From: Russ Cox Date: Fri, 5 Dec 2014 16:18:10 +0000 (-0500) Subject: [dev.cc] all: merge default (8d42099cdc23) into dev.cc X-Git-Tag: go1.5beta1~2688^2 X-Git-Url: http://www.git.cypherpunks.ru/?a=commitdiff_plain;h=829b286f2c5dc450394c526cd91ba8c8bea9bac8;p=gostls13.git [dev.cc] all: merge default (8d42099cdc23) into dev.cc TBR=austin CC=golang-codereviews https://golang.org/cl/178700044 --- 829b286f2c5dc450394c526cd91ba8c8bea9bac8 diff --cc src/runtime/proc1.go index 81b211d0d3,0000000000..aeded0e773 mode 100644,000000..100644 --- a/src/runtime/proc1.go +++ b/src/runtime/proc1.go @@@ -1,3170 -1,0 +1,3193 @@@ +// Copyright 2009 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 "unsafe" + +var ( + m0 m + g0 g +) + +// 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 http://golang.org/s/go11sched. + +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 +) + +/* +SchedT sched; +int32 gomaxprocs; +uint32 needextram; +bool iscgo; +M m0; +G g0; // idle goroutine for m0 +G* lastg; +M* allm; +M* extram; +P* allp[MaxGomaxprocs+1]; +int8* goos; +int32 ncpu; +int32 newprocs; + +Mutex allglock; // the following vars are protected by this lock or by stoptheworld +G** allg; +Slice allgs; +uintptr allglen; +ForceGCState forcegc; + +void mstart(void); +static void runqput(P*, G*); +static G* runqget(P*); +static bool runqputslow(P*, G*, uint32, uint32); +static G* runqsteal(P*, P*); +static void mput(M*); +static M* mget(void); +static void mcommoninit(M*); +static void schedule(void); +static void procresize(int32); +static void acquirep(P*); +static P* releasep(void); +static void newm(void(*)(void), P*); +static void stopm(void); +static void startm(P*, bool); +static void handoffp(P*); +static void wakep(void); +static void stoplockedm(void); +static void startlockedm(G*); +static void sysmon(void); +static uint32 retake(int64); +static void incidlelocked(int32); +static void checkdead(void); +static void exitsyscall0(G*); +void park_m(G*); +static void goexit0(G*); +static void gfput(P*, G*); +static G* gfget(P*); +static void gfpurge(P*); +static void globrunqput(G*); +static void globrunqputbatch(G*, G*, int32); +static G* globrunqget(P*, int32); +static P* pidleget(void); +static void pidleput(P*); +static void injectglist(G*); +static bool preemptall(void); +static bool preemptone(P*); +static bool exitsyscallfast(void); +static bool haveexperiment(int8*); +void allgadd(G*); +static void dropg(void); + +extern String buildVersion; +*/ + +// 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 = raceinit() + } + + sched.maxmcount = 10000 + + tracebackinit() + symtabinit() + stackinit() + mallocinit() + mcommoninit(_g_.m) + + goargs() + goenvs() + parsedebugvars() + gcinit() + + sched.lastpoll = uint64(nanotime()) + procs := 1 + if n := goatoi(gogetenv("GOMAXPROCS")); n > 0 { + if n > _MaxGomaxprocs { + n = _MaxGomaxprocs + } + procs = n + } + procresize(int32(procs)) + + if buildVersion == "" { + // Condition should never trigger. This code just serves + // to ensure runtime·buildVersion is kept in the resulting binary. + buildVersion = "unknown" + } +} + +func newsysmon() { + _newm(sysmon, nil) +} + +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") + gothrow("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[0], len(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) +} + +// Mark gp ready to run. +func ready(gp *g) { + 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) + gothrow("bad g->status in ready") + } + + // status is Gwaiting or Gscanwaiting, make Grunnable and put on runq + casgstatus(gp, _Gwaiting, _Grunnable) + runqput(_g_.m.p, gp) + if atomicload(&sched.npidle) != 0 && atomicload(&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 { + gothrow("gcprocs inconsistency") + } + mp.helpgc = n + mp.mcache = allp[pos].mcache + pos++ + notewakeup(&mp.park) + } + unlock(&sched.lock) +} + +// 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() { + if gomaxprocs == 1 { + return + } + // 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 = 0x7fffffff + atomicstore(&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 { + gothrow("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 atomicload(&gp.atomicstatus) +} + +// 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 { + case _Gscanrunnable, + _Gscanwaiting, + _Gscanrunning, + _Gscansyscall: + if newval == oldval&^_Gscan { + success = cas(&gp.atomicstatus, oldval, newval) + } + case _Gscanenqueue: + if newval == _Gwaiting { + success = cas(&gp.atomicstatus, oldval, newval) + } + } + if !success { + print("runtime: casfrom_Gscanstatus failed gp=", gp, ", oldval=", hex(oldval), ", newval=", hex(newval), "\n") + dumpgstatus(gp) + gothrow("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, + _Gwaiting, + _Gsyscall: + if newval == oldval|_Gscan { + return cas(&gp.atomicstatus, oldval, newval) + } + case _Grunning: + if newval == _Gscanrunning || newval == _Gscanenqueue { + return cas(&gp.atomicstatus, oldval, newval) + } + } + print("runtime: castogscanstatus oldval=", hex(oldval), " newval=", hex(newval), "\n") + gothrow("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("casgstatus: oldval=", hex(oldval), " newval=", hex(newval), "\n") + gothrow("casgstatus: bad incoming values") + }) + } + + // loop if gp->atomicstatus is in a scan state giving + // GC time to finish and change the state to oldval. + for !cas(&gp.atomicstatus, oldval, newval) { ++ if oldval == _Gwaiting && gp.atomicstatus == _Grunnable { ++ systemstack(func() { ++ gothrow("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) + }) + } + } +} + ++// 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 { ++ gothrow("copystack: bad status, not Gwaiting or Grunnable") ++ } ++ if cas(&gp.atomicstatus, oldstatus, _Gcopystack) { ++ return oldstatus ++ } ++ } ++} ++ +// stopg ensures that gp is stopped at a GC safe point where its stack can be scanned +// or in the context of a moving collector the pointers can be flipped from pointing +// to old object to pointing to new objects. +// If stopg returns true, the caller knows gp is at a GC safe point and will remain there until +// the caller calls restartg. +// If stopg returns false, the caller is not responsible for calling restartg. This can happen +// if another thread, either the gp itself or another GC thread is taking the responsibility +// to do the GC work related to this thread. +func stopg(gp *g) bool { + for { + if gp.gcworkdone { + return false + } + + switch s := readgstatus(gp); s { + default: + dumpgstatus(gp) + gothrow("stopg: gp->atomicstatus is not valid") + + case _Gdead: + return false + + case _Gcopystack: + // Loop until a new stack is in place. + + case _Grunnable, + _Gsyscall, + _Gwaiting: + // Claim goroutine by setting scan bit. + if !castogscanstatus(gp, s, s|_Gscan) { + break + } + // In scan state, do work. + gcphasework(gp) + return true + + case _Gscanrunnable, + _Gscanwaiting, + _Gscansyscall: + // Goroutine already claimed by another GC helper. + return false + + case _Grunning: + // Claim goroutine, so we aren't racing with a status + // transition away from Grunning. + if !castogscanstatus(gp, _Grunning, _Gscanrunning) { + break + } + + // Mark gp for preemption. + if !gp.gcworkdone { + gp.preemptscan = true + gp.preempt = true + gp.stackguard0 = stackPreempt + } + + // Unclaim. + casfrom_Gscanstatus(gp, _Gscanrunning, _Grunning) + return false + } + } +} + +// 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) + gothrow("restartg: unexpected status") + + case _Gdead: + // ok + + case _Gscanrunnable, + _Gscanwaiting, + _Gscansyscall: + casfrom_Gscanstatus(gp, s, s&^_Gscan) + + // Scan is now completed. + // Goroutine now needs to be made runnable. + // We put it on the global run queue; ready blocks on the global scheduler lock. + case _Gscanenqueue: + casfrom_Gscanstatus(gp, _Gscanenqueue, _Gwaiting) + if gp != getg().m.curg { + gothrow("processing Gscanenqueue on wrong m") + } + dropg() + ready(gp) + } +} + +func stopscanstart(gp *g) { + _g_ := getg() + if _g_ == gp { + gothrow("GC not moved to G0") + } + if stopg(gp) { + if !isscanstatus(readgstatus(gp)) { + dumpgstatus(gp) + gothrow("GC not in scan state") + } + restartg(gp) + } +} + +// Runs on g0 and does the actual work after putting the g back on the run queue. +func mquiesce(gpmaster *g) { + activeglen := len(allgs) + // enqueue the calling goroutine. + restartg(gpmaster) + for i := 0; i < activeglen; i++ { + gp := allgs[i] + if readgstatus(gp) == _Gdead { + gp.gcworkdone = true // noop scan. + } else { + gp.gcworkdone = false + } + stopscanstart(gp) + } + + // Check that the G's gcwork (such as scanning) has been done. If not do it now. + // You can end up doing work here if the page trap on a Grunning Goroutine has + // not been sprung or in some race situations. For example a runnable goes dead + // and is started up again with a gp->gcworkdone set to false. + for i := 0; i < activeglen; i++ { + gp := allgs[i] + for !gp.gcworkdone { + status := readgstatus(gp) + if status == _Gdead { + //do nothing, scan not needed. + gp.gcworkdone = true // scan is a noop + break + } + if status == _Grunning && gp.stackguard0 == uintptr(stackPreempt) && notetsleep(&sched.stopnote, 100*1000) { // nanosecond arg + noteclear(&sched.stopnote) + } else { + stopscanstart(gp) + } + } + } + + for i := 0; i < activeglen; i++ { + gp := allgs[i] + status := readgstatus(gp) + if isscanstatus(status) { + print("mstopandscang:bottom: post scan bad status gp=", gp, " has status ", hex(status), "\n") + dumpgstatus(gp) + } + if !gp.gcworkdone && status != _Gdead { + print("mstopandscang:bottom: post scan gp=", gp, "->gcworkdone still false\n") + dumpgstatus(gp) + } + } + + schedule() // Never returns. +} + +// quiesce moves all the goroutines to a GC safepoint which for now is a at preemption point. +// If the global gcphase is GCmark quiesce will ensure that all of the goroutine's stacks +// have been scanned before it returns. +func quiesce(mastergp *g) { + castogscanstatus(mastergp, _Grunning, _Gscanenqueue) + // Now move this to the g0 (aka m) stack. + // g0 will potentially scan this thread and put mastergp on the runqueue + mcall(mquiesce) +} + +// This is used by the GC as well as the routines that do stack dumps. In the case +// of GC all the routines can be reliably stopped. This is not always the case +// when the system is in panic or being exited. +func stoptheworld() { + _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 { + gothrow("stoptheworld: holding locks") + } + + lock(&sched.lock) + sched.stopwait = gomaxprocs + atomicstore(&sched.gcwaiting, 1) + preemptall() + // stop current P + _g_.m.p.