runtime, cmd: support race detector on darwin/arm64

https://reviews.llvm.org/D90435 is the counterpart in LLVM TSAN.

race_linux_arm64.syso is built with LLVM commit
00da38ce2d36c07f12c287dc515d37bb7bc410e9 on a macOS/ARM64 machine.
(It is not built on a builder with golang.org/x/build/cmd/racebuild
as we don't have darwin/arm64 builder for now.)

Updates #38485.

Change-Id: I391efdacd9480197e308370bfccd05777deb4aee
Reviewed-on: https://go-review.googlesource.com/c/go/+/266373
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

9 files changed, 35 insertions, 7 deletions

diff --git a/src/cmd/dist/test.go b/src/cmd/dist/test.go
index 7c454dd38d..9c25392cc0 100644
--- a/src/cmd/dist/test.go
+++ b/src/cmd/dist/test.go
@@ -1619,7 +1619,9 @@ func raceDetectorSupported(goos, goarch string) bool {
 	switch goos {
 	case "linux":
 		return goarch == "amd64" || goarch == "ppc64le" || goarch == "arm64"
-	case "darwin", "freebsd", "netbsd", "windows":
+	case "darwin":
+		return goarch == "amd64" || goarch == "arm64"
+	case "freebsd", "netbsd", "windows":
 		return goarch == "amd64"
 	default:
 		return false
diff --git a/src/cmd/go/internal/work/init.go b/src/cmd/go/internal/work/init.go
index d65c076c6a..102def4838 100644
--- a/src/cmd/go/internal/work/init.go
+++ b/src/cmd/go/internal/work/init.go
@@ -79,7 +79,7 @@ func instrumentInit() {
 	}
 	if cfg.BuildRace {
 		if !sys.RaceDetectorSupported(cfg.Goos, cfg.Goarch) {
-			fmt.Fprintf(os.Stderr, "go %s: -race is only supported on linux/amd64, linux/ppc64le, linux/arm64, freebsd/amd64, netbsd/amd64, darwin/amd64 and windows/amd64\n", flag.Args()[0])
+			fmt.Fprintf(os.Stderr, "go %s: -race is only supported on linux/amd64, linux/ppc64le, linux/arm64, freebsd/amd64, netbsd/amd64, darwin/amd64, darwin/arm64, and windows/amd64\n", flag.Args()[0])
 			base.SetExitStatus(2)
 			base.Exit()
 		}
diff --git a/src/cmd/internal/sys/supported.go b/src/cmd/internal/sys/supported.go
index 3c750774ed..69d7591440 100644
--- a/src/cmd/internal/sys/supported.go
+++ b/src/cmd/internal/sys/supported.go
@@ -13,7 +13,9 @@ func RaceDetectorSupported(goos, goarch string) bool {
 	switch goos {
 	case "linux":
 		return goarch == "amd64" || goarch == "ppc64le" || goarch == "arm64"
-	case "darwin", "freebsd", "netbsd", "windows":
+	case "darwin":
+		return goarch == "amd64" || goarch == "arm64"
+	case "freebsd", "netbsd", "windows":
 		return goarch == "amd64"
 	default:
 		return false
diff --git a/src/cmd/link/internal/ld/config.go b/src/cmd/link/internal/ld/config.go
index 0cb3cc25c0..cd64d86a4a 100644
--- a/src/cmd/link/internal/ld/config.go
+++ b/src/cmd/link/internal/ld/config.go
@@ -206,7 +206,7 @@ func mustLinkExternal(ctxt *Link) (res bool, reason string) {
 	// When the race flag is set, the LLVM tsan relocatable file is linked
 	// into the final binary, which means external linking is required because
 	// internal linking does not support it.
-	if *flagRace && ctxt.Arch.InFamily(sys.PPC64) {
+	if *flagRace && (ctxt.Arch.InFamily(sys.PPC64) || ctxt.IsDarwin() && ctxt.IsARM64()) {
 		return true, "race on " + objabi.GOARCH
 	}
 
diff --git a/src/cmd/link/internal/loadmacho/ldmacho.go b/src/cmd/link/internal/loadmacho/ldmacho.go
index d12f2bc2ac..d26869e23a 100644
--- a/src/cmd/link/internal/loadmacho/ldmacho.go
+++ b/src/cmd/link/internal/loadmacho/ldmacho.go
@@ -47,7 +47,7 @@ THE SOFTWARE.
 const (
 	MACHO_X86_64_RELOC_UNSIGNED = 0
 	MACHO_X86_64_RELOC_SIGNED   = 1
-	MACHO_FAKE_GOTPCREL         = 100
+	MACHO_ARM64_RELOC_ADDEND    = 10
 )
 
 type ldMachoObj struct {
@@ -707,11 +707,11 @@ func Load(l *loader.Loader, arch *sys.Arch, localSymVersion int, f *bio.Reader,
 		}
 
 		sb := l.MakeSymbolUpdater(sect.sym)
+		var rAdd int64
 		for j := uint32(0); j < sect.nreloc; j++ {
 			var (
 				rOff  int32
 				rSize uint8
-				rAdd  int64
 				rType objabi.RelocType
 				rSym  loader.Sym
 			)
@@ -722,6 +722,14 @@ func Load(l *loader.Loader, arch *sys.Arch, localSymVersion int, f *bio.Reader,
 				return errorf("%v: unexpected scattered relocation", s)
 			}
 
