diff --git a/src/runtime/crash_cgo_test.go b/src/runtime/crash_cgo_test.go
index 6da8341e841..c1dd757797d 100644
--- a/src/runtime/crash_cgo_test.go
+++ b/src/runtime/crash_cgo_test.go
@@ -263,7 +263,7 @@ func TestCgoTracebackContext(t *testing.T) {
 	}
 }
 
-func testCgoPprof(t *testing.T, buildArg, runArg string) {
+func testCgoPprof(t *testing.T, buildArg, runArg, top, bottom string) {
 	t.Parallel()
 	if runtime.GOOS != "linux" || (runtime.GOARCH != "amd64" && runtime.GOARCH != "ppc64le") {
 		t.Skipf("not yet supported on %s/%s", runtime.GOOS, runtime.GOARCH)
@@ -287,7 +287,7 @@ func testCgoPprof(t *testing.T, buildArg, runArg string) {
 	defer os.Remove(fn)
 
 	for try := 0; try < 2; try++ {
-		cmd := testenv.CleanCmdEnv(exec.Command(testenv.GoToolPath(t), "tool", "pprof", "-top", "-nodecount=1"))
+		cmd := testenv.CleanCmdEnv(exec.Command(testenv.GoToolPath(t), "tool", "pprof", "-traces"))
 		// Check that pprof works both with and without explicit executable on command line.
 		if try == 0 {
 			cmd.Args = append(cmd.Args, exe, fn)
@@ -307,30 +307,38 @@ func testCgoPprof(t *testing.T, buildArg, runArg string) {
 			cmd.Env = append(cmd.Env, "PPROF_TMPDIR="+os.TempDir())
 		}
 
-		top, err := cmd.CombinedOutput()
-		t.Logf("%s:\n%s", cmd.Args, top)
+		out, err := cmd.CombinedOutput()
+		t.Logf("%s:\n%s", cmd.Args, out)
 		if err != nil {
 			t.Error(err)
-		} else if !bytes.Contains(top, []byte("cpuHog")) {
-			t.Error("missing cpuHog in pprof output")
+			continue
+		}
+
+		trace := findTrace(string(out), top)
+		if len(trace) == 0 {
+			t.Errorf("%s traceback missing.", top)
+			continue
+		}
+		if trace[len(trace)-1] != bottom {
+			t.Errorf("invalid traceback origin: got=%v; want=[%s ... %s]", trace, top, bottom)
 		}
 	}
 }
 
 func TestCgoPprof(t *testing.T) {
-	testCgoPprof(t, "", "CgoPprof")
+	testCgoPprof(t, "", "CgoPprof", "cpuHog", "runtime.main")
 }
 
 func TestCgoPprofPIE(t *testing.T) {
-	testCgoPprof(t, "-buildmode=pie", "CgoPprof")
+	testCgoPprof(t, "-buildmode=pie", "CgoPprof", "cpuHog", "runtime.main")
 }
 
 func TestCgoPprofThread(t *testing.T) {
-	testCgoPprof(t, "", "CgoPprofThread")
+	testCgoPprof(t, "", "CgoPprofThread", "cpuHogThread", "cpuHogThread2")
 }
 
 func TestCgoPprofThreadNoTraceback(t *testing.T) {
-	testCgoPprof(t, "", "CgoPprofThreadNoTraceback")
+	testCgoPprof(t, "", "CgoPprofThreadNoTraceback", "cpuHogThread", "runtime._ExternalCode")
 }
 