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 && cas(&p.status, s, _Pgcstop) { + 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 { + gothrow("stoptheworld: not stopped") + } + for i := 0; i < int(gomaxprocs); i++ { + p := allp[i] + if p.status != _Pgcstop { + gothrow("stoptheworld: not stopped") + } + } +} + +func mhelpgc() { + _g_ := getg() + _g_.m.helpgc = -1 +} + +func starttheworld() { + _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) + if newprocs != 0 { + procresize(newprocs) + newprocs = 0 + } else { + procresize(gomaxprocs) + } + sched.gcwaiting = 0 + + var p1 *p + for { + p := pidleget() + if p == nil { + break + } + // procresize() puts p's with work at the beginning of the list. + // Once we reach a p without a run queue, the rest don't have one either. + if p.runqhead == p.runqtail { + pidleput(p) + break + } + p.m = mget() + p.link = p1 + p1 = p + } + if sched.sysmonwait != 0 { + sched.sysmonwait = 0 + notewakeup(&sched.sysmonnote) + } + unlock(&sched.lock) + + for p1 != nil { + p := p1 + p1 = p1.link + if p.m != nil { + mp := p.m + p.m = nil + if mp.nextp != nil { + gothrow("starttheworld: inconsistent mp->nextp") + } + mp.nextp = p + notewakeup(&mp.park) + } else { + // Start M to run P. Do not start another M below. + _newm(nil, p) + add = false + } + } + + 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 + } + _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 { + gothrow("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 { + initsig() + } + + if _g_.m.mstartfn != nil { + fn := *(*func())(unsafe.Pointer(&_g_.m.mstartfn)) + fn() + } + + if _g_.m.helpgc != 0 { + _g_.m.helpgc = 0 + stopm() + } else if _g_.m != &m0 { + acquirep(_g_.m.nextp) + _g_.m.nextp = nil + } + schedule() + + // TODO(brainman): This point is never reached, because scheduler + // does not release os threads at the moment. But once this path + // is enabled, we must remove our seh here. +} + +// 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 *g + tls *uint64 + fn unsafe.Pointer +} + +// Allocate a new m unassociated with any thread. +// Can use p for allocation context if needed. +func allocm(_p_ *p) *m { + _g_ := getg() + _g_.m.locks++ // disable GC because it can be called from sysmon + if _g_.m.p == nil { + acquirep(_p_) // temporarily borrow p for mallocs in this function + } + mp := newM() + 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) + } + mp.g0.m = mp + + if _p_ == _g_.m.p { + 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 +} + +func allocg() *g { + return newG() +} + +// 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 needextram != 0 { + // Can happen if C/C++ code calls Go from a global ctor. + // Can not throw, because scheduler is not initialized yet. + // XXX + // write(2, unsafe.Pointer("fatal error: cgo callback before cgo call\n"), sizeof("fatal error: cgo callback before cgo call\n") - 1) + 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 == nil + unlockextra(mp.schedlink) + + // 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() +} + +// newextram allocates an m and puts it 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() { + // 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) + gp := malg(4096) + gp.sched.pc = funcPC(goexit) + _PCQuantum + gp.sched.sp = gp.stack.hi + gp.sched.sp -= 4 * regSize // extra space in case of reads slightly beyond frame + gp.sched.lr = 0 + gp.sched.g = gp + gp.syscallpc = gp.sched.pc + gp.syscallsp = gp.sched.sp + // 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(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 = 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() { + // Undo whatever initialization minit did during needm. + unminit() + + // Clear m and g, and return m to the extra list. + // After the call to setmg we can only call nosplit functions. + mp := getg().m + setg(nil) + + mnext := lockextra(true) + mp.schedlink = mnext + unlockextra(mp) +} + +var extram uintptr + +// 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 + + for { + old := atomicloaduintptr(&extram) + if old == locked { + yield := osyield + yield() + continue + } + if old == 0 && !nilokay { + usleep(1) + continue + } + if casuintptr(&extram, old, locked) { + return (*m)(unsafe.Pointer(old)) + } + yield := osyield + yield() + continue + } +} + +//go:nosplit +func unlockextra(mp *m) { + atomicstoreuintptr(&extram, uintptr(unsafe.Pointer(mp))) +} + +// Create a new m. It will start off with a call to fn, or else the scheduler. +func _newm(fn func(), _p_ *p) { + mp := allocm(_p_) + mp.nextp = _p_ + mp.mstartfn = *(*unsafe.Pointer)(unsafe.Pointer(&fn)) + + if iscgo { + var ts cgothreadstart + if _cgo_thread_start == nil { + gothrow("_cgo_thread_start missing") + } + ts.g = mp.g0 + ts.tls = (*uint64)(unsafe.Pointer(&mp.tls[0])) + ts.fn = unsafe.Pointer(funcPC(mstart)) + 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 { + gothrow("stopm holding locks") + } + if _g_.m.p != nil { + gothrow("stopm holding p") + } + if _g_.m.spinning { + _g_.m.spinning = false + xadd(&sched.nmspinning, -1) + } + +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 + goto retry + } + acquirep(_g_.m.nextp) + _g_.m.nextp = nil +} + +func mspinning() { + 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. +func startm(_p_ *p, spinning bool) { + lock(&sched.lock) + if _p_ == nil { + _p_ = pidleget() + if _p_ == nil { + unlock(&sched.lock) + if spinning { + xadd(&sched.nmspinning, -1) + } + return + } + } + mp := mget() + unlock(&sched.lock) + if mp == nil { + var fn func() + if spinning { + fn = mspinning + } + _newm(fn, _p_) + return + } + if mp.spinning { + gothrow("startm: m is spinning") + } + if mp.nextp != nil { + gothrow("startm: m has p") + } + mp.spinning = spinning + mp.nextp = _p_ + notewakeup(&mp.park) +} + +// Hands off P from syscall or locked M. +func handoffp(_p_ *p) { + // if it has local work, start it straight away + if _p_.runqhead != _p_.runqtail || sched.runqsize != 0 { + startm(_p_, false) + return + } + // no local work, check that there are no spinning/idle M's, + // otherwise our help is not required + if atomicload(&sched.nmspinning)+atomicload(&sched.npidle) == 0 && 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 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) && atomicload64(&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 !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 { + gothrow("stoplockedm: inconsistent locking") + } + if _g_.m.p != nil { + // 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_) + gothrow("stoplockedm: not runnable") + } + acquirep(_g_.m.nextp) + _g_.m.nextp = nil +} + +// Schedules the locked m to run the locked gp. +func startlockedm(gp *g) { + _g_ := getg() + + mp := gp.lockedm + if mp == _g_.m { + gothrow("startlockedm: locked to me") + } + if mp.nextp != nil { + gothrow("startlockedm: m has p") + } + // directly handoff current P to the locked m + incidlelocked(-1) + _p_ := releasep() + mp.nextp = _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 { + gothrow("gcstopm: not waiting for gc") + } + if _g_.m.spinning { + _g_.m.spinning = false + xadd(&sched.nmspinning, -1) + } + _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. +// Never returns. +func execute(gp *g) { + _g_ := getg() + + casgstatus(gp, _Grunnable, _Grunning) + gp.waitsince = 0 + gp.preempt = false + gp.stackguard0 = gp.stack.lo + _StackGuard + _g_.m.p.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) + } + + 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() *g { + _g_ := getg() + +top: + if sched.gcwaiting != 0 { + gcstopm() + goto top + } + if fingwait && fingwake { + if gp := wakefing(); gp != nil { + ready(gp) + } + } + + // local runq + if gp := runqget(_g_.m.p); gp != nil { + return gp + } + + // global runq + if sched.runqsize != 0 { + lock(&sched.lock) + gp := globrunqget(_g_.m.p, 0) + unlock(&sched.lock) + if gp != nil { + return gp + } + } + + // poll network - returns list of goroutines + if gp := netpoll(false); gp != nil { // non-blocking + injectglist(gp.schedlink) + casgstatus(gp, _Gwaiting, _Grunnable) + return gp + } + + // 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*atomicload(&sched.nmspinning) >= uint32(gomaxprocs)-atomicload(&sched.npidle) { // TODO: fast atomic + goto stop + } + if !