Re: [PATCH] x86: Generate cmpxchg build failures

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Mon, 2009-10-05 at 12:16 -0700, Linus Torvalds wrote:
> 
> On Mon, 5 Oct 2009, Peter Zijlstra wrote:
> 
> > On Tue, 2009-09-29 at 14:17 -0700, Linus Torvalds wrote:
> > 
> > > And regardless, we should fix the silent cmpxchg failure, even if it's 
> > > just a link-time failure or something.
> > 
> > Something like the below?
> 
> Looks good to me. This is also one of the cases where a macro will do 
> better than an inline function, since the return value depends on the size 
> of the pointer, and it doesn't do the whole 'unsigned long' thing any 
> more.

Indeed, I was glad to see that go.

> It should also be fairly easy to make this now just do a cmpxchg8b for the 
> 64-bit case (again, that wouldn't have worked sanely due to the fixed type 
> in the inline function version - expanding that size to 64-bit would have 
> been insane). But that's a separate issue (and maybe we don't want to do 
> it, due to the whole "it's not DMA-atomic" etc issue - we may be better 
> off with a build failure, and forcing people who really want 64-bit 
> accesses to use the explicit cmpxchg64 thing)

Right, that would be a second patch if we think its a sane thing to do.

> That said, I think that you should merge the insane three versions of this 
> macro into one. Having separate versions for "__[sync_|local_|]cmpxchg()" 
> is disgusting. I bet you can do it with a single macro, and just pass in 
> the LOCK_PREFIX (or empty, or "lock;") to that as needed. Rather than 
> duplicating it three times.

You're right, what was I thinking doing all this copy/paste stuff.

Here's a new one, still seems to build i386 and x86_64 defconfigs.


Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 arch/x86/include/asm/cmpxchg_32.h |  218 ++++++++++++++---------------------
 arch/x86/include/asm/cmpxchg_64.h |  234 ++++++++++++++----------------------
 2 files changed, 177 insertions(+), 275 deletions(-)

diff --git a/arch/x86/include/asm/cmpxchg_32.h b/arch/x86/include/asm/cmpxchg_32.h
index ee1931b..720c3d4 100644
--- a/arch/x86/include/asm/cmpxchg_32.h
+++ b/arch/x86/include/asm/cmpxchg_32.h
@@ -8,14 +8,50 @@
  *       you need to test for the feature in boot_cpu_data.
  */
 
-#define xchg(ptr, v)							\
-	((__typeof__(*(ptr)))__xchg((unsigned long)(v), (ptr), sizeof(*(ptr))))
+extern void __xchg_wrong_size(void);
+
+/*
+ * Note: no "lock" prefix even on SMP: xchg always implies lock anyway
+ * Note 2: xchg has side effect, so that attribute volatile is necessary,
+ *	  but generally the primitive is invalid, *ptr is output argument. --ANK
+ */
 
 struct __xchg_dummy {
 	unsigned long a[100];
 };
 #define __xg(x) ((struct __xchg_dummy *)(x))
 
+#define __xchg(x, ptr, size)						\
+({									\
+ 	__typeof(*(ptr)) __x = (x);					\
+	switch (size) {							\
+ 	case 1:								\
+		asm volatile("xchgb %b0,%1"				\
+			     : "=q" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	case 2:								\
+		asm volatile("xchgw %w0,%1"				\
+			     : "=r" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	case 4:								\
+		asm volatile("xchgl %0,%1"				\
+			     : "=r" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	default:							\
+		__xchg_wrong_size();					\
+	}								\
+	__x;								\
+})
+
+#define xchg(ptr, v)							\
+	__xchg((v), (ptr), sizeof(*ptr))
+
 /*
  * The semantics of XCHGCMP8B are a bit strange, this is why
  * there is a loop and the loading of %%eax and %%edx has to
@@ -71,57 +107,63 @@ static inline void __set_64bit_var(unsigned long long *ptr,
 		       (unsigned int)((value) >> 32))			\
 	 : __set_64bit(ptr, ll_low((value)), ll_high((value))))
 
-/*
- * Note: no "lock" prefix even on SMP: xchg always implies lock anyway
- * Note 2: xchg has side effect, so that attribute volatile is necessary,
- *	  but generally the primitive is invalid, *ptr is output argument. --ANK
- */
-static inline unsigned long __xchg(unsigned long x, volatile void *ptr,
-				   int size)
-{
-	switch (size) {
-	case 1:
-		asm volatile("xchgb %b0,%1"
-			     : "=q" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	case 2:
-		asm volatile("xchgw %w0,%1"
-			     : "=r" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	case 4:
-		asm volatile("xchgl %0,%1"
-			     : "=r" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	}
-	return x;
-}
+extern void __cmpxchg_wrong_size(void);
 
