blob: e693f7e05f868a715e95094c8709f9d7d4e9a91a [file] [log] [blame]
// Copyright 2014 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package runtime
import (
"runtime/internal/atomic"
"runtime/internal/sys"
"unsafe"
)
var buildVersion = sys.TheVersion
// Goroutine scheduler
// The scheduler's job is to distribute ready-to-run goroutines over worker threads.
//
// The main concepts are:
// G - goroutine.
// M - worker thread, or machine.
// P - processor, a resource that is required to execute Go code.
// M must have an associated P to execute Go code, however it can be
// blocked or in a syscall w/o an associated P.
//
// Design doc at https://golang.org/s/go11sched.
// Worker thread parking/unparking.
// We need to balance between keeping enough running worker threads to utilize
// available hardware parallelism and parking excessive running worker threads
// to conserve CPU resources and power. This is not simple for two reasons:
// (1) scheduler state is intentionally distributed (in particular, per-P work
// queues), so it is not possible to compute global predicates on fast paths;
// (2) for optimal thread management we would need to know the future (don't park
// a worker thread when a new goroutine will be readied in near future).
//
// Three rejected approaches that would work badly:
// 1. Centralize all scheduler state (would inhibit scalability).
// 2. Direct goroutine handoff. That is, when we ready a new goroutine and there
// is a spare P, unpark a thread and handoff it the thread and the goroutine.
// This would lead to thread state thrashing, as the thread that readied the
// goroutine can be out of work the very next moment, we will need to park it.
// Also, it would destroy locality of computation as we want to preserve
// dependent goroutines on the same thread; and introduce additional latency.
// 3. Unpark an additional thread whenever we ready a goroutine and there is an
// idle P, but don't do handoff. This would lead to excessive thread parking/
// unparking as the additional threads will instantly park without discovering
// any work to do.
//
// The current approach:
// We unpark an additional thread when we ready a goroutine if (1) there is an
// idle P and there are no "spinning" worker threads. A worker thread is considered
// spinning if it is out of local work and did not find work in global run queue/
// netpoller; the spinning state is denoted in m.spinning and in sched.nmspinning.
// Threads unparked this way are also considered spinning; we don't do goroutine
// handoff so such threads are out of work initially. Spinning threads do some
// spinning looking for work in per-P run queues before parking. If a spinning
// thread finds work it takes itself out of the spinning state and proceeds to
// execution. If it does not find work it takes itself out of the spinning state
// and then parks.
// If there is at least one spinning thread (sched.nmspinning>1), we don't unpark
// new threads when readying goroutines. To compensate for that, if the last spinning
// thread finds work and stops spinning, it must unpark a new spinning thread.
// This approach smooths out unjustified spikes of thread unparking,
// but at the same time guarantees eventual maximal CPU parallelism utilization.
//
// The main implementation complication is that we need to be very careful during
// spinning->non-spinning thread transition. This transition can race with submission
// of a new goroutine, and either one part or another needs to unpark another worker
// thread. If they both fail to do that, we can end up with semi-persistent CPU
// underutilization. The general pattern for goroutine readying is: submit a goroutine
// to local work queue, #StoreLoad-style memory barrier, check sched.nmspinning.
// The general pattern for spinning->non-spinning transition is: decrement nmspinning,
// #StoreLoad-style memory barrier, check all per-P work queues for new work.
// Note that all this complexity does not apply to global run queue as we are not
// sloppy about thread unparking when submitting to global queue. Also see comments
// for nmspinning manipulation.
var (
m0 m
g0 g
raceprocctx0 uintptr
)
//go:linkname runtime_init runtime.init
func runtime_init()
//go:linkname main_init main.init
func main_init()
// main_init_done is a signal used by cgocallbackg that initialization
// has been completed. It is made before _cgo_notify_runtime_init_done,
// so all cgo calls can rely on it existing. When main_init is complete,
// it is closed, meaning cgocallbackg can reliably receive from it.
var main_init_done chan bool
//go:linkname main_main main.main
func main_main()
// runtimeInitTime is the nanotime() at which the runtime started.
var runtimeInitTime int64
// Value to use for signal mask for newly created M's.
var initSigmask sigset
// The main goroutine.
func main() {
g := getg()
// Racectx of m0->g0 is used only as the parent of the main goroutine.
// It must not be used for anything else.
g.m.g0.racectx = 0
// Max stack size is 1 GB on 64-bit, 250 MB on 32-bit.
// Using decimal instead of binary GB and MB because
// they look nicer in the stack overflow failure message.
if sys.PtrSize == 8 {
maxstacksize = 1000000000
} else {
maxstacksize = 250000000
}
// Record when the world started.
runtimeInitTime = nanotime()
systemstack(func() {
newm(sysmon, nil)
})
// Lock the main goroutine onto this, the main OS thread,
// during initialization. Most programs won't care, but a few
// do require certain calls to be made by the main thread.
// Those can arrange for main.main to run in the main thread
// by calling runtime.LockOSThread during initialization
// to preserve the lock.
lockOSThread()
if g.m != &m0 {
throw("runtime.main not on m0")
}
runtime_init() // must be before defer
// Defer unlock so that runtime.Goexit during init does the unlock too.
needUnlock := true
defer func() {
if needUnlock {
unlockOSThread()
}
}()
gcenable()
main_init_done = make(chan bool)
if iscgo {
if _cgo_thread_start == nil {
throw("_cgo_thread_start missing")
}
if GOOS != "windows" {
if _cgo_setenv == nil {
throw("_cgo_setenv missing")
}
if _cgo_unsetenv == nil {
throw("_cgo_unsetenv missing")
}
}
if _cgo_notify_runtime_init_done == nil {
throw("_cgo_notify_runtime_init_done missing")
}
cgocall(_cgo_notify_runtime_init_done, nil)
}
main_init()
close(main_init_done)
needUnlock = false
unlockOSThread()
if isarchive || islibrary {
// A program compiled with -buildmode=c-archive or c-shared
// has a main, but it is not executed.
return
}
main_main()
if raceenabled {
racefini()
}
// Make racy client program work: if panicking on
// another goroutine at the same time as main returns,
// let the other goroutine finish printing the panic trace.
// Once it does, it will exit. See issue 3934.
if panicking != 0 {
gopark(nil, nil, "panicwait", traceEvGoStop, 1)
}
exit(0)
for {
var x *int32
*x = 0
}
}
// os_beforeExit is called from os.Exit(0).
//go:linkname os_beforeExit os.runtime_beforeExit
func os_beforeExit() {
if raceenabled {
racefini()
}
}
// start forcegc helper goroutine
func init() {
go forcegchelper()
}
func forcegchelper() {
forcegc.g = getg()
for {
lock(&forcegc.lock)
if forcegc.idle != 0 {
throw("forcegc: phase error")
}
atomic.Store(&forcegc.idle, 1)
goparkunlock(&forcegc.lock, "force gc (idle)", traceEvGoBlock, 1)
// this goroutine is explicitly resumed by sysmon
if debug.gctrace > 0 {
println("GC forced")
}
gcStart(gcBackgroundMode, true)
}
}
//go:nosplit
// Gosched yields the processor, allowing other goroutines to run. It does not
// suspend the current goroutine, so execution resumes automatically.
func Gosched() {
mcall(gosched_m)
}
// Puts the current goroutine into a waiting state and calls unlockf.
// If unlockf returns false, the goroutine is resumed.
// unlockf must not access this G's stack, as it may be moved between
// the call to gopark and the call to unlockf.
func gopark(unlockf func(*g, unsafe.Pointer) bool, lock unsafe.Pointer, reason string, traceEv byte, traceskip int) {
mp := acquirem()
gp := mp.curg
status := readgstatus(gp)
if status != _Grunning && status != _Gscanrunning {
throw("gopark: bad g status")
}
mp.waitlock = lock
mp.waitunlockf = *(*unsafe.Pointer)(unsafe.Pointer(&unlockf))
gp.waitreason = reason
mp.waittraceev = traceEv
mp.waittraceskip = traceskip
releasem(mp)
// can't do anything that might move the G between Ms here.
mcall(park_m)
}
// Puts the current goroutine into a waiting state and unlocks the lock.
// The goroutine can be made runnable again by calling goready(gp).
func goparkunlock(lock *mutex, reason string, traceEv byte, traceskip int) {
gopark(parkunlock_c, unsafe.Pointer(lock), reason, traceEv, traceskip)
}
func goready(gp *g, traceskip int) {
systemstack(func() {
ready(gp, traceskip, true)
})
}
//go:nosplit
func acquireSudog() *sudog {
// Delicate dance: the semaphore implementation calls
// acquireSudog, acquireSudog calls new(sudog),
// new calls malloc, malloc can call the garbage collector,
// and the garbage collector calls the semaphore implementation
// in stopTheWorld.
// Break the cycle by doing acquirem/releasem around new(sudog).
// The acquirem/releasem increments m.locks during new(sudog),
// which keeps the garbage collector from being invoked.
mp := acquirem()
pp := mp.p.ptr()
if len(pp.sudogcache) == 0 {
lock(&sched.sudoglock)
// First, try to grab a batch from central cache.
for len(pp.sudogcache) < cap(pp.sudogcache)/2 && sched.sudogcache != nil {
s := sched.sudogcache
sched.sudogcache = s.next
s.next = nil
pp.sudogcache = append(pp.sudogcache, s)
}
unlock(&sched.sudoglock)
// If the central cache is empty, allocate a new one.
if len(pp.sudogcache) == 0 {
pp.sudogcache = append(pp.sudogcache, new(sudog))
}
}
n := len(pp.sudogcache)
s := pp.sudogcache[n-1]
pp.sudogcache[n-1] = nil
pp.sudogcache = pp.sudogcache[:n-1]
if s.elem != nil {
throw("acquireSudog: found s.elem != nil in cache")
}
releasem(mp)
return s
}
//go:nosplit
func releaseSudog(s *sudog) {
if s.elem != nil {
throw("runtime: sudog with non-nil elem")
}
if s.selectdone != nil {
throw("runtime: sudog with non-nil selectdone")
}
if s.next != nil {
throw("runtime: sudog with non-nil next")
}
if s.prev != nil {
throw("runtime: sudog with non-nil prev")
}
if s.waitlink != nil {
throw("runtime: sudog with non-nil waitlink")
}
if s.c != nil {
throw("runtime: sudog with non-nil c")
}
gp := getg()
if gp.param != nil {
throw("runtime: releaseSudog with non-nil gp.param")
}
mp := acquirem() // avoid rescheduling to another P
pp := mp.p.ptr()
if len(pp.sudogcache) == cap(pp.sudogcache) {
// Transfer half of local cache to the central cache.
var first, last *sudog
for len(pp.sudogcache) > cap(pp.sudogcache)/2 {
n := len(pp.sudogcache)
p := pp.sudogcache[n-1]
pp.sudogcache[n-1] = nil
pp.sudogcache = pp.sudogcache[:n-1]
if first == nil {
first = p
} else {
last.next = p
}
last = p
}
lock(&sched.sudoglock)
last.next = sched.sudogcache
sched.sudogcache = first
unlock(&sched.sudoglock)
}
pp.sudogcache = append(pp.sudogcache, s)
releasem(mp)
}
// funcPC returns the entry PC of the function f.
// It assumes that f is a func value. Otherwise the behavior is undefined.
//go:nosplit
func funcPC(f interface{}) uintptr {
return **(**uintptr)(add(unsafe.Pointer(&f), sys.PtrSize))
}
// called from assembly
func badmcall(fn func(*g)) {
throw("runtime: mcall called on m->g0 stack")
}
func badmcall2(fn func(*g)) {
throw("runtime: mcall function returned")
}
func badreflectcall() {
panic(plainError("arg size to reflect.call more than 1GB"))
}
func lockedOSThread() bool {
gp := getg()
return gp.lockedm != nil && gp.m.lockedg != nil
}
var (
allgs []*g
allglock mutex
)
func allgadd(gp *g) {
if readgstatus(gp) == _Gidle {
throw("allgadd: bad status Gidle")
}
lock(&allglock)
allgs = append(allgs, gp)
allglen = uintptr(len(allgs))
// Grow GC rescan list if necessary.
if len(allgs) > cap(work.rescan.list) {
lock(&work.rescan.lock)
l := work.rescan.list
// Let append do the heavy lifting, but keep the
// length the same.
work.rescan.list = append(l[:cap(l)], 0)[:len(l)]
unlock(&work.rescan.lock)
}
unlock(&allglock)
}
const (
// Number of goroutine ids to grab from sched.goidgen to local per-P cache at once.
// 16 seems to provide enough amortization, but other than that it's mostly arbitrary number.
_GoidCacheBatch = 16
)
// The bootstrap sequence is:
//
// call osinit
// call schedinit
// make & queue new G
// call runtime·mstart
//
// The new G calls runtime·main.
func schedinit() {
// raceinit must be the first call to race detector.
// In particular, it must be done before mallocinit below calls racemapshadow.