_g_.m.spinning { + _g_.m.spinning = true + xadd(&sched.nmspinning, 1) + } + // random steal from other P's + for i := 0; i < int(2*gomaxprocs); i++ { + if sched.gcwaiting != 0 { + goto top + } + _p_ := allp[fastrand1()%uint32(gomaxprocs)] + var gp *g + if _p_ == _g_.m.p { + gp = runqget(_p_) + } else { + gp = runqsteal(_g_.m.p, _p_) + } + if gp != nil { + return gp + } + } +stop: + + // return P and block + lock(&sched.lock) + if sched.gcwaiting != 0 { + unlock(&sched.lock) + goto top + } + if sched.runqsize != 0 { + gp := globrunqget(_g_.m.p, 0) + unlock(&sched.lock) + return gp + } + _p_ := releasep() + pidleput(_p_) + unlock(&sched.lock) + if _g_.m.spinning { + _g_.m.spinning = false + xadd(&sched.nmspinning, -1) + } + + // check all runqueues once again + for i := 0; i < int(gomaxprocs); i++ { + _p_ := allp[i] + if _p_ != nil && _p_.runqhead != _p_.runqtail { + lock(&sched.lock) + _p_ = pidleget() + unlock(&sched.lock) + if _p_ != nil { + acquirep(_p_) + goto top + } + break + } + } + + // poll network + if xchg64(&sched.lastpoll, 0) != 0 { + if _g_.m.p != nil { + gothrow("findrunnable: netpoll with p") + } + if _g_.m.spinning { + gothrow("findrunnable: netpoll with spinning") + } + gp := netpoll(true) // block until new work is available + atomicstore64(&sched.lastpoll, uint64(nanotime())) + if gp != nil { + lock(&sched.lock) + _p_ = pidleget() + unlock(&sched.lock) + if _p_ != nil { + acquirep(_p_) + injectglist(gp.schedlink) + casgstatus(gp, _Gwaiting, _Grunnable) + return gp + } + injectglist(gp) + } + } + stopm() + goto top +} + +func resetspinning() { + _g_ := getg() + + var nmspinning uint32 + if _g_.m.spinning { + _g_.m.spinning = false + nmspinning = xadd(&sched.nmspinning, -1) + if nmspinning < 0 { + gothrow("findrunnable: negative nmspinning") + } + } else { + nmspinning = atomicload(&sched.nmspinning) + } + + // M wakeup policy is deliberately somewhat conservative (see nmspinning handling), + // so see if we need to wakeup another P here. + if nmspinning == 0 && atomicload(&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 + } + lock(&sched.lock) + var n int + for n = 0; glist != nil; n++ { + gp := glist + glist = gp.schedlink + 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 { + gothrow("schedule: holding locks") + } + + if _g_.m.lockedg != nil { + stoplockedm() + execute(_g_.m.lockedg) // Never returns. + } + +top: + if sched.gcwaiting != 0 { + gcstopm() + goto top + } + + var gp *g + // 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. + tick := _g_.m.p.schedtick + // This is a fancy way to say tick%61==0, + // it uses 2 MUL instructions instead of a single DIV and so is faster on modern processors. + if uint64(tick)-((uint64(tick)*0x4325c53f)>>36)*61 == 0 && sched.runqsize > 0 { + lock(&sched.lock) + gp = globrunqget(_g_.m.p, 1) + unlock(&sched.lock) + if gp != nil { + resetspinning() + } + } + if gp == nil { + gp = runqget(_g_.m.p) + if gp != nil && _g_.m.spinning { + gothrow("schedule: spinning with local work") + } + } + if gp == nil { + gp = findrunnable() // blocks until work is available + 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) +} + +// 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() + + if _g_.m.lockedg == nil { + _g_.m.curg.m = nil + _g_.m.curg = nil + } +} + +// Puts the current goroutine into a waiting state and calls unlockf. +// If unlockf returns false, the goroutine is resumed. +func park(unlockf func(*g, unsafe.Pointer) bool, lock unsafe.Pointer, reason string) { + _g_ := getg() + + _g_.m.waitlock = lock + _g_.m.waitunlockf = *(*unsafe.Pointer)(unsafe.Pointer(&unlockf)) + _g_.waitreason = reason + mcall(park_m) +} + +func parkunlock_c(gp *g, lock unsafe.Pointer) bool { + unlock((*mutex)(lock)) + return true +} + +// Puts the current goroutine into a waiting state and unlocks the lock. +// The goroutine can be made runnable again by calling ready(gp). +func parkunlock(lock *mutex, reason string) { + park(parkunlock_c, unsafe.Pointer(lock), reason) +} + +// park continuation on g0. +func park_m(gp *g) { + _g_ := getg() + + 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 { + casgstatus(gp, _Gwaiting, _Grunnable) + execute(gp) // Schedule it back, never returns. + } + } + schedule() +} + +// Gosched continuation on g0. +func gosched_m(gp *g) { + status := readgstatus(gp) + if status&^_Gscan != _Grunning { + dumpgstatus(gp) + gothrow("bad g status") + } + casgstatus(gp, _Grunning, _Grunnable) + dropg() + lock(&sched.lock) + globrunqput(gp) + unlock(&sched.lock) + + schedule() +} + +// Finishes execution of the current goroutine. +// Must be NOSPLIT because it is called from Go. (TODO - probably not anymore) +//go:nosplit +func goexit1() { + if raceenabled { + racegoend() + } + mcall(goexit0) +} + +// goexit continuation on g0. +func goexit0(gp *g) { + _g_ := getg() + + casgstatus(gp, _Grunning, _Gdead) + 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 + + dropg() + + if _g_.m.locked&^_LockExternal != 0 { + print("invalid m->locked = ", _g_.m.locked, "\n") + gothrow("internal lockOSThread error") + } + _g_.m.locked = 0 + gfput(_g_.m.p, gp) + schedule() +} + +//go:nosplit +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 = _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 +// +// 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. +//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(entersyscall_bad) + } + + if atomicload(&sched.sysmonwait) != 0 { // TODO: fast atomic + systemstack(entersyscall_sysmon) + save(pc, sp) + } + + _g_.m.mcache = nil + _g_.m.p.m = nil + atomicstore(&_g_.m.p.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_bad() { + var gp *g + gp = getg().m.curg + print("entersyscall inconsistent ", hex(gp.syscallsp), " [", hex(gp.stack.lo), ",", hex(gp.stack.hi), "]\n") + gothrow("entersyscall") +} + +func entersyscall_sysmon() { + lock(&sched.lock) + if atomicload(&sched.sysmonwait) != 0 { + atomicstore(&sched.sysmonwait, 0) + notewakeup(&sched.sysmonnote) + } + unlock(&sched.lock) +} + +func entersyscall_gcwait() { + _g_ := getg() + + lock(&sched.lock) + if sched.stopwait > 0 && cas(&_g_.m.p.status, _Psyscall, _Pgcstop) { + 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 + + // Leave SP around for GC and traceback. + save(getcallerpc(unsafe.Pointer(&dummy)), getcallersp(unsafe.Pointer(&dummy))) + _g_.syscallsp = _g_.sched.sp + _g_.syscallpc = _g_.sched.pc + casgstatus(_g_, _Grunning, _Gsyscall) + if _g_.syscallsp < _g_.stack.lo || _g_.stack.hi < _g_.syscallsp { + systemstack(entersyscall_bad) + } + + systemstack(entersyscallblock_handoff) + + // Resave for traceback during blocked call. + save(getcallerpc(unsafe.Pointer(&dummy)), getcallersp(unsafe.Pointer(&dummy))) + + _g_.m.locks-- +} + +func entersyscallblock_handoff() { + 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 +//go:nosplit +func exitsyscall(dummy int32) { + _g_ := getg() + + _g_.m.locks++ // see comment in entersyscall + if getcallersp(unsafe.Pointer(&dummy)) > _g_.syscallsp { + gothrow("exitsyscall: syscall frame is no longer valid") + } + + _g_.waitsince = 0 + if exitsyscallfast() { + if _g_.m.mcache == nil { + gothrow("lost mcache") + } + // There's a cpu for us, so we can run. + _g_.m.p.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_.m.locks-- + + // Call the scheduler. + mcall(exitsyscall0) + + if _g_.m.mcache == nil { + gothrow("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.syscalltick++ + _g_.throwsplit = false +} + +//go:nosplit +func exitsyscallfast() bool { + _g_ := getg() + + // Freezetheworld sets stopwait but does not retake P's. + if sched.stopwait != 0 { + _g_.m.p = nil + return false + } + + // Try to re-acquire the last P. + if _g_.m.p != nil && _g_.m.p.status == _Psyscall && cas(&_g_.m.p.status, _Psyscall, _Prunning) { + // There's a cpu for us, so we can run. + _g_.m.mcache = _g_.m.p.mcache + _g_.m.p.m = _g_.m + return true + } + + // Try to get any other idle P. + _g_.m.p = nil + if sched.pidle != nil { + var ok bool + systemstack(func() { + ok = exitsyscallfast_pidle() + }) + if ok { + return true + } + } + return false +} + +func exitsyscallfast_pidle() bool { + lock(&sched.lock) + _p_ := pidleget() + if _p_ != nil && atomicload(&sched.sysmonwait) != 0 { + atomicstore(&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 atomicload(&sched.sysmonwait) != 0 { + atomicstore(&sched.sysmonwait, 0) + notewakeup(&sched.sysmonnote) + } + unlock(&sched.lock) + if _p_ != nil { + acquirep(_p_) + execute(gp) // Never returns. + } + if _g_.m.lockedg != nil { + // Wait until another thread schedules gp and so m again. + stoplockedm() + execute(gp) // 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:nosplit +func syscall_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:nosplit +func syscall_AfterFork() { + systemstack(afterfork) +} + +// Allocate a new g, with a stack big enough for stacksize bytes. +func malg(stacksize int32) *g { + newg := allocg() + if stacksize >= 0 { + stacksize = round2(_StackSystem + stacksize) + systemstack(func() { + newg.stack = stackalloc(uint32(stacksize)) + }) + newg.stackguard0 = newg.stack.lo + _StackGuard + newg.stackguard1 = ^uintptr(0) + } + 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), ptrSize) + if hasLinkRegister { + argp = add(argp, ptrSize) // skip caller's saved LR + } + + 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 + gothrow("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*regSize-regSize { + gothrow("newproc: function arguments too large for new goroutine") + } + + _p_ := _g_.m.p + newg := gfget(_p_) + if newg == nil { + newg = malg(_StackMin) + casgstatus(newg, _Gidle, _Gdead) + allgadd(newg) // publishes with a g->status of Gdead so GC scanner doesn't look at uninitialized stack. + } + if newg.stack.hi == 0 { + gothrow("newproc1: newg missing stack") + } + + if readgstatus(newg) != _Gdead { + gothrow("newproc1: new g is not Gdead") + } + + sp := newg.stack.hi + sp -= 4 * regSize // extra space in case of reads slightly beyond frame + sp -= uintptr(siz) + memmove(unsafe.Pointer(sp), unsafe.Pointer(argp), uintptr(narg)) + if hasLinkRegister { + // caller's LR + sp -= ptrSize + *(*unsafe.Pointer)(unsafe.Pointer(sp)) = nil + } + + memclr(unsafe.Pointer(&newg.sched), unsafe.Sizeof(newg.sched)) + newg.sched.sp = sp + newg.sched.pc = funcPC(goexit) + _PCQuantum // +PCQuantum so that previous instruction is in same function + newg.sched.g = newg + gostartcallfn(&newg.sched, fn) + newg.gopc = callerpc + 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 = 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) + } + runqput(_p_, newg) + + if atomicload(&sched.npidle) != 0 && atomicload(&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 { + gothrow("gfput: bad status (not Gdead)") + } + + stksize := gp.stack.hi - gp.stack.lo + + if stksize != _FixedStack { + // non-standard stack size - free it. + stackfree(gp.stack) + gp.stack.lo = 0 + gp.stack.hi = 0 + gp.stackguard0 = 0 + } + + gp.schedlink = _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 + gp.schedlink = sched.gfree + sched.gfree = 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.gfree != nil { + lock(&sched.gflock) + for _p_.gfreecnt < 32 && sched.gfree != nil { + _p_.gfreecnt++ + gp = sched.gfree + sched.gfree = gp.schedlink + sched.ngfree-- + gp.schedlink = _p_.gfree + _p_.gfree = gp + } + unlock(&sched.gflock) + goto retry + } + if gp != nil { + _p_.gfree = gp.schedlink + _p_.gfreecnt-- + if gp.stack.lo == 0 { + // Stack was deallocated in gfput. Allocate a new one. + systemstack(func() { + gp.stack = stackalloc(_FixedStack) + }) + gp.stackguard0 = gp.stack.lo + _StackGuard + } else { + if raceenabled { + racemalloc(unsafe.Pointer(gp.stack.lo), gp.stack.hi-gp.stack.lo) + } + } + } + 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 + gp.schedlink = sched.gfree + sched.gfree = 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() { + gothrow("runtime: internal error: misuse of lockOSThread/unlockOSThread") +} + +func gcount() int32 { + n := int32(allglen) - sched.ngfree + 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() } + +var etext struct{} + +// Called if we receive a SIGPROF signal. +func sigprof(pc *uint8, sp *uint8, lr *uint8, gp *g, mp *m) { + var n int32 + var traceback bool + var stk [100]uintptr + + 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've only made sure that we + // can unwind user g's, so exclude the system g's. + // + // It is not quite as easy as testing gp == m->curg (the current user g) + // because 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 g is updated, we'll see a system g and not look closer. + // If SP 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, as in gogo. + // + // Because gogo is the only instance, we check whether the PC lies + // within that function, and if so, not ask for a traceback. This approach + // requires knowing the size of the gogo function, which we + // record in arch_*.h and check in runtime_test.go. + // + // There is another apparently viable approach, recorded here in case + // the "PC within gogo" 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 + usp := uintptr(unsafe.Pointer(sp)) + gogo := funcPC(gogo) + if gp == nil || gp != mp.curg || + usp < gp.stack.lo || gp.stack.hi < usp || + (gogo <= uintptr(unsafe.Pointer(pc)) && uintptr(unsafe.Pointer(pc)) < gogo+_RuntimeGogoBytes) { + traceback = false + } + + n = 0 + if traceback { + n = int32(gentraceback(uintptr(unsafe.Pointer(pc)), uintptr(unsafe.Pointer(sp)), uintptr(unsafe.Pointer(lr)), gp, 0, &stk[0], len(stk), nil, nil, _TraceTrap)) + } + if !traceback || n <= 0 { + // Normal traceback is impossible or has failed. + // See if it falls into several common cases. + n = 0 + if mp.ncgo > 0 && mp.curg != nil && mp.curg.syscallpc != 0 && mp.curg.syscallsp != 0 { + // Cgo, we can't unwind and symbolize arbitrary C code, + // so instead collect Go stack that leads to the cgo call. + // This is especially important on windows, since all syscalls are cgo calls. + n = int32(gentraceback(mp.curg.syscallpc, mp.curg.syscallsp, 0, mp.curg, 0, &stk[0], len(stk), nil, nil, 0)) + } + if GOOS == "windows" && n == 0 && mp.libcallg != nil && mp.libcallpc != 0 && mp.libcallsp != 0 { + // Libcall, i.e. runtime syscall on windows. + // Collect Go stack that leads to the call. + n = int32(gentraceback(mp.libcallpc, mp.libcallsp, 0, mp.libcallg, 0, &stk[0], len(stk), nil, nil, 0)) + } + 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 uintptr(unsafe.Pointer(pc)) > uintptr(unsafe.Pointer(&etext)) { + pc = (*uint8)(unsafe.Pointer(uintptr(funcPC(_ExternalCode) + _PCQuantum))) + } + stk[0] = uintptr(unsafe.Pointer(pc)) + if mp.gcing != 0 || mp.helpgc != 0 { + stk[1] = funcPC(_GC) + _PCQuantum + } else { + stk[1] = funcPC(_System) + _PCQuantum + } + } + } + + if prof.hz != 0 { + // Simple cas-lock to coordinate with setcpuprofilerate. + for !cas(&prof.lock, 0, 1) { + osyield() + } + if prof.hz != 0 { + cpuproftick(&stk[0], n) + } + atomicstore(&prof.lock, 0) + } + mp.mallocing-- +} + +// 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 !cas(&prof.lock, 0, 1) { + osyield() + } + prof.hz = hz + atomicstore(&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. +func procresize(new int32) { + old := gomaxprocs + if old < 0 || old > _MaxGomaxprocs || new <= 0 || new > _MaxGomaxprocs { + gothrow("procresize: invalid arg") + } + + // initialize new P's + for i := int32(0); i < new; i++ { + p := allp[i] + if p == nil { + p = newP() + p.id = i + p.status = _Pgcstop + atomicstorep(unsafe.Pointer(&allp[i]), unsafe.Pointer(p)) + } + if p.mcache == nil { + if old == 0 && i == 0 { + if getg().m.mcache == nil { + gothrow("missing mcache?") + } + p.mcache = getg().m.mcache // bootstrap + } else { + p.mcache = allocmcache() + } + } + } + + // redistribute runnable G's evenly + // collect all runnable goroutines in global queue preserving FIFO order + // FIFO order is required to ensure fairness even during frequent GCs + // see http://golang.org/issue/7126 + empty := false + for !empty { + empty = true + for i := int32(0); i < old; i++ { + p := allp[i] + if p.runqhead == p.runqtail { + continue + } + empty = false + // pop from tail of local queue + p.runqtail-- + gp := p.runq[p.runqtail%uint32(len(p.runq))] + // push onto head of global queue + gp.schedlink = sched.runqhead + sched.runqhead = gp + if sched.runqtail == nil { + sched.runqtail = gp + } + sched.runqsize++ + } + } + + // fill local queues with at most len(p.runq)/2 goroutines + // start at 1 because current M already executes some G and will acquire allp[0] below, + // so if we have a spare G we want to put it into allp[1]. + var _p_ p + for i := int32(1); i < new*int32(len(_p_.runq))/2 && sched.runqsize > 0; i++ { + gp := sched.runqhead + sched.runqhead = gp.schedlink + if sched.runqhead == nil { + sched.runqtail = nil + } + sched.runqsize-- + runqput(allp[i%new], gp) + } + + // free unused P's + for i := new; i < old; i++ { + p := allp[i] + freemcache(p.mcache) + p.mcache = nil + gfpurge(p) + p.status = _Pdead + // can't free P itself because it can be referenced by an M in syscall + } + + _g_ := getg() + if _g_.m.p != nil { + _g_.m.p.m = nil + } + _g_.m.p = nil + _g_.m.mcache = nil + p := allp[0] + p.m = nil + p.status = _Pidle + acquirep(p) + for i := new - 1; i > 0; i-- { + p := allp[i] + p.status = _Pidle + pidleput(p) + } + var int32p *int32 = &gomaxprocs // make compiler check that gomaxprocs is an int32 + atomicstore((*uint32)(unsafe.Pointer(int32p)), uint32(new)) +} + +// Associate p and the current m. +func acquirep(_p_ *p) { + _g_ := getg() + + if _g_.m.p != nil || _g_.m.mcache != nil { + gothrow("acquirep: already in go") + } + if _p_.m != nil || _p_.status != _Pidle { + id := int32(0) + if _p_.m != nil { + id = _p_.m.id + } + print("acquirep: p->m=", _p_.m, "(", id, ") p->status=", _p_.status, "\n") + gothrow("acquirep: invalid p state") + } + _g_.m.mcache = _p_.mcache + _g_.m.p = _p_ + _p_.m = _g_.m + _p_.status = _Prunning +} + +// Disassociate p and the current m. +func releasep() *p { + _g_ := getg() + + if _g_.m.p == nil || _g_.m.mcache == nil { + gothrow("releasep: invalid arg") + } + _p_ := _g_.m.p + if _p_.m != _g_.m || _p_.mcache != _g_.m.mcache || _p_.status != _Prunning { + print("releasep: m=", _g_.m, " m->p=", _g_.m.p, " p->m=", _p_.m, " m->mcache=", _g_.m.mcache, " p->mcache=", _p_.mcache, " p->status=", _p_.status, "\n") + gothrow("releasep: