Merge tag 'locking-core-2023-05-05' of git://git.kernel.org/pub/scm/l…

…inux/kernel/git/tip/tip

Pull locking updates from Ingo Molnar:

 - Introduce local{,64}_try_cmpxchg() - a slightly more optimal
   primitive, which will be used in perf events ring-buffer code

 - Simplify/modify rwsems on PREEMPT_RT, to address writer starvation

 - Misc cleanups/fixes

* tag 'locking-core-2023-05-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/atomic: Correct (cmp)xchg() instrumentation
  locking/x86: Define arch_try_cmpxchg_local()
  locking/arch: Wire up local_try_cmpxchg()
  locking/generic: Wire up local{,64}_try_cmpxchg()
  locking/atomic: Add generic try_cmpxchg{,64}_local() support
  locking/rwbase: Mitigate indefinite writer starvation
  locking/arch: Rename all internal __xchg() names to __arch_xchg()

arch/alpha/include/asm/cmpxchg.h

@@ -6,15 +6,15 @@
  * Atomic exchange routines.
  */
 
-#define ____xchg(type, args...)		__xchg ## type ## _local(args)
+#define ____xchg(type, args...)		__arch_xchg ## type ## _local(args)
 #define ____cmpxchg(type, args...)	__cmpxchg ## type ## _local(args)
 #include <asm/xchg.h>
 
 #define xchg_local(ptr, x)						\
 ({									\
 	__typeof__(*(ptr)) _x_ = (x);					\
-	(__typeof__(*(ptr))) __xchg_local((ptr), (unsigned long)_x_,	\
-				       sizeof(*(ptr)));			\
+	(__typeof__(*(ptr))) __arch_xchg_local((ptr), (unsigned long)_x_,\
+					       sizeof(*(ptr)));		\
 })
 
 #define arch_cmpxchg_local(ptr, o, n)					\
@@ -34,7 +34,7 @@
 
 #undef ____xchg
 #undef ____cmpxchg
-#define ____xchg(type, args...)		__xchg ##type(args)
+#define ____xchg(type, args...)		__arch_xchg ##type(args)
 #define ____cmpxchg(type, args...)	__cmpxchg ##type(args)
 #include <asm/xchg.h>
 
@@ -48,7 +48,7 @@
 	__typeof__(*(ptr)) _x_ = (x);					\
 	smp_mb();							\
 	__ret = (__typeof__(*(ptr)))					\
-		__xchg((ptr), (unsigned long)_x_, sizeof(*(ptr)));	\
+		__arch_xchg((ptr), (unsigned long)_x_, sizeof(*(ptr)));	\
 	smp_mb();							\
 	__ret;								\
 })

arch/alpha/include/asm/local.h

@@ -52,8 +52,16 @@ static __inline__ long local_sub_return(long i, local_t * l)
 	return result;
 }
 
-#define local_cmpxchg(l, o, n) \
-	(cmpxchg_local(&((l)->a.counter), (o), (n)))
+static __inline__ long local_cmpxchg(local_t *l, long old, long new)
+{
+	return cmpxchg_local(&l->a.counter, old, new);
+}
+
+static __inline__ bool local_try_cmpxchg(local_t *l, long *old, long new)
+{
+	return try_cmpxchg_local(&l->a.counter, (s64 *)old, new);
+}
+
 #define local_xchg(l, n) (xchg_local(&((l)->a.counter), (n)))
 
 /**

arch/arc/include/asm/cmpxchg.h

@@ -85,7 +85,7 @@
  */
 #ifdef CONFIG_ARC_HAS_LLSC
 
-#define __xchg(ptr, val)						\
+#define __arch_xchg(ptr, val)						\
 ({									\
 	__asm__ __volatile__(						\
 	"	ex  %0, [%1]	\n"	/* set new value */	        \
@@ -102,7 +102,7 @@
 									\
 	switch(sizeof(*(_p_))) {					\
 	case 4:								\
-		_val_ = __xchg(_p_, _val_);				\
+		_val_ = __arch_xchg(_p_, _val_);			\
 		break;							\
 	default:							\
 		BUILD_BUG();						\

arch/arm/include/asm/cmpxchg.h

@@ -25,7 +25,8 @@
 #define swp_is_buggy
 #endif
 
-static inline unsigned long __xchg(unsigned long x, volatile void *ptr, int size)
+static inline unsigned long
+__arch_xchg(unsigned long x, volatile void *ptr, int size)
 {
 	extern void __bad_xchg(volatile void *, int);
 	unsigned long ret;
@@ -115,8 +116,8 @@ static inline unsigned long __xchg(unsigned long x, volatile void *ptr, int size
 }
 