_g_ := getg()
if raceenabled {
_g_.racectx, raceprocctx0 = raceinit()
}
sched.maxmcount = 10000
tracebackinit()
moduledataverify()
stackinit()
mallocinit()
mcommoninit(_g_.m)
alginit() // maps must not be used before this call
typelinksinit() // uses maps
itabsinit()
msigsave(_g_.m)
initSigmask = _g_.m.sigmask
goargs()
goenvs()
parsedebugvars()
gcinit()
sched.lastpoll = uint64(nanotime())
procs := int(ncpu)
if procs > _MaxGomaxprocs {
procs = _MaxGomaxprocs
}
if n := atoi(gogetenv("GOMAXPROCS")); n > 0 {
if n > _MaxGomaxprocs {
n = _MaxGomaxprocs
}
procs = n
}
if procresize(int32(procs)) != nil {
throw("unknown runnable goroutine during bootstrap")
}
if buildVersion == "" {
// Condition should never trigger. This code just serves
// to ensure runtime·buildVersion is kept in the resulting binary.
buildVersion = "unknown"
}
}
func dumpgstatus(gp *g) {
_g_ := getg()
print("runtime: gp: gp=", gp, ", goid=", gp.goid, ", gp->atomicstatus=", readgstatus(gp), "\n")
print("runtime: g: g=", _g_, ", goid=", _g_.goid, ", g->atomicstatus=", readgstatus(_g_), "\n")
}
func checkmcount() {
// sched lock is held
if sched.mcount > sched.maxmcount {
print("runtime: program exceeds ", sched.maxmcount, "-thread limit\n")
throw("thread exhaustion")
}
}
func mcommoninit(mp *m) {
_g_ := getg()
// g0 stack won't make sense for user (and is not necessary unwindable).
if _g_ != _g_.m.g0 {
callers(1, mp.createstack[:])
}
mp.fastrand = 0x49f6428a + uint32(mp.id) + uint32(cputicks())
if mp.fastrand == 0 {
mp.fastrand = 0x49f6428a
}
lock(&sched.lock)
mp.id = sched.mcount
sched.mcount++
checkmcount()
mpreinit(mp)
if mp.gsignal != nil {
mp.gsignal.stackguard1 = mp.gsignal.stack.lo + _StackGuard
}
// Add to allm so garbage collector doesn't free g->m
// when it is just in a register or thread-local storage.
mp.alllink = allm
// NumCgoCall() iterates over allm w/o schedlock,
// so we need to publish it safely.
atomicstorep(unsafe.Pointer(&allm), unsafe.Pointer(mp))
unlock(&sched.lock)
// Allocate memory to hold a cgo traceback if the cgo call crashes.
if iscgo || GOOS == "solaris" || GOOS == "windows" {
mp.cgoCallers = new(cgoCallers)
}
}
// Mark gp ready to run.
func ready(gp *g, traceskip int, next bool) {
if trace.enabled {
traceGoUnpark(gp, traceskip)
}
status := readgstatus(gp)
// Mark runnable.
_g_ := getg()
_g_.m.locks++ // disable preemption because it can be holding p in a local var
if status&^_Gscan != _Gwaiting {
dumpgstatus(gp)
throw("bad g->status in ready")
}
// status is Gwaiting or Gscanwaiting, make Grunnable and put on runq
casgstatus(gp, _Gwaiting, _Grunnable)
runqput(_g_.m.p.ptr(), gp, next)
if atomic.Load(&sched.npidle) != 0 && atomic.Load(&sched.nmspinning) == 0 { // TODO: fast atomic
wakep()
}
_g_.m.locks--
if _g_.m.locks == 0 && _g_.preempt { // restore the preemption request in Case we've cleared it in newstack
_g_.stackguard0 = stackPreempt
}
}
func gcprocs() int32 {
// Figure out how many CPUs to use during GC.
// Limited by gomaxprocs, number of actual CPUs, and MaxGcproc.
lock(&sched.lock)
n := gomaxprocs
if n > ncpu {
n = ncpu
}
if n > _MaxGcproc {
n = _MaxGcproc
}
if n > sched.nmidle+1 { // one M is currently running
n = sched.nmidle + 1
}
unlock(&sched.lock)
return n
}
func needaddgcproc() bool {
lock(&sched.lock)
n := gomaxprocs
if n > ncpu {
n = ncpu
}
if n > _MaxGcproc {
n = _MaxGcproc
}
n -= sched.nmidle + 1 // one M is currently running
unlock(&sched.lock)
return n > 0
}
func helpgc(nproc int32) {
_g_ := getg()
lock(&sched.lock)
pos := 0
for n := int32(1); n < nproc; n++ { // one M is currently running
if allp[pos].mcache == _g_.m.mcache {
pos++
}
mp := mget()
if mp == nil {
throw("gcprocs inconsistency")
}
mp.helpgc = n
mp.p.set(allp[pos])
mp.mcache = allp[pos].mcache
pos++
notewakeup(&mp.park)
}
unlock(&sched.lock)
}
// freezeStopWait is a large value that freezetheworld sets
// sched.stopwait to in order to request that all Gs permanently stop.
const freezeStopWait = 0x7fffffff
// Similar to stopTheWorld but best-effort and can be called several times.
// There is no reverse operation, used during crashing.
// This function must not lock any mutexes.
func freezetheworld() {
// stopwait and preemption requests can be lost
// due to races with concurrently executing threads,
// so try several times
for i := 0; i < 5; i++ {
// this should tell the scheduler to not start any new goroutines
sched.stopwait = freezeStopWait
atomic.Store(&sched.gcwaiting, 1)
// this should stop running goroutines
if !preemptall() {
break // no running goroutines
}
usleep(1000)
}
// to be sure
usleep(1000)
preemptall()
usleep(1000)
}
func isscanstatus(status uint32) bool {
if status == _Gscan {
throw("isscanstatus: Bad status Gscan")
}
return status&_Gscan == _Gscan
}
// All reads and writes of g's status go through readgstatus, casgstatus
// castogscanstatus, casfrom_Gscanstatus.
//go:nosplit
func readgstatus(gp *g) uint32 {
return atomic.Load(&gp.atomicstatus)
}
// Ownership of gcscanvalid:
//
// If gp is running (meaning status == _Grunning or _Grunning|_Gscan),
// then gp owns gp.gcscanvalid, and other goroutines must not modify it.
//
// Otherwise, a second goroutine can lock the scan state by setting _Gscan
// in the status bit and then modify gcscanvalid, and then unlock the scan state.
//
// Note that the first condition implies an exception to the second:
// if a second goroutine changes gp's status to _Grunning|_Gscan,
// that second goroutine still does not have the right to modify gcscanvalid.
// The Gscanstatuses are acting like locks and this releases them.
// If it proves to be a performance hit we should be able to make these
// simple atomic stores but for now we are going to throw if
// we see an inconsistent state.
func casfrom_Gscanstatus(gp *g, oldval, newval uint32) {
success := false
// Check that transition is valid.
switch oldval {
default:
print("runtime: casfrom_Gscanstatus bad oldval gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n")
dumpgstatus(gp)
throw("casfrom_Gscanstatus:top gp->status is not in scan state")
case _Gscanrunnable,
_Gscanwaiting,
_Gscanrunning,
_Gscansyscall:
if newval == oldval&^_Gscan {
success = atomic.Cas(&gp.atomicstatus, oldval, newval)
}
}
if !success {
print("runtime: casfrom_Gscanstatus failed gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n")
dumpgstatus(gp)
throw("casfrom_Gscanstatus: gp->status is not in scan state")
}
}
// This will return false if the gp is not in the expected status and the cas fails.
// This acts like a lock acquire while the casfromgstatus acts like a lock release.
func castogscanstatus(gp *g, oldval, newval uint32) bool {
switch oldval {
case _Grunnable,
_Grunning,
_Gwaiting,
_Gsyscall:
if newval == oldval|_Gscan {
return atomic.Cas(&gp.atomicstatus, oldval, newval)
}
}
print("runtime: castogscanstatus oldval=", hex(oldval), " newval=", hex(newval), "\n")
throw("castogscanstatus")
panic("not reached")
}
// If asked to move to or from a Gscanstatus this will throw. Use the castogscanstatus
// and casfrom_Gscanstatus instead.
// casgstatus will loop if the g->atomicstatus is in a Gscan status until the routine that
// put it in the Gscan state is finished.
//go:nosplit
func casgstatus(gp *g, oldval, newval uint32) {
if (oldval&_Gscan != 0) || (newval&_Gscan != 0) || oldval == newval {
systemstack(func() {
print("runtime: casgstatus: oldval=", hex(oldval), " newval=", hex(newval), "\n")
throw("casgstatus: bad incoming values")
})
}
if oldval == _Grunning && gp.gcscanvalid {
// If oldvall == _Grunning, then the actual status must be
// _Grunning or _Grunning|_Gscan; either way,
// we own gp.gcscanvalid, so it's safe to read.
// gp.gcscanvalid must not be true when we are running.
print("runtime: casgstatus ", hex(oldval), "->", hex(newval), " gp.status=", hex(gp.atomicstatus), " gp.gcscanvalid=true\n")
throw("casgstatus")
}
// See http://golang.org/cl/21503 for justification of the yield delay.
const yieldDelay = 5 * 1000
var nextYield int64
// loop if gp->atomicstatus is in a scan state giving
// GC time to finish and change the state to oldval.
for i := 0; !atomic.Cas(&gp.atomicstatus, oldval, newval); i++ {
if oldval == _Gwaiting && gp.atomicstatus == _Grunnable {
systemstack(func() {
throw("casgstatus: waiting for Gwaiting but is Grunnable")
})
}
// Help GC if needed.
// if gp.preemptscan && !gp.gcworkdone && (oldval == _Grunning || oldval == _Gsyscall) {
// gp.preemptscan = false
// systemstack(func() {
// gcphasework(gp)
// })
// }
// But meanwhile just yield.
if i == 0 {
nextYield = nanotime() + yieldDelay
}
if nanotime() < nextYield {
for x := 0; x < 10 && gp.atomicstatus != oldval; x++ {
procyield(1)
}
} else {
osyield()
nextYield = nanotime() + yieldDelay/2
}
}
if newval == _Grunning && gp.gcscanvalid {
// Run queueRescan on the system stack so it has more space.
systemstack(func() { queueRescan(gp) })
}
}
// casgstatus(gp, oldstatus, Gcopystack), assuming oldstatus is Gwaiting or Grunnable.
// Returns old status. Cannot call casgstatus directly, because we are racing with an
// async wakeup that might come in from netpoll. If we see Gwaiting from the readgstatus,
// it might have become Grunnable by the time we get to the cas. If we called casgstatus,
// it would loop waiting for the status to go back to Gwaiting, which it never will.
//go:nosplit
func casgcopystack(gp *g) uint32 {
for {
oldstatus := readgstatus(gp) &^ _Gscan
if oldstatus != _Gwaiting && oldstatus != _Grunnable {
throw("copystack: bad status, not Gwaiting or Grunnable")
}
if atomic.Cas(&gp.atomicstatus, oldstatus, _Gcopystack) {
return oldstatus
}
}
}
// scang blocks until gp's stack has been scanned.
// It might be scanned by scang or it might be scanned by the goroutine itself.
// Either way, the stack scan has completed when scang returns.
func scang(gp *g, gcw *gcWork) {
// Invariant; we (the caller, markroot for a specific goroutine) own gp.gcscandone.
// Nothing is racing with us now, but gcscandone might be set to true left over
// from an earlier round of stack scanning (we scan twice per GC).
// We use gcscandone to record whether the scan has been done during this round.
// It is important that the scan happens exactly once: if called twice,
// the installation of stack barriers will detect the double scan and die.
gp.gcscandone = false
// See http://golang.org/cl/21503 for justification of the yield delay.
const yieldDelay = 10 * 1000
var nextYield int64
// Endeavor to get gcscandone set to true,
// either by doing the stack scan ourselves or by coercing gp to scan itself.
// gp.gcscandone can transition from false to true when we're not looking
// (if we asked for preemption), so any time we lock the status using
// castogscanstatus we have to double-check that the scan is still not done.
loop:
for i := 0; !gp.gcscandone; i++ {
switch s := readgstatus(gp); s {
default:
dumpgstatus(gp)
throw("stopg: invalid status")
case _Gdead:
// No stack.
gp.gcscandone = true
break loop
case _Gcopystack:
// Stack being switched. Go around again.
case _Grunnable, _Gsyscall, _Gwaiting:
// Claim goroutine by setting scan bit.
// Racing with execution or readying of gp.
// The scan bit keeps them from running
// the goroutine until we're done.
if castogscanstatus(gp, s, s|_Gscan) {
if !gp.gcscandone {
scanstack(gp, gcw)
gp.gcscandone = true
}
restartg(gp)
break loop
}
case _Gscanwaiting:
// newstack is doing a scan for us right now. Wait.
case _Grunning:
// Goroutine running. Try to preempt execution so it can scan itself.
// The preemption handler (in newstack) does the actual scan.
// Optimization: if there is already a pending preemption request
// (from the previous loop iteration), don't bother with the atomics.
if gp.preemptscan && gp.preempt && gp.stackguard0 == stackPreempt {
break
}
// Ask for preemption and self scan.
if castogscanstatus(gp, _Grunning, _Gscanrunning) {
if !gp.gcscandone {
gp.preemptscan = true
gp.preempt = true
gp.stackguard0 = stackPreempt
}
casfrom_Gscanstatus(gp, _Gscanrunning, _Grunning)
}
}
if i == 0 {
nextYield = nanotime() + yieldDelay
}
if nanotime() < nextYield {
procyield(10)
} else {
osyield()
nextYield = nanotime() + yieldDelay/2
}
}
gp.preemptscan = false // cancel scan request if no longer needed
}
// The GC requests that this routine be moved from a scanmumble state to a mumble state.
func restartg(gp *g) {
s := readgstatus(gp)
switch s {
default:
dumpgstatus(gp)
throw("restartg: unexpected status")
case _Gdead:
// ok
case _Gscanrunnable,
_Gscanwaiting,
_Gscansyscall:
casfrom_Gscanstatus(gp, s, s&^_Gscan)
}
}
// stopTheWorld stops all P's from executing goroutines, interrupting
// all goroutines at GC safe points and records reason as the reason
// for the stop. On return, only the current goroutine's P is running.
// stopTheWorld must not be called from a system stack and the caller
// must not hold worldsema. The caller must call startTheWorld when
// other P's should resume execution.
//
// stopTheWorld is safe for multiple goroutines to call at the
// same time. Each will execute its own stop, and the stops will
// be serialized.
//
// This is also used by routines that do stack dumps. If the system is
// in panic or being exited, this may not reliably stop all
// goroutines.
func stopTheWorld(reason string) {
semacquire(&worldsema, false)
getg().m.preemptoff = reason
systemstack(stopTheWorldWithSema)
}
// startTheWorld undoes the effects of stopTheWorld.
func startTheWorld() {
systemstack(startTheWorldWithSema)
// worldsema must be held over startTheWorldWithSema to ensure
// gomaxprocs cannot change while worldsema is held.
semrelease(&worldsema)
getg().m.preemptoff = ""
}
// Holding worldsema grants an M the right to try to stop the world
// and prevents gomaxprocs from changing concurrently.
var worldsema uint32 = 1
// stopTheWorldWithSema is the core implementation of stopTheWorld.
