runtime: check that masks and shifts are correct aligned

We need a runtime check because the original issue is encountered
when running cross compiled windows program from linux. It's better
to give a meaningful crash message earlier than to segfault later.

The added test should not impose any measurable overhead to Go
programs.

For #12415.

Change-Id: Ib4a24ef560c09c0585b351d62eefd157b6b7f04c
Reviewed-on: https://go-review.googlesource.com/14207Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

src/runtime/asm_386.s

@@ -1188,6 +1188,15 @@ DATA shifts<>+0xfc(SB)/4, $0xff0f0e0d
 
 GLOBL shifts<>(SB),RODATA,$256
 
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	// check that masks<>(SB) and shifts<>(SB) are aligned to 16-byte
+	MOVL	$masks<>(SB), AX
+	MOVL	$shifts<>(SB), BX
+	ORL	BX, AX
+	TESTL	$15, AX
+	SETEQ	ret+0(FP)
+	RET
+
 TEXT runtime·memeq(SB),NOSPLIT,$0-13
 	MOVL	a+0(FP), SI
 	MOVL	b+4(FP), DI

src/runtime/asm_amd64.s

@@ -1222,6 +1222,15 @@ DATA masks<>+0xf0(SB)/8, $0xffffffffffffffff
 DATA masks<>+0xf8(SB)/8, $0x00ffffffffffffff
 GLOBL masks<>(SB),RODATA,$256
 
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	// check that masks<>(SB) and shifts<>(SB) are aligned to 16-byte
+	MOVQ	$masks<>(SB), AX
+	MOVQ	$shifts<>(SB), BX
+	ORQ	BX, AX
+	TESTQ	$15, AX
+	SETEQ	ret+0(FP)
+	RET
+
 // these are arguments to pshufb.  They move data down from
 // the high bytes of the register to the low bytes of the register.
 // index is how many bytes to move.

src/runtime/asm_amd64p32.s

@@ -1012,3 +1012,7 @@ TEXT runtime·prefetchnta(SB),NOSPLIT,$0-4
 	MOVL	addr+0(FP), AX
 	PREFETCHNTA	(AX)
 	RET
+
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	MOVB	$1, ret+0(FP)
+	RET

src/runtime/asm_arm.s

@@ -1030,3 +1030,8 @@ TEXT runtime·addmoduledata(SB),NOSPLIT,$0-4
 	MOVW	saver9-4(SP), R9
 	RET
 #endif
+
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	MOVW	$1, R3
+	MOVB	R3, ret+0(FP)
+	RET

src/runtime/asm_arm64.s

@@ -996,4 +996,8 @@ TEXT runtime·addmoduledata(SB),NOSPLIT,$0-0
 	MOVD	R0, runtime·lastmoduledatap(SB)
 	MOVD	8(RSP), R27
 	ADD	$0x10, RSP
+
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	MOVW	$1, R3
+	MOVB	R3, ret+0(FP)
 	RET

src/runtime/asm_mips64x.s

@@ -818,3 +818,8 @@ TEXT runtime·prefetcht2(SB),NOSPLIT,$0-8
 
 TEXT runtime·prefetchnta(SB),NOSPLIT,$0-8
 	RET
+
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	MOVW	$1, R1
+	MOVB	R1, ret+0(FP)
+	RET

src/runtime/asm_ppc64x.s

@@ -1079,4 +1079,8 @@ TEXT runtime·addmoduledata(SB),NOSPLIT|NOFRAME,$0-0
 	MOVD	R3, runtime·lastmoduledatap(SB)
 	MOVD	0(R1), R31
 	ADD	$8, R1
+
+TEXT ·checkASM(SB),NOSPLIT,$0-1
+	MOVW	$1, R3
+	MOVB	R3, ret+0(FP)
 	RET

src/runtime/runtime1.go

@@ -296,6 +296,10 @@ func check() {
 	if _FixedStack != round2(_FixedStack) {
 		throw("FixedStack is not power-of-2")
 	}
+
+	if !checkASM() {
+		throw("assembly checks failed")
+	}
 }
 
 type dbgVar struct {

src/runtime/stubs.go

@@ -270,3 +270,6 @@ func unixnanotime() int64 {
 func round(n, a uintptr) uintptr {
 	return (n + a - 1) &^ (a - 1)
 }
+
+// checkASM returns whether assembly runtime checks have passed.
+func checkASM() bool