From f3c1070fa4cb02c55b47b874076fe74879288a4c Mon Sep 17 00:00:00 2001
From: Dmitriy Vyukov <dvyukov@google.com>
Date: Mon, 12 Aug 2013 21:46:33 +0400
Subject: [PATCH] sync/atomic: specify argsize for asm routines Fixes #6098.

R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/12717043
---
 src/pkg/sync/atomic/asm_386.s         | 46 +++++++++++++--------------
 src/pkg/sync/atomic/asm_amd64.s       | 46 +++++++++++++--------------
 src/pkg/sync/atomic/asm_arm.s         | 16 +++++-----
 src/pkg/sync/atomic/asm_freebsd_arm.s |  4 +--
 src/pkg/sync/atomic/asm_linux_arm.s   | 16 +++++-----
 src/pkg/sync/atomic/asm_netbsd_arm.s  |  4 +--
 src/pkg/sync/atomic/atomic_test.go    | 37 +++++++++++++++++++++
 7 files changed, 103 insertions(+), 66 deletions(-)

diff --git a/src/pkg/sync/atomic/asm_386.s b/src/pkg/sync/atomic/asm_386.s
index 5313a095913..75379f5fcf0 100644
--- a/src/pkg/sync/atomic/asm_386.s
+++ b/src/pkg/sync/atomic/asm_386.s
@@ -6,10 +6,10 @@
 
 #include "../../../cmd/ld/textflag.h"
 
-TEXT ·CompareAndSwapInt32(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapInt32(SB),NOSPLIT,$0-13
 	JMP	·CompareAndSwapUint32(SB)
 
-TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0-13
 	MOVL	addr+0(FP), BP
 	MOVL	old+4(FP), AX
 	MOVL	new+8(FP), CX
@@ -19,16 +19,16 @@ TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0
 	SETEQ	swapped+12(FP)
 	RET
 
-TEXT ·CompareAndSwapUintptr(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUintptr(SB),NOSPLIT,$0-13
 	JMP	·CompareAndSwapUint32(SB)
 
-TEXT ·CompareAndSwapPointer(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapPointer(SB),NOSPLIT,$0-13
 	JMP	·CompareAndSwapUint32(SB)
 
-TEXT ·CompareAndSwapInt64(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapInt64(SB),NOSPLIT,$0-21
 	JMP	·CompareAndSwapUint64(SB)
 
-TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$0-21
 	MOVL	addr+0(FP), BP
 	TESTL	$7, BP
 	JZ	2(PC)
@@ -43,10 +43,10 @@ TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$0
 	SETEQ	swapped+20(FP)
 	RET
 
-TEXT ·AddInt32(SB),NOSPLIT,$0
+TEXT ·AddInt32(SB),NOSPLIT,$0-12
 	JMP	·AddUint32(SB)
 
-TEXT ·AddUint32(SB),NOSPLIT,$0
+TEXT ·AddUint32(SB),NOSPLIT,$0-12
 	MOVL	addr+0(FP), BP
 	MOVL	delta+4(FP), AX
 	MOVL	AX, CX
@@ -57,13 +57,13 @@ TEXT ·AddUint32(SB),NOSPLIT,$0
 	MOVL	CX, new+8(FP)
 	RET
 
-TEXT ·AddUintptr(SB),NOSPLIT,$0
+TEXT ·AddUintptr(SB),NOSPLIT,$0-12
 	JMP	·AddUint32(SB)
 
-TEXT ·AddInt64(SB),NOSPLIT,$0
+TEXT ·AddInt64(SB),NOSPLIT,$0-20
 	JMP	·AddUint64(SB)
 
-TEXT ·AddUint64(SB),NOSPLIT,$0
+TEXT ·AddUint64(SB),NOSPLIT,$0-20
 	// no XADDQ so use CMPXCHG8B loop
 	MOVL	addr+0(FP), BP
 	TESTL	$7, BP
@@ -99,19 +99,19 @@ addloop:
 	MOVL	CX, new_hi+16(FP)
 	RET
 
-TEXT ·LoadInt32(SB),NOSPLIT,$0
+TEXT ·LoadInt32(SB),NOSPLIT,$0-8
 	JMP	·LoadUint32(SB)
 