// The caller is responsible for acquiring worldsema and disabling
// preemption first and then should stopTheWorldWithSema on the system
// stack:
//
// semacquire(&worldsema, false)
// m.preemptoff = "reason"
// systemstack(stopTheWorldWithSema)
//
// When finished, the caller must either call startTheWorld or undo
// these three operations separately:
//
// m.preemptoff = ""
// systemstack(startTheWorldWithSema)
// semrelease(&worldsema)
//
// It is allowed to acquire worldsema once and then execute multiple
// startTheWorldWithSema/stopTheWorldWithSema pairs.
// Other P's are able to execute between successive calls to
// startTheWorldWithSema and stopTheWorldWithSema.
// Holding worldsema causes any other goroutines invoking
// stopTheWorld to block.
func stopTheWorldWithSema() {
_g_ := getg()
// If we hold a lock, then we won't be able to stop another M
// that is blocked trying to acquire the lock.
if _g_.m.locks > 0 {
throw("stopTheWorld: holding locks")
}
lock(&sched.lock)
sched.stopwait = gomaxprocs
atomic.Store(&sched.gcwaiting, 1)
preemptall()
// stop current P
_g_.m.p.ptr().status = _Pgcstop // Pgcstop is only diagnostic.
sched.stopwait--
// try to retake all P's in Psyscall status
for i := 0; i < int(gomaxprocs); i++ {
p := allp[i]
s := p.status
if s == _Psyscall && atomic.Cas(&p.status, s, _Pgcstop) {
if trace.enabled {
traceGoSysBlock(p)
traceProcStop(p)
}
p.syscalltick++
sched.stopwait--
}
}
// stop idle P's
for {
p := pidleget()
if p == nil {
break
}
p.status = _Pgcstop
sched.stopwait--
}
wait := sched.stopwait > 0
unlock(&sched.lock)
// wait for remaining P's to stop voluntarily
if wait {
for {
// wait for 100us, then try to re-preempt in case of any races
if notetsleep(&sched.stopnote, 100*1000) {
noteclear(&sched.stopnote)
break
}
preemptall()
}
}
if sched.stopwait != 0 {
throw("stopTheWorld: not stopped")
}
for i := 0; i < int(gomaxprocs); i++ {
p := allp[i]
if p.status != _Pgcstop {
throw("stopTheWorld: not stopped")
}
}
}
func mhelpgc() {
_g_ := getg()
_g_.m.helpgc = -1
}
func startTheWorldWithSema() {
_g_ := getg()
_g_.m.locks++ // disable preemption because it can be holding p in a local var
gp := netpoll(false) // non-blocking
injectglist(gp)
add := needaddgcproc()
lock(&sched.lock)
procs := gomaxprocs
if newprocs != 0 {
procs = newprocs
newprocs = 0
}
p1 := procresize(procs)
sched.gcwaiting = 0
if sched.sysmonwait != 0 {
sched.sysmonwait = 0
notewakeup(&sched.sysmonnote)
}
unlock(&sched.lock)
for p1 != nil {
p := p1
p1 = p1.link.ptr()
if p.m != 0 {
mp := p.m.ptr()
p.m = 0
if mp.nextp != 0 {
throw("startTheWorld: inconsistent mp->nextp")
}
mp.nextp.set(p)
notewakeup(&mp.park)
} else {
// Start M to run P. Do not start another M below.
newm(nil, p)
add = false
}
}
// Wakeup an additional proc in case we have excessive runnable goroutines
// in local queues or in the global queue. If we don't, the proc will park itself.
// If we have lots of excessive work, resetspinning will unpark additional procs as necessary.
if atomic.Load(&sched.npidle) != 0 && atomic.Load(&sched.nmspinning) == 0 {
wakep()
}
if add {
// If GC could have used another helper proc, start one now,
// in the hope that it will be available next time.
// It would have been even better to start it before the collection,
// but doing so requires allocating memory, so it's tricky to
// coordinate. This lazy approach works out in practice:
// we don't mind if the first couple gc rounds don't have quite
// the maximum number of procs.
newm(mhelpgc, nil)
}
_g_.m.locks--
if _g_.m.locks == 0 && _g_.preempt { // restore the preemption request in case we've cleared it in newstack
_g_.stackguard0 = stackPreempt
}
}
// Called to start an M.
//go:nosplit
func mstart() {
_g_ := getg()
if _g_.stack.lo == 0 {
// Initialize stack bounds from system stack.
// Cgo may have left stack size in stack.hi.
size := _g_.stack.hi
if size == 0 {
size = 8192 * sys.StackGuardMultiplier
}
_g_.stack.hi = uintptr(noescape(unsafe.Pointer(&size)))
_g_.stack.lo = _g_.stack.hi - size + 1024
}
// Initialize stack guards so that we can start calling
// both Go and C functions with stack growth prologues.
_g_.stackguard0 = _g_.stack.lo + _StackGuard
_g_.stackguard1 = _g_.stackguard0
mstart1()
}
func mstart1() {
_g_ := getg()
if _g_ != _g_.m.g0 {
throw("bad runtime·mstart")
}
// Record top of stack for use by mcall.
// Once we call schedule we're never coming back,
// so other calls can reuse this stack space.
gosave(&_g_.m.g0.sched)
_g_.m.g0.sched.pc = ^uintptr(0) // make sure it is never used
asminit()
minit()
// Install signal handlers; after minit so that minit can
// prepare the thread to be able to handle the signals.
if _g_.m == &m0 {
// Create an extra M for callbacks on threads not created by Go.
if iscgo && !cgoHasExtraM {
cgoHasExtraM = true
newextram()
}
initsig(false)
}
if fn := _g_.m.mstartfn; fn != nil {
fn()
}
if _g_.m.helpgc != 0 {
_g_.m.helpgc = 0
stopm()
} else if _g_.m != &m0 {
acquirep(_g_.m.nextp.ptr())
_g_.m.nextp = 0
}
schedule()
}
// forEachP calls fn(p) for every P p when p reaches a GC safe point.
// If a P is currently executing code, this will bring the P to a GC
// safe point and execute fn on that P. If the P is not executing code
// (it is idle or in a syscall), this will call fn(p) directly while
// preventing the P from exiting its state. This does not ensure that
// fn will run on every CPU executing Go code, but it acts as a global
// memory barrier. GC uses this as a "ragged barrier."
//
// The caller must hold worldsema.
//
//go:systemstack
func forEachP(fn func(*p)) {
mp := acquirem()
_p_ := getg().m.p.ptr()
lock(&sched.lock)
if sched.safePointWait != 0 {
throw("forEachP: sched.safePointWait != 0")
}
sched.safePointWait = gomaxprocs - 1
sched.safePointFn = fn
// Ask all Ps to run the safe point function.
for _, p := range allp[:gomaxprocs] {
if p != _p_ {
atomic.Store(&p.runSafePointFn, 1)
}
}
preemptall()
// Any P entering _Pidle or _Psyscall from now on will observe
// p.runSafePointFn == 1 and will call runSafePointFn when
// changing its status to _Pidle/_Psyscall.
// Run safe point function for all idle Ps. sched.pidle will
// not change because we hold sched.lock.
for p := sched.pidle.ptr(); p != nil; p = p.link.ptr() {
if atomic.Cas(&p.runSafePointFn, 1, 0) {
fn(p)
sched.safePointWait--
}
}
wait := sched.safePointWait > 0
unlock(&sched.lock)
// Run fn for the current P.
fn(_p_)
// Force Ps currently in _Psyscall into _Pidle and hand them
// off to induce safe point function execution.
for i := 0; i < int(gomaxprocs); i++ {
p := allp[i]
s := p.status
if s == _Psyscall && p.runSafePointFn == 1 && atomic.Cas(&p.status, s, _Pidle) {
if trace.enabled {
traceGoSysBlock(p)
traceProcStop(p)
}
p.syscalltick++
handoffp(p)
}
}
// Wait for remaining Ps to run fn.
if wait {
for {
// Wait for 100us, then try to re-preempt in
// case of any races.
//
// Requires system stack.
if notetsleep(&sched.safePointNote, 100*1000) {
noteclear(&sched.safePointNote)
break
}
preemptall()
}
}
if sched.safePointWait != 0 {
throw("forEachP: not done")
}
for i := 0; i < int(gomaxprocs); i++ {
p := allp[i]
if p.runSafePointFn != 0 {
throw("forEachP: P did not run fn")
}
}
lock(&sched.lock)
sched.safePointFn = nil
unlock(&sched.lock)
releasem(mp)
}
// runSafePointFn runs the safe point function, if any, for this P.
// This should be called like
//
// if getg().m.p.runSafePointFn != 0 {
// runSafePointFn()
// }
//
// runSafePointFn must be checked on any transition in to _Pidle or
// _Psyscall to avoid a race where forEachP sees that the P is running
// just before the P goes into _Pidle/_Psyscall and neither forEachP
// nor the P run the safe-point function.
func runSafePointFn() {
p := getg().m.p.ptr()
// Resolve the race between forEachP running the safe-point
// function on this P's behalf and this P running the
// safe-point function directly.
if !atomic.Cas(&p.runSafePointFn, 1, 0) {
return
}
sched.safePointFn(p)
lock(&sched.lock)
sched.safePointWait--
if sched.safePointWait == 0 {
notewakeup(&sched.safePointNote)
}
unlock(&sched.lock)
}
// When running with cgo, we call _cgo_thread_start
// to start threads for us so that we can play nicely with
// foreign code.
var cgoThreadStart unsafe.Pointer
type cgothreadstart struct {
g guintptr
tls *uint64
fn unsafe.Pointer
}
// Allocate a new m unassociated with any thread.
// Can use p for allocation context if needed.
// fn is recorded as the new m's m.mstartfn.
//
// This function it known to the compiler to inhibit the
// go:nowritebarrierrec annotation because it uses P for allocation.
func allocm(_p_ *p, fn func()) *m {
_g_ := getg()
_g_.m.locks++ // disable GC because it can be called from sysmon
if _g_.m.p == 0 {
acquirep(_p_) // temporarily borrow p for mallocs in this function
}
mp := new(m)
mp.mstartfn = fn
mcommoninit(mp)
// In case of cgo or Solaris, pthread_create will make us a stack.
// Windows and Plan 9 will layout sched stack on OS stack.
if iscgo || GOOS == "solaris" || GOOS == "windows" || GOOS == "plan9" {
mp.g0 = malg(-1)
} else {
mp.g0 = malg(8192 * sys.StackGuardMultiplier)
}
mp.g0.m = mp
if _p_ == _g_.m.p.ptr() {
releasep()
}
_g_.m.locks--
if _g_.m.locks == 0 && _g_.preempt { // restore the preemption request in case we've cleared it in newstack
_g_.stackguard0 = stackPreempt
}
return mp
}
// needm is called when a cgo callback happens on a
// thread without an m (a thread not created by Go).
// In this case, needm is expected to find an m to use
// and return with m, g initialized correctly.
// Since m and g are not set now (likely nil, but see below)
// needm is limited in what routines it can call. In particular
// it can only call nosplit functions (textflag 7) and cannot
// do any scheduling that requires an m.
//
// In order to avoid needing heavy lifting here, we adopt
// the following strategy: there is a stack of available m's
// that can be stolen. Using compare-and-swap
// to pop from the stack has ABA races, so we simulate
// a lock by doing an exchange (via casp) to steal the stack
// head and replace the top pointer with MLOCKED (1).
// This serves as a simple spin lock that we can use even
// without an m. The thread that locks the stack in this way
// unlocks the stack by storing a valid stack head pointer.
//
// In order to make sure that there is always an m structure
// available to be stolen, we maintain the invariant that there
// is always one more than needed. At the beginning of the
// program (if cgo is in use) the list is seeded with a single m.
// If needm finds that it has taken the last m off the list, its job
// is - once it has installed its own m so that it can do things like
// allocate memory - to create a spare m and put it on the list.
//
// Each of these extra m's also has a g0 and a curg that are
// pressed into service as the scheduling stack and current
// goroutine for the duration of the cgo callback.
//
// When the callback is done with the m, it calls dropm to
// put the m back on the list.
//go:nosplit
func needm(x byte) {
if iscgo && !cgoHasExtraM {
// Can happen if C/C++ code calls Go from a global ctor.
// Can not throw, because scheduler is not initialized yet.
write(2, unsafe.Pointer(&earlycgocallback[0]), int32(len(earlycgocallback)))
exit(1)
}
// Lock extra list, take head, unlock popped list.
// nilokay=false is safe here because of the invariant above,
// that the extra list always contains or will soon contain
// at least one m.
mp := lockextra(false)
// Set needextram when we've just emptied the list,
// so that the eventual call into cgocallbackg will
// allocate a new m for the extra list. We delay the
// allocation until then so that it can be done
// after exitsyscall makes sure it is okay to be
// running at all (that is, there's no garbage collection
// running right now).
mp.needextram = mp.schedlink == 0
unlockextra(mp.schedlink.ptr())
// Save and block signals before installing g.
// Once g is installed, any incoming signals will try to execute,
// but we won't have the sigaltstack settings and other data
// set up appropriately until the end of minit, which will
// unblock the signals. This is the same dance as when
// starting a new m to run Go code via newosproc.
msigsave(mp)
sigblock()
// Install g (= m->g0) and set the stack bounds
// to match the current stack. We don't actually know
// how big the stack is, like we don't know how big any
// scheduling stack is, but we assume there's at least 32 kB,
// which is more than enough for us.
setg(mp.g0)
_g_ := getg()
_g_.stack.hi = uintptr(noescape(unsafe.Pointer(&x))) + 1024
_g_.stack.lo = uintptr(noescape(unsafe.Pointer(&x))) - 32*1024
_g_.stackguard0 = _g_.stack.lo + _StackGuard
// Initialize this thread to use the m.
asminit()
minit()
}
var earlycgocallback = []byte("fatal error: cgo callback before cgo call\n")
// newextram allocates m's and puts them on the extra list.
// It is called with a working local m, so that it can do things
// like call schedlock and allocate.
func newextram() {
c := atomic.Xchg(&extraMWaiters, 0)
if c > 0 {
for i := uint32(0); i < c; i++ {
oneNewExtraM()
}
} else {
// Make sure there is at least one extra M.
mp := lockextra(true)
unlockextra(mp)
if mp == nil {
oneNewExtraM()
}
}
}
// oneNewExtraM allocates an m and puts it on the extra list.
func oneNewExtraM() {
// Create extra goroutine locked to extra m.
// The goroutine is the context in which the cgo callback will run.
