func traceStack(skip int, gp *g, gen uintptr) uint64 {
	var pcBuf [traceStackSize]uintptr

	// Figure out gp and mp for the backtrace.
	var mp *m
	if gp == nil {
		mp = getg().m
		gp = mp.curg
	}

	// Double-check that we own the stack we're about to trace.
	if debug.traceCheckStackOwnership != 0 && gp != nil {
		status := readgstatus(gp)
		// If the scan bit is set, assume we're the ones that acquired it.

		var tagPtr *unsafe.Pointer
		if gp != nil && gp.m != nil && gp.m.curg != nil {
			tagPtr = &gp.m.curg.labels
		}
		cpuprof.add(tagPtr, stk[:n])

		gprof := gp
		var mp *m
		var pp *p
		if gp != nil && gp.m != nil {
			if gp.m.curg != nil {
				gprof = gp.m.curg
			}
			mp = gp.m
			pp = gp.m.p.ptr()
		}
		traceCPUSample(gprof, mp, pp, stk[:n])

// taken from the frame structure, records the results in the frame,
// and returns to runtime.asmcgocall.
//
// After it regains control, runtime.asmcgocall switches back to the
// original g (m->curg)'s stack and returns to runtime.cgocall.
//
// After it regains control, runtime.cgocall calls exitsyscall, which blocks
// until this m can run Go code without violating the $GOMAXPROCS limit,
// and then unlocks g from m.
//

// See racecallback for command codes.
TEXT	runtime·racecallbackthunk(SB), NOSPLIT|NOFRAME, $0-0
	// Handle command raceGetProcCmd (0) here.
	// First, code below assumes that we are on curg, while raceGetProcCmd
	// can be executed on g0. Second, it is called frequently, so will
	// benefit from this fast path.
	CMPQ	RARG0, $0
	JNE	rest
	get_tls(RARG0)
	MOVQ	g(RARG0), RARG0
	MOVQ	g_m(RARG0), RARG0

// function call with return PC pc. If not, it returns a string
// explaining why.
//
//go:nosplit
func debugCallCheck(pc uintptr) string {
	// No user calls from the system stack.
	if getg() != getg().m.curg {
		return debugCallSystemStack
	}
	if sp := getcallersp(); !(getg().stack.lo < sp && sp <= getg().stack.hi) {
		// Fast syscalls (nanotime) and racecall switch to the
		// g0 stack without switching g. We can't safely make

	if tracefpunwindoff() || gp.m.hasCgoOnStack() {
		mp.profStack[0] = logicalStackSentinel
		if gp.m.curg == nil || gp.m.curg == gp {
			nstk = callers(skip, mp.profStack[1:])
		} else {
			nstk = gcallers(gp.m.curg, skip, mp.profStack[1:])
		}
	} else {
		mp.profStack[0] = uintptr(skip)
		if gp.m.curg == nil || gp.m.curg == gp {
			if skip > 0 {
				// We skip one fewer frame than the provided value for frame

}

// Create a new deferred function fn, which has no arguments and results.
// The compiler turns a defer statement into a call to this.
func deferproc(fn func()) {
	gp := getg()
	if gp.m.curg != gp {
		// go code on the system stack can't defer
		throw("defer on system stack")
	}

	d := newdefer()
	d.link = gp._defer
	gp._defer = d
	d.fn = fn
	d.pc = getcallerpc()

// See racecallback for command codes.
TEXT	runtime·racecallbackthunk(SB), NOSPLIT|NOFRAME, $0
	// Handle command raceGetProcCmd (0) here.
	// First, code below assumes that we are on curg, while raceGetProcCmd
	// can be executed on g0. Second, it is called frequently, so will
	// benefit from this fast path.
	CBNZ	R0, rest
	MOVD	g, R13
#ifdef TLS_darwin

	gp := getg()
	casGToWaiting(gp.m.curg, _Grunning, waitReasonDumpingHeap)

	// Set dump file.
	dumpfd = fd

	// Call dump routine.
	mdump(m)

	// Reset dump file.
	dumpfd = 0
	if tmpbuf != nil {
		sysFree(unsafe.Pointer(&tmpbuf[0]), uintptr(len(tmpbuf)), &memstats.other_sys)
		tmpbuf = nil
	}

	casgstatus(gp.m.curg, _Gwaiting, _Grunning)
}

	w := tl.writer()
	if pp := tl.mp.p.ptr(); pp != nil && !pp.trace.statusWasTraced(tl.gen) && pp.trace.acquireStatus(tl.gen) {
		w = w.writeProcStatus(uint64(pp.id), procStatus, pp.trace.inSweep)
	}
	if gp := tl.mp.curg; gp != nil && !gp.trace.statusWasTraced(tl.gen) && gp.trace.acquireStatus(tl.gen) {
		w = w.writeGoStatus(uint64(gp.goid), int64(tl.mp.procid), goStatus, gp.inMarkAssist, 0 /* no stack */)
	}
	return traceEventWriter{w}
}