+			if arch.Family == sys.ARM64 && rel.type_ == MACHO_ARM64_RELOC_ADDEND {
+				// Two relocations. This addend will be applied to the next one.
+				rAdd = int64(rel.symnum)
+				continue
+			} else {
+				rAdd = 0
+			}
+
 			rSize = rel.length
 			rType = objabi.MachoRelocOffset + (objabi.RelocType(rel.type_) << 1) + objabi.RelocType(rel.pcrel)
 			rOff = int32(rel.addr)
diff --git a/src/runtime/race/README b/src/runtime/race/README
index b36d82ccfd..178ab94ab5 100644
--- a/src/runtime/race/README
+++ b/src/runtime/race/README
@@ -11,3 +11,4 @@ race_linux_ppc64le.syso built with LLVM 89f7ccea6f6488c443655880229c54db1f180153
 race_netbsd_amd64.syso built with LLVM 89f7ccea6f6488c443655880229c54db1f180153 and Go f62d3202bf9dbb3a00ad2a2c63ff4fa4188c5d3b.
 race_windows_amd64.syso built with LLVM 89f7ccea6f6488c443655880229c54db1f180153 and Go f62d3202bf9dbb3a00ad2a2c63ff4fa4188c5d3b.
 race_linux_arm64.syso built with LLVM 89f7ccea6f6488c443655880229c54db1f180153 and Go f62d3202bf9dbb3a00ad2a2c63ff4fa4188c5d3b.
+race_darwin_arm64.syso built with LLVM 00da38ce2d36c07f12c287dc515d37bb7bc410e9 and Go fe70a3a0fd31441bcbb9932ecab11a6083cf2119.
diff --git a/src/runtime/race/race.go b/src/runtime/race/race.go
index c894de5f72..d6a14b79e7 100644
--- a/src/runtime/race/race.go
+++ b/src/runtime/race/race.go
@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build race,linux,amd64 race,freebsd,amd64 race,netbsd,amd64 race,darwin,amd64 race,windows,amd64 race,linux,ppc64le race,linux,arm64
+// +build race,linux,amd64 race,freebsd,amd64 race,netbsd,amd64 race,darwin,amd64 race,windows,amd64 race,linux,ppc64le race,linux,arm64 race,darwin,arm64
 
 package race
 
diff --git a/src/runtime/race/race_darwin_arm64.syso b/src/runtime/race/race_darwin_arm64.syso
new file mode 100644
index 0000000000..f6eaa62ae3
--- /dev/null
+++ b/src/runtime/race/race_darwin_arm64.syso
diff --git a/src/runtime/race_arm64.s b/src/runtime/race_arm64.s
index 9b909ac021..59373a9f3f 100644
--- a/src/runtime/race_arm64.s
+++ b/src/runtime/race_arm64.s
@@ -25,8 +25,15 @@
 
 // The race ctx, ThreadState *thr below, is passed in R0 and loaded in racecalladdr.
 
+#ifdef TLS_darwin
+#define TP_ALIGN	AND	$~7, R0
+#else
+#define TP_ALIGN
+#endif
+
 #define load_g \
 	MRS_TPIDR_R0 \
+	TP_ALIGN \
 	MOVD    runtime·tls_g(SB), R11 \
 	ADD     R11, R0 \
 	MOVD    0(R0), g
@@ -423,7 +430,13 @@ TEXT	runtime·racecallbackthunk(SB), NOSPLIT|NOFRAME, $0
 	// benefit from this fast path.
 	CBNZ	R0, rest
 	MOVD	g, R13
+#ifdef TLS_darwin
+	MOVD	R27, R12 // save R27 a.k.a. REGTMP (callee-save in C). load_g clobbers it
+#endif
 	load_g
+#ifdef TLS_darwin
+	MOVD	R12, R27
+#endif
 	MOVD	g_m(g), R0
 	MOVD	m_p(R0), R0
 	MOVD	p_raceprocctx(R0), R0
@@ -477,5 +490,7 @@ noswitch:
 	BL	runtime·racecallback(SB)
 	JMP	ret
 
+#ifndef TLSG_IS_VARIABLE
 // tls_g, g value for each thread in TLS
 GLOBL runtime·tls_g+0(SB), TLSBSS+DUPOK, $8
+#endif