// The sched.pc will never be returned to, but setting it to
// goexit makes clear to the traceback routines where
// the goroutine stack ends.
mp := allocm(nil, nil)
gp := malg(4096)
gp.sched.pc = funcPC(goexit) + sys.PCQuantum
gp.sched.sp = gp.stack.hi
gp.sched.sp -= 4 * sys.RegSize // extra space in case of reads slightly beyond frame
gp.sched.lr = 0
gp.sched.g = guintptr(unsafe.Pointer(gp))
gp.syscallpc = gp.sched.pc
gp.syscallsp = gp.sched.sp
gp.stktopsp = gp.sched.sp
gp.gcscanvalid = true // fresh G, so no dequeueRescan necessary
gp.gcRescan = -1
// malg returns status as Gidle, change to Gsyscall before adding to allg
// where GC will see it.
casgstatus(gp, _Gidle, _Gsyscall)
gp.m = mp
mp.curg = gp
mp.locked = _LockInternal
mp.lockedg = gp
gp.lockedm = mp
gp.goid = int64(atomic.Xadd64(&sched.goidgen, 1))
if raceenabled {
gp.racectx = racegostart(funcPC(newextram))
}
// put on allg for garbage collector
allgadd(gp)
// Add m to the extra list.
mnext := lockextra(true)
mp.schedlink.set(mnext)
unlockextra(mp)
}
// dropm is called when a cgo callback has called needm but is now
// done with the callback and returning back into the non-Go thread.
// It puts the current m back onto the extra list.
//
// The main expense here is the call to signalstack to release the
// m's signal stack, and then the call to needm on the next callback
// from this thread. It is tempting to try to save the m for next time,
// which would eliminate both these costs, but there might not be
// a next time: the current thread (which Go does not control) might exit.
// If we saved the m for that thread, there would be an m leak each time
// such a thread exited. Instead, we acquire and release an m on each
// call. These should typically not be scheduling operations, just a few
// atomics, so the cost should be small.
//
// TODO(rsc): An alternative would be to allocate a dummy pthread per-thread
// variable using pthread_key_create. Unlike the pthread keys we already use
// on OS X, this dummy key would never be read by Go code. It would exist
// only so that we could register at thread-exit-time destructor.
// That destructor would put the m back onto the extra list.
// This is purely a performance optimization. The current version,
// in which dropm happens on each cgo call, is still correct too.
// We may have to keep the current version on systems with cgo
// but without pthreads, like Windows.
func dropm() {
// Clear m and g, and return m to the extra list.
// After the call to setg we can only call nosplit functions
// with no pointer manipulation.
mp := getg().m
// Block signals before unminit.
// Unminit unregisters the signal handling stack (but needs g on some systems).
// Setg(nil) clears g, which is the signal handler's cue not to run Go handlers.
// It's important not to try to handle a signal between those two steps.
sigmask := mp.sigmask
sigblock()
unminit()
mnext := lockextra(true)
mp.schedlink.set(mnext)
setg(nil)
// Commit the release of mp.
unlockextra(mp)
msigrestore(sigmask)
}
// A helper function for EnsureDropM.
func getm() uintptr {
return uintptr(unsafe.Pointer(getg().m))
}
var extram uintptr
var extraMWaiters uint32
// lockextra locks the extra list and returns the list head.
// The caller must unlock the list by storing a new list head
// to extram. If nilokay is true, then lockextra will
// return a nil list head if that's what it finds. If nilokay is false,
// lockextra will keep waiting until the list head is no longer nil.
//go:nosplit
func lockextra(nilokay bool) *m {
const locked = 1
incr := false
for {
old := atomic.Loaduintptr(&extram)
if old == locked {
yield := osyield
yield()
continue
}
if old == 0 && !nilokay {
if !incr {
// Add 1 to the number of threads
// waiting for an M.
// This is cleared by newextram.
atomic.Xadd(&extraMWaiters, 1)
incr = true
}
usleep(1)
continue
}
if atomic.Casuintptr(&extram, old, locked) {
return (*m)(unsafe.Pointer(old))
}
yield := osyield
yield()
continue
}
}
//go:nosplit
func unlockextra(mp *m) {
atomic.Storeuintptr(&extram, uintptr(unsafe.Pointer(mp)))
}
// Create a new m. It will start off with a call to fn, or else the scheduler.
// fn needs to be static and not a heap allocated closure.
// May run with m.p==nil, so write barriers are not allowed.
//go:nowritebarrier
func newm(fn func(), _p_ *p) {
mp := allocm(_p_, fn)
mp.nextp.set(_p_)
mp.sigmask = initSigmask
if iscgo {
var ts cgothreadstart
if _cgo_thread_start == nil {
throw("_cgo_thread_start missing")
}
ts.g.set(mp.g0)
ts.tls = (*uint64)(unsafe.Pointer(&mp.tls[0]))
ts.fn = unsafe.Pointer(funcPC(mstart))
if msanenabled {
msanwrite(unsafe.Pointer(&ts), unsafe.Sizeof(ts))
}
asmcgocall(_cgo_thread_start, unsafe.Pointer(&ts))
return
}
newosproc(mp, unsafe.Pointer(mp.g0.stack.hi))
}
// Stops execution of the current m until new work is available.
// Returns with acquired P.
func stopm() {
_g_ := getg()
if _g_.m.locks != 0 {
throw("stopm holding locks")
}
if _g_.m.p != 0 {
throw("stopm holding p")
}
if _g_.m.spinning {
throw("stopm spinning")
}
retry:
lock(&sched.lock)
mput(_g_.m)
unlock(&sched.lock)
notesleep(&_g_.m.park)
noteclear(&_g_.m.park)
if _g_.m.helpgc != 0 {
gchelper()
_g_.m.helpgc = 0
_g_.m.mcache = nil
_g_.m.p = 0
goto retry
}
acquirep(_g_.m.nextp.ptr())
_g_.m.nextp = 0
}
func mspinning() {
// startm's caller incremented nmspinning. Set the new M's spinning.
getg().m.spinning = true
}
// Schedules some M to run the p (creates an M if necessary).
// If p==nil, tries to get an idle P, if no idle P's does nothing.
// May run with m.p==nil, so write barriers are not allowed.
// If spinning is set, the caller has incremented nmspinning and startm will
// either decrement nmspinning or set m.spinning in the newly started M.
//go:nowritebarrier
func startm(_p_ *p, spinning bool) {
lock(&sched.lock)
if _p_ == nil {
_p_ = pidleget()
if _p_ == nil {
unlock(&sched.lock)
if spinning {
// The caller incremented nmspinning, but there are no idle Ps,
// so it's okay to just undo the increment and give up.
if int32(atomic.Xadd(&sched.nmspinning, -1)) < 0 {
throw("startm: negative nmspinning")
}
}
return
}
}
mp := mget()
unlock(&sched.lock)
if mp == nil {
var fn func()
if spinning {
// The caller incremented nmspinning, so set m.spinning in the new M.
fn = mspinning
}
newm(fn, _p_)
return
}
if mp.spinning {
throw("startm: m is spinning")
}
if mp.nextp != 0 {
throw("startm: m has p")
}
if spinning && !runqempty(_p_) {
throw("startm: p has runnable gs")
}
// The caller incremented nmspinning, so set m.spinning in the new M.
mp.spinning = spinning
mp.nextp.set(_p_)
notewakeup(&mp.park)
}
// Hands off P from syscall or locked M.
// Always runs without a P, so write barriers are not allowed.
//go:nowritebarrier
func handoffp(_p_ *p) {
// handoffp must start an M in any situation where
// findrunnable would return a G to run on _p_.
// if it has local work, start it straight away
if !runqempty(_p_) || sched.runqsize != 0 {
startm(_p_, false)
return
}
// if it has GC work, start it straight away
if gcBlackenEnabled != 0 && gcMarkWorkAvailable(_p_) {
startm(_p_, false)
return
}
// no local work, check that there are no spinning/idle M's,
// otherwise our help is not required
if atomic.Load(&sched.nmspinning)+atomic.Load(&sched.npidle) == 0 && atomic.Cas(&sched.nmspinning, 0, 1) { // TODO: fast atomic
startm(_p_, true)
return
}
lock(&sched.lock)
if sched.gcwaiting != 0 {
_p_.status = _Pgcstop
sched.stopwait--
if sched.stopwait == 0 {
notewakeup(&sched.stopnote)
}
unlock(&sched.lock)
return
}
if _p_.runSafePointFn != 0 && atomic.Cas(&_p_.runSafePointFn, 1, 0) {
sched.safePointFn(_p_)
sched.safePointWait--
if sched.safePointWait == 0 {
notewakeup(&sched.safePointNote)
}
}
if sched.runqsize != 0 {
unlock(&sched.lock)
startm(_p_, false)
return
}
// If this is the last running P and nobody is polling network,
// need to wakeup another M to poll network.
if sched.npidle == uint32(gomaxprocs-1) && atomic.Load64(&sched.lastpoll) != 0 {
unlock(&sched.lock)
startm(_p_, false)
return
}
pidleput(_p_)
unlock(&sched.lock)
}
// Tries to add one more P to execute G's.
// Called when a G is made runnable (newproc, ready).
func wakep() {
// be conservative about spinning threads
if !atomic.Cas(&sched.nmspinning, 0, 1) {
return
}
startm(nil, true)
}
// Stops execution of the current m that is locked to a g until the g is runnable again.
// Returns with acquired P.
func stoplockedm() {
_g_ := getg()
if _g_.m.lockedg == nil || _g_.m.lockedg.lockedm != _g_.m {
throw("stoplockedm: inconsistent locking")
}
if _g_.m.p != 0 {
// Schedule another M to run this p.
_p_ := releasep()
handoffp(_p_)
}
incidlelocked(1)
// Wait until another thread schedules lockedg again.
notesleep(&_g_.m.park)
noteclear(&_g_.m.park)
status := readgstatus(_g_.m.lockedg)
if status&^_Gscan != _Grunnable {
print("runtime:stoplockedm: g is not Grunnable or Gscanrunnable\n")
dumpgstatus(_g_)
throw("stoplockedm: not runnable")
}
acquirep(_g_.m.nextp.ptr())
_g_.m.nextp = 0
}
// Schedules the locked m to run the locked gp.
// May run during STW, so write barriers are not allowed.
//go:nowritebarrier
func startlockedm(gp *g) {
_g_ := getg()
mp := gp.lockedm
if mp == _g_.m {
throw("startlockedm: locked to me")
}
if mp.nextp != 0 {
throw("startlockedm: m has p")
}
// directly handoff current P to the locked m
incidlelocked(-1)
_p_ := releasep()
mp.nextp.set(_p_)
notewakeup(&mp.park)
stopm()
}
// Stops the current m for stopTheWorld.
// Returns when the world is restarted.
func gcstopm() {
_g_ := getg()
if sched.gcwaiting == 0 {
throw("gcstopm: not waiting for gc")
}
if _g_.m.spinning {
_g_.m.spinning = false
// OK to just drop nmspinning here,
// startTheWorld will unpark threads as necessary.
if int32(atomic.Xadd(&sched.nmspinning, -1)) < 0 {
throw("gcstopm: negative nmspinning")
}
}
_p_ := releasep()
lock(&sched.lock)
_p_.status = _Pgcstop
sched.stopwait--
if sched.stopwait == 0 {
notewakeup(&sched.stopnote)
}
unlock(&sched.lock)
stopm()
}
// Schedules gp to run on the current M.
// If inheritTime is true, gp inherits the remaining time in the
// current time slice. Otherwise, it starts a new time slice.
// Never returns.
func execute(gp *g, inheritTime bool) {
_g_ := getg()
casgstatus(gp, _Grunnable, _Grunning)
gp.waitsince = 0
gp.preempt = false
gp.stackguard0 = gp.stack.lo + _StackGuard
if !inheritTime {
_g_.m.p.ptr().schedtick++
}
_g_.m.curg = gp
gp.m = _g_.m
// Check whether the profiler needs to be turned on or off.
hz := sched.profilehz
if _g_.m.profilehz != hz {
resetcpuprofiler(hz)
}
if trace.enabled {
// GoSysExit has to happen when we have a P, but before GoStart.
// So we emit it here.
if gp.syscallsp != 0 && gp.sysblocktraced {
traceGoSysExit(gp.sysexitticks)
}
traceGoStart()
}
gogo(&gp.sched)
}
// Finds a runnable goroutine to execute.
// Tries to steal from other P's, get g from global queue, poll network.
func findrunnable() (gp *g, inheritTime bool) {
_g_ := getg()
// The conditions here and in handoffp must agree: if
// findrunnable would return a G to run, handoffp must start
// an M.
top:
_p_ := _g_.m.p.ptr()
if sched.gcwaiting != 0 {
gcstopm()
goto top
}
if _p_.runSafePointFn != 0 {
runSafePointFn()
}
if fingwait && fingwake {
if gp := wakefing(); gp != nil {
ready(gp, 0, true)
}
}
// local runq
if gp, inheritTime := runqget(_p_); gp != nil {
return gp, inheritTime
}
// global runq
if sched.runqsize != 0 {
lock(&sched.lock)
gp := globrunqget(_p_, 0)
unlock(&sched.lock)
if gp != nil {
return gp, false
}
}
// Poll network.
// This netpoll is only an optimization before we resort to stealing.
// We can safely skip it if there a thread blocked in netpoll already.
// If there is any kind of logical race with that blocked thread
// (e.g. it has already returned from netpoll, but does not set lastpoll yet),
// this thread will do blocking netpoll below anyway.
if netpollinited() && sched.lastpoll != 0 {
if gp := netpoll(false); gp != nil { // non-blocking
// netpoll returns list of goroutines linked by schedlink.
injectglist(gp.schedlink.ptr())
casgstatus(gp, _Gwaiting, _Grunnable)
if trace.enabled {
traceGoUnpark(gp, 0)
}
return gp, false
}
}
// Steal work from other P's.
procs := uint32(gomaxprocs)
if atomic.Load(&sched.npidle) == procs-1 {
// Either GOMAXPROCS=1 or everybody, except for us, is idle already.
// New work can appear from returning syscall/cgocall, network or timers.
// Neither of that submits to local run queues, so no point in stealing.
goto stop
}
// If number of spinning M's >= number of busy P's, block.
