diff --git a/misc/cgo/test/issue7978.go b/misc/cgo/test/issue7978.go
index 094ccc1839..e8f340f8b8 100644
--- a/misc/cgo/test/issue7978.go
+++ b/misc/cgo/test/issue7978.go
@@ -110,13 +110,13 @@ func test7978(t *testing.T) {
 	go issue7978go()
 	// test in c code, before callback
 	issue7978wait(0, 1)
-	issue7978check(t, "runtime.cgocall(", "", 1)
+	issue7978check(t, "_Cfunc_issue7978c(", "", 1)
 	// test in go code, during callback
 	issue7978wait(2, 3)
 	issue7978check(t, "test.issue7978cb(", "test.issue7978go", 3)
 	// test in c code, after callback
 	issue7978wait(4, 5)
-	issue7978check(t, "runtime.cgocall(", "runtime.cgocallback", 1)
+	issue7978check(t, "_Cfunc_issue7978c(", "_cgoexpwrap", 1)
 	// test in go code, after return from cgo
 	issue7978wait(6, 7)
 	issue7978check(t, "test.issue7978go(", "", 3)
diff --git a/src/runtime/mprof.go b/src/runtime/mprof.go
index 684ab0b055..eb7231aec2 100644
--- a/src/runtime/mprof.go
+++ b/src/runtime/mprof.go
@@ -576,12 +576,17 @@ func Stack(buf []byte, all bool) int {
 		pc := getcallerpc(unsafe.Pointer(&buf))
 		systemstack(func() {
 			g0 := getg()
+			// Force traceback=1 to override GOTRACEBACK setting,
+			// so that Stack's results are consistent.
+			// GOTRACEBACK is only about crash dumps.
+			g0.m.traceback = 1
 			g0.writebuf = buf[0:0:len(buf)]
 			goroutineheader(gp)
 			traceback(pc, sp, 0, gp)
 			if all {
 				tracebackothers(gp)
 			}
+			g0.m.traceback = 0
 			n = len(g0.writebuf)
 			g0.writebuf = nil
 		})
diff --git a/src/runtime/proc.go b/src/runtime/proc.go
index 545e134cc2..be1bb815d5 100644
--- a/src/runtime/proc.go
+++ b/src/runtime/proc.go
@@ -2167,6 +2167,9 @@ func goexit0(gp *g) {
 	_g_ := getg()
 
 	casgstatus(gp, _Grunning, _Gdead)
+	if isSystemGoroutine(gp) {
+		atomic.Xadd(&sched.ngsys, -1)
+	}
 	gp.m = nil
 	gp.lockedm = nil
 	_g_.m.lockedg = nil
@@ -2698,6 +2701,9 @@ func newproc1(fn *funcval, argp *uint8, narg int32, nret int32, callerpc uintptr
 	gostartcallfn(&newg.sched, fn)
 	newg.gopc = callerpc
 	newg.startpc = fn.fn
+	if isSystemGoroutine(newg) {
+		atomic.Xadd(&sched.ngsys, +1)
+	}
 	casgstatus(newg, _Gdead, _Grunnable)
 
 	if _p_.goidcache == _p_.goidcacheend {
@@ -2890,7 +2896,7 @@ func badunlockosthread() {
 }
 
 func gcount() int32 {
-	n := int32(allglen) - sched.ngfree
+	n := int32(allglen) - sched.ngfree - int32(atomic.Load(&sched.ngsys))
 	for i := 0; ; i++ {
 		_p_ := allp[i]
 		if _p_ == nil {
diff --git a/src/runtime/proc_test.go b/src/runtime/proc_test.go
index 30798f723d..f3e90bcbd7 100644
--- a/src/runtime/proc_test.go
+++ b/src/runtime/proc_test.go
@@ -9,6 +9,7 @@ import (
 	"net"
 	"runtime"
 	"runtime/debug"
+	"strings"
 	"sync"
 	"sync/atomic"
 	"syscall"
@@ -336,6 +337,23 @@ func TestGCFairness(t *testing.T) {
 	}
 }
 
+func TestNumGoroutine(t *testing.T) {
+	output := runTestProg(t, "testprog", "NumGoroutine")
+	want := "1\n"
+	if output != want {
+		t.Fatalf("want %q, got %q", want, output)
+	}
+
+	buf := make([]byte, 1<<20)
+	buf = buf[:runtime.Stack(buf, true)]
+
+	n := runtime.NumGoroutine()
+
+	if nstk := strings.Count(string(buf), "goroutine "); n != nstk {
+		t.Fatalf("NumGoroutine=%d, but found %d goroutines in stack dump", n, nstk)
+	}
+}
+
 func TestPingPongHog(t *testing.T) {
 	if testing.Short() {
 		t.Skip("skipping in -short mode")
diff --git a/src/runtime/runtime2.go b/src/runtime/runtime2.go
index d9a449b68b..54c4686f79 100644
--- a/src/runtime/runtime2.go
+++ b/src/runtime/runtime2.go
@@ -408,9 +408,11 @@ const (
 )
 
 type schedt struct {
-	lock mutex
+	// accessed atomically. keep at top to ensure alignment on 32-bit systems.
+	goidgen  uint64
+	lastpoll uint64
 
-	goidgen uint64
+	lock mutex
 
 	midle        muintptr // idle m's waiting for work
 	nmidle       int32    // number of idle m's waiting for work
@@ -418,6 +420,8 @@ type schedt struct {
 	mcount       int32    // number of m's that have been created
 	maxmcount    int32    // maximum number of m's allowed (or die)
 
+	ngsys uint32 // number of system goroutines; updated atomically
+
 	pidle      puintptr // idle p's
 	npidle     uint32
 	nmspinning uint32 // See "Worker thread parking/unparking" comment in proc.go.
@@ -445,7 +449,6 @@ type schedt struct {
 	stopnote   note
 	sysmonwait uint32
 	sysmonnote note
-	lastpoll   uint64
 
 	// safepointFn should be called on each P at the next GC
 	// safepoint if p.runSafePointFn is set.
diff --git a/src/runtime/testdata/testprog/misc.go b/src/runtime/testdata/testprog/misc.go
new file mode 100644
index 0000000000..237680fc87
--- /dev/null
+++ b/src/runtime/testdata/testprog/misc.go
@@ -0,0 +1,15 @@
+// Copyright 2016 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 main
+
+import "runtime"
+
+func init() {
+	register("NumGoroutine", NumGoroutine)
+}
+
+func NumGoroutine() {
+	println(runtime.NumGoroutine())
+}