diff --git a/src/runtime/race/testdata/atomic_test.go b/src/runtime/race/testdata/atomic_test.go
index 769c8d7..4ce7260 100644
--- a/src/runtime/race/testdata/atomic_test.go
+++ b/src/runtime/race/testdata/atomic_test.go
@@ -299,3 +299,27 @@
 	}()
 	atomic.AddInt32(nilptr, 1)
 }
+
+func TestNoRaceDeferAtomicStore(t *testing.T) {
+	// Test that when an atomic function is deferred directly, the
+	// GC scans it correctly. See issue 42599.
+	type foo struct {
+		bar int64
+	}
+
+	var doFork func(f *foo, depth int)
+	doFork = func(f *foo, depth int) {
+		atomic.StoreInt64(&f.bar, 1)
+		defer atomic.StoreInt64(&f.bar, 0)
+		if depth > 0 {
+			for i := 0; i < 2; i++ {
+				f2 := &foo{}
+				go doFork(f2, depth-1)
+			}
+		}
+		runtime.GC()
+	}
+
+	f := &foo{}
+	doFork(f, 11)
+}
diff --git a/src/runtime/race_amd64.s b/src/runtime/race_amd64.s
index 4a86b33..9818bc6 100644
--- a/src/runtime/race_amd64.s
+++ b/src/runtime/race_amd64.s
@@ -207,110 +207,136 @@
 // Atomic operations for sync/atomic package.
 
 // Load
-TEXT	sync∕atomic·LoadInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadInt32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic32_load(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·LoadInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadInt64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic64_load(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·LoadUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadUint32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	JMP	sync∕atomic·LoadInt32(SB)
 
-TEXT	sync∕atomic·LoadUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadUint64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	JMP	sync∕atomic·LoadInt64(SB)
 
-TEXT	sync∕atomic·LoadUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadUintptr(SB), NOSPLIT, $0-16
+	GO_ARGS
 	JMP	sync∕atomic·LoadInt64(SB)
 
-TEXT	sync∕atomic·LoadPointer(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadPointer(SB), NOSPLIT, $0-16
+	GO_ARGS
 	JMP	sync∕atomic·LoadInt64(SB)
 
 // Store
-TEXT	sync∕atomic·StoreInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreInt32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic32_store(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·StoreInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreInt64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic64_store(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·StoreUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreUint32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	JMP	sync∕atomic·StoreInt32(SB)
 
-TEXT	sync∕atomic·StoreUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreUint64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	JMP	sync∕atomic·StoreInt64(SB)
 
-TEXT	sync∕atomic·StoreUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreUintptr(SB), NOSPLIT, $0-16
+	GO_ARGS
 	JMP	sync∕atomic·StoreInt64(SB)
 
 // Swap
-TEXT	sync∕atomic·SwapInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapInt32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic32_exchange(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·SwapInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapInt64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic64_exchange(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·SwapUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapUint32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	JMP	sync∕atomic·SwapInt32(SB)
 
-TEXT	sync∕atomic·SwapUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapUint64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	JMP	sync∕atomic·SwapInt64(SB)
 
-TEXT	sync∕atomic·SwapUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapUintptr(SB), NOSPLIT, $0-24
+	GO_ARGS
 	JMP	sync∕atomic·SwapInt64(SB)
 
 // Add
-TEXT	sync∕atomic·AddInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddInt32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic32_fetch_add(SB), AX
 	CALL	racecallatomic<>(SB)
 	MOVL	add+8(FP), AX	// convert fetch_add to add_fetch
 	ADDL	AX, ret+16(FP)
 	RET
 
-TEXT	sync∕atomic·AddInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddInt64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic64_fetch_add(SB), AX
 	CALL	racecallatomic<>(SB)
 	MOVQ	add+8(FP), AX	// convert fetch_add to add_fetch
 	ADDQ	AX, ret+16(FP)
 	RET
 
-TEXT	sync∕atomic·AddUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddUint32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	JMP	sync∕atomic·AddInt32(SB)
 
-TEXT	sync∕atomic·AddUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddUint64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	JMP	sync∕atomic·AddInt64(SB)
 
-TEXT	sync∕atomic·AddUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddUintptr(SB), NOSPLIT, $0-24
+	GO_ARGS
 	JMP	sync∕atomic·AddInt64(SB)
 