// This is necessary to prevent excessive CPU consumption
// when GOMAXPROCS>>1 but the program parallelism is low.
if !_g_.m.spinning && 2*atomic.Load(&sched.nmspinning) >= procs-atomic.Load(&sched.npidle) { // TODO: fast atomic
goto stop
}
if !_g_.m.spinning {
_g_.m.spinning = true
atomic.Xadd(&sched.nmspinning, 1)
}
for i := 0; i < 4; i++ {
for enum := stealOrder.start(fastrand1()); !enum.done(); enum.next() {
if sched.gcwaiting != 0 {
goto top
}
stealRunNextG := i > 2 // first look for ready queues with more than 1 g
if gp := runqsteal(_p_, allp[enum.position()], stealRunNextG); gp != nil {
return gp, false
}
}
}
stop:
// We have nothing to do. If we're in the GC mark phase, can
// safely scan and blacken objects, and have work to do, run
// idle-time marking rather than give up the P.
if gcBlackenEnabled != 0 && _p_.gcBgMarkWorker != 0 && gcMarkWorkAvailable(_p_) {
_p_.gcMarkWorkerMode = gcMarkWorkerIdleMode
gp := _p_.gcBgMarkWorker.ptr()
casgstatus(gp, _Gwaiting, _Grunnable)
if trace.enabled {
traceGoUnpark(gp, 0)
}
return gp, false
}
// return P and block
lock(&sched.lock)
if sched.gcwaiting != 0 || _p_.runSafePointFn != 0 {
unlock(&sched.lock)
goto top
}
if sched.runqsize != 0 {
gp := globrunqget(_p_, 0)
unlock(&sched.lock)
return gp, false
}
if releasep() != _p_ {
throw("findrunnable: wrong p")
}
pidleput(_p_)
unlock(&sched.lock)
// Delicate dance: thread transitions from spinning to non-spinning state,
// potentially concurrently with submission of new goroutines. We must
// drop nmspinning first and then check all per-P queues again (with
// #StoreLoad memory barrier in between). If we do it the other way around,
// another thread can submit a goroutine after we've checked all run queues
// but before we drop nmspinning; as the result nobody will unpark a thread
// to run the goroutine.
// If we discover new work below, we need to restore m.spinning as a signal
// for resetspinning to unpark a new worker thread (because there can be more
// than one starving goroutine). However, if after discovering new work
// we also observe no idle Ps, it is OK to just park the current thread:
// the system is fully loaded so no spinning threads are required.
// Also see "Worker thread parking/unparking" comment at the top of the file.
wasSpinning := _g_.m.spinning
if _g_.m.spinning {
_g_.m.spinning = false
if int32(atomic.Xadd(&sched.nmspinning, -1)) < 0 {
throw("findrunnable: negative nmspinning")
}
}
// check all runqueues once again
for i := 0; i < int(gomaxprocs); i++ {
_p_ := allp[i]
if _p_ != nil && !runqempty(_p_) {
lock(&sched.lock)
_p_ = pidleget()
unlock(&sched.lock)
if _p_ != nil {
acquirep(_p_)
if wasSpinning {
_g_.m.spinning = true
atomic.Xadd(&sched.nmspinning, 1)
}
goto top
}
break
}
}
// poll network
if netpollinited() && atomic.Xchg64(&sched.lastpoll, 0) != 0 {
if _g_.m.p != 0 {
throw("findrunnable: netpoll with p")
}
if _g_.m.spinning {
throw("findrunnable: netpoll with spinning")
}
gp := netpoll(true) // block until new work is available
atomic.Store64(&sched.lastpoll, uint64(nanotime()))
if gp != nil {
lock(&sched.lock)
_p_ = pidleget()
unlock(&sched.lock)
if _p_ != nil {
acquirep(_p_)
injectglist(gp.schedlink.ptr())
casgstatus(gp, _Gwaiting, _Grunnable)
if trace.enabled {
traceGoUnpark(gp, 0)
}
return gp, false
}
injectglist(gp)
}
}
stopm()
goto top
}
func resetspinning() {
_g_ := getg()
if !_g_.m.spinning {
throw("resetspinning: not a spinning m")
}
_g_.m.spinning = false
nmspinning := atomic.Xadd(&sched.nmspinning, -1)
if int32(nmspinning) < 0 {
throw("findrunnable: negative nmspinning")
}
// M wakeup policy is deliberately somewhat conservative, so check if we
// need to wakeup another P here. See "Worker thread parking/unparking"
// comment at the top of the file for details.
if nmspinning == 0 && atomic.Load(&sched.npidle) > 0 {
wakep()
}
}
// Injects the list of runnable G's into the scheduler.
// Can run concurrently with GC.
func injectglist(glist *g) {
if glist == nil {
return
}
if trace.enabled {
for gp := glist; gp != nil; gp = gp.schedlink.ptr() {
traceGoUnpark(gp, 0)
}
}
lock(&sched.lock)
var n int
for n = 0; glist != nil; n++ {
gp := glist
glist = gp.schedlink.ptr()
casgstatus(gp, _Gwaiting, _Grunnable)
globrunqput(gp)
}
unlock(&sched.lock)
for ; n != 0 && sched.npidle != 0; n-- {
startm(nil, false)
}
}
// One round of scheduler: find a runnable goroutine and execute it.
// Never returns.
func schedule() {
_g_ := getg()
if _g_.m.locks != 0 {
throw("schedule: holding locks")
}
if _g_.m.lockedg != nil {
stoplockedm()
execute(_g_.m.lockedg, false) // Never returns.
}
top:
if sched.gcwaiting != 0 {
gcstopm()
goto top
}
if _g_.m.p.ptr().runSafePointFn != 0 {
runSafePointFn()
}
var gp *g
var inheritTime bool
if trace.enabled || trace.shutdown {
gp = traceReader()
if gp != nil {
casgstatus(gp, _Gwaiting, _Grunnable)
traceGoUnpark(gp, 0)
}
}
if gp == nil && gcBlackenEnabled != 0 {
gp = gcController.findRunnableGCWorker(_g_.m.p.ptr())
}
if gp == nil {
// Check the global runnable queue once in a while to ensure fairness.
// Otherwise two goroutines can completely occupy the local runqueue
// by constantly respawning each other.
if _g_.m.p.ptr().schedtick%61 == 0 && sched.runqsize > 0 {
lock(&sched.lock)
gp = globrunqget(_g_.m.p.ptr(), 1)
unlock(&sched.lock)
}
}
if gp == nil {
gp, inheritTime = runqget(_g_.m.p.ptr())
if gp != nil && _g_.m.spinning {
throw("schedule: spinning with local work")
}
}
if gp == nil {
gp, inheritTime = findrunnable() // blocks until work is available
}
// This thread is going to run a goroutine and is not spinning anymore,
// so if it was marked as spinning we need to reset it now and potentially
// start a new spinning M.
if _g_.m.spinning {
resetspinning()
}
if gp.lockedm != nil {
// Hands off own p to the locked m,
// then blocks waiting for a new p.
startlockedm(gp)
goto top
}
execute(gp, inheritTime)
}
// dropg removes the association between m and the current goroutine m->curg (gp for short).
// Typically a caller sets gp's status away from Grunning and then
// immediately calls dropg to finish the job. The caller is also responsible
// for arranging that gp will be restarted using ready at an
// appropriate time. After calling dropg and arranging for gp to be
// readied later, the caller can do other work but eventually should
// call schedule to restart the scheduling of goroutines on this m.
func dropg() {
_g_ := getg()
_g_.m.curg.m = nil
_g_.m.curg = nil
}
func parkunlock_c(gp *g, lock unsafe.Pointer) bool {
unlock((*mutex)(lock))
return true
}
// park continuation on g0.
func park_m(gp *g) {
_g_ := getg()
if trace.enabled {
traceGoPark(_g_.m.waittraceev, _g_.m.waittraceskip, gp)
}
casgstatus(gp, _Grunning, _Gwaiting)
dropg()
if _g_.m.waitunlockf != nil {
fn := *(*func(*g, unsafe.Pointer) bool)(unsafe.Pointer(&_g_.m.waitunlockf))
ok := fn(gp, _g_.m.waitlock)
_g_.m.waitunlockf = nil
_g_.m.waitlock = nil
if !ok {
if trace.enabled {
traceGoUnpark(gp, 2)
}
casgstatus(gp, _Gwaiting, _Grunnable)
execute(gp, true) // Schedule it back, never returns.
}
}
schedule()
}
func goschedImpl(gp *g) {
status := readgstatus(gp)
if status&^_Gscan != _Grunning {
dumpgstatus(gp)
throw("bad g status")
}
casgstatus(gp, _Grunning, _Grunnable)
dropg()
lock(&sched.lock)
globrunqput(gp)
unlock(&sched.lock)
schedule()
}
// Gosched continuation on g0.
func gosched_m(gp *g) {
if trace.enabled {
traceGoSched()
}
goschedImpl(gp)
}
func gopreempt_m(gp *g) {
if trace.enabled {
traceGoPreempt()
}
goschedImpl(gp)
}
// Finishes execution of the current goroutine.
func goexit1() {
if raceenabled {
racegoend()
}
if trace.enabled {
traceGoEnd()
}
mcall(goexit0)
}
// goexit continuation on g0.
func goexit0(gp *g) {
_g_ := getg()
casgstatus(gp, _Grunning, _Gdead)
if isSystemGoroutine(gp) {
atomic.Xadd(&sched.ngsys, -1)
}
gp.m = nil
gp.lockedm = nil
_g_.m.lockedg = nil
gp.paniconfault = false
gp._defer = nil // should be true already but just in case.
gp._panic = nil // non-nil for Goexit during panic. points at stack-allocated data.
gp.writebuf = nil
gp.waitreason = ""
gp.param = nil
// Note that gp's stack scan is now "valid" because it has no
// stack. We could dequeueRescan, but that takes a lock and
// isn't really necessary.
gp.gcscanvalid = true
dropg()
if _g_.m.locked&^_LockExternal != 0 {
print("invalid m->locked = ", _g_.m.locked, "\n")
throw("internal lockOSThread error")
}
_g_.m.locked = 0
gfput(_g_.m.p.ptr(), gp)
schedule()
}
//go:nosplit
//go:nowritebarrier
func save(pc, sp uintptr) {
_g_ := getg()
_g_.sched.pc = pc
_g_.sched.sp = sp
_g_.sched.lr = 0
_g_.sched.ret = 0
_g_.sched.ctxt = nil
_g_.sched.g = guintptr(unsafe.Pointer(_g_))
}
// The goroutine g is about to enter a system call.
// Record that it's not using the cpu anymore.
// This is called only from the go syscall library and cgocall,
// not from the low-level system calls used by the runtime.
//
// Entersyscall cannot split the stack: the gosave must
// make g->sched refer to the caller's stack segment, because
// entersyscall is going to return immediately after.
//
// Nothing entersyscall calls can split the stack either.
// We cannot safely move the stack during an active call to syscall,
// because we do not know which of the uintptr arguments are
// really pointers (back into the stack).
// In practice, this means that we make the fast path run through
// entersyscall doing no-split things, and the slow path has to use systemstack
// to run bigger things on the system stack.
//
// reentersyscall is the entry point used by cgo callbacks, where explicitly
// saved SP and PC are restored. This is needed when exitsyscall will be called
// from a function further up in the call stack than the parent, as g->syscallsp
// must always point to a valid stack frame. entersyscall below is the normal
// entry point for syscalls, which obtains the SP and PC from the caller.
//
// Syscall tracing:
// At the start of a syscall we emit traceGoSysCall to capture the stack trace.
// If the syscall does not block, that is it, we do not emit any other events.
// If the syscall blocks (that is, P is retaken), retaker emits traceGoSysBlock;
// when syscall returns we emit traceGoSysExit and when the goroutine starts running
// (potentially instantly, if exitsyscallfast returns true) we emit traceGoStart.
// To ensure that traceGoSysExit is emitted strictly after traceGoSysBlock,
// we remember current value of syscalltick in m (_g_.m.syscalltick = _g_.m.p.ptr().syscalltick),
// whoever emits traceGoSysBlock increments p.syscalltick afterwards;
// and we wait for the increment before emitting traceGoSysExit.
// Note that the increment is done even if tracing is not enabled,
// because tracing can be enabled in the middle of syscall. We don't want the wait to hang.
//
//go:nosplit
func reentersyscall(pc, sp uintptr) {
_g_ := getg()
// Disable preemption because during this function g is in Gsyscall status,
// but can have inconsistent g->sched, do not let GC observe it.
_g_.m.locks++
// Entersyscall must not call any function that might split/grow the stack.
// (See details in comment above.)
// Catch calls that might, by replacing the stack guard with something that
// will trip any stack check and leaving a flag to tell newstack to die.
_g_.stackguard0 = stackPreempt
_g_.throwsplit = true
// Leave SP around for GC and traceback.
save(pc, sp)
_g_.syscallsp = sp
_g_.syscallpc = pc
casgstatus(_g_, _Grunning, _Gsyscall)
if _g_.syscallsp < _g_.stack.lo || _g_.stack.hi < _g_.syscallsp {
systemstack(func() {
print("entersyscall inconsistent ", hex(_g_.syscallsp), " [", hex(_g_.stack.lo), ",", hex(_g_.stack.hi), "]\n")
throw("entersyscall")
})
}
if trace.enabled {
systemstack(traceGoSysCall)
// systemstack itself clobbers g.sched.{pc,sp} and we might
// need them later when the G is genuinely blocked in a
// syscall
save(pc, sp)
}
if atomic.Load(&sched.sysmonwait) != 0 { // TODO: fast atomic
systemstack(entersyscall_sysmon)
save(pc, sp)
}
if _g_.m.p.ptr().runSafePointFn != 0 {
// runSafePointFn may stack split if run on this stack
systemstack(runSafePointFn)
save(pc, sp)
}
_g_.m.syscalltick = _g_.m.p.ptr().syscalltick
_g_.sysblocktraced = true
_g_.m.mcache = nil
_g_.m.p.ptr().m = 0
atomic.Store(&_g_.m.p.ptr().status, _Psyscall)
if sched.gcwaiting != 0 {
systemstack(entersyscall_gcwait)
save(pc, sp)
}
// Goroutines must not split stacks in Gsyscall status (it would corrupt g->sched).
// We set _StackGuard to StackPreempt so that first split stack check calls morestack.