 func TestRaceProf(t *testing.T) {
@@ -509,3 +517,35 @@ func TestBigStackCallbackCgo(t *testing.T) {
 		t.Errorf("expected %q got %v", want, got)
 	}
 }
+
+func nextTrace(lines []string) ([]string, []string) {
+	var trace []string
+	for n, line := range lines {
+		if strings.HasPrefix(line, "---") {
+			return trace, lines[n+1:]
+		}
+		fields := strings.Fields(strings.TrimSpace(line))
+		if len(fields) == 0 {
+			continue
+		}
+		// Last field contains the function name.
+		trace = append(trace, fields[len(fields)-1])
+	}
+	return nil, nil
+}
+
+func findTrace(text, top string) []string {
+	lines := strings.Split(text, "\n")
+	_, lines = nextTrace(lines) // Skip the header.
+	for len(lines) > 0 {
+		var t []string
+		t, lines = nextTrace(lines)
+		if len(t) == 0 {
+			continue
+		}
+		if t[0] == top {
+			return t
+		}
+	}
+	return nil
+}
diff --git a/src/runtime/proc.go b/src/runtime/proc.go
index 409869fd107..fc77a964b6e 100644
--- a/src/runtime/proc.go
+++ b/src/runtime/proc.go
@@ -3742,6 +3742,9 @@ func sigprof(pc, sp, lr uintptr, gp *g, mp *m) {
 
 		// Collect Go stack that leads to the cgo call.
 		n = gentraceback(mp.curg.syscallpc, mp.curg.syscallsp, 0, mp.curg, 0, &stk[cgoOff], len(stk)-cgoOff, nil, nil, 0)
+		if n > 0 {
+			n += cgoOff
+		}
 	} else if traceback {
 		n = gentraceback(pc, sp, lr, gp, 0, &stk[0], len(stk), nil, nil, _TraceTrap|_TraceJumpStack)
 	}
diff --git a/src/runtime/testdata/testprogcgo/pprof.go b/src/runtime/testdata/testprogcgo/pprof.go
index 4460b9304e5..00f2c42e93c 100644
--- a/src/runtime/testdata/testprogcgo/pprof.go
+++ b/src/runtime/testdata/testprogcgo/pprof.go
@@ -26,6 +26,9 @@ void cpuHog() {
 	salt2 = foo;
 }
 
+void cpuHog2() {
+}
+
 static int cpuHogCount;
 
 struct cgoTracebackArg {
@@ -37,10 +40,13 @@ struct cgoTracebackArg {
 
 // pprofCgoTraceback is passed to runtime.SetCgoTraceback.
 // For testing purposes it pretends that all CPU hits in C code are in cpuHog.
+// Issue #29034: At least 2 frames are required to verify all frames are captured
+// since runtime/pprof ignores the runtime.goexit base frame if it exists.
 void pprofCgoTraceback(void* parg) {
 	struct cgoTracebackArg* arg = (struct cgoTracebackArg*)(parg);
 	arg->buf[0] = (uintptr_t)(cpuHog) + 0x10;
-	arg->buf[1] = 0;
+	arg->buf[1] = (uintptr_t)(cpuHog2) + 0x4;
+	arg->buf[2] = 0;
 	++cpuHogCount;
 }
 
diff --git a/src/runtime/testdata/testprogcgo/threadpprof.go b/src/runtime/testdata/testprogcgo/threadpprof.go
index 3da82961b9b..37a2a1ab659 100644
--- a/src/runtime/testdata/testprogcgo/threadpprof.go
+++ b/src/runtime/testdata/testprogcgo/threadpprof.go
@@ -30,6 +30,9 @@ void cpuHogThread() {
 	threadSalt2 = foo;
 }
 
+void cpuHogThread2() {
+}
+
 static int cpuHogThreadCount;
 
 struct cgoTracebackArg {
@@ -44,7 +47,8 @@ struct cgoTracebackArg {
 void pprofCgoThreadTraceback(void* parg) {
 	struct cgoTracebackArg* arg = (struct cgoTracebackArg*)(parg);
 	arg->buf[0] = (uintptr_t)(cpuHogThread) + 0x10;
-	arg->buf[1] = 0;
+	arg->buf[1] = (uintptr_t)(cpuHogThread2) + 0x4;
+	arg->buf[2] = 0;
 	__sync_add_and_fetch(&cpuHogThreadCount, 1);
 }