //go:yeswritebarrierrec
5145 func acquirep(pp *p) {
5146 // Do the part that isn't allowed to have write barriers.
5149 // Have p; write barriers now allowed.
5151 // Perform deferred mcache flush before this P can allocate
5152 // from a potentially stale mcache.
5153 pp.mcache.prepareForSweep()
5160 // wirep is the first step of acquirep, which actually associates the
5161 // current M to pp. This is broken out so we can disallow write
5162 // barriers for this part, since we don't yet have a P.
5164 //go:nowritebarrierrec
5170 throw("wirep: already in go")
5172 if pp.m != 0 || pp.status != _Pidle {
5177 print("wirep: p->m=", pp.m, "(", id, ") p->status=", pp.status, "\n")
5178 throw("wirep: invalid p state")
5182 pp.status = _Prunning
5185 // Disassociate p and the current m.
5186 func releasep() *p {
5190 throw("releasep: invalid arg")
5193 if pp.m.ptr() != gp.m || pp.status != _Prunning {
5194 print("releasep: m=", gp.m, " m->p=", gp.m.p.ptr(), " p->m=", hex(pp.m), " p->status=", pp.status, "\n")
5195 throw("releasep: invalid p state")
5198 traceProcStop(gp.m.p.ptr())
5206 func incidlelocked(v int32) {
5208 sched.nmidlelocked += v
5215 // Check for deadlock situation.
5216 // The check is based on number of running M's, if 0 -> deadlock.
5217 // sched.lock must be held.
5219 assertLockHeld(&sched.lock)
5221 // For -buildmode=c-shared or -buildmode=c-archive it's OK if
5222 // there are no running goroutines. The calling program is
5223 // assumed to be running.
5224 if islibrary || isarchive {
5228 // If we are dying because of a signal caught on an already idle thread,
5229 // freezetheworld will cause all running threads to block.
5230 // And runtime will essentially enter into deadlock state,
5231 // except that there is a thread that will call exit soon.
5232 if panicking.Load() > 0 {
5236 // If we are not running under cgo, but we have an extra M then account
5237 // for it. (It is possible to have an extra M on Windows without cgo to
5238 // accommodate callbacks created by syscall.NewCallback. See issue #6751
5241 if !iscgo && cgoHasExtraM {
5242 mp := lockextra(true)
5243 haveExtraM := extraMCount > 0
5250 run := mcount() - sched.nmidle - sched.nmidlelocked - sched.nmsys
5255 print("runtime: checkdead: nmidle=", sched.nmidle, " nmidlelocked=", sched.nmidlelocked, " mcount=", mcount(), " nmsys=", sched.nmsys, "\n")
5256 throw("checkdead: inconsistent counts")
5260 forEachG(func(gp *g) {
5261 if isSystemGoroutine(gp, false) {
5264 s := readgstatus(gp)
5265 switch s &^ _Gscan {
5272 print("runtime: checkdead: find g ", gp.goid, " in status ", s, "\n")
5273 throw("checkdead: runnable g")
5276 if grunning == 0 { // possible if main goroutine calls runtime·Goexit()
5277 unlock(&sched.lock) // unlock so that GODEBUG=scheddetail=1 doesn't hang
5278 fatal("no goroutines (main called runtime.Goexit) - deadlock!")
5281 // Maybe jump time forward for playground.
5283 if when := timeSleepUntil(); when < maxWhen {
5286 // Start an M to steal the timer.
5287 pp, _ := pidleget(faketime)
5289 // There should always be a free P since
5290 // nothing is running.
5291 throw("checkdead: no p for timer")
5295 // There should always be a free M since
5296 // nothing is running.
5297 throw("checkdead: no m for timer")
5299 // M must be spinning to steal. We set this to be
5300 // explicit, but since this is the only M it would
5301 // become spinning on its own anyways.
5302 sched.nmspinning.Add(1)
5305 notewakeup(&mp.park)
5310 // There are no goroutines running, so we can look at the P's.
5311 for _, pp := range allp {
5312 if len(pp.timers) > 0 {
5317 unlock(&sched.lock) // unlock so that GODEBUG=scheddetail=1 doesn't hang
5318 fatal("all goroutines are asleep - deadlock!")
5321 // forcegcperiod is the maximum time in nanoseconds between garbage
5322 // collections. If we go this long without a garbage collection, one
5323 // is forced to run.
5325 // This is a variable for testing purposes. It normally doesn't change.
5326 var forcegcperiod int64 = 2 * 60 * 1e9
5328 // needSysmonWorkaround is true if the workaround for
5329 // golang.org/issue/42515 is needed on NetBSD.
5330 var needSysmonWorkaround bool = false
5332 // Always runs without a P, so write barriers are not allowed.