// Morestack detects this case and throws.
_g_.stackguard0 = stackPreempt
_g_.m.locks--
}
// Standard syscall entry used by the go syscall library and normal cgo calls.
//go:nosplit
func entersyscall(dummy int32) {
reentersyscall(getcallerpc(unsafe.Pointer(&dummy)), getcallersp(unsafe.Pointer(&dummy)))
}
func entersyscall_sysmon() {
lock(&sched.lock)
if atomic.Load(&sched.sysmonwait) != 0 {
atomic.Store(&sched.sysmonwait, 0)
notewakeup(&sched.sysmonnote)
}
unlock(&sched.lock)
}
func entersyscall_gcwait() {
_g_ := getg()
_p_ := _g_.m.p.ptr()
lock(&sched.lock)
if sched.stopwait > 0 && atomic.Cas(&_p_.status, _Psyscall, _Pgcstop) {
if trace.enabled {
traceGoSysBlock(_p_)
traceProcStop(_p_)
}
_p_.syscalltick++
if sched.stopwait--; sched.stopwait == 0 {
notewakeup(&sched.stopnote)
}
}
unlock(&sched.lock)
}
// The same as entersyscall(), but with a hint that the syscall is blocking.
//go:nosplit
func entersyscallblock(dummy int32) {
_g_ := getg()
_g_.m.locks++ // see comment in entersyscall
_g_.throwsplit = true
_g_.stackguard0 = stackPreempt // see comment in entersyscall
_g_.m.syscalltick = _g_.m.p.ptr().syscalltick
_g_.sysblocktraced = true
_g_.m.p.ptr().syscalltick++
// Leave SP around for GC and traceback.
pc := getcallerpc(unsafe.Pointer(&dummy))
sp := getcallersp(unsafe.Pointer(&dummy))
save(pc, sp)
_g_.syscallsp = _g_.sched.sp
_g_.syscallpc = _g_.sched.pc
if _g_.syscallsp < _g_.stack.lo || _g_.stack.hi < _g_.syscallsp {
sp1 := sp
sp2 := _g_.sched.sp
sp3 := _g_.syscallsp
systemstack(func() {
print("entersyscallblock inconsistent ", hex(sp1), " ", hex(sp2), " ", hex(sp3), " [", hex(_g_.stack.lo), ",", hex(_g_.stack.hi), "]\n")
throw("entersyscallblock")
})
}
casgstatus(_g_, _Grunning, _Gsyscall)
if _g_.syscallsp < _g_.stack.lo || _g_.stack.hi < _g_.syscallsp {
systemstack(func() {
print("entersyscallblock inconsistent ", hex(sp), " ", hex(_g_.sched.sp), " ", hex(_g_.syscallsp), " [", hex(_g_.stack.lo), ",", hex(_g_.stack.hi), "]\n")
throw("entersyscallblock")
})
}
systemstack(entersyscallblock_handoff)
// Resave for traceback during blocked call.
save(getcallerpc(unsafe.Pointer(&dummy)), getcallersp(unsafe.Pointer(&dummy)))
_g_.m.locks--
}
func entersyscallblock_handoff() {
if trace.enabled {
traceGoSysCall()
traceGoSysBlock(getg().m.p.ptr())
}
handoffp(releasep())
}
// The goroutine g exited its system call.
// Arrange for it to run on a cpu again.
// This is called only from the go syscall library, not
// from the low-level system calls used by the runtime.
//go:nosplit
func exitsyscall(dummy int32) {
_g_ := getg()
_g_.m.locks++ // see comment in entersyscall
if getcallersp(unsafe.Pointer(&dummy)) > _g_.syscallsp {
// throw calls print which may try to grow the stack,
// but throwsplit == true so the stack can not be grown;
// use systemstack to avoid that possible problem.
systemstack(func() {
throw("exitsyscall: syscall frame is no longer valid")
})
}
_g_.waitsince = 0
oldp := _g_.m.p.ptr()
if exitsyscallfast() {
if _g_.m.mcache == nil {
throw("lost mcache")
}
if trace.enabled {
if oldp != _g_.m.p.ptr() || _g_.m.syscalltick != _g_.m.p.ptr().syscalltick {
systemstack(traceGoStart)
}
}
// There's a cpu for us, so we can run.
_g_.m.p.ptr().syscalltick++
// We need to cas the status and scan before resuming...
casgstatus(_g_, _Gsyscall, _Grunning)
// Garbage collector isn't running (since we are),
// so okay to clear syscallsp.
_g_.syscallsp = 0
_g_.m.locks--
if _g_.preempt {
// restore the preemption request in case we've cleared it in newstack
_g_.stackguard0 = stackPreempt
} else {
// otherwise restore the real _StackGuard, we've spoiled it in entersyscall/entersyscallblock
_g_.stackguard0 = _g_.stack.lo + _StackGuard
}
_g_.throwsplit = false
return
}
_g_.sysexitticks = 0
if trace.enabled {
// Wait till traceGoSysBlock event is emitted.
// This ensures consistency of the trace (the goroutine is started after it is blocked).
for oldp != nil && oldp.syscalltick == _g_.m.syscalltick {
osyield()
}
// We can't trace syscall exit right now because we don't have a P.
// Tracing code can invoke write barriers that cannot run without a P.
// So instead we remember the syscall exit time and emit the event
// in execute when we have a P.
_g_.sysexitticks = cputicks()
}
_g_.m.locks--
// Call the scheduler.
mcall(exitsyscall0)
if _g_.m.mcache == nil {
throw("lost mcache")
}
// Scheduler returned, so we're allowed to run now.
// Delete the syscallsp information that we left for
// the garbage collector during the system call.
// Must wait until now because until gosched returns
// we don't know for sure that the garbage collector
// is not running.
_g_.syscallsp = 0
_g_.m.p.ptr().syscalltick++
_g_.throwsplit = false
}
//go:nosplit
func exitsyscallfast() bool {
_g_ := getg()
// Freezetheworld sets stopwait but does not retake P's.
if sched.stopwait == freezeStopWait {
_g_.m.mcache = nil
_g_.m.p = 0
return false
}
// Try to re-acquire the last P.
if _g_.m.p != 0 && _g_.m.p.ptr().status == _Psyscall && atomic.Cas(&_g_.m.p.ptr().status, _Psyscall, _Prunning) {
// There's a cpu for us, so we can run.
_g_.m.mcache = _g_.m.p.ptr().mcache
_g_.m.p.ptr().m.set(_g_.m)
if _g_.m.syscalltick != _g_.m.p.ptr().syscalltick {
if trace.enabled {
// The p was retaken and then enter into syscall again (since _g_.m.syscalltick has changed).
// traceGoSysBlock for this syscall was already emitted,
// but here we effectively retake the p from the new syscall running on the same p.
systemstack(func() {
// Denote blocking of the new syscall.
traceGoSysBlock(_g_.m.p.ptr())
// Denote completion of the current syscall.
traceGoSysExit(0)
})
}
_g_.m.p.ptr().syscalltick++
}
return true
}
// Try to get any other idle P.
oldp := _g_.m.p.ptr()
_g_.m.mcache = nil
_g_.m.p = 0
if sched.pidle != 0 {
var ok bool
systemstack(func() {
ok = exitsyscallfast_pidle()
if ok && trace.enabled {
if oldp != nil {
// Wait till traceGoSysBlock event is emitted.
// This ensures consistency of the trace (the goroutine is started after it is blocked).
for oldp.syscalltick == _g_.m.syscalltick {
osyield()
}
}
traceGoSysExit(0)
}
})
if ok {
return true
}
}
return false
}
func exitsyscallfast_pidle() bool {
lock(&sched.lock)
_p_ := pidleget()
if _p_ != nil && atomic.Load(&sched.sysmonwait) != 0 {
atomic.Store(&sched.sysmonwait, 0)
notewakeup(&sched.sysmonnote)
}
unlock(&sched.lock)
if _p_ != nil {
acquirep(_p_)
return true
}
return false
}
// exitsyscall slow path on g0.
// Failed to acquire P, enqueue gp as runnable.
func exitsyscall0(gp *g) {
_g_ := getg()
casgstatus(gp, _Gsyscall, _Grunnable)
dropg()
lock(&sched.lock)
_p_ := pidleget()
if _p_ == nil {
globrunqput(gp)
} else if atomic.Load(&sched.sysmonwait) != 0 {
atomic.Store(&sched.sysmonwait, 0)
notewakeup(&sched.sysmonnote)
}
unlock(&sched.lock)
if _p_ != nil {
acquirep(_p_)
execute(gp, false) // Never returns.
}
if _g_.m.lockedg != nil {
// Wait until another thread schedules gp and so m again.
stoplockedm()
execute(gp, false) // Never returns.
}
stopm()
schedule() // Never returns.
}
func beforefork() {
gp := getg().m.curg
// Fork can hang if preempted with signals frequently enough (see issue 5517).
// Ensure that we stay on the same M where we disable profiling.
gp.m.locks++
if gp.m.profilehz != 0 {
resetcpuprofiler(0)
}
// This function is called before fork in syscall package.
// Code between fork and exec must not allocate memory nor even try to grow stack.
// Here we spoil g->_StackGuard to reliably detect any attempts to grow stack.
// runtime_AfterFork will undo this in parent process, but not in child.
gp.stackguard0 = stackFork
}
// Called from syscall package before fork.
//go:linkname syscall_runtime_BeforeFork syscall.runtime_BeforeFork
//go:nosplit
func syscall_runtime_BeforeFork() {
systemstack(beforefork)
}
func afterfork() {
gp := getg().m.curg
// See the comment in beforefork.
gp.stackguard0 = gp.stack.lo + _StackGuard
hz := sched.profilehz
if hz != 0 {
resetcpuprofiler(hz)
}
gp.m.locks--
}
// Called from syscall package after fork in parent.
//go:linkname syscall_runtime_AfterFork syscall.runtime_AfterFork
//go:nosplit
func syscall_runtime_AfterFork() {
systemstack(afterfork)
}
// Allocate a new g, with a stack big enough for stacksize bytes.
func malg(stacksize int32) *g {
newg := new(g)
if stacksize >= 0 {
stacksize = round2(_StackSystem + stacksize)
systemstack(func() {
newg.stack, newg.stkbar = stackalloc(uint32(stacksize))
})
newg.stackguard0 = newg.stack.lo + _StackGuard
newg.stackguard1 = ^uintptr(0)
newg.stackAlloc = uintptr(stacksize)
}
return newg
}
// Create a new g running fn with siz bytes of arguments.
// Put it on the queue of g's waiting to run.
// The compiler turns a go statement into a call to this.
// Cannot split the stack because it assumes that the arguments
// are available sequentially after &fn; they would not be
// copied if a stack split occurred.
//go:nosplit
func newproc(siz int32, fn *funcval) {
argp := add(unsafe.Pointer(&fn), sys.PtrSize)
pc := getcallerpc(unsafe.Pointer(&siz))
systemstack(func() {
newproc1(fn, (*uint8)(argp), siz, 0, pc)
})
}
// Create a new g running fn with narg bytes of arguments starting
// at argp and returning nret bytes of results. callerpc is the
// address of the go statement that created this. The new g is put
// on the queue of g's waiting to run.
func newproc1(fn *funcval, argp *uint8, narg int32, nret int32, callerpc uintptr) *g {
_g_ := getg()
if fn == nil {
_g_.m.throwing = -1 // do not dump full stacks
throw("go of nil func value")
}
_g_.m.locks++ // disable preemption because it can be holding p in a local var
siz := narg + nret
siz = (siz + 7) &^ 7
// We could allocate a larger initial stack if necessary.
// Not worth it: this is almost always an error.
// 4*sizeof(uintreg): extra space added below
// sizeof(uintreg): caller's LR (arm) or return address (x86, in gostartcall).
if siz >= _StackMin-4*sys.RegSize-sys.RegSize {
throw("newproc: function arguments too large for new goroutine")
}
_p_ := _g_.m.p.ptr()
newg := gfget(_p_)
if newg == nil {
newg = malg(_StackMin)
casgstatus(newg, _Gidle, _Gdead)
newg.gcRescan = -1
allgadd(newg) // publishes with a g->status of Gdead so GC scanner doesn't look at uninitialized stack.
}
if newg.stack.hi == 0 {
throw("newproc1: newg missing stack")
}
if readgstatus(newg) != _Gdead {
throw("newproc1: new g is not Gdead")
}
totalSize := 4*sys.RegSize + uintptr(siz) + sys.MinFrameSize // extra space in case of reads slightly beyond frame
totalSize += -totalSize & (sys.SpAlign - 1) // align to spAlign
sp := newg.stack.hi - totalSize
spArg := sp
if usesLR {
// caller's LR
*(*unsafe.Pointer)(unsafe.Pointer(sp)) = nil
prepGoExitFrame(sp)
spArg += sys.MinFrameSize
}
memmove(unsafe.Pointer(spArg), unsafe.Pointer(argp), uintptr(narg))
memclr(unsafe.Pointer(&newg.sched), unsafe.Sizeof(newg.sched))
newg.sched.sp = sp
newg.stktopsp = sp
newg.sched.pc = funcPC(goexit) + sys.PCQuantum // +PCQuantum so that previous instruction is in same function
newg.sched.g = guintptr(unsafe.Pointer(newg))
gostartcallfn(&newg.sched, fn)
newg.gopc = callerpc
newg.startpc = fn.fn
if isSystemGoroutine(newg) {
atomic.Xadd(&sched.ngsys, +1)
}
// The stack is dirty from the argument frame, so queue it for
// scanning. Do this before setting it to runnable so we still
// own the G. If we're recycling a G, it may already be on the
// rescan list.
if newg.gcRescan == -1 {
queueRescan(newg)
} else {
// The recycled G is already on the rescan list. Just
// mark the stack dirty.
newg.gcscanvalid = false
}
casgstatus(newg, _Gdead, _Grunnable)
if _p_.goidcache == _p_.goidcacheend {
// Sched.goidgen is the last allocated id,
// this batch must be [sched.goidgen+1, sched.goidgen+GoidCacheBatch].
// At startup sched.goidgen=0, so main goroutine receives goid=1.