-TEXT ·LoadUint32(SB),NOSPLIT,$0
+TEXT ·LoadUint32(SB),NOSPLIT,$0-8
 	MOVL	addr+0(FP), AX
 	MOVL	0(AX), AX
 	MOVL	AX, val+4(FP)
 	RET
 
-TEXT ·LoadInt64(SB),NOSPLIT,$0
+TEXT ·LoadInt64(SB),NOSPLIT,$0-16
 	JMP	·LoadUint64(SB)
 
-TEXT ·LoadUint64(SB),NOSPLIT,$0
+TEXT ·LoadUint64(SB),NOSPLIT,$0-16
 	MOVL	addr+0(FP), AX
 	TESTL	$7, AX
 	JZ	2(PC)
@@ -124,25 +124,25 @@ TEXT ·LoadUint64(SB),NOSPLIT,$0
 	EMMS
 	RET
 
-TEXT ·LoadUintptr(SB),NOSPLIT,$0
+TEXT ·LoadUintptr(SB),NOSPLIT,$0-8
 	JMP	·LoadUint32(SB)
 
-TEXT ·LoadPointer(SB),NOSPLIT,$0
+TEXT ·LoadPointer(SB),NOSPLIT,$0-8
 	JMP	·LoadUint32(SB)
 
-TEXT ·StoreInt32(SB),NOSPLIT,$0
+TEXT ·StoreInt32(SB),NOSPLIT,$0-8
 	JMP	·StoreUint32(SB)
 
-TEXT ·StoreUint32(SB),NOSPLIT,$0
+TEXT ·StoreUint32(SB),NOSPLIT,$0-8
 	MOVL	addr+0(FP), BP
 	MOVL	val+4(FP), AX
 	XCHGL	AX, 0(BP)
 	RET
 
-TEXT ·StoreInt64(SB),NOSPLIT,$0
+TEXT ·StoreInt64(SB),NOSPLIT,$0-16
 	JMP	·StoreUint64(SB)
 
-TEXT ·StoreUint64(SB),NOSPLIT,$0
+TEXT ·StoreUint64(SB),NOSPLIT,$0-16
 	MOVL	addr+0(FP), AX
 	TESTL	$7, AX
 	JZ	2(PC)
@@ -160,8 +160,8 @@ TEXT ·StoreUint64(SB),NOSPLIT,$0
 	XADDL	AX, (SP)
 	RET
 
-TEXT ·StoreUintptr(SB),NOSPLIT,$0
+TEXT ·StoreUintptr(SB),NOSPLIT,$0-8
 	JMP	·StoreUint32(SB)
 
-TEXT ·StorePointer(SB),NOSPLIT,$0
+TEXT ·StorePointer(SB),NOSPLIT,$0-8
 	JMP	·StoreUint32(SB)
diff --git a/src/pkg/sync/atomic/asm_amd64.s b/src/pkg/sync/atomic/asm_amd64.s
index c254614a578..4e062cc20db 100644
--- a/src/pkg/sync/atomic/asm_amd64.s
+++ b/src/pkg/sync/atomic/asm_amd64.s
@@ -6,10 +6,10 @@
 
 #include "../../../cmd/ld/textflag.h"
 
-TEXT ·CompareAndSwapInt32(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapInt32(SB),NOSPLIT,$0-17
 	JMP	·CompareAndSwapUint32(SB)
 
-TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0-17
 	MOVQ	addr+0(FP), BP
 	MOVL	old+8(FP), AX
 	MOVL	new+12(FP), CX
@@ -18,16 +18,16 @@ TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0
 	SETEQ	swapped+16(FP)
 	RET
 
-TEXT ·CompareAndSwapUintptr(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUintptr(SB),NOSPLIT,$0-25
 	JMP	·CompareAndSwapUint64(SB)
 
-TEXT ·CompareAndSwapPointer(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapPointer(SB),NOSPLIT,$0-25
 	JMP	·CompareAndSwapUint64(SB)
 
-TEXT ·CompareAndSwapInt64(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapInt64(SB),NOSPLIT,$0-25
 	JMP	·CompareAndSwapUint64(SB)
 
-TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$0-25
 	MOVQ	addr+0(FP), BP
 	MOVQ	old+8(FP), AX
 	MOVQ	new+16(FP), CX
@@ -36,10 +36,10 @@ TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$0
 	SETEQ	swapped+24(FP)
 	RET
 