5334 //go:nowritebarrierrec
5341 lasttrace := int64(0)
5342 idle := 0 // how many cycles in succession we had not wokeup somebody
5346 if idle == 0 { // start with 20us sleep...
5348 } else if idle > 50 { // start doubling the sleep after 1ms...
5351 if delay > 10*1000 { // up to 10ms
5356 // sysmon should not enter deep sleep if schedtrace is enabled so that
5357 // it can print that information at the right time.
5359 // It should also not enter deep sleep if there are any active P's so
5360 // that it can retake P's from syscalls, preempt long running G's, and
5361 // poll the network if all P's are busy for long stretches.
5363 // It should wakeup from deep sleep if any P's become active either due
5364 // to exiting a syscall or waking up due to a timer expiring so that it
5365 // can resume performing those duties. If it wakes from a syscall it
5366 // resets idle and delay as a bet that since it had retaken a P from a
5367 // syscall before, it may need to do it again shortly after the
5368 // application starts work again. It does not reset idle when waking
5369 // from a timer to avoid adding system load to applications that spend
5370 // most of their time sleeping.
5372 if debug.schedtrace <= 0 && (sched.gcwaiting.Load() || sched.npidle.Load() == gomaxprocs) {
5374 if sched.gcwaiting.Load() || sched.npidle.Load() == gomaxprocs {
5375 syscallWake := false
5376 next := timeSleepUntil()
5378 sched.sysmonwait.Store(true)
5380 // Make wake-up period small enough
5381 // for the sampling to be correct.
5382 sleep := forcegcperiod / 2
5383 if next-now < sleep {
5386 shouldRelax := sleep >= osRelaxMinNS
5390 syscallWake = notetsleep(&sched.sysmonnote, sleep)
5395 sched.sysmonwait.Store(false)
5396 noteclear(&sched.sysmonnote)
5406 lock(&sched.sysmonlock)
5407 // Update now in case we blocked on sysmonnote or spent a long time
5408 // blocked on schedlock or sysmonlock above.
5411 // trigger libc interceptors if needed
5412 if *cgo_yield != nil {
5413 asmcgocall(*cgo_yield, nil)
5415 // poll network if not polled for more than 10ms
5416 lastpoll := sched.lastpoll.Load()
5417 if netpollinited() && lastpoll != 0 && lastpoll+10*1000*1000 < now {
5418 sched.lastpoll.CompareAndSwap(lastpoll, now)
5419 list := netpoll(0) // non-blocking - returns list of goroutines
5421 // Need to decrement number of idle locked M's
5422 // (pretending that one more is running) before injectglist.
5423 // Otherwise it can lead to the following situation:
5424 // injectglist grabs all P's but before it starts M's to run the P's,
5425 // another M returns from syscall, finishes running its G,
5426 // observes that there is no work to do and no other running M's
5427 // and reports deadlock.
5433 if GOOS == "netbsd" && needSysmonWorkaround {
5434 // netpoll is responsible for waiting for timer
5435 // expiration, so we typically don't have to worry
5436 // about starting an M to service timers. (Note that
5437 // sleep for timeSleepUntil above simply ensures sysmon
5438 // starts running again when that timer expiration may
5439 // cause Go code to run again).
5441 // However, netbsd has a kernel bug that sometimes
5442 // misses netpollBreak wake-ups, which can lead to
5443 // unbounded delays servicing timers. If we detect this
5444 // overrun, then startm to get something to handle the
5447 // See issue 42515 and
5448 // https://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=50094.
5449 if next := timeSleepUntil(); next < now {
5453 if scavenger.sysmonWake.Load() != 0 {
5454 // Kick the scavenger awake if someone requested it.
5457 // retake P's blocked in syscalls
5458 // and preempt long running G's
5459 if retake(now) != 0 {
5464 // check if we need to force a GC
5465 if t := (gcTrigger{kind: gcTriggerTime, now: now}); t.test() && forcegc.idle.Load() {
5467 forcegc.idle.Store(false)
5469 list.push(forcegc.g)
5471 unlock(&forcegc.lock)
5473 if debug.schedtrace > 0 && lasttrace+int64(debug.schedtrace)*1000000 <= now {
5475 schedtrace(debug.scheddetail > 0)
5477 unlock(&sched.sysmonlock)
5481 type sysmontick struct {
5488 // forcePreemptNS is the time slice given to a G before it is
5490 const forcePreemptNS = 10 * 1000 * 1000 // 10ms
5492 func retake(now int64) uint32 {
5494 // Prevent allp slice changes. This lock will be completely
5495 // uncontended unless we're already stopping the world.
5497 // We can't use a range loop over allp because we may
5498 // temporarily drop the allpLock. Hence, we need to re-fetch
5499 // allp each time around the loop.