 /*
  * Atomic compare and exchange.  Compare OLD with MEM, if identical,
  * store NEW in MEM.  Return the initial value in MEM.  Success is
  * indicated by comparing RETURN with OLD.
  */
+#define __raw_cmpxchg(ptr, old, new, size, lock)			\
+({									\
+ 	__typeof__(*(ptr)) __ret;					\
+ 	__typeof__(*(ptr)) __old = (old);				\
+ 	__typeof__(*(ptr)) __new = (new);				\
+	switch (size) {							\
+	case 1:								\
+		asm volatile(lock "cmpxchgb %b1,%2"			\
+			     : "=a"(__ret)				\
+			     : "q"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+ 		break;							\
+	case 2:								\
+		asm volatile(lock "cmpxchgw %w1,%2"			\
+			     : "=a"(__ret)				\
+			     : "r"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+ 		break;							\
+	case 4:								\
+		asm volatile(lock "cmpxchgl %1,%2"			\
+			     : "=a"(__ret)				\
+			     : "r"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+ 		break;							\
+	default:							\
+		__cmpxchg_wrong_size();					\
+	}								\
+	__ret;								\
+})
+
+#define __cmpxchg(ptr, old, new, size)					\
+	__raw_cmpxchg((ptr), (old), (new), (size), LOCK_PREFIX)
+
+#define __sync_cmpxchg(ptr, old, new, size)				\
+	__raw_cmpxchg((ptr), (old), (new), (size), "lock; ")
+
+#define __cmpxchg_local(ptr, old, new, size)				\
+	__raw_cmpxchg((ptr), (old), (new), (size), "")
 
 #ifdef CONFIG_X86_CMPXCHG
 #define __HAVE_ARCH_CMPXCHG 1
-#define cmpxchg(ptr, o, n)						\
-	((__typeof__(*(ptr)))__cmpxchg((ptr), (unsigned long)(o),	\
-				       (unsigned long)(n),		\
-				       sizeof(*(ptr))))
-#define sync_cmpxchg(ptr, o, n)						\
-	((__typeof__(*(ptr)))__sync_cmpxchg((ptr), (unsigned long)(o),	\
-					    (unsigned long)(n),		\
-					    sizeof(*(ptr))))
-#define cmpxchg_local(ptr, o, n)					\
-	((__typeof__(*(ptr)))__cmpxchg_local((ptr), (unsigned long)(o),	\
-					     (unsigned long)(n),	\
-					     sizeof(*(ptr))))
+
+#define cmpxchg(ptr, old, new)						\
+	__cmpxchg((ptr), (old), (new), sizeof(*ptr))
+
+#define sync_cmpxchg(ptr, old, new)					\
+	__sync_cmpxchg((ptr), (old), (new), sizeof(*ptr))
+
+#define cmpxchg_local(ptr, old, new)					\
+	__cmpxchg_local((ptr), (old), (new), sizeof(*ptr))
 #endif
 
 #ifdef CONFIG_X86_CMPXCHG64
@@ -133,94 +175,6 @@ static inline unsigned long __xchg(unsigned long x, volatile void *ptr,
 					       (unsigned long long)(n)))
 #endif
 