-TEXT ·AddInt32(SB),NOSPLIT,$0
+TEXT ·AddInt32(SB),NOSPLIT,$0-20
 	JMP	·AddUint32(SB)
 
-TEXT ·AddUint32(SB),NOSPLIT,$0
+TEXT ·AddUint32(SB),NOSPLIT,$0-20
 	MOVQ	addr+0(FP), BP
 	MOVL	delta+8(FP), AX
 	MOVL	AX, CX
@@ -49,13 +49,13 @@ TEXT ·AddUint32(SB),NOSPLIT,$0
 	MOVL	CX, new+16(FP)
 	RET
 
-TEXT ·AddUintptr(SB),NOSPLIT,$0
+TEXT ·AddUintptr(SB),NOSPLIT,$0-24
 	JMP	·AddUint64(SB)
 
-TEXT ·AddInt64(SB),NOSPLIT,$0
+TEXT ·AddInt64(SB),NOSPLIT,$0-24
 	JMP	·AddUint64(SB)
 
-TEXT ·AddUint64(SB),NOSPLIT,$0
+TEXT ·AddUint64(SB),NOSPLIT,$0-24
 	MOVQ	addr+0(FP), BP
 	MOVQ	delta+8(FP), AX
 	MOVQ	AX, CX
@@ -65,55 +65,55 @@ TEXT ·AddUint64(SB),NOSPLIT,$0
 	MOVQ	CX, new+16(FP)
 	RET
 
-TEXT ·LoadInt32(SB),NOSPLIT,$0
+TEXT ·LoadInt32(SB),NOSPLIT,$0-12
 	JMP	·LoadUint32(SB)
 
-TEXT ·LoadUint32(SB),NOSPLIT,$0
+TEXT ·LoadUint32(SB),NOSPLIT,$0-12
 	MOVQ	addr+0(FP), AX
 	MOVL	0(AX), AX
 	MOVL	AX, val+8(FP)
 	RET
 
-TEXT ·LoadInt64(SB),NOSPLIT,$0
+TEXT ·LoadInt64(SB),NOSPLIT,$0-16
 	JMP	·LoadUint64(SB)
 
-TEXT ·LoadUint64(SB),NOSPLIT,$0
+TEXT ·LoadUint64(SB),NOSPLIT,$0-16
 	MOVQ	addr+0(FP), AX
 	MOVQ	0(AX), AX
 	MOVQ	AX, val+8(FP)
 	RET
 
-TEXT ·LoadUintptr(SB),NOSPLIT,$0
+TEXT ·LoadUintptr(SB),NOSPLIT,$0-16
 	JMP	·LoadPointer(SB)
 
-TEXT ·LoadPointer(SB),NOSPLIT,$0
+TEXT ·LoadPointer(SB),NOSPLIT,$0-16
 	MOVQ	addr+0(FP), AX
 	MOVQ	0(AX), AX
 	MOVQ	AX, val+8(FP)
 	RET
 
-TEXT ·StoreInt32(SB),NOSPLIT,$0
+TEXT ·StoreInt32(SB),NOSPLIT,$0-12
 	JMP	·StoreUint32(SB)
 
-TEXT ·StoreUint32(SB),NOSPLIT,$0
+TEXT ·StoreUint32(SB),NOSPLIT,$0-12
 	MOVQ	addr+0(FP), BP
 	MOVL	val+8(FP), AX
 	XCHGL	AX, 0(BP)
 	RET
 
-TEXT ·StoreInt64(SB),NOSPLIT,$0
+TEXT ·StoreInt64(SB),NOSPLIT,$0-16
 	JMP	·StoreUint64(SB)
 
-TEXT ·StoreUint64(SB),NOSPLIT,$0
+TEXT ·StoreUint64(SB),NOSPLIT,$0-16
 	MOVQ	addr+0(FP), BP
 	MOVQ	val+8(FP), AX
 	XCHGQ	AX, 0(BP)
 	RET
 
-TEXT ·StoreUintptr(SB),NOSPLIT,$0
+TEXT ·StoreUintptr(SB),NOSPLIT,$0-16
 	JMP	·StorePointer(SB)
 
-TEXT ·StorePointer(SB),NOSPLIT,$0
+TEXT ·StorePointer(SB),NOSPLIT,$0-16
 	MOVQ	addr+0(FP), BP
 	MOVQ	val+8(FP), AX
 	XCHGQ	AX, 0(BP)
diff --git a/src/pkg/sync/atomic/asm_arm.s b/src/pkg/sync/atomic/asm_arm.s
index 2d974b625c6..70db48967bd 100644
--- a/src/pkg/sync/atomic/asm_arm.s
+++ b/src/pkg/sync/atomic/asm_arm.s
@@ -8,7 +8,7 @@
 
 // ARM atomic operations, for use by asm_$(GOOS)_arm.s.
 