5500 for i := 0; i < len(allp); i++ {
5503 // This can happen if procresize has grown
5504 // allp but not yet created new Ps.
5507 pd := &pp.sysmontick
5510 if s == _Prunning || s == _Psyscall {
5511 // Preempt G if it's running for too long.
5512 t := int64(pp.schedtick)
5513 if int64(pd.schedtick) != t {
5514 pd.schedtick = uint32(t)
5516 } else if pd.schedwhen+forcePreemptNS <= now {
5518 // In case of syscall, preemptone() doesn't
5519 // work, because there is no M wired to P.
5524 // Retake P from syscall if it's there for more than 1 sysmon tick (at least 20us).
5525 t := int64(pp.syscalltick)
5526 if !sysretake && int64(pd.syscalltick) != t {
5527 pd.syscalltick = uint32(t)
5528 pd.syscallwhen = now
5531 // On the one hand we don't want to retake Ps if there is no other work to do,
5532 // but on the other hand we want to retake them eventually
5533 // because they can prevent the sysmon thread from deep sleep.
5534 if runqempty(pp) && sched.nmspinning.Load()+sched.npidle.Load() > 0 && pd.syscallwhen+10*1000*1000 > now {
5537 // Drop allpLock so we can take sched.lock.
5539 // Need to decrement number of idle locked M's
5540 // (pretending that one more is running) before the CAS.
5541 // Otherwise the M from which we retake can exit the syscall,
5542 // increment nmidle and report deadlock.
5544 if atomic.Cas(&pp.status, s, _Pidle) {
5561 // Tell all goroutines that they have been preempted and they should stop.
5562 // This function is purely best-effort. It can fail to inform a goroutine if a
5563 // processor just started running it.
5564 // No locks need to be held.
5565 // Returns true if preemption request was issued to at least one goroutine.
5566 func preemptall() bool {
5568 for _, pp := range allp {
5569 if pp.status != _Prunning {
5579 // Tell the goroutine running on processor P to stop.
5580 // This function is purely best-effort. It can incorrectly fail to inform the
5581 // goroutine. It can inform the wrong goroutine. Even if it informs the
5582 // correct goroutine, that goroutine might ignore the request if it is
5583 // simultaneously executing newstack.
5584 // No lock needs to be held.
5585 // Returns true if preemption request was issued.
5586 // The actual preemption will happen at some point in the future
5587 // and will be indicated by the gp->status no longer being
5589 func preemptone(pp *p) bool {
5591 if mp == nil || mp == getg().m {
5595 if gp == nil || gp == mp.g0 {
5601 // Every call in a goroutine checks for stack overflow by
5602 // comparing the current stack pointer to gp->stackguard0.
5603 // Setting gp->stackguard0 to StackPreempt folds
5604 // preemption into the normal stack overflow check.
5605 gp.stackguard0 = stackPreempt
5607 // Request an async preemption of this P.
5608 if preemptMSupported && debug.asyncpreemptoff == 0 {
5618 func schedtrace(detailed bool) {
5625 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)
5627 print(" gcwaiting=", sched.gcwaiting.Load(), " nmidlelocked=", sched.nmidlelocked, " stopwait=", sched.stopwait, " sysmonwait=", sched.sysmonwait.Load(), "\n")
5629 // We must be careful while reading data from P's, M's and G's.
5630 // Even if we hold schedlock, most data can be changed concurrently.
5631 // E.g. (p->m ? p->m->id : -1) can crash if p->m changes from non-nil to nil.
5632 for i, pp := range allp {
5634 h := atomic.Load(&pp.runqhead)
5635 t := atomic.Load(&pp.runqtail)
5637 print(" P", i, ": status=", pp.status, " schedtick=", pp.schedtick, " syscalltick=", pp.syscalltick, " m=")
5643 print(" runqsize=", t-h, " gfreecnt=", pp.gFree.n, " timerslen=", len(pp.timers), "\n")
5645 // In non-detailed mode format lengths of per-P run queues as:
5646 // [len1 len2 len3 len4]
5652 if i == len(allp)-1 {
5663 for mp := allm; mp != nil; mp = mp.alllink {
5665 print(" M", mp.id, ": p=")
5677 print(" mallocing=", mp.mallocing, " throwing=", mp.throwing, " preemptoff=", mp.preemptoff, " locks=", mp.locks, " dying=", mp.dying, " spinning=", mp.spinning, " blocked=", mp.blocked, " lockedg=")
5678 if lockedg := mp.lockedg.ptr(); lockedg != nil {
5686 forEachG(func(gp *g) {
5687 print(" G", gp.goid, ": status=", readgstatus(gp), "(", gp.waitreason.String(), ") m=")
5694 if lockedm := gp.lockedm.ptr(); lockedm != nil {
5704 // schedEnableUser enables or disables the scheduling of user
5707 // This does not stop already running user goroutines, so the caller
5708 // should first stop the world when disabling user goroutines.