_p_.goidcache = atomic.Xadd64(&sched.goidgen, _GoidCacheBatch)
_p_.goidcache -= _GoidCacheBatch - 1
_p_.goidcacheend = _p_.goidcache + _GoidCacheBatch
}
newg.goid = int64(_p_.goidcache)
_p_.goidcache++
if raceenabled {
newg.racectx = racegostart(callerpc)
}
if trace.enabled {
traceGoCreate(newg, newg.startpc)
}
runqput(_p_, newg, true)
if atomic.Load(&sched.npidle) != 0 && atomic.Load(&sched.nmspinning) == 0 && unsafe.Pointer(fn.fn) != unsafe.Pointer(funcPC(main)) { // TODO: fast atomic
wakep()
}
_g_.m.locks--
if _g_.m.locks == 0 && _g_.preempt { // restore the preemption request in case we've cleared it in newstack
_g_.stackguard0 = stackPreempt
}
return newg
}
// Put on gfree list.
// If local list is too long, transfer a batch to the global list.
func gfput(_p_ *p, gp *g) {
if readgstatus(gp) != _Gdead {
throw("gfput: bad status (not Gdead)")
}
stksize := gp.stackAlloc
if stksize != _FixedStack {
// non-standard stack size - free it.
stackfree(gp.stack, gp.stackAlloc)
gp.stack.lo = 0
gp.stack.hi = 0
gp.stackguard0 = 0
gp.stkbar = nil
gp.stkbarPos = 0
} else {
// Reset stack barriers.
gp.stkbar = gp.stkbar[:0]
gp.stkbarPos = 0
}
gp.schedlink.set(_p_.gfree)
_p_.gfree = gp
_p_.gfreecnt++
if _p_.gfreecnt >= 64 {
lock(&sched.gflock)
for _p_.gfreecnt >= 32 {
_p_.gfreecnt--
gp = _p_.gfree
_p_.gfree = gp.schedlink.ptr()
if gp.stack.lo == 0 {
gp.schedlink.set(sched.gfreeNoStack)
sched.gfreeNoStack = gp
} else {
gp.schedlink.set(sched.gfreeStack)
sched.gfreeStack = gp
}
sched.ngfree++
}
unlock(&sched.gflock)
}
}
// Get from gfree list.
// If local list is empty, grab a batch from global list.
func gfget(_p_ *p) *g {
retry:
gp := _p_.gfree
if gp == nil && (sched.gfreeStack != nil || sched.gfreeNoStack != nil) {
lock(&sched.gflock)
for _p_.gfreecnt < 32 {
if sched.gfreeStack != nil {
// Prefer Gs with stacks.
gp = sched.gfreeStack
sched.gfreeStack = gp.schedlink.ptr()
} else if sched.gfreeNoStack != nil {
gp = sched.gfreeNoStack
sched.gfreeNoStack = gp.schedlink.ptr()
} else {
break
}
_p_.gfreecnt++
sched.ngfree--
gp.schedlink.set(_p_.gfree)
_p_.gfree = gp
}
unlock(&sched.gflock)
goto retry
}
if gp != nil {
_p_.gfree = gp.schedlink.ptr()
_p_.gfreecnt--
if gp.stack.lo == 0 {
// Stack was deallocated in gfput. Allocate a new one.
systemstack(func() {
gp.stack, gp.stkbar = stackalloc(_FixedStack)
})
gp.stackguard0 = gp.stack.lo + _StackGuard
gp.stackAlloc = _FixedStack
} else {
if raceenabled {
racemalloc(unsafe.Pointer(gp.stack.lo), gp.stackAlloc)
}
if msanenabled {
msanmalloc(unsafe.Pointer(gp.stack.lo), gp.stackAlloc)
}
}
}
return gp
}
// Purge all cached G's from gfree list to the global list.
func gfpurge(_p_ *p) {
lock(&sched.gflock)
for _p_.gfreecnt != 0 {
_p_.gfreecnt--
gp := _p_.gfree
_p_.gfree = gp.schedlink.ptr()
if gp.stack.lo == 0 {
gp.schedlink.set(sched.gfreeNoStack)
sched.gfreeNoStack = gp
} else {
gp.schedlink.set(sched.gfreeStack)
sched.gfreeStack = gp
}
sched.ngfree++
}
unlock(&sched.gflock)
}
// Breakpoint executes a breakpoint trap.
func Breakpoint() {
breakpoint()
}
// dolockOSThread is called by LockOSThread and lockOSThread below
// after they modify m.locked. Do not allow preemption during this call,
// or else the m might be different in this function than in the caller.
//go:nosplit
func dolockOSThread() {
_g_ := getg()
_g_.m.lockedg = _g_
_g_.lockedm = _g_.m
}
//go:nosplit
// LockOSThread wires the calling goroutine to its current operating system thread.
// Until the calling goroutine exits or calls UnlockOSThread, it will always
// execute in that thread, and no other goroutine can.
func LockOSThread() {
getg().m.locked |= _LockExternal
dolockOSThread()
}
//go:nosplit
func lockOSThread() {
getg().m.locked += _LockInternal
dolockOSThread()
}
// dounlockOSThread is called by UnlockOSThread and unlockOSThread below
// after they update m->locked. Do not allow preemption during this call,
// or else the m might be in different in this function than in the caller.
//go:nosplit
func dounlockOSThread() {
_g_ := getg()
if _g_.m.locked != 0 {
return
}
_g_.m.lockedg = nil
_g_.lockedm = nil
}
//go:nosplit
// UnlockOSThread unwires the calling goroutine from its fixed operating system thread.
// If the calling goroutine has not called LockOSThread, UnlockOSThread is a no-op.
func UnlockOSThread() {
getg().m.locked &^= _LockExternal
dounlockOSThread()
}
//go:nosplit
func unlockOSThread() {
_g_ := getg()
if _g_.m.locked < _LockInternal {
systemstack(badunlockosthread)
}
_g_.m.locked -= _LockInternal
dounlockOSThread()
}
func badunlockosthread() {
throw("runtime: internal error: misuse of lockOSThread/unlockOSThread")
}
func gcount() int32 {
n := int32(allglen) - sched.ngfree - int32(atomic.Load(&sched.ngsys))
for i := 0; ; i++ {
_p_ := allp[i]
if _p_ == nil {
break
}
n -= _p_.gfreecnt
}
// All these variables can be changed concurrently, so the result can be inconsistent.
// But at least the current goroutine is running.
if n < 1 {
n = 1
}
return n
}
func mcount() int32 {
return sched.mcount
}
var prof struct {
lock uint32
hz int32
}
func _System() { _System() }
func _ExternalCode() { _ExternalCode() }
func _GC() { _GC() }
// Called if we receive a SIGPROF signal.
// Called by the signal handler, may run during STW.
//go:nowritebarrierrec
func sigprof(pc, sp, lr uintptr, gp *g, mp *m) {
if prof.hz == 0 {
return
}
// Profiling runs concurrently with GC, so it must not allocate.
mp.mallocing++
// Define that a "user g" is a user-created goroutine, and a "system g"
// is one that is m->g0 or m->gsignal.
//
// We might be interrupted for profiling halfway through a
// goroutine switch. The switch involves updating three (or four) values:
// g, PC, SP, and (on arm) LR. The PC must be the last to be updated,
// because once it gets updated the new g is running.
//
// When switching from a user g to a system g, LR is not considered live,
// so the update only affects g, SP, and PC. Since PC must be last, there
// the possible partial transitions in ordinary execution are (1) g alone is updated,
// (2) both g and SP are updated, and (3) SP alone is updated.
// If SP or g alone is updated, we can detect the partial transition by checking
// whether the SP is within g's stack bounds. (We could also require that SP
// be changed only after g, but the stack bounds check is needed by other
// cases, so there is no need to impose an additional requirement.)
//
// There is one exceptional transition to a system g, not in ordinary execution.
// When a signal arrives, the operating system starts the signal handler running
// with an updated PC and SP. The g is updated last, at the beginning of the
// handler. There are two reasons this is okay. First, until g is updated the
// g and SP do not match, so the stack bounds check detects the partial transition.
// Second, signal handlers currently run with signals disabled, so a profiling
// signal cannot arrive during the handler.
//
// When switching from a system g to a user g, there are three possibilities.
//
// First, it may be that the g switch has no PC update, because the SP
// either corresponds to a user g throughout (as in asmcgocall)
// or because it has been arranged to look like a user g frame
// (as in cgocallback_gofunc). In this case, since the entire
// transition is a g+SP update, a partial transition updating just one of
// those will be detected by the stack bounds check.
//
// Second, when returning from a signal handler, the PC and SP updates
// are performed by the operating system in an atomic update, so the g
// update must be done before them. The stack bounds check detects
// the partial transition here, and (again) signal handlers run with signals
// disabled, so a profiling signal cannot arrive then anyway.
//
// Third, the common case: it may be that the switch updates g, SP, and PC
// separately. If the PC is within any of the functions that does this,
// we don't ask for a traceback. C.F. the function setsSP for more about this.
//
// There is another apparently viable approach, recorded here in case
// the "PC within setsSP function" check turns out not to be usable.
// It would be possible to delay the update of either g or SP until immediately
// before the PC update instruction. Then, because of the stack bounds check,
// the only problematic interrupt point is just before that PC update instruction,
// and the sigprof handler can detect that instruction and simulate stepping past
// it in order to reach a consistent state. On ARM, the update of g must be made
// in two places (in R10 and also in a TLS slot), so the delayed update would
// need to be the SP update. The sigprof handler must read the instruction at
// the current PC and if it was the known instruction (for example, JMP BX or
// MOV R2, PC), use that other register in place of the PC value.
// The biggest drawback to this solution is that it requires that we can tell
// whether it's safe to read from the memory pointed at by PC.
// In a correct program, we can test PC == nil and otherwise read,
// but if a profiling signal happens at the instant that a program executes
// a bad jump (before the program manages to handle the resulting fault)
// the profiling handler could fault trying to read nonexistent memory.
//
// To recap, there are no constraints on the assembly being used for the
// transition. We simply require that g and SP match and that the PC is not
// in gogo.
traceback := true
if gp == nil || sp < gp.stack.lo || gp.stack.hi < sp || setsSP(pc) {
traceback = false
}
var stk [maxCPUProfStack]uintptr
var haveStackLock *g
n := 0
if mp.ncgo > 0 && mp.curg != nil && mp.curg.syscallpc != 0 && mp.curg.syscallsp != 0 {
cgoOff := 0
// Check cgoCallersUse to make sure that we are not
// interrupting other code that is fiddling with
// cgoCallers. We are running in a signal handler
// with all signals blocked, so we don't have to worry
// about any other code interrupting us.
if atomic.Load(&mp.cgoCallersUse) == 0 && mp.cgoCallers != nil && mp.cgoCallers[0] != 0 {
for cgoOff < len(mp.cgoCallers) && mp.cgoCallers[cgoOff] != 0 {
cgoOff++
}
copy(stk[:], mp.cgoCallers[:cgoOff])
mp.cgoCallers[0] = 0
}
// Collect Go stack that leads to the cgo call.
if gcTryLockStackBarriers(mp.curg) {
haveStackLock = mp.curg
n = gentraceback(mp.curg.syscallpc, mp.curg.syscallsp, 0, mp.curg, 0, &stk[cgoOff], len(stk)-cgoOff, nil, nil, 0)
}
} else if traceback {
var flags uint = _TraceTrap
if gp.m.curg != nil && gcTryLockStackBarriers(gp.m.curg) {
// It's safe to traceback the user stack.
haveStackLock = gp.m.curg
flags |= _TraceJumpStack
}
// Traceback is safe if we're on the system stack (if
// necessary, flags will stop it before switching to
// the user stack), or if we locked the user stack.
if gp != gp.m.curg || haveStackLock != nil {
n = gentraceback(pc, sp, lr, gp, 0, &stk[0], len(stk), nil, nil, flags)
}
}
if haveStackLock != nil {
gcUnlockStackBarriers(haveStackLock)
}
if n <= 0 {
// Normal traceback is impossible or has failed.
// See if it falls into several common cases.
n = 0
if GOOS == "windows" && mp.libcallg != 0 && mp.libcallpc != 0 && mp.libcallsp != 0 {
// Libcall, i.e. runtime syscall on windows.
// Collect Go stack that leads to the call.
if gcTryLockStackBarriers(mp.libcallg.ptr()) {
n = gentraceback(mp.libcallpc, mp.libcallsp, 0, mp.libcallg.ptr(), 0, &stk[0], len(stk), nil, nil, 0)
gcUnlockStackBarriers(mp.libcallg.ptr())
}
}
if n == 0 {
// If all of the above has failed, account it against abstract "System" or "GC".
n = 2
// "ExternalCode" is better than "etext".
if pc > firstmoduledata.etext {
pc = funcPC(_ExternalCode) + sys.PCQuantum
}
stk[0] = pc
if mp.preemptoff != "" || mp.helpgc != 0 {
stk[1] = funcPC(_GC) + sys.PCQuantum
} else {
stk[1] = funcPC(_System) + sys.PCQuantum
}
}
}
if prof.hz != 0 {
// Simple cas-lock to coordinate with setcpuprofilerate.
for !atomic.Cas(&prof.lock, 0, 1) {
osyield()
}
if prof.hz != 0 {
cpuprof.add(stk[:n])
}
atomic.Store(&prof.lock, 0)
}
mp.mallocing--
}
// If the signal handler receives a SIGPROF signal on a non-Go thread,
// it tries to collect a traceback into sigprofCallers.
// sigprofCallersUse is set to non-zero while sigprofCallers holds a traceback.
var sigprofCallers cgoCallers
var sigprofCallersUse uint32
// Called if we receive a SIGPROF signal on a non-Go thread.
// When this is called, sigprofCallersUse will be non-zero.
// g is nil, and what we can do is very limited.
//go:nosplit
//go:nowritebarrierrec
func sigprofNonGo() {
if prof.hz != 0 {
n := 0
for n < len(sigprofCallers) && sigprofCallers[n] != 0 {
n++
}
// Simple cas-lock to coordinate with setcpuprofilerate.
if atomic.Cas(&prof.lock, 0, 1) {
if prof.hz != 0 {
cpuprof.addNonGo(sigprofCallers[:n])
}
atomic.Store(&prof.lock, 0)
}
}
atomic.Store(&sigprofCallersUse, 0)
}
// Reports whether a function will set the SP
// to an absolute value. Important that
// we don't traceback when these are at the bottom
// of the stack since we can't be sure that we will
// find the caller.
