change to emit create event in exitsyscall.

src/runtime/cgocall.go

@@ -227,6 +227,9 @@ func cgocallbackg(fn, frame unsafe.Pointer, ctxt uintptr) {
 	savedpc := gp.syscallpc
 	exitsyscall() // coming out of cgo call
 	gp.m.incgo = false
+	if gp.m.isextra {
+		gp.m.cgolevel++
+	}
 
 	osPreemptExtExit(gp.m)
 
@@ -237,6 +240,12 @@ func cgocallbackg(fn, frame unsafe.Pointer, ctxt uintptr) {
 	// This is enforced by checking incgo in the schedule function.
 
 	gp.m.incgo = true
+	if gp.m.isextra {
+		gp.m.cgolevel--
+		if gp.m.cgolevel < 0 {
+			throw("unexpected negative cgolevel")
+		}
+	}
 
 	if gp.m != checkm {
 		throw("m changed unexpectedly in cgocallbackg")

src/runtime/proc.go

@@ -1917,11 +1917,6 @@ func oneNewExtraM() {
 	if raceenabled {
 		gp.racectx = racegostart(abi.FuncPCABIInternal(newextram) + sys.PCQuantum)
 	}
-	if trace.enabled {
-		traceGoCreate(gp, 0) // no start pc
-		gp.traceseq++
-		traceEvent(traceEvGoInSyscall, -1, uint64(gp.goid))
-	}
 	// put on allg for garbage collector
 	allgadd(gp)
 
@@ -2515,9 +2510,13 @@ func execute(gp *g, inheritTime bool) {
 	}
 
 	if trace.enabled {
-		// GoSysExit has to happen when we have a P, but before GoStart.
-		// So we emit it here.
-		if gp.syscallsp != 0 && gp.sysblocktraced {
+		if gp.m.isextra && gp.m.cgolevel == 0 {
+			// GoCreate happen in needm, but there is no P.
+			// So we emit it here.
+			traceGoCreate(gp, 0) // no start pc for locked g in extra M
+		} else if gp.syscallsp != 0 && gp.sysblocktraced {
+			// GoSysExit has to happen when we have a P, but before GoStart.
+			// So we emit it here.
 			traceGoSysExit(gp.sysexitticks)
 		}
 		traceGoStart()
@@ -3615,7 +3614,14 @@ func reentersyscall(pc, sp uintptr) {
 	}
 
 	if trace.enabled {
-		systemstack(traceGoSysCall)
+		if _g_.m.isextra && _g_.m.cgolevel == 0 {
+			systemstack(func() {
+				traceGoEnd()
+				traceProcStop(_g_.m.p.ptr())
+			})
+		} else {
+			systemstack(traceGoSysCall)
+		}
 		// systemstack itself clobbers g.sched.{pc,sp} and we might
 		// need them later when the G is genuinely blocked in a
 		// syscall
@@ -3639,10 +3645,19 @@ func reentersyscall(pc, sp uintptr) {
 	pp.m = 0
 	_g_.m.oldp.set(pp)
 	_g_.m.p = 0
-	atomic.Store(&pp.status, _Psyscall)
-	if sched.gcwaiting != 0 {
-		systemstack(entersyscall_gcwait)
+
+	if _g_.m.isextra && _g_.m.cgolevel == 0 {
+		atomic.Store(&pp.status, _Pidle)
+		systemstack(func() {
+			handoffp(pp)
+		})
 		save(pc, sp)
+	} else {
+		atomic.Store(&pp.status, _Psyscall)
+		if sched.gcwaiting != 0 {
+			systemstack(entersyscall_gcwait)
+			save(pc, sp)
+		}
 	}
 
 	_g_.m.locks--
@@ -3851,7 +3866,13 @@ func exitsyscallfast(oldp *p) bool {
 						osyield()
 					}
 				}
-				traceGoSysExit(0)
+				if _g_.m.isextra && _g_.m.cgolevel == 0 {
+					// GoCreate happen in needm, but there is no P.
+					// So we emit it here.
+					traceGoCreate(_g_, 0) // no start pc for locked g in extra M
+				} else {
+					traceGoSysExit(0)
+				}
 			}
 		})
 		if ok {
@@ -3874,10 +3895,16 @@ func exitsyscallfast_reacquired() {
 			// traceGoSysBlock for this syscall was already emitted,
 			// but here we effectively retake the p from the new syscall running on the same p.
 			systemstack(func() {
-				// Denote blocking of the new syscall.
-				traceGoSysBlock(_g_.m.p.ptr())
-				// Denote completion of the current syscall.
-				traceGoSysExit(0)
+				if _g_.m.isextra && _g_.m.cgolevel == 0 {
+					// GoCreate happen in needm, but there is no P.
+					// So we emit it here.
+					traceGoCreate(_g_, 0) // no start pc for locked g in extra M
+				} else {
+					// Denote blocking of the new syscall.
+					traceGoSysBlock(_g_.m.p.ptr())
+					// Denote completion of the current syscall.
+					traceGoSysExit(0)
+				}
 			})
 		}
 		_g_.m.p.ptr().syscalltick++

src/runtime/runtime2.go

@@ -540,6 +540,7 @@ type m struct {
 	printlock     int8
 	incgo         bool   // m is executing a cgo call
 	isextra       bool   // m is an extra m
+	cgolevel      int32  // level of cgo call
 	freeWait      uint32 // if == 0, safe to free g0 and delete m (atomic)
 	fastrand      uint64
 	needextram    bool

src/runtime/trace.go

@@ -225,17 +225,12 @@ func StartTrace() error {
 	// World is stopped, no need to lock.
 	forEachGRace(func(gp *g) {
 		status := readgstatus(gp)
-		if status != _Gdead || (gp.m != nil && gp.m.isextra) {
+		if status != _Gdead {
 			gp.traceseq = 0
 			gp.tracelastp = getg().m.p
 			// +PCQuantum because traceFrameForPC expects return PCs and subtracts PCQuantum.
 			id := trace.stackTab.put([]uintptr{startPCforTrace(gp.startpc) + sys.PCQuantum})
 			traceEvent(traceEvGoCreate, -1, uint64(gp.goid), uint64(id), stackID)
-
-			if status == _Gdead {
-				gp.traceseq++
-				traceEvent(traceEvGoInSyscall, -1, uint64(gp.goid))
-			}
 		}
 		if status == _Gwaiting {
 			// traceEvGoWaiting is implied to have seq=1.