5709 func schedEnableUser(enable bool) {
5711 if sched.disable.user == !enable {
5715 sched.disable.user = !enable
5717 n := sched.disable.n
5719 globrunqputbatch(&sched.disable.runnable, n)
5721 for ; n != 0 && sched.npidle.Load() != 0; n-- {
5729 // schedEnabled reports whether gp should be scheduled. It returns
5730 // false is scheduling of gp is disabled.
5732 // sched.lock must be held.
5733 func schedEnabled(gp *g) bool {
5734 assertLockHeld(&sched.lock)
5736 if sched.disable.user {
5737 return isSystemGoroutine(gp, true)
5742 // Put mp on midle list.
5743 // sched.lock must be held.
5744 // May run during STW, so write barriers are not allowed.
5746 //go:nowritebarrierrec
5748 assertLockHeld(&sched.lock)
5750 mp.schedlink = sched.midle
5756 // Try to get an m from midle list.
5757 // sched.lock must be held.
5758 // May run during STW, so write barriers are not allowed.
5760 //go:nowritebarrierrec
5762 assertLockHeld(&sched.lock)
5764 mp := sched.midle.ptr()
5766 sched.midle = mp.schedlink
5772 // Put gp on the global runnable queue.
5773 // sched.lock must be held.
5774 // May run during STW, so write barriers are not allowed.
5776 //go:nowritebarrierrec
5777 func globrunqput(gp *g) {
5778 assertLockHeld(&sched.lock)
5780 sched.runq.pushBack(gp)
5784 // Put gp at the head of the global runnable queue.
5785 // sched.lock must be held.
5786 // May run during STW, so write barriers are not allowed.
5788 //go:nowritebarrierrec
5789 func globrunqputhead(gp *g) {
5790 assertLockHeld(&sched.lock)
5796 // Put a batch of runnable goroutines on the global runnable queue.
5797 // This clears *batch.
5798 // sched.lock must be held.
5799 // May run during STW, so write barriers are not allowed.
5801 //go:nowritebarrierrec
5802 func globrunqputbatch(batch *gQueue, n int32) {
5803 assertLockHeld(&sched.lock)
5805 sched.runq.pushBackAll(*batch)
5810 // Try get a batch of G's from the global runnable queue.
5811 // sched.lock must be held.
5812 func globrunqget(pp *p, max int32) *g {
5813 assertLockHeld(&sched.lock)
5815 if sched.runqsize == 0 {
5819 n := sched.runqsize/gomaxprocs + 1
5820 if n > sched.runqsize {
5823 if max > 0 && n > max {
5826 if n > int32(len(pp.runq))/2 {
5827 n = int32(len(pp.runq)) / 2
5832 gp := sched.runq.pop()
5835 gp1 := sched.runq.pop()
5836 runqput(pp, gp1, false)
5841 // pMask is an atomic bitstring with one bit per P.
5844 // read returns true if P id's bit is set.
5845 func (p pMask) read(id uint32) bool {
5847 mask := uint32(1) << (id % 32)
5848 return (atomic.Load(&p[word]) & mask) != 0
5851 // set sets P id's bit.
5852 func (p pMask) set(id int32) {
5854 mask := uint32(1) << (id % 32)
5855 atomic.Or(&p[word], mask)
5858 // clear clears P id's bit.
5859 func (p pMask) clear(id int32) {
5861 mask := uint32(1) << (id % 32)
5862 atomic.And(&p[word], ^mask)
5865 // updateTimerPMask clears pp's timer mask if it has no timers on its heap.
5867 // Ideally, the timer mask would be kept immediately consistent on any timer
5868 // operations. Unfortunately, updating a shared global data structure in the
5869 // timer hot path adds too much overhead in applications frequently switching
5870 // between no timers and some timers.
5872 // As a compromise, the timer mask is updated only on pidleget / pidleput. A
5873 // running P (returned by pidleget) may add a timer at any time, so its mask
5874 // must be set. An idle P (passed to pidleput) cannot add new timers while
5875 // idle, so if it has no timers at that time, its mask may be cleared.
5877 // Thus, we get the following effects on timer-stealing in findrunnable:
5879 // - Idle Ps with no timers when they go idle are never checked in findrunnable
5880 // (for work- or timer-stealing; this is the ideal case).
5881 // - Running Ps must always be checked.
5882 // - Idle Ps whose timers are stolen must continue to be checked until they run
5883 // again, even after timer expiration.
5885 // When the P starts running again, the mask should be set, as a timer may be
5886 // added at any time.
5888 // TODO(prattmic): Additional targeted updates may improve the above cases.
5889 // e.g., updating the mask when stealing a timer.