-TEXT ·armCompareAndSwapUint32(SB),NOSPLIT,$0
+TEXT ·armCompareAndSwapUint32(SB),NOSPLIT,$0-13
 	MOVW	addr+0(FP), R1
 	MOVW	old+4(FP), R2
 	MOVW	new+8(FP), R3
@@ -28,7 +28,7 @@ casfail:
 	MOVBU	R0, ret+12(FP)
 	RET
 
-TEXT ·armCompareAndSwapUint64(SB),NOSPLIT,$0
+TEXT ·armCompareAndSwapUint64(SB),NOSPLIT,$0-21
 	BL	fastCheck64<>(SB)
 	MOVW	addr+0(FP), R1
 	// make unaligned atomic access panic
@@ -57,7 +57,7 @@ cas64fail:
 	MOVBU	R0, ret+20(FP)
 	RET
 
-TEXT ·armAddUint32(SB),NOSPLIT,$0
+TEXT ·armAddUint32(SB),NOSPLIT,$0-12
 	MOVW	addr+0(FP), R1
 	MOVW	delta+4(FP), R2
 addloop:
@@ -70,7 +70,7 @@ addloop:
 	MOVW	R3, ret+8(FP)
 	RET
 
-TEXT ·armAddUint64(SB),NOSPLIT,$0
+TEXT ·armAddUint64(SB),NOSPLIT,$0-20
 	BL	fastCheck64<>(SB)
 	MOVW	addr+0(FP), R1
 	// make unaligned atomic access panic
@@ -91,7 +91,7 @@ add64loop:
 	MOVW	R5, rethi+16(FP)
 	RET
 
-TEXT ·armLoadUint64(SB),NOSPLIT,$0
+TEXT ·armLoadUint64(SB),NOSPLIT,$0-12
 	BL	fastCheck64<>(SB)
 	MOVW	addr+0(FP), R1
 	// make unaligned atomic access panic
@@ -107,7 +107,7 @@ load64loop:
 	MOVW	R3, valhi+8(FP)
 	RET
 
-TEXT ·armStoreUint64(SB),NOSPLIT,$0
+TEXT ·armStoreUint64(SB),NOSPLIT,$0-12
 	BL	fastCheck64<>(SB)
 	MOVW	addr+0(FP), R1
 	// make unaligned atomic access panic
@@ -131,7 +131,7 @@ store64loop:
 // which will make uses of the 64-bit atomic operations loop forever.
 // If things are working, set okLDREXD to avoid future checks.
 // https://bugs.launchpad.net/qemu/+bug/670883.
-TEXT	check64<>(SB),NOSPLIT,$16
+TEXT	check64<>(SB),NOSPLIT,$16-0
 	MOVW	$10, R1
 	// 8-aligned stack address scratch space.
 	MOVW	$8(R13), R5
@@ -156,7 +156,7 @@ TEXT	fastCheck64<>(SB),NOSPLIT,$-4
 	RET.NE
 	B	slowCheck64<>(SB)
 
-TEXT slowCheck64<>(SB),NOSPLIT,$0
+TEXT slowCheck64<>(SB),NOSPLIT,$0-0
 	BL	check64<>(SB)
 	// Still here, must be okay.
 	MOVW	$1, R0
diff --git a/src/pkg/sync/atomic/asm_freebsd_arm.s b/src/pkg/sync/atomic/asm_freebsd_arm.s
index 013fdfc57db..813c1aa4f25 100644
--- a/src/pkg/sync/atomic/asm_freebsd_arm.s
+++ b/src/pkg/sync/atomic/asm_freebsd_arm.s
@@ -43,7 +43,7 @@ TEXT ·AddUint64(SB),NOSPLIT,$0
 TEXT ·LoadInt32(SB),NOSPLIT,$0
 	B ·LoadUint32(SB)
 