 // CompareAndSwap
-TEXT	sync∕atomic·CompareAndSwapInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapInt32(SB), NOSPLIT, $0-17
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic32_compare_exchange(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·CompareAndSwapInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapInt64(SB), NOSPLIT, $0-25
+	GO_ARGS
 	MOVQ	$__tsan_go_atomic64_compare_exchange(SB), AX
 	CALL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·CompareAndSwapUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapUint32(SB), NOSPLIT, $0-17
+	GO_ARGS
 	JMP	sync∕atomic·CompareAndSwapInt32(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapUint64(SB), NOSPLIT, $0-25
+	GO_ARGS
 	JMP	sync∕atomic·CompareAndSwapInt64(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapUintptr(SB), NOSPLIT, $0-25
+	GO_ARGS
 	JMP	sync∕atomic·CompareAndSwapInt64(SB)
 
 // Generic atomic operation implementation.
diff --git a/src/runtime/race_arm64.s b/src/runtime/race_arm64.s
index 6bc389f..8aa1774 100644
--- a/src/runtime/race_arm64.s
+++ b/src/runtime/race_arm64.s
@@ -200,86 +200,86 @@
 // R0, R1, R2 set in racecallatomic
 
 // Load
-TEXT	sync∕atomic·LoadInt32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·LoadInt32(SB), NOSPLIT, $0-12
 	GO_ARGS
 	MOVD	$__tsan_go_atomic32_load(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·LoadInt64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·LoadInt64(SB), NOSPLIT, $0-16
 	GO_ARGS
 	MOVD	$__tsan_go_atomic64_load(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·LoadUint32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·LoadUint32(SB), NOSPLIT, $0-12
 	GO_ARGS
 	JMP	sync∕atomic·LoadInt32(SB)
 
-TEXT	sync∕atomic·LoadUint64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·LoadUint64(SB), NOSPLIT, $0-16
 	GO_ARGS
 	JMP	sync∕atomic·LoadInt64(SB)
 
-TEXT	sync∕atomic·LoadUintptr(SB), NOSPLIT, $0
+TEXT	sync∕atomic·LoadUintptr(SB), NOSPLIT, $0-16
 	GO_ARGS
 	JMP	sync∕atomic·LoadInt64(SB)
 
-TEXT	sync∕atomic·LoadPointer(SB), NOSPLIT, $0
+TEXT	sync∕atomic·LoadPointer(SB), NOSPLIT, $0-16
 	GO_ARGS
 	JMP	sync∕atomic·LoadInt64(SB)
 
 // Store
-TEXT	sync∕atomic·StoreInt32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·StoreInt32(SB), NOSPLIT, $0-12
 	GO_ARGS
 	MOVD	$__tsan_go_atomic32_store(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·StoreInt64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·StoreInt64(SB), NOSPLIT, $0-16
 	GO_ARGS
 	MOVD	$__tsan_go_atomic64_store(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·StoreUint32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·StoreUint32(SB), NOSPLIT, $0-12
 	GO_ARGS
 	JMP	sync∕atomic·StoreInt32(SB)
 
-TEXT	sync∕atomic·StoreUint64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·StoreUint64(SB), NOSPLIT, $0-16
 	GO_ARGS
 	JMP	sync∕atomic·StoreInt64(SB)
 
-TEXT	sync∕atomic·StoreUintptr(SB), NOSPLIT, $0
+TEXT	sync∕atomic·StoreUintptr(SB), NOSPLIT, $0-16
 	GO_ARGS
 	JMP	sync∕atomic·StoreInt64(SB)
 
 // Swap
-TEXT	sync∕atomic·SwapInt32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·SwapInt32(SB), NOSPLIT, $0-20
 	GO_ARGS
 	MOVD	$__tsan_go_atomic32_exchange(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·SwapInt64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·SwapInt64(SB), NOSPLIT, $0-24
 	GO_ARGS
 	MOVD	$__tsan_go_atomic64_exchange(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·SwapUint32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·SwapUint32(SB), NOSPLIT, $0-20
 	GO_ARGS
 	JMP	sync∕atomic·SwapInt32(SB)
 
-TEXT	sync∕atomic·SwapUint64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·SwapUint64(SB), NOSPLIT, $0-24
 	GO_ARGS
 	JMP	sync∕atomic·SwapInt64(SB)
 