5890 func updateTimerPMask(pp *p) {
5891 if pp.numTimers.Load() > 0 {
5895 // Looks like there are no timers, however another P may transiently
5896 // decrement numTimers when handling a timerModified timer in
5897 // checkTimers. We must take timersLock to serialize with these changes.
5898 lock(&pp.timersLock)
5899 if pp.numTimers.Load() == 0 {
5900 timerpMask.clear(pp.id)
5902 unlock(&pp.timersLock)
5905 // pidleput puts p on the _Pidle list. now must be a relatively recent call
5906 // to nanotime or zero. Returns now or the current time if now was zero.
5908 // This releases ownership of p. Once sched.lock is released it is no longer
5911 // sched.lock must be held.
5913 // May run during STW, so write barriers are not allowed.
5915 //go:nowritebarrierrec
5916 func pidleput(pp *p, now int64) int64 {
5917 assertLockHeld(&sched.lock)
5920 throw("pidleput: P has non-empty run queue")
5925 updateTimerPMask(pp) // clear if there are no timers.
5926 idlepMask.set(pp.id)
5927 pp.link = sched.pidle
5930 if !pp.limiterEvent.start(limiterEventIdle, now) {
5931 throw("must be able to track idle limiter event")
5936 // pidleget tries to get a p from the _Pidle list, acquiring ownership.
5938 // sched.lock must be held.
5940 // May run during STW, so write barriers are not allowed.
5942 //go:nowritebarrierrec
5943 func pidleget(now int64) (*p, int64) {
5944 assertLockHeld(&sched.lock)
5946 pp := sched.pidle.ptr()
5948 // Timer may get added at any time now.
5952 timerpMask.set(pp.id)
5953 idlepMask.clear(pp.id)
5954 sched.pidle = pp.link
5955 sched.npidle.Add(-1)
5956 pp.limiterEvent.stop(limiterEventIdle, now)
5961 // pidlegetSpinning tries to get a p from the _Pidle list, acquiring ownership.
5962 // This is called by spinning Ms (or callers than need a spinning M) that have
5963 // found work. If no P is available, this must synchronized with non-spinning
5964 // Ms that may be preparing to drop their P without discovering this work.
5966 // sched.lock must be held.
5968 // May run during STW, so write barriers are not allowed.
5970 //go:nowritebarrierrec
5971 func pidlegetSpinning(now int64) (*p, int64) {
5972 assertLockHeld(&sched.lock)
5974 pp, now := pidleget(now)
5976 // See "Delicate dance" comment in findrunnable. We found work
5977 // that we cannot take, we must synchronize with non-spinning
5978 // Ms that may be preparing to drop their P.
5979 sched.needspinning.Store(1)
5986 // runqempty reports whether pp has no Gs on its local run queue.
5987 // It never returns true spuriously.
5988 func runqempty(pp *p) bool {
5989 // Defend against a race where 1) pp has G1 in runqnext but runqhead == runqtail,
5990 // 2) runqput on pp kicks G1 to the runq, 3) runqget on pp empties runqnext.
5991 // Simply observing that runqhead == runqtail and then observing that runqnext == nil
5992 // does not mean the queue is empty.
5994 head := atomic.Load(&pp.runqhead)
5995 tail := atomic.Load(&pp.runqtail)
5996 runnext := atomic.Loaduintptr((*uintptr)(unsafe.Pointer(&pp.runnext)))
5997 if tail == atomic.Load(&pp.runqtail) {
5998 return head == tail && runnext == 0
6003 // To shake out latent assumptions about scheduling order,
6004 // we introduce some randomness into scheduling decisions
6005 // when running with the race detector.
6006 // The need for this was made obvious by changing the
6007 // (deterministic) scheduling order in Go 1.5 and breaking
6008 // many poorly-written tests.
6009 // With the randomness here, as long as the tests pass
6010 // consistently with -race, they shouldn't have latent scheduling
6012 const randomizeScheduler = raceenabled
6014 // runqput tries to put g on the local runnable queue.
6015 // If next is false, runqput adds g to the tail of the runnable queue.
6016 // If next is true, runqput puts g in the pp.runnext slot.
6017 // If the run queue is full, runnext puts g on the global queue.
6018 // Executed only by the owner P.
6019 func runqput(pp *p, gp *g, next bool) {
6020 if randomizeScheduler && next && fastrandn(2) == 0 {
6026 oldnext := pp.runnext
6027 if !pp.runnext.cas(oldnext, guintptr(unsafe.Pointer(gp))) {
6033 // Kick the old runnext out to the regular run queue.
6038 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with consumers
6040 if t-h < uint32(len(pp.runq)) {
6041 pp.runq[t%uint32(len(pp.runq))].set(gp)
6042 atomic.StoreRel(&pp.runqtail, t+1) // store-release, makes the item available for consumption
6045 if runqputslow(pp, gp, h, t) {
6048 // the queue is not full, now the put above must succeed
6052 // Put g and a batch of work from local runnable queue on global queue.