-static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old,
-				      unsigned long new, int size)
-{
-	unsigned long prev;
-	switch (size) {
-	case 1:
-		asm volatile(LOCK_PREFIX "cmpxchgb %b1,%2"
-			     : "=a"(prev)
-			     : "q"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 2:
-		asm volatile(LOCK_PREFIX "cmpxchgw %w1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 4:
-		asm volatile(LOCK_PREFIX "cmpxchgl %1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	}
-	return old;
-}
-
-/*
- * Always use locked operations when touching memory shared with a
- * hypervisor, since the system may be SMP even if the guest kernel
- * isn't.
- */
-static inline unsigned long __sync_cmpxchg(volatile void *ptr,
-					   unsigned long old,
-					   unsigned long new, int size)
-{
-	unsigned long prev;
-	switch (size) {
-	case 1:
-		asm volatile("lock; cmpxchgb %b1,%2"
-			     : "=a"(prev)
-			     : "q"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 2:
-		asm volatile("lock; cmpxchgw %w1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 4:
-		asm volatile("lock; cmpxchgl %1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	}
-	return old;
-}
-
-static inline unsigned long __cmpxchg_local(volatile void *ptr,
-					    unsigned long old,
-					    unsigned long new, int size)
-{
-	unsigned long prev;
-	switch (size) {
-	case 1:
-		asm volatile("cmpxchgb %b1,%2"
-			     : "=a"(prev)
-			     : "q"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 2:
-		asm volatile("cmpxchgw %w1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 4:
-		asm volatile("cmpxchgl %1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	}
-	return old;
-}
-
 static inline unsigned long long __cmpxchg64(volatile void *ptr,
 					     unsigned long long old,
 					     unsigned long long new)
diff --git a/arch/x86/include/asm/cmpxchg_64.h b/arch/x86/include/asm/cmpxchg_64.h
index 52de72e..d565670 100644
--- a/arch/x86/include/asm/cmpxchg_64.h
+++ b/arch/x86/include/asm/cmpxchg_64.h
@@ -3,9 +3,6 @@
 
 #include <asm/alternative.h> /* Provides LOCK_PREFIX */
 
-#define xchg(ptr, v) ((__typeof__(*(ptr)))__xchg((unsigned long)(v), \
-						 (ptr), sizeof(*(ptr))))
-
 #define __xg(x) ((volatile long *)(x))
 
 static inline void set_64bit(volatile unsigned long *ptr, unsigned long val)
@@ -15,167 +12,118 @@ static inline void set_64bit(volatile unsigned long *ptr, unsigned long val)
 
 #define _set_64bit set_64bit
 
+extern void __xchg_wrong_size(void);
+extern void __cmpxchg_wrong_size(void);
+
 /*
  * Note: no "lock" prefix even on SMP: xchg always implies lock anyway
  * Note 2: xchg has side effect, so that attribute volatile is necessary,
  *	  but generally the primitive is invalid, *ptr is output argument. --ANK
  */
-static inline unsigned long __xchg(unsigned long x, volatile void *ptr,
-				   int size)
-{
-	switch (size) {
-	case 1:
-		asm volatile("xchgb %b0,%1"
-			     : "=q" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	case 2:
-		asm volatile("xchgw %w0,%1"
-			     : "=r" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	case 4:
-		asm volatile("xchgl %k0,%1"
-			     : "=r" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	case 8:
-		asm volatile("xchgq %0,%1"
-			     : "=r" (x)
-			     : "m" (*__xg(ptr)), "0" (x)
-			     : "memory");
-		break;
-	}
-	return x;
-}
+#define __xchg(x, ptr, size)						\
+({									\
+ 	__typeof(*(ptr)) __x = (x);					\
+	switch (size) {							\
+ 	case 1:								\
+		asm volatile("xchgb %b0,%1"				\
+			     : "=q" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	case 2:								\
+		asm volatile("xchgw %w0,%1"				\
+			     : "=r" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	case 4:								\
+		asm volatile("xchgl %k0,%1"				\
+			     : "=r" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	case 8:								\
+		asm volatile("xchgq %0,%1"				\
+			     : "=r" (__x)				\
+			     : "m" (*__xg(ptr)), "0" (__x)		\
+			     : "memory");				\
+		break;							\
+	default:							\
+		__xchg_wrong_size();					\
+	}								\
+	__x;								\
+})
+
+#define xchg(ptr, v)							\
+	__xchg((v), (ptr), sizeof(*ptr))
+
+#define __HAVE_ARCH_CMPXCHG 1
 
 /*
  * Atomic compare and exchange.  Compare OLD with MEM, if identical,
  * store NEW in MEM.  Return the initial value in MEM.  Success is
  * indicated by comparing RETURN with OLD.
  */
+#define __raw_cmpxchg(ptr, old, new, size, lock)			\
+({									\
+ 	__typeof__(*(ptr)) __ret;					\
+ 	__typeof__(*(ptr)) __old = (old);				\
+ 	__typeof__(*(ptr)) __new = (new);				\
+	switch (size) {							\
+	case 1:								\
+		asm volatile(lock "cmpxchgb %b1,%2"			\
+			     : "=a"(__ret)				\
+			     : "q"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+ 		break;							\
+	case 2:								\
+		asm volatile(lock "cmpxchgw %w1,%2"			\
+			     : "=a"(__ret)				\
+			     : "r"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+ 		break;							\
+	case 4:								\
+		asm volatile(lock "cmpxchgl %k1,%2"			\
+			     : "=a"(__ret)				\
+			     : "r"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+ 		break;							\
+	case 8:								\
+		asm volatile(lock "cmpxchgq %1,%2"			\
+			     : "=a"(__ret)				\
+			     : "r"(__new), "m"(*__xg(ptr)), "0"(__old)	\
+			     : "memory");				\
+		break;							\
+	default:							\
+		__cmpxchg_wrong_size();					\
+	}								\
+	__ret;								\
+})
 