//
// If the function is not on the bottom of the stack
// we assume that it will have set it up so that traceback will be consistent,
// either by being a traceback terminating function
// or putting one on the stack at the right offset.
func setsSP(pc uintptr) bool {
f := findfunc(pc)
if f == nil {
// couldn't find the function for this PC,
// so assume the worst and stop traceback
return true
}
switch f.entry {
case gogoPC, systemstackPC, mcallPC, morestackPC:
return true
}
return false
}
// Arrange to call fn with a traceback hz times a second.
func setcpuprofilerate_m(hz int32) {
// Force sane arguments.
if hz < 0 {
hz = 0
}
// Disable preemption, otherwise we can be rescheduled to another thread
// that has profiling enabled.
_g_ := getg()
_g_.m.locks++
// Stop profiler on this thread so that it is safe to lock prof.
// if a profiling signal came in while we had prof locked,
// it would deadlock.
resetcpuprofiler(0)
for !atomic.Cas(&prof.lock, 0, 1) {
osyield()
}
prof.hz = hz
atomic.Store(&prof.lock, 0)
lock(&sched.lock)
sched.profilehz = hz
unlock(&sched.lock)
if hz != 0 {
resetcpuprofiler(hz)
}
_g_.m.locks--
}
// Change number of processors. The world is stopped, sched is locked.
// gcworkbufs are not being modified by either the GC or
// the write barrier code.
// Returns list of Ps with local work, they need to be scheduled by the caller.
func procresize(nprocs int32) *p {
old := gomaxprocs
if old < 0 || old > _MaxGomaxprocs || nprocs <= 0 || nprocs > _MaxGomaxprocs {
throw("procresize: invalid arg")
}
if trace.enabled {
traceGomaxprocs(nprocs)
}
// update statistics
now := nanotime()
if sched.procresizetime != 0 {
sched.totaltime += int64(old) * (now - sched.procresizetime)
}
sched.procresizetime = now
// initialize new P's
for i := int32(0); i < nprocs; i++ {
pp := allp[i]
if pp == nil {
pp = new(p)
pp.id = i
pp.status = _Pgcstop
pp.sudogcache = pp.sudogbuf[:0]
for i := range pp.deferpool {
pp.deferpool[i] = pp.deferpoolbuf[i][:0]
}
atomicstorep(unsafe.Pointer(&allp[i]), unsafe.Pointer(pp))
}
if pp.mcache == nil {
if old == 0 && i == 0 {
if getg().m.mcache == nil {
throw("missing mcache?")
}
pp.mcache = getg().m.mcache // bootstrap
} else {
pp.mcache = allocmcache()
}
}
if raceenabled && pp.racectx == 0 {
if old == 0 && i == 0 {
pp.racectx = raceprocctx0
raceprocctx0 = 0 // bootstrap
} else {
pp.racectx = raceproccreate()
}
}
}
// free unused P's
for i := nprocs; i < old; i++ {
p := allp[i]
if trace.enabled {
if p == getg().m.p.ptr() {
// moving to p[0], pretend that we were descheduled
// and then scheduled again to keep the trace sane.
traceGoSched()
traceProcStop(p)
}
}
// move all runnable goroutines to the global queue
for p.runqhead != p.runqtail {
// pop from tail of local queue
p.runqtail--
gp := p.runq[p.runqtail%uint32(len(p.runq))].ptr()
// push onto head of global queue
globrunqputhead(gp)
}
if p.runnext != 0 {
globrunqputhead(p.runnext.ptr())
p.runnext = 0
}
// if there's a background worker, make it runnable and put
// it on the global queue so it can clean itself up
if gp := p.gcBgMarkWorker.ptr(); gp != nil {
casgstatus(gp, _Gwaiting, _Grunnable)
if trace.enabled {
traceGoUnpark(gp, 0)
}
globrunqput(gp)
// This assignment doesn't race because the
// world is stopped.
p.gcBgMarkWorker.set(nil)
}
for i := range p.sudogbuf {
p.sudogbuf[i] = nil
}
p.sudogcache = p.sudogbuf[:0]
for i := range p.deferpool {
for j := range p.deferpoolbuf[i] {
p.deferpoolbuf[i][j] = nil
}
p.deferpool[i] = p.deferpoolbuf[i][:0]
}
freemcache(p.mcache)
p.mcache = nil
gfpurge(p)
traceProcFree(p)
if raceenabled {
raceprocdestroy(p.racectx)
p.racectx = 0
}
p.status = _Pdead
// can't free P itself because it can be referenced by an M in syscall
}
_g_ := getg()
if _g_.m.p != 0 && _g_.m.p.ptr().id < nprocs {
// continue to use the current P
_g_.m.p.ptr().status = _Prunning
} else {
// release the current P and acquire allp[0]
if _g_.m.p != 0 {
_g_.m.p.ptr().m = 0
}
_g_.m.p = 0
_g_.m.mcache = nil
p := allp[0]
p.m = 0
p.status = _Pidle
acquirep(p)
if trace.enabled {
traceGoStart()
}
}
var runnablePs *p
for i := nprocs - 1; i >= 0; i-- {
p := allp[i]
if _g_.m.p.ptr() == p {
continue
}
p.status = _Pidle
if runqempty(p) {
pidleput(p)
} else {
p.m.set(mget())
p.link.set(runnablePs)
runnablePs = p
}
}
stealOrder.reset(uint32(nprocs))
var int32p *int32 = &gomaxprocs // make compiler check that gomaxprocs is an int32
atomic.Store((*uint32)(unsafe.Pointer(int32p)), uint32(nprocs))
return runnablePs
}
// Associate p and the current m.
func acquirep(_p_ *p) {
acquirep1(_p_)
// have p; write barriers now allowed
_g_ := getg()
_g_.m.mcache = _p_.mcache
if trace.enabled {
traceProcStart()
}
}
// May run during STW, so write barriers are not allowed.
//go:nowritebarrier
func acquirep1(_p_ *p) {
_g_ := getg()
if _g_.m.p != 0 || _g_.m.mcache != nil {
throw("acquirep: already in go")
}
if _p_.m != 0 || _p_.status != _Pidle {
id := int32(0)
if _p_.m != 0 {
id = _p_.m.ptr().id
}
print("acquirep: p->m=", _p_.m, "(", id, ") p->status=", _p_.status, "\n")
throw("acquirep: invalid p state")
}
_g_.m.p.set(_p_)
_p_.m.set(_g_.m)
_p_.status = _Prunning
}
// Disassociate p and the current m.
func releasep() *p {
_g_ := getg()
if _g_.m.p == 0 || _g_.m.mcache == nil {
throw("releasep: invalid arg")
}
_p_ := _g_.m.p.ptr()
if _p_.m.ptr() != _g_.m || _p_.mcache != _g_.m.mcache || _p_.status != _Prunning {
print("releasep: m=", _g_.m, " m->p=", _g_.m.p.ptr(), " p->m=", _p_.m, " m->mcache=", _g_.m.mcache, " p->mcache=", _p_.mcache, " p->status=", _p_.status, "\n")
throw("releasep: invalid p state")
}
if trace.enabled {
traceProcStop(_g_.m.p.ptr())
}
_g_.m.p = 0
_g_.m.mcache = nil
_p_.m = 0
_p_.status = _Pidle
return _p_
}
func incidlelocked(v int32) {
lock(&sched.lock)
sched.nmidlelocked += v
if v > 0 {
checkdead()
}
unlock(&sched.lock)
}
// Check for deadlock situation.
// The check is based on number of running M's, if 0 -> deadlock.
func checkdead() {
// For -buildmode=c-shared or -buildmode=c-archive it's OK if
// there are no running goroutines. The calling program is
// assumed to be running.
if islibrary || isarchive {
return
}
// If we are dying because of a signal caught on an already idle thread,
// freezetheworld will cause all running threads to block.
// And runtime will essentially enter into deadlock state,
// except that there is a thread that will call exit soon.
if panicking > 0 {
return
}
// -1 for sysmon
run := sched.mcount - sched.nmidle - sched.nmidlelocked - 1
if run > 0 {
return
}
if run < 0 {
print("runtime: checkdead: nmidle=", sched.nmidle, " nmidlelocked=", sched.nmidlelocked, " mcount=", sched.mcount, "\n")
throw("checkdead: inconsistent counts")
}
grunning := 0
lock(&allglock)
for i := 0; i < len(allgs); i++ {
gp := allgs[i]
if isSystemGoroutine(gp) {
continue
}
s := readgstatus(gp)
switch s &^ _Gscan {
case _Gwaiting:
grunning++
case _Grunnable,
_Grunning,
_Gsyscall:
unlock(&allglock)
print("runtime: checkdead: find g ", gp.goid, " in status ", s, "\n")
throw("checkdead: runnable g")
}
}
unlock(&allglock)
if grunning == 0 { // possible if main goroutine calls runtime·Goexit()
throw("no goroutines (main called runtime.Goexit) - deadlock!")
}
// Maybe jump time forward for playground.
gp := timejump()
if gp != nil {
casgstatus(gp, _Gwaiting, _Grunnable)
globrunqput(gp)
_p_ := pidleget()
if _p_ == nil {
throw("checkdead: no p for timer")
}
mp := mget()
if mp == nil {
// There should always be a free M since
// nothing is running.
throw("checkdead: no m for timer")
}
mp.nextp.set(_p_)
notewakeup(&mp.park)
return
}
getg().m.throwing = -1 // do not dump full stacks
throw("all goroutines are asleep - deadlock!")
}
// forcegcperiod is the maximum time in nanoseconds between garbage
// collections. If we go this long without a garbage collection, one
// is forced to run.
//
// This is a variable for testing purposes. It normally doesn't change.
var forcegcperiod int64 = 2 * 60 * 1e9
// Always runs without a P, so write barriers are not allowed.
//
//go:nowritebarrierrec
func sysmon() {
// If a heap span goes unused for 5 minutes after a garbage collection,
// we hand it back to the operating system.
scavengelimit := int64(5 * 60 * 1e9)
if debug.scavenge > 0 {
// Scavenge-a-lot for testing.
forcegcperiod = 10 * 1e6
scavengelimit = 20 * 1e6
}
lastscavenge := nanotime()
nscavenge := 0
lasttrace := int64(0)
idle := 0 // how many cycles in succession we had not wokeup somebody
delay := uint32(0)
for {
if idle == 0 { // start with 20us sleep...
delay = 20
} else if idle > 50 { // start doubling the sleep after 1ms...
delay *= 2
}
if delay > 10*1000 { // up to 10ms
delay = 10 * 1000
}
usleep(delay)
if debug.schedtrace <= 0 && (sched.gcwaiting != 0 || atomic.Load(&sched.npidle) == uint32(gomaxprocs)) { // TODO: fast atomic
lock(&sched.lock)
if atomic.Load(&sched.gcwaiting) != 0 || atomic.Load(&sched.npidle) == uint32(gomaxprocs) {
atomic.Store(&sched.sysmonwait, 1)
unlock(&sched.lock)
// Make wake-up period small enough
// for the sampling to be correct.
maxsleep := forcegcperiod / 2
if scavengelimit < forcegcperiod {
maxsleep = scavengelimit / 2
}
notetsleep(&sched.sysmonnote, maxsleep)
lock(&sched.lock)
atomic.Store(&sched.sysmonwait, 0)
noteclear(&sched.sysmonnote)
idle = 0
delay = 20
}
unlock(&sched.lock)
}
// poll network if not polled for more than 10ms
lastpoll := int64(atomic.Load64(&sched.lastpoll))
now := nanotime()
unixnow := unixnanotime()
if lastpoll != 0 && lastpoll+10*1000*1000 < now {
atomic.Cas64(&sched.lastpoll, uint64(lastpoll), uint64(now))
gp := netpoll(false) // non-blocking - returns list of goroutines
if gp != nil {
// Need to decrement number of idle locked M's
// (pretending that one more is running) before injectglist.
// Otherwise it can lead to the following situation:
// injectglist grabs all P's but before it starts M's to run the P's,
// another M returns from syscall, finishes running its G,
// observes that there is no work to do and no other running M's
// and reports deadlock.
incidlelocked(-1)
injectglist(gp)
incidlelocked(1)
}
}
// retake P's blocked in syscalls
// and preempt long running G's
if retake(now) != 0 {
idle = 0
} else {
idle++
}
// check if we need to force a GC
lastgc := int64(atomic.Load64(&memstats.last_gc))
if gcphase == _GCoff && lastgc != 0 && unixnow-lastgc > forcegcperiod && atomic.Load(&forcegc.idle) != 0 {
lock(&forcegc.lock)
forcegc.idle = 0
forcegc.g.schedlink = 0
injectglist(forcegc.g)
unlock(&forcegc.lock)
}
// scavenge heap once in a while
if lastscavenge+scavengelimit/2 < now {
mheap_.scavenge(int32(nscavenge), uint64(now), uint64(scavengelimit))
lastscavenge = now
nscavenge++
}
if debug.schedtrace > 0 && lasttrace+int64(debug.schedtrace)*1000000 <= now {
lasttrace = now
schedtrace(debug.scheddetail > 0)
}
}
}
var pdesc [_MaxGomaxprocs]struct {
schedtick uint32
schedwhen int64
syscalltick uint32
syscallwhen int64
}
// forcePreemptNS is the time slice given to a G before it is
// preempted.
const forcePreemptNS = 10 * 1000 * 1000 // 10ms
func retake(now int64) uint32 {
n := 0
for i := int32(0); i < gomaxprocs; i++ {
_p_ := allp[i]
if _p_ == nil {
continue
}
pd := &pdesc[i]
s := _p_.status
if s == _Psyscall {
// Retake P from syscall if it's there for more than 1 sysmon tick (at least 20us).
t := int64(_p_.syscalltick)
if int64(pd.syscalltick) != t {
pd.syscalltick = uint32(t)
pd.syscallwhen = now
continue
}
// On the one hand we don't want to retake Ps if there is no other work to do,
// but on the other hand we want to retake them eventually
// because they can prevent the sysmon thread from deep sleep.
if runqempty(_p_) && atomic.Load(&sched.nmspinning)+atomic.Load(&sched.npidle) > 0 && pd.syscallwhen+10*1000*1000 > now {
conti