6053 // Executed only by the owner P.
6054 func runqputslow(pp *p, gp *g, h, t uint32) bool {
6055 var batch [len(pp.runq)/2 + 1]*g
6057 // First, grab a batch from local queue.
6060 if n != uint32(len(pp.runq)/2) {
6061 throw("runqputslow: queue is not full")
6063 for i := uint32(0); i < n; i++ {
6064 batch[i] = pp.runq[(h+i)%uint32(len(pp.runq))].ptr()
6066 if !atomic.CasRel(&pp.runqhead, h, h+n) { // cas-release, commits consume
6071 if randomizeScheduler {
6072 for i := uint32(1); i <= n; i++ {
6073 j := fastrandn(i + 1)
6074 batch[i], batch[j] = batch[j], batch[i]
6078 // Link the goroutines.
6079 for i := uint32(0); i < n; i++ {
6080 batch[i].schedlink.set(batch[i+1])
6083 q.head.set(batch[0])
6084 q.tail.set(batch[n])
6086 // Now put the batch on global queue.
6088 globrunqputbatch(&q, int32(n+1))
6093 // runqputbatch tries to put all the G's on q on the local runnable queue.
6094 // If the queue is full, they are put on the global queue; in that case
6095 // this will temporarily acquire the scheduler lock.
6096 // Executed only by the owner P.
6097 func runqputbatch(pp *p, q *gQueue, qsize int) {
6098 h := atomic.LoadAcq(&pp.runqhead)
6101 for !q.empty() && t-h < uint32(len(pp.runq)) {
6103 pp.runq[t%uint32(len(pp.runq))].set(gp)
6109 if randomizeScheduler {
6110 off := func(o uint32) uint32 {
6111 return (pp.runqtail + o) % uint32(len(pp.runq))
6113 for i := uint32(1); i < n; i++ {
6114 j := fastrandn(i + 1)
6115 pp.runq[off(i)], pp.runq[off(j)] = pp.runq[off(j)], pp.runq[off(i)]
6119 atomic.StoreRel(&pp.runqtail, t)
6122 globrunqputbatch(q, int32(qsize))
6127 // Get g from local runnable queue.
6128 // If inheritTime is true, gp should inherit the remaining time in the
6129 // current time slice. Otherwise, it should start a new time slice.
6130 // Executed only by the owner P.
6131 func runqget(pp *p) (gp *g, inheritTime bool) {
6132 // If there's a runnext, it's the next G to run.
6134 // If the runnext is non-0 and the CAS fails, it could only have been stolen by another P,
6135 // because other Ps can race to set runnext to 0, but only the current P can set it to non-0.
6136 // Hence, there's no need to retry this CAS if it fails.
6137 if next != 0 && pp.runnext.cas(next, 0) {
6138 return next.ptr(), true
6142 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with other consumers
6147 gp := pp.runq[h%uint32(len(pp.runq))].ptr()
6148 if atomic.CasRel(&pp.runqhead, h, h+1) { // cas-release, commits consume
6154 // runqdrain drains the local runnable queue of pp and returns all goroutines in it.
6155 // Executed only by the owner P.
6156 func runqdrain(pp *p) (drainQ gQueue, n uint32) {
6157 oldNext := pp.runnext
6158 if oldNext != 0 && pp.runnext.cas(oldNext, 0) {
6159 drainQ.pushBack(oldNext.ptr())
6164 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with other consumers
6170 if qn > uint32(len(pp.runq)) { // read inconsistent h and t
6174 if !atomic.CasRel(&pp.runqhead, h, h+qn) { // cas-release, commits consume
6178 // We've inverted the order in which it gets G's from the local P's runnable queue
6179 // and then advances the head pointer because we don't want to mess up the statuses of G's
6180 // while runqdrain() and runqsteal() are running in parallel.
6181 // Thus we should advance the head pointer before draining the local P into a gQueue,
6182 // so that we can update any gp.schedlink only after we take the full ownership of G,
6183 // meanwhile, other P's can't access to all G's in local P's runnable queue and steal them.
6184 // See https://groups.google.com/g/golang-dev/c/0pTKxEKhHSc/m/6Q85QjdVBQAJ for more details.
6185 for i := uint32(0); i < qn; i++ {
6186 gp := pp.runq[(h+i)%uint32(len(pp.runq))].ptr()
6193 // Grabs a batch of goroutines from pp's runnable queue into batch.
6194 // Batch is a ring buffer starting at batchHead.
6195 // Returns number of grabbed goroutines.
6196 // Can be executed by any P.