-TEXT	sync∕atomic·SwapUintptr(SB), NOSPLIT, $0
+TEXT	sync∕atomic·SwapUintptr(SB), NOSPLIT, $0-24
 	GO_ARGS
 	JMP	sync∕atomic·SwapInt64(SB)
 
 // Add
-TEXT	sync∕atomic·AddInt32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·AddInt32(SB), NOSPLIT, $0-20
 	GO_ARGS
 	MOVD	$__tsan_go_atomic32_fetch_add(SB), R9
 	BL	racecallatomic<>(SB)
@@ -289,7 +289,7 @@
 	MOVW	R0, ret+16(FP)
 	RET
 
-TEXT	sync∕atomic·AddInt64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·AddInt64(SB), NOSPLIT, $0-24
 	GO_ARGS
 	MOVD	$__tsan_go_atomic64_fetch_add(SB), R9
 	BL	racecallatomic<>(SB)
@@ -299,40 +299,40 @@
 	MOVD	R0, ret+16(FP)
 	RET
 
-TEXT	sync∕atomic·AddUint32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·AddUint32(SB), NOSPLIT, $0-20
 	GO_ARGS
 	JMP	sync∕atomic·AddInt32(SB)
 
-TEXT	sync∕atomic·AddUint64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·AddUint64(SB), NOSPLIT, $0-24
 	GO_ARGS
 	JMP	sync∕atomic·AddInt64(SB)
 
-TEXT	sync∕atomic·AddUintptr(SB), NOSPLIT, $0
+TEXT	sync∕atomic·AddUintptr(SB), NOSPLIT, $0-24
 	GO_ARGS
 	JMP	sync∕atomic·AddInt64(SB)
 
 // CompareAndSwap
-TEXT	sync∕atomic·CompareAndSwapInt32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·CompareAndSwapInt32(SB), NOSPLIT, $0-17
 	GO_ARGS
 	MOVD	$__tsan_go_atomic32_compare_exchange(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·CompareAndSwapInt64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·CompareAndSwapInt64(SB), NOSPLIT, $0-25
 	GO_ARGS
 	MOVD	$__tsan_go_atomic64_compare_exchange(SB), R9
 	BL	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·CompareAndSwapUint32(SB), NOSPLIT, $0
+TEXT	sync∕atomic·CompareAndSwapUint32(SB), NOSPLIT, $0-17
 	GO_ARGS
 	JMP	sync∕atomic·CompareAndSwapInt32(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUint64(SB), NOSPLIT, $0
+TEXT	sync∕atomic·CompareAndSwapUint64(SB), NOSPLIT, $0-25
 	GO_ARGS
 	JMP	sync∕atomic·CompareAndSwapInt64(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUintptr(SB), NOSPLIT, $0
+TEXT	sync∕atomic·CompareAndSwapUintptr(SB), NOSPLIT, $0-25
 	GO_ARGS
 	JMP	sync∕atomic·CompareAndSwapInt64(SB)
 
diff --git a/src/runtime/race_ppc64le.s b/src/runtime/race_ppc64le.s
index 7421d53..8961254 100644
--- a/src/runtime/race_ppc64le.s
+++ b/src/runtime/race_ppc64le.s
@@ -207,78 +207,95 @@
 // R3, R4, R5 set in racecallatomic
 
 // Load atomic in tsan
-TEXT	sync∕atomic·LoadInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadInt32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	// void __tsan_go_atomic32_load(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic32_load(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·LoadInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadInt64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	// void __tsan_go_atomic64_load(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic64_load(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 	RET
 
-TEXT	sync∕atomic·LoadUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadUint32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	BR	sync∕atomic·LoadInt32(SB)
 
-TEXT	sync∕atomic·LoadUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadUint64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	BR	sync∕atomic·LoadInt64(SB)
 
-TEXT	sync∕atomic·LoadUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadUintptr(SB), NOSPLIT, $0-16
+	GO_ARGS
 	BR	sync∕atomic·LoadInt64(SB)
 
-TEXT	sync∕atomic·LoadPointer(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·LoadPointer(SB), NOSPLIT, $0-16
+	GO_ARGS
 	BR	sync∕atomic·LoadInt64(SB)
 
 // Store atomic in tsan
-TEXT	sync∕atomic·StoreInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreInt32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	// void __tsan_go_atomic32_store(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic32_store(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 
-TEXT	sync∕atomic·StoreInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreInt64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	// void __tsan_go_atomic64_store(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic64_store(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 
-TEXT	sync∕atomic·StoreUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreUint32(SB), NOSPLIT, $0-12
+	GO_ARGS
 	BR	sync∕atomic·StoreInt32(SB)
 