-TEXT ·LoadUint32(SB),NOSPLIT,$0
+TEXT ·LoadUint32(SB),NOSPLIT,$0-8
 	MOVW addr+0(FP), R1
 load32loop:
 	LDREX (R1), R2		// loads R2
@@ -68,7 +68,7 @@ TEXT ·LoadPointer(SB),NOSPLIT,$0
 TEXT ·StoreInt32(SB),NOSPLIT,$0
 	B ·StoreUint32(SB)
 
-TEXT ·StoreUint32(SB),NOSPLIT,$0
+TEXT ·StoreUint32(SB),NOSPLIT,$0-8
 	MOVW addr+0(FP), R1
 	MOVW val+4(FP), R2
 storeloop:
diff --git a/src/pkg/sync/atomic/asm_linux_arm.s b/src/pkg/sync/atomic/asm_linux_arm.s
index b78c7fa038a..31c86ccbe3d 100644
--- a/src/pkg/sync/atomic/asm_linux_arm.s
+++ b/src/pkg/sync/atomic/asm_linux_arm.s
@@ -30,7 +30,7 @@ TEXT ·CompareAndSwapInt32(SB),NOSPLIT,$0
 	B	·CompareAndSwapUint32(SB)
 
 // Implement using kernel cas for portability.
-TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0
+TEXT ·CompareAndSwapUint32(SB),NOSPLIT,$0-13
 	MOVW	addr+0(FP), R2
 	MOVW	old+4(FP), R0
 casagain:
@@ -61,7 +61,7 @@ TEXT ·AddInt32(SB),NOSPLIT,$0
 	B	·AddUint32(SB)
 
 // Implement using kernel cas for portability.
-TEXT ·AddUint32(SB),NOSPLIT,$0
+TEXT ·AddUint32(SB),NOSPLIT,$0-12
 	MOVW	addr+0(FP), R2
 	MOVW	delta+4(FP), R4
 addloop1:
@@ -79,7 +79,7 @@ TEXT ·AddUintptr(SB),NOSPLIT,$0
 TEXT cas64<>(SB),NOSPLIT,$0
 	MOVW	$0xffff0f60, PC // __kuser_cmpxchg64: Linux-3.1 and above
 
-TEXT kernelCAS64<>(SB),NOSPLIT,$0
+TEXT kernelCAS64<>(SB),NOSPLIT,$0-21
 	// int (*__kuser_cmpxchg64_t)(const int64_t *oldval, const int64_t *newval, volatile int64_t *ptr);
 	MOVW	addr+0(FP), R2 // ptr
 	// make unaligned atomic access panic
@@ -94,7 +94,7 @@ TEXT kernelCAS64<>(SB),NOSPLIT,$0
 	MOVW	R0, 20(FP)
 	RET
 
-TEXT generalCAS64<>(SB),NOSPLIT,$20
+TEXT generalCAS64<>(SB),NOSPLIT,$20-21
 	// bool runtime·cas64(uint64 volatile *addr, uint64 *old, uint64 new)
 	MOVW	addr+0(FP), R0
 	// make unaligned atomic access panic
@@ -114,7 +114,7 @@ TEXT generalCAS64<>(SB),NOSPLIT,$20
 
 GLOBL armCAS64(SB), $4
 
-TEXT setupAndCallCAS64<>(SB),NOSPLIT,$-4
+TEXT setupAndCallCAS64<>(SB),NOSPLIT,$-21
 	MOVW	$0xffff0ffc, R0 // __kuser_helper_version
 	MOVW	(R0), R0
 	// __kuser_cmpxchg64 only present if helper version >= 5
@@ -136,7 +136,7 @@ TEXT setupAndCallCAS64<>(SB),NOSPLIT,$-4
 TEXT ·CompareAndSwapInt64(SB),NOSPLIT,$0
 	B   	·CompareAndSwapUint64(SB)
 
-TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$-4
+TEXT ·CompareAndSwapUint64(SB),NOSPLIT,$-21
 	MOVW	armCAS64(SB), R0
 	CMP 	$0, R0
 	MOVW.NE	R0, PC
@@ -151,7 +151,7 @@ TEXT ·AddUint64(SB),NOSPLIT,$0
 TEXT ·LoadInt32(SB),NOSPLIT,$0
 	B	·LoadUint32(SB)
 