6197 func runqgrab(pp *p, batch *[256]guintptr, batchHead uint32, stealRunNextG bool) uint32 {
6199 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with other consumers
6200 t := atomic.LoadAcq(&pp.runqtail) // load-acquire, synchronize with the producer
6205 // Try to steal from pp.runnext.
6206 if next := pp.runnext; next != 0 {
6207 if pp.status == _Prunning {
6208 // Sleep to ensure that pp isn't about to run the g
6209 // we are about to steal.
6210 // The important use case here is when the g running
6211 // on pp ready()s another g and then almost
6212 // immediately blocks. Instead of stealing runnext
6213 // in this window, back off to give pp a chance to
6214 // schedule runnext. This will avoid thrashing gs
6215 // between different Ps.
6216 // A sync chan send/recv takes ~50ns as of time of
6217 // writing, so 3us gives ~50x overshoot.
6218 if GOOS != "windows" && GOOS != "openbsd" && GOOS != "netbsd" {
6221 // On some platforms system timer granularity is
6222 // 1-15ms, which is way too much for this
6223 // optimization. So just yield.
6227 if !pp.runnext.cas(next, 0) {
6230 batch[batchHead%uint32(len(batch))] = next
6236 if n > uint32(len(pp.runq)/2) { // read inconsistent h and t
6239 for i := uint32(0); i < n; i++ {
6240 g := pp.runq[(h+i)%uint32(len(pp.runq))]
6241 batch[(batchHead+i)%uint32(len(batch))] = g
6243 if atomic.CasRel(&pp.runqhead, h, h+n) { // cas-release, commits consume
6249 // Steal half of elements from local runnable queue of p2
6250 // and put onto local runnable queue of p.
6251 // Returns one of the stolen elements (or nil if failed).
6252 func runqsteal(pp, p2 *p, stealRunNextG bool) *g {
6254 n := runqgrab(p2, &pp.runq, t, stealRunNextG)
6259 gp := pp.runq[(t+n)%uint32(len(pp.runq))].ptr()
6263 h := atomic.LoadAcq(&pp.runqhead) // load-acquire, synchronize with consumers
6264 if t-h+n >= uint32(len(pp.runq)) {
6265 throw("runqsteal: runq overflow")
6267 atomic.StoreRel(&pp.runqtail, t+n) // store-release, makes the item available for consumption
6271 // A gQueue is a dequeue of Gs linked through g.schedlink. A G can only
6272 // be on one gQueue or gList at a time.
6273 type gQueue struct {
6278 // empty reports whether q is empty.
6279 func (q *gQueue) empty() bool {
6283 // push adds gp to the head of q.
6284 func (q *gQueue) push(gp *g) {
6285 gp.schedlink = q.head
6292 // pushBack adds gp to the tail of q.
6293 func (q *gQueue) pushBack(gp *g) {
6296 q.tail.ptr().schedlink.set(gp)
6303 // pushBackAll adds all Gs in q2 to the tail of q. After this q2 must
6305 func (q *gQueue) pushBackAll(q2 gQueue) {
6309 q2.tail.ptr().schedlink = 0
6311 q.tail.ptr().schedlink = q2.head
6318 // pop removes and returns the head of queue q. It returns nil if
6320 func (q *gQueue) pop() *g {
6323 q.head = gp.schedlink
6331 // popList takes all Gs in q and returns them as a gList.
6332 func (q *gQueue) popList() gList {
6333 stack := gList{q.head}
6338 // A gList is a list of Gs linked through g.schedlink. A G can only be
6339 // on one gQueue or gList at a time.
6344 // empty reports whether l is empty.
6345 func (l *gList) empty() bool {
6349 // push adds gp to the head of l.
6350 func (l *gList) push(gp *g) {
6351 gp.schedlink = l.head
6355 // pushAll prepends all Gs in q to l.
6356 func (l *gList) pushAll(q gQueue) {
6358 q.tail.ptr().schedlink = l.head
6363 // pop removes and returns the head of l. If l is empty, it returns nil.
6364 func (l *gList) pop() *g {
6367 l.head = gp.schedlink
6372 //go:linkname setMaxThreads runtime/debug.setMaxThreads
6373 func setMaxThreads(in int) (out int) {
6375 out = int(sched.maxmcount)
6376 if in > 0x7fffffff { // MaxInt32
6377 sched.maxmcount = 0x7fffffff
6379 sched.maxmcount = int32(in)
6387 func procPin() int {
6392 return int(mp.p.ptr().id)
6401 //go:linkname sync_runtime_procPin sync.runtime_procPin
6403 func sync_runtime_procPin() int {
6407 //go:linkname sync_runtime_procUnpin sync.runtime_procUnpin
6409 func sync_runtime_procUnpin() {
6413 //go:linkname sync_atomic_runtime_procPin sync/atomic.runtime_procPin
6415 func sync_atomic_runtime_procPin() int {
6419 //go:linkname sync_atomic_runtime_procUnpin sync/atomic.runtime_procUnpin
6421 func sync_atomic_runtime_procUnpin() {
6425 // Active spinning for sync.Mutex.