invalid p state") + } + _g_.m.p = nil + _g_.m.mcache = nil + _p_.m = nil + _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() { + // 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") + gothrow("checkdead: inconsistent counts") + } + + grunning := 0 + lock(&allglock) + for i := 0; i < len(allgs); i++ { + gp := allgs[i] + if gp.issystem { + 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") + gothrow("checkdead: runnable g") + } + } + unlock(&allglock) + if grunning == 0 { // possible if main goroutine calls runtime·Goexit() + gothrow("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 { + gothrow("checkdead: no p for timer") + } + mp := mget() + if mp == nil { + _newm(nil, _p_) + } else { + mp.nextp = _p_ + notewakeup(&mp.park) + } + return + } + + getg().m.throwing = -1 // do not dump full stacks + gothrow("all goroutines are asleep - deadlock!") +} + +func sysmon() { + // If we go two minutes without a garbage collection, force one to run. + forcegcperiod := int64(2 * 60 * 1e9) + + // 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 + + // Make wake-up period small enough for the sampling to be correct. + maxsleep := forcegcperiod / 2 + if scavengelimit < forcegcperiod { + maxsleep = scavengelimit / 2 + } + + 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 || atomicload(&sched.npidle) == uint32(gomaxprocs)) { // TODO: fast atomic + lock(&sched.lock) + if atomicload(&sched.gcwaiting) != 0 || atomicload(&sched.npidle) == uint32(gomaxprocs) { + atomicstore(&sched.sysmonwait, 1) + unlock(&sched.lock) + notetsleep(&sched.sysmonnote, maxsleep) + lock(&sched.lock) + atomicstore(&sched.sysmonwait, 0) + noteclear(&sched.sysmonnote) + idle = 0 + delay = 20 + } + unlock(&sched.lock) + } + // poll network if not polled for more than 10ms + lastpoll := int64(atomicload64(&sched.lastpoll)) + now := nanotime() + unixnow := unixnanotime() + if lastpoll != 0 && lastpoll+10*1000*1000 < now { + 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(atomicload64(&memstats.last_gc)) + if lastgc != 0 && unixnow-lastgc > forcegcperiod && atomicload(&forcegc.idle) != 0 { + lock(&forcegc.lock) + forcegc.idle = 0 + forcegc.g.schedlink = nil + 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 +} + +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 _p_.runqhead == _p_.runqtail && atomicload(&sched.nmspinning)+atomicload(&sched.npidle) > 0 && pd.syscallwhen+10*1000*1000 > now { + continue + } + // Need to decrement number of idle locked M's + // (pretending that one more is running) before the CAS. + // Otherwise the M from which we retake can exit the syscall, + // increment nmidle and report deadlock. + incidlelocked(-1) + if cas(&_p_.status, s, _Pidle) { + n++ + handoffp(_p_) + } + incidlelocked(1) + } else if s == _Prunning { + // Preempt G if it's running for more than 10ms. + t := int64(_p_.schedtick) + if int64(pd.schedtick) != t { + pd.schedtick = uint32(t) + pd.schedwhen = now + continue + } + if pd.schedwhen+10*1000*1000 > now { + continue + } + preemptone(_p_) + } + } + return uint32(n) +} + +// Tell all goroutines that they have been preempted and they should stop. +// This function is purely best-effort. It can fail to inform a goroutine if a +// processor just started running it. +// No locks need to be held. +// Returns true if preemption request was issued to at least one goroutine. +func preemptall() bool { + res := false + for i := int32(0); i < gomaxprocs; i++ { + _p_ := allp[i] + if _p_ == nil || _p_.status != _Prunning { + continue + } + if preemptone(_p_) { + res = true + } + } + return res +} + +// Tell the goroutine running on processor P to stop. +// This function is purely best-effort. It can incorrectly fail to inform the +// goroutine. It can send inform the wrong goroutine. Even if it informs the +// correct goroutine, that goroutine might ignore the request if it is +// simultaneously executing newstack. +// No lock needs to be held. +// Returns true if preemption request was issued. +// The actual preemption will happen at some point in the future +// and will be indicated by the gp->status no longer being +// Grunning +func preemptone(_p_ *p) bool { + mp := _p_.m + if mp == nil || mp == getg().m { + return false + } + gp := mp.curg + if gp == nil || gp == mp.g0 { + return false + } + + gp.preempt = true + + // Every call in a go routine checks for stack overflow by + // comparing the current stack pointer to gp->stackguard0. + // Setting gp->stackguard0 to StackPreempt folds + // preemption into the normal stack overflow check. + gp.stackguard0 = stackPreempt + return true +} + +var starttime int64 + +func schedtrace(detailed bool) { + now := nanotime() + if starttime == 0 { + starttime = now + } + + lock(&sched.lock) + print("SCHED ", (now-starttime)/1e6, "ms: gomaxprocs=", gomaxprocs, " idleprocs=", sched.npidle, " threads=", sched.mcount, " spinningthreads=", sched.nmspinning, " idlethreads=", sched.nmidle, " runqueue=", sched.runqsize) + if detailed { + print(" gcwaiting=", sched.gcwaiting, " nmidlelocked=", sched.nmidlelocked, " stopwait=", sched.stopwait, " sysmonwait=", sched.sysmonwait, "\n") + } + // We must be careful while reading data from P's, M's and G's. + // Even if we hold schedlock, most data can be changed concurrently. + // E.g. (p->m ? p->m->id : -1) can crash if p->m changes from non-nil to nil. + for i := int32(0); i < gomaxprocs; i++ { + _p_ := allp[i] + if _p_ == nil { + continue + } + mp := _p_.m + h := atomicload(&_p_.runqhead) + t := atomicload(&_p_.runqtail) + if detailed { + id := int32(-1) + if mp != nil { + id = mp.id + } + print(" P", i, ": status=", _p_.status, " schedtick=", _p_.schedtick, " syscalltick=", _p_.syscalltick, " m=", id, " runqsize=", t-h, " gfreecnt=", _p_.gfreecnt, "\n") + } else { + // In non-detailed mode format lengths of per-P run queues as: + // [len1 len2 len3 len4] + print(" ") + if i == 0 { + print("[") + } + print(t - h) + if i == gomaxprocs-1 { + print("]\n") + } + } + } + + if !detailed { + unlock(&sched.lock) + return + } + + for mp := allm; mp != nil; mp = mp.alllink { + _p_ := mp.p + gp := mp.curg + lockedg := mp.lockedg + id1 := int32(-1) + if _p_ != nil { + id1 = _p_.id + } + id2 := int64(-1) + if gp != nil { + id2 = gp.goid + } + id3 := int64(-1) + if lockedg != nil { + id3 = lockedg.goid + } + print(" M", mp.id, ": p=", id1, " curg=", id2, " mallocing=", mp.mallocing, " throwing=", mp.throwing, " gcing=", mp.gcing, ""+" locks=", mp.locks, " dying=", mp.dying, " helpgc=", mp.helpgc, " spinning=", mp.spinning, " blocked=", getg().m.blocked, " lockedg=", id3, "\n") + } + + lock(&allglock) + for gi := 0; gi < len(allgs); gi++ { + gp := allgs[gi] + mp := gp.m + lockedm := gp.lockedm + id1 := int32(-1) + if mp != nil { + id1 = mp.id + } + id2 := int32(-1) + if lockedm != nil { + id2 = lockedm.id + } + print(" G", gp.goid, ": status=", readgstatus(gp), "(", gp.waitreason, ") m=", id1, " lockedm=", id2, "\n") + } + unlock(&allglock) + unlock(&sched.lock) +} + +// Put mp on midle list. +// Sched must be locked. +func mput(mp *m) { + mp.schedlink = sched.midle + sched.midle = mp + sched.nmidle++ + checkdead() +} + +// Try to get an m from midle list. +// Sched must be locked. +func mget() *m { + mp := sched.midle + if mp != nil { + sched.midle = mp.schedlink + sched.nmidle-- + } + return mp +} + +// Put gp on the global runnable queue. +// Sched must be locked. +func globrunqput(gp *g) { + gp.schedlink = nil + if sched.runqtail != nil { + sched.runqtail.schedlink = gp + } else { + sched.runqhead = gp + } + sched.runqtail = gp + sched.runqsize++ +} + +// Put a batch of runnable goroutines on the global runnable queue. +// Sched must be locked. +func globrunqputbatch(ghead *g, gtail *g, n int32) { + gtail.schedlink = nil + if sched.runqtail != nil { + sched.runqtail.schedlink = ghead + } else { + sched.runqhead = ghead + } + sched.runqtail = gtail + sched.runqsize += n +} + +// Try get a batch of G's from the global runnable queue. +// Sched must be locked. +func globrunqget(_p_ *p, max int32) *g { + if sched.runqsize == 0 { + return nil + } + + n := sched.runqsize/gomaxprocs + 1 + if n > sched.runqsize { + n = sched.runqsize + } + if max > 0 && n > max { + n = max + } + if n > int32(len(_p_.runq))/2 { + n = int32(len(_p_.runq)) / 2 + } + + sched.runqsize -= n + if sched.runqsize == 0 { + sched.runqtail = nil + } + + gp := sched.runqhead + sched.runqhead = gp.schedlink + n-- + for ; n > 0; n-- { + gp1 := sched.runqhead + sched.runqhead = gp1.schedlink + runqput(_p_, gp1) + } + return gp +} + +// Put p to on _Pidle list. +// Sched must be locked. +func pidleput(_p_ *p) { + _p_.link = sched.pidle + sched.pidle = _p_ + xadd(&sched.npidle, 1) // TODO: fast atomic +} + +// Try get a p from _Pidle list. +// Sched must be locked. +func pidleget() *p { + _p_ := sched.pidle + if _p_ != nil { + sched.pidle = _p_.link + xadd(&sched.npidle, -1) // TODO: fast atomic + } + return _p_ +} + +// Try to put g on local runnable queue. +// If it's full, put onto global queue. +// Executed only by the owner P. +func runqput(_p_ *p, gp *g) { +retry: + h := atomicload(&_p_.runqhead) // load-acquire, synchronize with consumers + t := _p_.runqtail + if t-h < uint32(len(_p_.runq)) { + _p_.runq[t%uint32(len(_p_.runq))] = gp + atomicstore(&_p_.runqtail, t+1) // store-release, makes