 #define arch_xchg_relaxed(ptr, x) ({					\
-	(__typeof__(*(ptr)))__xchg((unsigned long)(x), (ptr),		\
-				   sizeof(*(ptr)));			\
+	(__typeof__(*(ptr)))__arch_xchg((unsigned long)(x), (ptr),	\
+					sizeof(*(ptr)));		\
 })
 
 #include <asm-generic/cmpxchg-local.h>

arch/arm64/include/asm/cmpxchg.h

@@ -62,9 +62,8 @@ __XCHG_CASE( ,  ,  mb_, 64, dmb ish, nop,  , a, l, "memory")
 #undef __XCHG_CASE
 
 #define __XCHG_GEN(sfx)							\
-static __always_inline  unsigned long __xchg##sfx(unsigned long x,	\
-					volatile void *ptr,		\
-					int size)			\
+static __always_inline unsigned long					\
+__arch_xchg##sfx(unsigned long x, volatile void *ptr, int size)		\
 {									\
 	switch (size) {							\
 	case 1:								\
@@ -93,7 +92,7 @@ __XCHG_GEN(_mb)
 ({									\
 	__typeof__(*(ptr)) __ret;					\
 	__ret = (__typeof__(*(ptr)))					\
-		__xchg##sfx((unsigned long)(x), (ptr), sizeof(*(ptr))); \
+		__arch_xchg##sfx((unsigned long)(x), (ptr), sizeof(*(ptr))); \
 	__ret;								\
 })
 

arch/hexagon/include/asm/cmpxchg.h

@@ -9,7 +9,7 @@
 #define _ASM_CMPXCHG_H
 
 /*
- * __xchg - atomically exchange a register and a memory location
+ * __arch_xchg - atomically exchange a register and a memory location
  * @x: value to swap
  * @ptr: pointer to memory
  * @size:  size of the value
@@ -19,8 +19,8 @@
  * Note:  there was an errata for V2 about .new's and memw_locked.
  *
  */
-static inline unsigned long __xchg(unsigned long x, volatile void *ptr,
-				   int size)
+static inline unsigned long
+__arch_xchg(unsigned long x, volatile void *ptr, int size)
 {
 	unsigned long retval;
 
@@ -42,8 +42,8 @@ static inline unsigned long __xchg(unsigned long x, volatile void *ptr,
  * Atomically swap the contents of a register with memory.  Should be atomic
  * between multiple CPU's and within interrupts on the same CPU.
  */
-#define arch_xchg(ptr, v) ((__typeof__(*(ptr)))__xchg((unsigned long)(v), (ptr), \
-	sizeof(*(ptr))))
+#define arch_xchg(ptr, v) ((__typeof__(*(ptr)))__arch_xchg((unsigned long)(v), (ptr), \
+							   sizeof(*(ptr))))
 
 /*
  *  see rt-mutex-design.txt; cmpxchg supposedly checks if *ptr == A and swaps.

arch/ia64/include/asm/cmpxchg.h

@@ -5,7 +5,7 @@
 #include <uapi/asm/cmpxchg.h>
 
 #define arch_xchg(ptr, x)	\
-({(__typeof__(*(ptr))) __xchg((unsigned long) (x), (ptr), sizeof(*(ptr)));})
+({(__typeof__(*(ptr))) __arch_xchg((unsigned long) (x), (ptr), sizeof(*(ptr)));})
 
 #define arch_cmpxchg(ptr, o, n)		cmpxchg_acq((ptr), (o), (n))
 #define arch_cmpxchg64(ptr, o, n)	cmpxchg_acq((ptr), (o), (n))

arch/ia64/include/uapi/asm/cmpxchg.h

@@ -23,7 +23,7 @@
  */
 extern void ia64_xchg_called_with_bad_pointer(void);
 
-#define __xchg(x, ptr, size)						\
+#define __arch_xchg(x, ptr, size)					\
 ({									\
 	unsigned long __xchg_result;					\
 									\
@@ -51,7 +51,7 @@ extern void ia64_xchg_called_with_bad_pointer(void);
 