6427 //go:linkname sync_runtime_canSpin sync.runtime_canSpin
6429 func sync_runtime_canSpin(i int) bool {
6430 // sync.Mutex is cooperative, so we are conservative with spinning.
6431 // Spin only few times and only if running on a multicore machine and
6432 // GOMAXPROCS>1 and there is at least one other running P and local runq is empty.
6433 // As opposed to runtime mutex we don't do passive spinning here,
6434 // because there can be work on global runq or on other Ps.
6435 if i >= active_spin || ncpu <= 1 || gomaxprocs <= sched.npidle.Load()+sched.nmspinning.Load()+1 {
6438 if p := getg().m.p.ptr(); !runqempty(p) {
6444 //go:linkname sync_runtime_doSpin sync.runtime_doSpin
6446 func sync_runtime_doSpin() {
6447 procyield(active_spin_cnt)
6450 var stealOrder randomOrder
6452 // randomOrder/randomEnum are helper types for randomized work stealing.
6453 // They allow to enumerate all Ps in different pseudo-random orders without repetitions.
6454 // The algorithm is based on the fact that if we have X such that X and GOMAXPROCS
6455 // are coprime, then a sequences of (i + X) % GOMAXPROCS gives the required enumeration.
6456 type randomOrder struct {
6461 type randomEnum struct {
6468 func (ord *randomOrder) reset(count uint32) {
6470 ord.coprimes = ord.coprimes[:0]
6471 for i := uint32(1); i <= count; i++ {
6472 if gcd(i, count) == 1 {
6473 ord.coprimes = append(ord.coprimes, i)
6478 func (ord *randomOrder) start(i uint32) randomEnum {
6482 inc: ord.coprimes[i/ord.count%uint32(len(ord.coprimes))],
6486 func (enum *randomEnum) done() bool {
6487 return enum.i == enum.count
6490 func (enum *randomEnum) next() {
6492 enum.pos = (enum.pos + enum.inc) % enum.count
6495 func (enum *randomEnum) position() uint32 {
6499 func gcd(a, b uint32) uint32 {
6506 // An initTask represents the set of initializations that need to be done for a package.
6507 // Keep in sync with ../../test/initempty.go:initTask
6508 type initTask struct {
6509 // TODO: pack the first 3 fields more tightly?
6510 state uintptr // 0 = uninitialized, 1 = in progress, 2 = done
6513 // followed by ndeps instances of an *initTask, one per package depended on
6514 // followed by nfns pcs, one per init function to run
6517 // inittrace stores statistics for init functions which are
6518 // updated by malloc and newproc when active is true.
6519 var inittrace tracestat
6521 type tracestat struct {
6522 active bool // init tracing activation status
6523 id uint64 // init goroutine id
6524 allocs uint64 // heap allocations
6525 bytes uint64 // heap allocated bytes
6528 func doInit(t *initTask) {
6530 case 2: // fully initialized
6532 case 1: // initialization in progress
6533 throw("recursive call during initialization - linker skew")
6534 default: // not initialized yet
6535 t.state = 1 // initialization in progress
6537 for i := uintptr(0); i < t.ndeps; i++ {
6538 p := add(unsafe.Pointer(t), (3+i)*goarch.PtrSize)
6539 t2 := *(**initTask)(p)
6544 t.state = 2 // initialization done
6553 if inittrace.active {
6555 // Load stats non-atomically since tracinit is updated only by this init goroutine.
6559 firstFunc := add(unsafe.Pointer(t), (3+t.ndeps)*goarch.PtrSize)
6560 for i := uintptr(0); i < t.nfns; i++ {
6561 p := add(firstFunc, i*goarch.PtrSize)
6562 f := *(*func())(unsafe.Pointer(&p))
6566 if inittrace.active {
6568 // Load stats non-atomically since tracinit is updated only by this init goroutine.
6571 f := *(*func())(unsafe.Pointer(&firstFunc))
6572 pkg := funcpkgpath(findfunc(abi.FuncPCABIInternal(f)))
6575 print("init ", pkg, " @")
6576 print(string(fmtNSAsMS(sbuf[:], uint64(start-runtimeInitTime))), " ms, ")
6577 print(string(fmtNSAsMS(sbuf[:], uint64(end-start))), " ms clock, ")
6578 print(string(itoa(sbuf[:], after.bytes-before.bytes)), " bytes, ")
6579 print(string(itoa(sbuf[:], after.allocs-before.allocs)), " allocs")
6583 t.state = 2 // initialization done