[release-branch.go1.16] runtime: set vdsoSP to caller's SP consistently

m.vdsoSP should be set to the SP of the caller of nanotime1,
instead of the SP of nanotime1 itself, which matches m.vdsoPC.
Otherwise the unmatched vdsoPC and vdsoSP would make the stack
trace look like recursive.

We already do it correctly on AMD64, 386, and RISCV64. This CL
fixes the rest.

Also incorporate CL 352509, skipping a flaky test.

Updates #47324, #50772.
Fixes #50780.

Change-Id: I98b6fcfbe9fc6bdd28b8fe2a1299b7c505371dd4
Reviewed-on: https://go-review.googlesource.com/c/go/+/337590
Trust: Cherry Mui <cherryyz@google.com>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
(cherry picked from commit 217507eb03)
Reviewed-on: https://go-review.googlesource.com/c/go/+/380716
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

src/runtime/pprof/pprof_test.go

@@ -1458,3 +1458,38 @@ func TestTryAdd(t *testing.T) {
 		})
 	}
 }
+
+func TestTimeVDSO(t *testing.T) {
+	// Test that time functions have the right stack trace. In particular,
+	// it shouldn't be recursive.
+
+	if runtime.GOOS == "android" {
+		// Flaky on Android, issue 48655. VDSO may not be enabled.
+		testenv.SkipFlaky(t, 48655)
+	}
+
+	p := testCPUProfile(t, stackContains, []string{"time.now"}, avoidFunctions(), func(dur time.Duration) {
+		t0 := time.Now()
+		for {
+			t := time.Now()
+			if t.Sub(t0) >= dur {
+				return
+			}
+		}
+	})
+
+	// Check for recursive time.now sample.
+	for _, sample := range p.Sample {
+		var seenNow bool
+		for _, loc := range sample.Location {
+			for _, line := range loc.Line {
+				if line.Function.Name == "time.now" {
+					if seenNow {
+						t.Fatalf("unexpected recursive time.now")
+					}
+					seenNow = true
+				}
+			}
+		}
+	}
+}

src/runtime/sys_linux_arm.s

@@ -259,8 +259,9 @@ TEXT runtime·walltime1(SB),NOSPLIT,$8-12
 	MOVW	R1, 4(R13)
 	MOVW	R2, 8(R13)
 
+	MOVW	$ret-4(FP), R2 // caller's SP
 	MOVW	LR, m_vdsoPC(R5)
-	MOVW	R13, m_vdsoSP(R5)
+	MOVW	R2, m_vdsoSP(R5)
 
 	MOVW	m_curg(R5), R0
 
@@ -351,8 +352,9 @@ TEXT runtime·nanotime1(SB),NOSPLIT,$8-8
 	MOVW	R1, 4(R13)
 	MOVW	R2, 8(R13)
 
+	MOVW	$ret-4(FP), R2 // caller's SP
 	MOVW	LR, m_vdsoPC(R5)
-	MOVW	R13, m_vdsoSP(R5)
+	MOVW	R2, m_vdsoSP(R5)
 
 	MOVW	m_curg(R5), R0
 

src/runtime/sys_linux_arm64.s

@@ -221,8 +221,9 @@ TEXT runtime·walltime1(SB),NOSPLIT,$24-12
 	MOVD	R2, 8(RSP)
 	MOVD	R3, 16(RSP)
 
+	MOVD	$ret-8(FP), R2 // caller's SP
 	MOVD	LR, m_vdsoPC(R21)
-	MOVD	R20, m_vdsoSP(R21)
+	MOVD	R2, m_vdsoSP(R21)
 
 	MOVD	m_curg(R21), R0
 	CMP	g, R0
@@ -304,8 +305,9 @@ TEXT runtime·nanotime1(SB),NOSPLIT,$24-8
 	MOVD	R2, 8(RSP)
 	MOVD	R3, 16(RSP)
 
+	MOVD	$ret-8(FP), R2 // caller's SP
 	MOVD	LR, m_vdsoPC(R21)
-	MOVD	R20, m_vdsoSP(R21)
+	MOVD	R2, m_vdsoSP(R21)
 
 	MOVD	m_curg(R21), R0
 	CMP	g, R0

src/runtime/sys_linux_mips64x.s

@@ -228,8 +228,9 @@ TEXT runtime·walltime1(SB),NOSPLIT,$16-12
 	MOVV	R2, 8(R29)
 	MOVV	R3, 16(R29)
 
+	MOVV	$ret-8(FP), R2 // caller's SP
 	MOVV	R31, m_vdsoPC(R17)
-	MOVV	R29, m_vdsoSP(R17)
+	MOVV	R2, m_vdsoSP(R17)
 
 	MOVV	m_curg(R17), R4
 	MOVV	g, R5
@@ -297,8 +298,9 @@ TEXT runtime·nanotime1(SB),NOSPLIT,$16-8
 	MOVV	R2, 8(R29)
 	MOVV	R3, 16(R29)
 
+	MOVV	$ret-8(FP), R2 // caller's SP
 	MOVV	R31, m_vdsoPC(R17)
-	MOVV	R29, m_vdsoSP(R17)
+	MOVV	R2, m_vdsoSP(R17)
 
 	MOVV	m_curg(R17), R4
 	MOVV	g, R5

src/runtime/sys_linux_ppc64x.s

@@ -204,8 +204,9 @@ TEXT runtime·walltime1(SB),NOSPLIT,$16-12
 	MOVD	R5, 40(R1)
 
 	MOVD	LR, R14
+	MOVD	$ret-FIXED_FRAME(FP), R5 // caller's SP
 	MOVD	R14, m_vdsoPC(R21)
-	MOVD	R15, m_vdsoSP(R21)
+	MOVD	R5, m_vdsoSP(R21)
 
 	MOVD	m_curg(R21), R6
 	CMP	g, R6
@@ -296,9 +297,10 @@ TEXT runtime·nanotime1(SB),NOSPLIT,$16-8
 	MOVD	R4, 32(R1)
 	MOVD	R5, 40(R1)
 
-	MOVD	LR, R14		// R14 is unchanged by C code
+	MOVD	LR, R14				// R14 is unchanged by C code
+	MOVD	$ret-FIXED_FRAME(FP), R5	// caller's SP
 	MOVD	R14, m_vdsoPC(R21)
-	MOVD	R15, m_vdsoSP(R21)
+	MOVD	R5, m_vdsoSP(R21)
 
 	MOVD	m_curg(R21), R6
 	CMP	g, R6