internal/runtime/atomic: add Xchg8 for s390x and wasm

This makes the single-byte atomic.Xchg8 operation available on all
GOARCHes, including those without direct / single-instruction support.

Fixes #69735

Change-Id: Icb6aff8f907257db81ea440dc4d29f96b3cff6c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/657936
Reviewed-by: Mauri de Souza Meneguzzo <mauri870@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: David Chase <drchase@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Rhys Hiltner <rhys.hiltner@gmail.com>
TryBot-Result: Gopher Robot <gobot@golang.org>

src/internal/runtime/atomic/atomic_arm.go

@@ -77,24 +77,6 @@ func Xchg(addr *uint32, v uint32) uint32 {
 //go:noescape
 func Xchg8(addr *uint8, v uint8) uint8
 
-//go:nosplit
-func goXchg8(addr *uint8, v uint8) uint8 {
-	// Align down to 4 bytes and use 32-bit CAS.
-	addr32 := (*uint32)(unsafe.Pointer(uintptr(unsafe.Pointer(addr)) &^ 3))
-	shift := (uintptr(unsafe.Pointer(addr)) & 3) * 8 // little endian
-	word := uint32(v) << shift
-	mask := uint32(0xFF) << shift
-
-	for {
-		old := *addr32 // Read the old 32-bit value
-		// Clear the old 8 bits then insert the new value
-		if Cas(addr32, old, (old&^mask)|word) {
-			// Return the old 8-bit value
-			return uint8((old & mask) >> shift)
-		}
-	}
-}
-
 //go:nosplit
 func Xchguintptr(addr *uintptr, v uintptr) uintptr {
 	return uintptr(Xchg((*uint32)(unsafe.Pointer(addr)), uint32(v)))

src/internal/runtime/atomic/atomic_s390x.go

@@ -128,6 +128,11 @@ func Xadduintptr(ptr *uintptr, delta uintptr) uintptr
 //go:noescape
 func Xchg(ptr *uint32, new uint32) uint32
 
+//go:nosplit
+func Xchg8(addr *uint8, v uint8) uint8 {
+	return goXchg8(addr, v)
+}
+
 //go:noescape
 func Xchg64(ptr *uint64, new uint64) uint64
 

src/internal/runtime/atomic/atomic_wasm.go

@@ -114,6 +114,11 @@ func Xchg(ptr *uint32, new uint32) uint32 {
 	return old
 }
 
+//go:nosplit
+func Xchg8(addr *uint8, v uint8) uint8 {
+	return goXchg8(addr, v)
+}
+
 //go:nosplit
 //go:noinline
 func Xchg64(ptr *uint64, new uint64) uint64 {

src/internal/runtime/atomic/xchg8.go

@@ -0,0 +1,32 @@
+// Copyright 2025 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 atomic
+
+import (
+	"internal/goarch"
+	"unsafe"
+)
+
+//go:nosplit
+func goXchg8(addr *uint8, v uint8) uint8 {
+	// Align down to 4 bytes and use 32-bit CAS.
+	addr32 := (*uint32)(unsafe.Pointer(uintptr(unsafe.Pointer(addr)) &^ 3))
+	shift := (uintptr(unsafe.Pointer(addr)) & 3)
+	if goarch.BigEndian {
+		shift = shift ^ 3
+	}
+	shift = shift * 8
+	word := uint32(v) << shift
+	mask := uint32(0xFF) << shift
+
+	for {
+		old := *addr32 // Read the old 32-bit value
+		// Clear the old 8 bits then insert the new value
+		if Cas(addr32, old, (old&^mask)|word) {
+			// Return the old 8-bit value
+			return uint8((old & mask) >> shift)
+		}
+	}
+}

src/internal/runtime/atomic/xchg8_test.go

@@ -2,8 +2,6 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build 386 || amd64 || arm || arm64 || loong64 || mips || mipsle || mips64 || mips64le || ppc64 || ppc64le || riscv64
-
 package atomic_test
 
 import (