the item available for consumption + return + } + if runqputslow(_p_, gp, h, t) { + return + } + // the queue is not full, now the put above must suceed + goto retry +} + +// Put g and a batch of work from local runnable queue on global queue. +// Executed only by the owner P. +func runqputslow(_p_ *p, gp *g, h, t uint32) bool { + var batch [len(_p_.runq)/2 + 1]*g + + // First, grab a batch from local queue. + n := t - h + n = n / 2 + if n != uint32(len(_p_.runq)/2) { + gothrow("runqputslow: queue is not full") + } + for i := uint32(0); i < n; i++ { + batch[i] = _p_.runq[(h+i)%uint32(len(_p_.runq))] + } + if !cas(&_p_.runqhead, h, h+n) { // cas-release, commits consume + return false + } + batch[n] = gp + + // Link the goroutines. + for i := uint32(0); i < n; i++ { + batch[i].schedlink = batch[i+1] + } + + // Now put the batch on global queue. + lock(&sched.lock) + globrunqputbatch(batch[0], batch[n], int32(n+1)) + unlock(&sched.lock) + return true +} + +// Get g from local runnable queue. +// Executed only by the owner P. +func runqget(_p_ *p) *g { + for { + h := atomicload(&_p_.runqhead) // load-acquire, synchronize with other consumers + t := _p_.runqtail + if t == h { + return nil + } + gp := _p_.runq[h%uint32(len(_p_.runq))] + if cas(&_p_.runqhead, h, h+1) { // cas-release, commits consume + return gp + } + } +} + +// Grabs a batch of goroutines from local runnable queue. +// batch array must be of size len(p->runq)/2. Returns number of grabbed goroutines. +// Can be executed by any P. +func runqgrab(_p_ *p, batch []*g) uint32 { + for { + h := atomicload(&_p_.runqhead) // load-acquire, synchronize with other consumers + t := atomicload(&_p_.runqtail) // load-acquire, synchronize with the producer + n := t - h + n = n - n/2 + if n == 0 { + return 0 + } + if n > uint32(len(_p_.runq)/2) { // read inconsistent h and t + continue + } + for i := uint32(0); i < n; i++ { + batch[i] = _p_.runq[(h+i)%uint32(len(_p_.runq))] + } + if cas(&_p_.runqhead, h, h+n) { // cas-release, commits consume + return n + } + } +} + +// Steal half of elements from local runnable queue of p2 +// and put onto local runnable queue of p. +// Returns one of the stolen elements (or nil if failed). +func runqsteal(_p_, p2 *p) *g { + var batch [len(_p_.runq) / 2]*g + + n := runqgrab(p2, batch[:]) + if n == 0 { + return nil + } + n-- + gp := batch[n] + if n == 0 { + return gp + } + h := atomicload(&_p_.runqhead) // load-acquire, synchronize with consumers + t := _p_.runqtail + if t-h+n >= uint32(len(_p_.runq)) { + gothrow("runqsteal: runq overflow") + } + for i := uint32(0); i < n; i++ { + _p_.runq[(t+i)%uint32(len(_p_.runq))] = batch[i] + } + atomicstore(&_p_.runqtail, t+n) // store-release, makes the item available for consumption + return gp +} + +func testSchedLocalQueue() { + _p_ := new(p) + gs := make([]g, len(_p_.runq)) + for i := 0; i < len(_p_.runq); i++ { + if runqget(_p_) != nil { + gothrow("runq is not empty initially") + } + for j := 0; j < i; j++ { + runqput(_p_, &gs[i]) + } + for j := 0; j < i; j++ { + if runqget(_p_) != &gs[i] { + print("bad element at iter ", i, "/", j, "\n") + gothrow("bad element") + } + } + if runqget(_p_) != nil { + gothrow("runq is not empty afterwards") + } + } +} + +func testSchedLocalQueueSteal() { + p1 := new(p) + p2 := new(p) + gs := make([]g, len(p1.runq)) + for i := 0; i < len(p1.runq); i++ { + for j := 0; j < i; j++ { + gs[j].sig = 0 + runqput(p1, &gs[j]) + } + gp := runqsteal(p2, p1) + s := 0 + if gp != nil { + s++ + gp.sig++ + } + for { + gp = runqget(p2) + if gp == nil { + break + } + s++ + gp.sig++ + } + for { + gp = runqget(p1) + if gp == nil { + break + } + gp.sig++ + } + for j := 0; j < i; j++ { + if gs[j].sig != 1 { + print("bad element ", j, "(", gs[j].sig, ") at iter ", i, "\n") + gothrow("bad element") + } + } + if s != i/2 && s != i/2+1 { + print("bad steal ", s, ", want ", i/2, " or ", i/2+1, ", iter ", i, "\n") + gothrow("bad steal") + } + } +} + +func setMaxThreads(in int) (out int) { + lock(&sched.lock) + out = int(sched.maxmcount) + sched.maxmcount = int32(in) + checkmcount() + unlock(&sched.lock) + return +} + +var goexperiment string = "GOEXPERIMENT" // TODO: defined in zaexperiment.h + +func haveexperiment(name string) bool { + x := goexperiment + for x != "" { + xname := "" + i := index(x, ",") + if i < 0 { + xname, x = x, "" + } else { + xname, x = x[:i], x[i+1:] + } + if xname == name { + return true + } + } + return false +} + +//go:nosplit +func sync_procPin() int { + _g_ := getg() + mp := _g_.m + + mp.locks++ + return int(mp.p.id) +} + +//go:nosplit +func sync_procUnpin() { + _g_ := getg() + _g_.m.locks-- +} diff --cc src/runtime/stack1.go index ad83e58951,0000000000..1fd61ce1a2 mode 100644,000000..100644 --- a/src/runtime/stack1.go +++ b/src/runtime/stack1.go @@@ -1,807 -1,0 +1,801 @@@ +// Copyright 2013 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 "unsafe" + +const ( + // StackDebug == 0: no logging + // == 1: logging of per-stack operations + // == 2: logging of per-frame operations + // == 3: logging of per-word updates + // == 4: logging of per-word reads + stackDebug = 0 + stackFromSystem = 0 // allocate stacks from system memory instead of the heap + stackFaultOnFree = 0 // old stacks are mapped noaccess to detect use after free + stackPoisonCopy = 0 // fill stack that should not be accessed with garbage, to detect bad dereferences during copy + + stackCache = 1 +) + +const ( + uintptrMask = 1<<(8*ptrSize) - 1 + poisonGC = uintptrMask & 0xf969696969696969 + poisonStack = uintptrMask & 0x6868686868686868 + + // Goroutine preemption request. + // Stored into g->stackguard0 to cause split stack check failure. + // Must be greater than any real sp. + // 0xfffffade in hex. + stackPreempt = uintptrMask & -1314 + + // Thread is forking. + // Stored into g->stackguard0 to cause split stack check failure. + // Must be greater than any real sp. + stackFork = uintptrMask & -1234 +) + +// Global pool of spans that have free stacks. +// Stacks are assigned an order according to size. +// order = log_2(size/FixedStack) +// There is a free list for each order. +// TODO: one lock per order? +var stackpool [_NumStackOrders]mspan +var stackpoolmu mutex + +var stackfreequeue stack + +func stackinit() { + if _StackCacheSize&_PageMask != 0 { + gothrow("cache size must be a multiple of page size") + } + for i := range stackpool { + mSpanList_Init(&stackpool[i]) + } +} + +// Allocates a stack from the free pool. Must be called with +// stackpoolmu held. +func stackpoolalloc(order uint8) *mlink { + list := &stackpool[order] + s := list.next + if s == list { + // no free stacks. Allocate another span worth. + s = mHeap_AllocStack(&mheap_, _StackCacheSize>>_PageShift) + if s == nil { + gothrow("out of memory") + } + if s.ref != 0 { + gothrow("bad ref") + } + if s.freelist != nil { + gothrow("bad freelist") + } + for i := uintptr(0); i < _StackCacheSize; i += _FixedStack << order { + x := (*mlink)(unsafe.Pointer(uintptr(s.start)<<_PageShift + i)) + x.next = s.freelist + s.freelist = x + } + mSpanList_Insert(list, s) + } + x := s.freelist + if x == nil { + gothrow("span has no free stacks") + } + s.freelist = x.next + s.ref++ + if s.freelist == nil { + // all stacks in s are allocated. + mSpanList_Remove(s) + } + return x +} + +// Adds stack x to the free pool. Must be called with stackpoolmu held. +func stackpoolfree(x *mlink, order uint8) { + s := mHeap_Lookup(&mheap_, (unsafe.Pointer)(x)) + if s.state != _MSpanStack { + gothrow("freeing stack not in a stack span") + } + if s.freelist == nil { + // s will now have a free stack + mSpanList_Insert(&stackpool[order], s) + } + x.next = s.freelist + s.freelist = x + s.ref-- + if s.ref == 0 { + // span is completely free - return to heap + mSpanList_Remove(s) + s.freelist = nil + mHeap_FreeStack(&mheap_, s) + } +} + +// stackcacherefill/stackcacherelease implement a global pool of stack segments. +// The pool is required to prevent unlimited growth of per-thread caches. +func stackcacherefill(c *mcache, order uint8) { + if stackDebug >= 1 { + print("stackcacherefill order=", order, "\n") + } + + // Grab some stacks from the global cache. + // Grab half of the allowed capacity (to prevent thrashing). + var list *mlink + var size uintptr + lock(&stackpoolmu) + for size < _StackCacheSize/2 { + x := stackpoolalloc(order) + x.next = list + list = x + size += _FixedStack << order + } + unlock(&stackpoolmu) + c.stackcache[order].list = list + c.stackcache[order].size = size +} + +func stackcacherelease(c *mcache, order uint8) { + if stackDebug >= 1 { + print("stackcacherelease order=", order, "\n") + } + x := c.stackcache[order].list + size := c.stackcache[order].size + lock(&stackpoolmu) + for size > _StackCacheSize/2 { + y := x.next + stackpoolfree(x, order) + x = y + size -= _FixedStack << order + } + unlock(&stackpoolmu) + c.stackcache[order].list = x + c.stackcache[order].size = size +} + +func stackcache_clear(c *mcache) { + if stackDebug >= 1 { + print("stackcache clear\n") + } + lock(&stackpoolmu) + for order := uint8(0); order < _NumStackOrders; order++ { + x := c.stackcache[order].list + for x != nil { + y := x.next + stackpoolfree(x, order) + x = y + } + c.stackcache[order].list = nil + c.stackcache[order].size = 0 + } + unlock(&stackpoolmu) +} + +func stackalloc(n uint32) stack { + // Stackalloc must be called on scheduler stack, so that we + // never try to grow the stack during the code that stackalloc runs. + // Doing so would cause a deadlock (issue 1547). + thisg := getg() + if thisg != thisg.m.g0 { + gothrow("stackalloc not on scheduler stack") + } + if n&(n-1) != 0 { + gothrow("stack size not a power of 2") + } + if stackDebug >= 1 { + print("stackalloc ", n, "\n") + } + + if debug.efence != 0 || stackFromSystem != 0 { + v := sysAlloc(round(uintptr(n), _PageSize), &memstats.stacks_sys) + if v == nil { + gothrow("out of memory (stackalloc)") + } + return stack{uintptr(v), uintptr(v) + uintptr(n)} + } + + // Small stacks are allocated with a fixed-size free-list allocator. + // If we need a stack of a bigger size, we fall back on allocating + // a dedicated span. + var v unsafe.Pointer + if stackCache != 0 && n < _FixedStack<<_NumStackOrders && n < _StackCacheSize { + order := uint8(0) + n2 := n + for n2 > _FixedStack { + order++ + n2 >>= 1 + } + var x *mlink + c := thisg.m.mcache + if c == nil || thisg.m.gcing != 0 || thisg.m.helpgc != 0 { + // c == nil can happen in the guts of exitsyscall or + // procresize. Just get a stack from the global pool. + // Also don't touch stackcache during gc + // as it's flushed concurrently. + lock(&stackpoolmu) + x = stackpoolalloc(order) + unlock(&stackpoolmu) + } else { + x = c.stackcache[order].list + if x == nil { + stackcacherefill(c, order) + x = c.stackcache[order].list + } + c.stackcache[order].list = x.next + c.stackcache[order].size -= uintptr(n) + } + v = (unsafe.Pointer)(x) + } else { + s := mHeap_AllocStack(&mheap_, round(uintptr(n), _PageSize)>>_PageShift) + if s == nil { + gothrow("out of memory") + } + v = (unsafe.Pointer)(s.start << _PageShift) + } + + if raceenabled { + racemalloc(v, uintptr(n)) + } + if stackDebug >= 1 { + print(" allocated ", v, "\n") + } + return stack{uintptr(v), uintptr(v) + uintptr(n)} +} + +func stackfree(stk stack) { + gp := getg() + n := stk.hi - stk.lo + v := (unsafe.Pointer)(stk.lo) + if n&(n-1) != 0 { + gothrow("stack not a power of 2") + } + if stackDebug >= 1 { + println("stackfree", v, n) + memclr(v, n) // for testing, clobber stack data + } + if debug.efence != 0 || stackFromSystem != 0 { + if debug.efence != 0 || stackFaultOnFree != 0 { + sysFault(v, n) + } else { + sysFree(v, n, &memstats.stacks_sys) + } + return + } + if stackCache != 0 && n < _FixedStack<<_NumStackOrders && n < _StackCacheSize { + order := uint8(0) + n2 := n + for n2 > _FixedStack { + order++ + n2 >>= 1 + } + x := (*mlink)(v) + c := gp.m.mcache + if c == nil || gp.m.gcing != 0 || gp.m.helpgc != 0 { + lock(&stackpoolmu) + stackpoolfree(x, order) + unlock(&stackpoolmu) + } else { + if c.stackcache[order].size >= _StackCacheSize { + stackcacherelease(c, order) + } + x.next = c.stackcache[order].list + c.stackcache[order].list = x + c.stackcache[order].size += n + } + } else { + s := mHeap_Lookup(&mheap_, v) + if s.state != _MSpanStack { + println(hex(s.start<<_PageShift), v) + gothrow("bad span state") + } + mHeap_FreeStack(&mheap_, s) + } +} + +var maxstacksize uintptr = 1 << 20 // enough until runtime.main sets it for real + +var mapnames = []string{ + _BitsDead: "---", + _BitsScalar: "scalar", + _BitsPointer: "ptr", +} + +// Stack frame layout +// +// (x86) +// +------------------+ +// | args from caller | +// +------------------+ <- frame->argp +// | return address | +// +------------------+ <- frame->varp +// | locals | +// +------------------+ +// | args to callee | +// +------------------+ <- frame->sp +// +// (arm) +// +------------------+ +// | args from caller | +// +------------------+ <- frame->argp +// | caller's retaddr | +// +------------------+ <- frame->varp +// | locals | +// +------------------+ +// | args to callee | +// +------------------+ +// | return address | +// +------------------+ <- frame->sp + +type adjustinfo struct { + old stack + delta uintptr // ptr distance from old to new stack (newbase - oldbase) +} + +// Adjustpointer checks whether *vpp is in the old stack described by adjinfo. +// If so, it rewrites *vpp to point into the new stack. +func adjustpointer(adjinfo *adjustinfo, vpp unsafe.Pointer) { + pp := (*unsafe.Pointer)(vpp) + p := *pp + if stackDebug >= 4 { + print(" ", pp, ":", p, "\n") + } + if adjinfo.old.lo <= uintptr(p) && uintptr(p) < adjinfo.old.hi { + *pp = add(p, adjinfo.delta) + if stackDebug >= 3 { + print(" adjust ptr ", pp, ":", p, " -> ", *pp, "\n") + } + } +} + +type gobitvector struct { + n uintptr + bytedata []uint8 +} + +func gobv(bv bitvector) gobitvector { + return gobitvector{ + uintptr(bv.n), + (*[1 << 30]byte)(unsafe.Pointer(bv.bytedata))[:(bv.n+7)/8], + } +} + +func ptrbits(bv *gobitvector, i uintptr) uint8 { + return (bv.bytedata[i/4] >> ((i & 3) * 2)) & 3 +} + +// bv describes the memory starting at address scanp. +// Adjust any pointers contained therein. +func adjustpointers(scanp unsafe.Pointer, cbv *bitvector, adjinfo *adjustinfo, f *_func) { + bv := gobv(*cbv) + minp := adjinfo.old.lo + maxp := adjinfo.old.hi + delta := adjinfo.delta + num := uintptr(bv.n / _BitsPerPointer) + for i := uintptr(0); i < num; i++ { + if stackDebug >= 4 { + print(" ", add(scanp, i*ptrSize), ":", mapnames[ptrbits(&bv, i)], ":", hex(*(*uintptr)(add(scanp, i*ptrSize))), " # ", i, " ", bv.bytedata[i/4], "\n") + } + switch ptrbits(&bv, i) { + default: + gothrow("unexpected pointer bits") + case _BitsDead: + if debug.gcdead != 0 { + *(*unsafe.Pointer)(add(scanp, i*ptrSize)) = unsafe.Pointer(uintptr(poisonStack)) + } + case _BitsScalar: + // ok + case _BitsPointer: + p := *(*unsafe.Pointer)(add(scanp, i*ptrSize)) + up := uintptr(p) + if f != nil && 0 < up && up < _PageSize && invalidptr != 0 || up == poisonGC || up == poisonStack { + // Looks like a junk value in a pointer slot. + // Live analysis wrong? + getg().m.traceback = 2 + print("runtime: bad pointer in frame ", gofuncname(f), " at ", add(scanp, i*ptrSize), ": ", p, "\n") + gothrow("invalid stack pointer") + } + if minp <= up && up < maxp { + if stackDebug >= 3 { + print("adjust ptr ", p, " ", gofuncname(f), "\n") + } + *(*unsafe.Pointer)(add(scanp, i*ptrSize)) = unsafe.Pointer(up + delta) + } + } + } +} + +// Note: the argument/return area is adjusted by the callee. +func adjustframe(frame *stkframe, arg unsafe.Pointer) bool { + adjinfo := (*adjustinfo)(arg) + targetpc := frame.continpc + if targetpc == 0 { + // Frame is dead. + return true + } + f := frame.fn + if stackDebug >= 2 { + print(" adjusting ", funcname(f), " frame=[", hex(frame.sp), ",", hex(frame.fp), "] pc=", hex(frame.pc), " continpc=", hex(frame.continpc), "\n") + } + if f.entry == systemstack_switchPC { + // A special routine at the bottom of stack of a goroutine that does an systemstack call. + // We will allow it to be copied even though we don't + // have full GC info for it (because it is written in asm). + return true + } + if targetpc != f.entry { + targetpc-- + } + pcdata := pcdatavalue(f, _PCDATA_StackMapIndex, targetpc) + if pcdata == -1 { + pcdata = 0 // in prologue + } + + // Adjust local variables if stack frame has been allocated. + size := frame.varp - frame.sp + var minsize uintptr + if thechar != '6' && thechar != '8' { + minsize = ptrSize + } else { + minsize = 0 + } + if size > minsize { + var bv bitvector + stackmap := (*stackmap)(funcdata(f, _FUNCDATA_LocalsPointerMaps)) + if stackmap == nil || stackmap.n <= 0 { + print("runtime: frame ", funcname(f), " untyped locals ", hex(frame.varp-size), "+", hex(size), "\n") + gothrow("missing stackmap") + } + // Locals bitmap information, scan just the pointers in locals. + if pcdata < 0 || pcdata >= stackmap.n { + // don't know where we are + print("runtime: pcdata is ", pcdata, " and ", stackmap.n, " locals stack map entries for ", funcname(f), " (targetpc=", targetpc, ")\n") + gothrow("bad symbol table") + } + bv = stackmapdata(stackmap, pcdata) + size = (uintptr(bv.n) * ptrSize) / _BitsPerPointer + if stackDebug >= 3 { + print(" locals ", pcdata, "/", stackmap.n, " ", size/ptrSize, " words ", bv.bytedata, "\n") + } + adjustpointers(unsafe.Pointer(frame.varp-size), &bv, adjinfo, f) + } + + // Adjust arguments. + if frame.arglen > 0 { + var bv bitvector + if frame.argmap != nil { + bv = *frame.argmap + } else { + stackmap := (*stackmap)(funcdata(f, _FUNCDATA_ArgsPointerMaps)) + if stackmap == nil || stackmap.n <= 0 { + print("runtime: frame ", funcname(f), " untyped args ", frame.argp, "+", uintptr(frame.arglen), "\n") + gothrow("missing stackmap") + } + if pcdata < 0 || pcdata >= stackmap.n { + // don't know where we are + print("runtime: pcdata is ", pcdata, " and ", stackmap.n, " args stack map entries for ", funcname(f), " (targetpc=", targetpc, ")\n") + gothrow("bad symbol table") + } + bv = stackmapdata(stackmap, pcdata) + } + if stackDebug >= 3 { + print(" args\n") + } + adjustpointers(unsafe.Pointer(frame.argp), &bv, adjinfo, nil) + } + return true +} + +func adjustctxt(gp *g, adjinfo *adjustinfo) { + adjustpointer(adjinfo, (unsafe.Pointer)(&gp.sched.ctxt)) +} + +func adjustdefers(gp *g, adjinfo *adjustinfo) { + // Adjust defer argument blocks the same way we adjust active stack frames. + tracebackdefers(gp, adjustframe, noescape(unsafe.Pointer(adjinfo))) + + // Adjust pointers in the Defer structs. + // Defer structs themselves are never on the stack. + for d := gp._defer; d != nil; d = d.link { + adjustpointer(adjinfo, (unsafe.Pointer)(&d.fn)) + adjustpointer(adjinfo, (unsafe.Pointer)(&d.argp)) + adjustpointer(adjinfo, (unsafe.Pointer)(&d._panic)) + } +} + +func adjustpanics(gp *g, adjinfo *adjustinfo) { + // Panics are on stack and already adjusted. + // Update pointer to head of list in G. + adjustpointer(adjinfo, (unsafe.Pointer)(&gp._panic)) +} + +func adjustsudogs(gp *g, adjinfo *adjustinfo) { + // the data elements pointed to by a SudoG structure + // might be in the stack. + for s := gp.waiting; s != nil; s = s.waitlink { + adjustpointer(adjinfo, (unsafe.Pointer)(&s.elem)) + adjustpointer(adjinfo, (unsafe.Pointer)(&s.selectdone)) + } +} + +func fillstack(stk stack, b byte) { + for p := stk.lo; p < stk.hi; p++ { + *(*byte)(unsafe.Pointer(p)) = b + } +} + +// Copies gp's stack to a new stack of a different size. +func copystack(gp *g, newsize uintptr) { + if gp.syscallsp != 0 { + gothrow("stack growth not allowed in system call") + } + old := gp.stack + if old.lo == 0 { + gothrow("nil stackbase") + } + used := old.hi - gp.sched.sp + + // allocate new stack + new := stackalloc(uint32(newsize)) + if stackPoisonCopy != 0 { + fillstack(new, 0xfd) + } + if stackDebug >= 1 { + print("copystack gp=", gp, " [", hex(old.lo), " ", hex(old.hi-used), " ", hex(old.hi), "]/", old.hi-old.lo, " -> [", hex(new.lo), " ", hex(new.hi-used), " ", hex(new.hi), "]/", newsize, "\n") + } + + // adjust pointers in the to-be-copied frames + var adjinfo adjustinfo + adjinfo.old = old + adjinfo.delta = new.hi - old.hi + gentraceback(^uintptr(0), ^uintptr(0), 0, gp, 0, nil, 0x7fffffff, adjustframe, noescape(unsafe.Pointer(&adjinfo)), 0) + + // adjust other miscellaneous things that have pointers into stacks. + adjustctxt(gp, &adjinfo) + adjustdefers(gp, &adjinfo) + adjustpanics(gp, &adjinfo) + adjustsudogs(gp, &adjinfo) + + // copy the stack to the new location + if stackPoisonCopy != 0 { + fillstack(new, 0xfb) + } + memmove(unsafe.Pointer(new.hi-used), unsafe.Pointer(old.hi-used), used) + - oldstatus := readgstatus(gp) - oldstatus &^= _Gscan - if oldstatus == _Gwaiting || oldstatus == _Grunnable { - casgstatus(gp, oldstatus, _Gcopystack) // oldstatus is Gwaiting or Grunnable - } else { - gothrow("copystack: bad status, not Gwaiting or Grunnable") - } ++ oldstatus := casgcopystack(gp) // cas from Gwaiting or Grunnable to Gcopystack, return old status + + // Swap out old stack for new one + gp.stack = new + gp.stackguard0 = new.lo + _StackGuard // NOTE: might clobber a preempt request + gp.sched.sp = new.hi - used + + casgstatus(gp, _Gcopystack, oldstatus) // oldstatus is Gwaiting or Grunnable + + // free old stack + if stackPoisonCopy != 0 { + fillstack(old, 0xfc) + } + if newsize > old.hi-old.lo { + // growing, free stack immediately + stackfree(old) + } else { + // shrinking, queue up free operation. We can't actually free the stack + // just yet because we might run into the following situation: + // 1) GC starts, scans a SudoG but does not yet mark the SudoG.elem pointer + // 2) The stack that pointer points to is shrunk + // 3) The old stack is freed + // 4) The containing span is marked free + // 5) GC attempts to mark the SudoG.elem pointer. The marking fails because + // the pointer looks like a pointer into a free span. + // By not freeing, we prevent step #4 until GC is done. + lock(&stackpoolmu) + *(*stack)(unsafe.Pointer(old.lo)) = stackfreequeue + stackfreequeue = old + unlock(&stackpoolmu) + } +} + +// round x up to a power of 2. +func round2(x int32) int32 { + s := uint(0) + for 1<atomicstatus will be Grunning or Gscanrunning upon entry. +// If the GC is trying to stop this g then it will set preemptscan to true. +func newstack() { + thisg := getg() + // TODO: double check all gp. shouldn't be getg(). + if thisg.m.morebuf.g.stackguard0 == stackFork { + gothrow("stack growth after fork") + } + if thisg.m.morebuf.g != thisg.m.curg { + print("runtime: newstack called from g=", thisg.m.morebuf.g, "\n"+"\tm=", thisg.m, " m->curg=", thisg.m.curg, " m->g0=", thisg.m.g0, " m->gsignal=", thisg.m.gsignal, "\n") + morebuf := thisg.m.morebuf + traceback(morebuf.pc, morebuf.sp, morebuf.lr, morebuf.g) + gothrow("runtime: wrong goroutine in newstack") + } + if thisg.m.curg.throwsplit { + gp := thisg.m.curg + // Update syscallsp, syscallpc in case traceback uses them. + morebuf := thisg.m.morebuf + gp.syscallsp = morebuf.sp + gp.syscallpc = morebuf.pc + print("runtime: newstack sp=", hex(gp.sched.sp), " stack=[", hex(gp.stack.lo), ", ", hex(gp.stack.hi), "]\n", + "\tmorebuf={pc:", hex(morebuf.pc), " sp:", hex(morebuf.sp), " lr:", hex(morebuf.lr), "}\n", + "\tsched={pc:", hex(gp.sched.pc), " sp:", hex(gp.sched.sp), " lr:", hex(gp.sched.lr), " ctxt:", gp.sched.ctxt, "}\n") + gothrow("runtime: stack split at bad time") + } + + // The goroutine must be executing in order to call newstack, + // so it must be Grunning or Gscanrunning. + + gp := thisg.m.curg + morebuf := thisg.m.morebuf + thisg.m.morebuf.pc = 0 + thisg.m.morebuf.lr = 0 + thisg.m.morebuf.sp = 0 + thisg.m.morebuf.g = nil + + casgstatus(gp, _Grunning, _Gwaiting) + gp.waitreason = "stack growth" + + rewindmorestack(&gp.sched) + + if gp.stack.lo == 0 { + gothrow("missing stack in newstack") + } + sp := gp.sched.sp + if thechar == '6' || thechar == '8' { + // The call to morestack cost a word. + sp -= ptrSize + } + if stackDebug >= 1 || sp < gp.stack.lo { + print("runtime: newstack sp=", hex(sp), " stack=[", hex(gp.stack.lo), ", ", hex(gp.stack.hi), "]\n", + "\tmorebuf={pc:", hex(morebuf.pc), " sp:", hex(morebuf.sp), " lr:", hex(morebuf.lr), "}\n", + "\tsched={pc:", hex(gp.sched.pc), " sp:", hex(gp.sched.sp), " lr:", hex(gp.sched.lr), " ctxt:", gp.sched.ctxt, "}\n") + } + if sp < gp.stack.lo { + print("runtime: gp=", gp, ", gp->status=", hex(readgstatus(gp)), "\n ") + print("runtime: split stack overflow: ", hex(sp), " < ", hex(gp.stack.lo), "\n") + gothrow("runtime: split stack overflow") + } + + if gp.stackguard0 == stackPreempt { + if gp == thisg.m.g0 { + gothrow("runtime: preempt g0") + } + if thisg.m.p == nil && thisg.m.locks == 0 { + gothrow("runtime: g is running but p is not") + } + if gp.preemptscan { + gcphasework(gp) + casgstatus(gp, _Gwaiting, _Grunning) + gp.stackguard0 = gp.stack.lo + _StackGuard + gp.preempt = false + gp.preemptscan = false // Tells the GC premption was successful. + gogo(&gp.sched) // never return + } + + // Be conservative about where we preempt. + // We are interested in preempting user Go code, not runtime code. + if thisg.m.locks != 0 || thisg.m.mallocing != 0 || thisg.m.gcing != 0 || thisg.m.p.status != _Prunning { + // Let the goroutine keep running for now. + // gp->preempt is set, so it will be preempted next time. + gp.stackguard0 = gp.stack.lo + _StackGuard + casgstatus(gp, _Gwaiting, _Grunning) + gogo(&gp.sched) // never return + } + + // Act like goroutine called runtime.Gosched. + casgstatus(gp, _Gwaiting, _Grunning) + gosched_m(gp) // never return + } + + // Allocate a bigger segment and move the stack. + oldsize := int(gp.stack.hi - gp.stack.lo) + newsize := oldsize * 2 + if uintptr(newsize) > maxstacksize { + print("runtime: goroutine stack exceeds ", maxstacksize, "-byte limit\n") + gothrow("stack overflow") + } + + // Note that the concurrent GC might be scanning the stack as we try to replace it. + // copystack takes care of the appropriate coordination with the stack scanner. + copystack(gp, uintptr(newsize)) + if stackDebug >= 1 { + print("stack grow done\n") + } + casgstatus(gp, _Gwaiting, _Grunning) + gogo(&gp.sched) +} + +//go:nosplit +func nilfunc() { + *(*uint8)(nil) = 0 +} + +// adjust Gobuf as if it executed a call to fn +// and then did an immediate gosave. +func gostartcallfn(gobuf *gobuf, fv *funcval) { + var fn unsafe.Pointer + if fv != nil { + fn = (unsafe.Pointer)(fv.fn) + } else { + fn = unsafe.Pointer(funcPC(nilfunc)) + } + gostartcall(gobuf, fn, (unsafe.Pointer)(fv)) +} + +// Maybe shrink the stack being used by gp. +// Called at garbage collection time. +func shrinkstack(gp *g) { + if readgstatus(gp) == _Gdead { + if gp.stack.lo != 0 { + // Free whole stack - it will get reallocated + // if G is used again. + stackfree(gp.stack) + gp.stack.lo = 0 + gp.stack.hi = 0 + } + return + } + if gp.stack.lo == 0 { + gothrow("missing stack in shrinkstack") + } + + oldsize := gp.stack.hi - gp.stack.lo + newsize := oldsize / 2 + if newsize < _FixedStack { + return // don't shrink below the minimum-sized stack + } + used := gp.stack.hi - gp.sched.sp + if used >= oldsize/4 { + return // still using at least 1/4 of the segment. + } + + // We can't copy the stack if we're in a syscall. + // The syscall might have pointers into the stack. + if gp.syscallsp != 0 { + return + } + + /* TODO + if goos_windows && gp.m != nil && gp.m.libcallsp != 0 { + return + } + */ + + if stackDebug > 0 { + print("shrinking stack ", oldsize, "->", newsize, "\n") + } + copystack(gp, newsize) +} + +// Do any delayed stack freeing that was queued up during GC. +func shrinkfinish() { + lock(&stackpoolmu) + s := stackfreequeue + stackfreequeue = stack{} + unlock(&stackpoolmu) + for s.lo != 0 { + t := *(*stack)(unsafe.Pointer(s.lo)) + stackfree(s) + s = t + } +} + +//go:nosplit +func morestackc() { + systemstack(func() { + gothrow("attempt to execute C code on Go stack") + }) +}