-TEXT	sync∕atomic·StoreUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreUint64(SB), NOSPLIT, $0-16
+	GO_ARGS
 	BR	sync∕atomic·StoreInt64(SB)
 
-TEXT	sync∕atomic·StoreUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·StoreUintptr(SB), NOSPLIT, $0-16
+	GO_ARGS
 	BR	sync∕atomic·StoreInt64(SB)
 
 // Swap in tsan
-TEXT	sync∕atomic·SwapInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapInt32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	// void __tsan_go_atomic32_exchange(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic32_exchange(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 
-TEXT	sync∕atomic·SwapInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapInt64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	// void __tsan_go_atomic64_exchange(ThreadState *thr, uptr cpc, uptr pc, u8 *a)
 	MOVD	$__tsan_go_atomic64_exchange(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 
-TEXT	sync∕atomic·SwapUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapUint32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	BR	sync∕atomic·SwapInt32(SB)
 
-TEXT	sync∕atomic·SwapUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapUint64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	BR	sync∕atomic·SwapInt64(SB)
 
-TEXT	sync∕atomic·SwapUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·SwapUintptr(SB), NOSPLIT, $0-24
+	GO_ARGS
 	BR	sync∕atomic·SwapInt64(SB)
 
 // Add atomic in tsan
-TEXT	sync∕atomic·AddInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddInt32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	// void __tsan_go_atomic32_fetch_add(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic32_fetch_add(SB), R8
 	ADD	$64, R1, R6	// addr of caller's 1st arg
@@ -291,7 +308,8 @@
 	MOVW	R3, ret+16(FP)
 	RET
 
-TEXT	sync∕atomic·AddInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddInt64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	// void __tsan_go_atomic64_fetch_add(ThreadState *thr, uptr cpc, uptr pc, u8 *a);
 	MOVD	$__tsan_go_atomic64_fetch_add(SB), R8
 	ADD	$64, R1, R6	// addr of caller's 1st arg
@@ -304,37 +322,45 @@
 	MOVD	R3, ret+16(FP)
 	RET
 
-TEXT	sync∕atomic·AddUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddUint32(SB), NOSPLIT, $0-20
+	GO_ARGS
 	BR	sync∕atomic·AddInt32(SB)
 
-TEXT	sync∕atomic·AddUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddUint64(SB), NOSPLIT, $0-24
+	GO_ARGS
 	BR	sync∕atomic·AddInt64(SB)
 
-TEXT	sync∕atomic·AddUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·AddUintptr(SB), NOSPLIT, $0-24
+	GO_ARGS
 	BR	sync∕atomic·AddInt64(SB)
 
 // CompareAndSwap in tsan
-TEXT	sync∕atomic·CompareAndSwapInt32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapInt32(SB), NOSPLIT, $0-17
+	GO_ARGS
 	// void __tsan_go_atomic32_compare_exchange(
 	//   ThreadState *thr, uptr cpc, uptr pc, u8 *a)
 	MOVD	$__tsan_go_atomic32_compare_exchange(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 
-TEXT	sync∕atomic·CompareAndSwapInt64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapInt64(SB), NOSPLIT, $0-25
+	GO_ARGS
 	// void __tsan_go_atomic32_compare_exchange(
 	//   ThreadState *thr, uptr cpc, uptr pc, u8 *a)
 	MOVD	$__tsan_go_atomic64_compare_exchange(SB), R8
 	ADD	$32, R1, R6	// addr of caller's 1st arg
 	BR	racecallatomic<>(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUint32(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapUint32(SB), NOSPLIT, $0-17
+	GO_ARGS
 	BR	sync∕atomic·CompareAndSwapInt32(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUint64(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapUint64(SB), NOSPLIT, $0-25
+	GO_ARGS
 	BR	sync∕atomic·CompareAndSwapInt64(SB)
 
-TEXT	sync∕atomic·CompareAndSwapUintptr(SB), NOSPLIT, $0-0
+TEXT	sync∕atomic·CompareAndSwapUintptr(SB), NOSPLIT, $0-25
+	GO_ARGS
 	BR	sync∕atomic·CompareAndSwapInt64(SB)
 
 // Common function used to call tsan's atomic functions