-#define __HAVE_ARCH_CMPXCHG 1
+#define __cmpxchg(ptr, old, new, size)					\
+	__raw_cmpxchg((ptr), (old), (new), (size), LOCK_PREFIX)
 
-static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old,
-				      unsigned long new, int size)
-{
-	unsigned long prev;
-	switch (size) {
-	case 1:
-		asm volatile(LOCK_PREFIX "cmpxchgb %b1,%2"
-			     : "=a"(prev)
-			     : "q"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 2:
-		asm volatile(LOCK_PREFIX "cmpxchgw %w1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 4:
-		asm volatile(LOCK_PREFIX "cmpxchgl %k1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 8:
-		asm volatile(LOCK_PREFIX "cmpxchgq %1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	}
-	return old;
-}
+#define __sync_cmpxchg(ptr, old, new, size)				\
+	__raw_cmpxchg((ptr), (old), (new), (size), "lock; ")
 
-/*
- * Always use locked operations when touching memory shared with a
- * hypervisor, since the system may be SMP even if the guest kernel
- * isn't.
- */
-static inline unsigned long __sync_cmpxchg(volatile void *ptr,
-					   unsigned long old,
-					   unsigned long new, int size)
-{
-	unsigned long prev;
-	switch (size) {
-	case 1:
-		asm volatile("lock; cmpxchgb %b1,%2"
-			     : "=a"(prev)
-			     : "q"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 2:
-		asm volatile("lock; cmpxchgw %w1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 4:
-		asm volatile("lock; cmpxchgl %1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	}
-	return old;
-}
+#define __cmpxchg_local(ptr, old, new, size)				\
+	__raw_cmpxchg((ptr), (old), (new), (size), "")
 
-static inline unsigned long __cmpxchg_local(volatile void *ptr,
-					    unsigned long old,
-					    unsigned long new, int size)
-{
-	unsigned long prev;
-	switch (size) {
-	case 1:
-		asm volatile("cmpxchgb %b1,%2"
-			     : "=a"(prev)
-			     : "q"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 2:
-		asm volatile("cmpxchgw %w1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 4:
-		asm volatile("cmpxchgl %k1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	case 8:
-		asm volatile("cmpxchgq %1,%2"
-			     : "=a"(prev)
-			     : "r"(new), "m"(*__xg(ptr)), "0"(old)
-			     : "memory");
-		return prev;
-	}
-	return old;
-}
+#define cmpxchg(ptr, old, new)						\
+	__cmpxchg((ptr), (old), (new), sizeof(*ptr))
+
+#define sync_cmpxchg(ptr, old, new)					\
+	__sync_cmpxchg((ptr), (old), (new), sizeof(*ptr))
+
+#define cmpxchg_local(ptr, old, new)					\
+	__cmpxchg_local((ptr), (old), (new), sizeof(*ptr))
 
-#define cmpxchg(ptr, o, n)						\
-	((__typeof__(*(ptr)))__cmpxchg((ptr), (unsigned long)(o),	\
-				       (unsigned long)(n), sizeof(*(ptr))))
 #define cmpxchg64(ptr, o, n)						\
 ({									\
 	BUILD_BUG_ON(sizeof(*(ptr)) != 8);				\
 	cmpxchg((ptr), (o), (n));					\
 })
-#define cmpxchg_local(ptr, o, n)					\
-	((__typeof__(*(ptr)))__cmpxchg_local((ptr), (unsigned long)(o),	\
-					     (unsigned long)(n),	\
-					     sizeof(*(ptr))))
-#define sync_cmpxchg(ptr, o, n)						\
-	((__typeof__(*(ptr)))__sync_cmpxchg((ptr), (unsigned long)(o),	\
-					    (unsigned long)(n),		\
-					    sizeof(*(ptr))))
+
 #define cmpxchg64_local(ptr, o, n)					\
 ({									\
 	BUILD_BUG_ON(sizeof(*(ptr)) != 8);				\