-TEXT ·LoadUint32(SB),NOSPLIT,$0
+TEXT ·LoadUint32(SB),NOSPLIT,$0-8
 	MOVW	addr+0(FP), R2
 loadloop1:
 	MOVW	0(R2), R0
@@ -176,7 +176,7 @@ TEXT ·LoadPointer(SB),NOSPLIT,$0
 TEXT ·StoreInt32(SB),NOSPLIT,$0
 	B	·StoreUint32(SB)
 
-TEXT ·StoreUint32(SB),NOSPLIT,$0
+TEXT ·StoreUint32(SB),NOSPLIT,$0-8
 	MOVW	addr+0(FP), R2
 	MOVW	val+4(FP), R1
 storeloop1:
diff --git a/src/pkg/sync/atomic/asm_netbsd_arm.s b/src/pkg/sync/atomic/asm_netbsd_arm.s
index e2eea9f8db6..a4d1f4e828d 100644
--- a/src/pkg/sync/atomic/asm_netbsd_arm.s
+++ b/src/pkg/sync/atomic/asm_netbsd_arm.s
@@ -43,7 +43,7 @@ TEXT ·AddUint64(SB),NOSPLIT,$0
 TEXT ·LoadInt32(SB),NOSPLIT,$0
 	B ·LoadUint32(SB)
 
-TEXT ·LoadUint32(SB),NOSPLIT,$0
+TEXT ·LoadUint32(SB),NOSPLIT,$0-8
 	MOVW addr+0(FP), R1
 load32loop:
 	LDREX (R1), R2		// loads R2
@@ -68,7 +68,7 @@ TEXT ·LoadPointer(SB),NOSPLIT,$0
 TEXT ·StoreInt32(SB),NOSPLIT,$0
 	B ·StoreUint32(SB)
 
-TEXT ·StoreUint32(SB),NOSPLIT,$0
+TEXT ·StoreUint32(SB),NOSPLIT,$0-8
 	MOVW addr+0(FP), R1
 	MOVW val+4(FP), R2
 storeloop:
diff --git a/src/pkg/sync/atomic/atomic_test.go b/src/pkg/sync/atomic/atomic_test.go
index b392df595e4..cec81c4626b 100644
--- a/src/pkg/sync/atomic/atomic_test.go
+++ b/src/pkg/sync/atomic/atomic_test.go
@@ -1203,3 +1203,40 @@ func TestUnaligned64(t *testing.T) {
 	shouldPanic(t, "CompareAndSwapUint64", func() { CompareAndSwapUint64(p, 1, 2) })
 	shouldPanic(t, "AddUint64", func() { AddUint64(p, 3) })
 }
+
+func TestNilDeref(t *testing.T) {
+	funcs := [...]func(){
+		func() { CompareAndSwapInt32(nil, 0, 0) },
+		func() { CompareAndSwapInt64(nil, 0, 0) },
+		func() { CompareAndSwapUint32(nil, 0, 0) },
+		func() { CompareAndSwapUint64(nil, 0, 0) },
+		func() { CompareAndSwapUintptr(nil, 0, 0) },
+		func() { CompareAndSwapPointer(nil, nil, nil) },
+		func() { AddInt32(nil, 0) },
+		func() { AddUint32(nil, 0) },
+		func() { AddInt64(nil, 0) },
+		func() { AddUint64(nil, 0) },
+		func() { AddUintptr(nil, 0) },
+		func() { LoadInt32(nil) },
+		func() { LoadInt64(nil) },
+		func() { LoadUint32(nil) },
+		func() { LoadUint64(nil) },
+		func() { LoadUintptr(nil) },
+		func() { LoadPointer(nil) },
+		func() { StoreInt32(nil, 0) },
+		func() { StoreInt64(nil, 0) },
+		func() { StoreUint32(nil, 0) },
+		func() { StoreUint64(nil, 0) },
+		func() { StoreUintptr(nil, 0) },
+		func() { StorePointer(nil, nil) },
+	}
+	for _, f := range funcs {
+		func() {
+			defer func() {
+				runtime.GC()
+				recover()
+			}()
+			f()
+		}()
+	}
+}