 #ifndef __KERNEL__
 #define xchg(ptr, x)							\
-({(__typeof__(*(ptr))) __xchg((unsigned long) (x), (ptr), sizeof(*(ptr)));})
+({(__typeof__(*(ptr))) __arch_xchg((unsigned long) (x), (ptr), sizeof(*(ptr)));})
 #endif
 
 /*

arch/loongarch/include/asm/cmpxchg.h

@@ -62,7 +62,7 @@ static inline unsigned int __xchg_small(volatile void *ptr, unsigned int val,
 }
 
 static __always_inline unsigned long
-__xchg(volatile void *ptr, unsigned long x, int size)
+__arch_xchg(volatile void *ptr, unsigned long x, int size)
 {
 	switch (size) {
 	case 1:
@@ -87,7 +87,7 @@ __xchg(volatile void *ptr, unsigned long x, int size)
 	__typeof__(*(ptr)) __res;					\
 									\
 	__res = (__typeof__(*(ptr)))					\
-		__xchg((ptr), (unsigned long)(x), sizeof(*(ptr)));	\
+		__arch_xchg((ptr), (unsigned long)(x), sizeof(*(ptr)));	\
 									\
 	__res;								\
 })

arch/loongarch/include/asm/local.h

@@ -56,8 +56,17 @@ static inline long local_sub_return(long i, local_t *l)
 	return result;
 }
 
-#define local_cmpxchg(l, o, n) \
-	((long)cmpxchg_local(&((l)->a.counter), (o), (n)))
+static inline long local_cmpxchg(local_t *l, long old, long new)
+{
+	return cmpxchg_local(&l->a.counter, old, new);
+}
+
+static inline bool local_try_cmpxchg(local_t *l, long *old, long new)
+{
+	typeof(l->a.counter) *__old = (typeof(l->a.counter) *) old;
+	return try_cmpxchg_local(&l->a.counter, __old, new);
+}
+
 #define local_xchg(l, n) (atomic_long_xchg((&(l)->a), (n)))
 
 /**

arch/m68k/include/asm/cmpxchg.h

@@ -9,7 +9,7 @@
 extern unsigned long __invalid_xchg_size(unsigned long, volatile void *, int);
 
 #ifndef CONFIG_RMW_INSNS
-static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int size)
+static inline unsigned long __arch_xchg(unsigned long x, volatile void * ptr, int size)
 {
 	unsigned long flags, tmp;
 
@@ -40,7 +40,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int siz
 	return x;
 }
 #else
-static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int size)
+static inline unsigned long __arch_xchg(unsigned long x, volatile void * ptr, int size)
 {
 	switch (size) {
 	case 1:
@@ -75,7 +75,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void * ptr, int siz
 }
 #endif
 
-#define arch_xchg(ptr,x) ({(__typeof__(*(ptr)))__xchg((unsigned long)(x),(ptr),sizeof(*(ptr)));})
+#define arch_xchg(ptr,x) ({(__typeof__(*(ptr)))__arch_xchg((unsigned long)(x),(ptr),sizeof(*(ptr)));})
 
 #include <asm-generic/cmpxchg-local.h>
 

arch/mips/include/asm/cmpxchg.h

@@ -68,7 +68,7 @@ extern unsigned long __xchg_small(volatile void *ptr, unsigned long val,
 				  unsigned int size);
 
 static __always_inline
-unsigned long __xchg(volatile void *ptr, unsigned long x, int size)
+unsigned long __arch_xchg(volatile void *ptr, unsigned long x, int size)
 {
 	switch (size) {
 	case 1:
@@ -102,7 +102,7 @@ unsigned long __xchg(volatile void *ptr, unsigned long x, int size)
 		smp_mb__before_llsc();					\
 									\
 	__res = (__typeof__(*(ptr)))					\
-		__xchg((ptr), (unsigned long)(x), sizeof(*(ptr)));	\
+		__arch_xchg((ptr), (unsigned long)(x), sizeof(*(ptr)));	\
 									\
 	smp_llsc_mb();							\
 									\

arch/mips/include/asm/local.h

@@ -94,8 +94,17 @@ static __inline__ long local_sub_return(long i, local_t * l)
 	return result;
 }
 
-#define local_cmpxchg(l, o, n) \
-	((long)cmpxchg_local(&((l)->a.counter), (o), (n)))
+static __inline__ long local_cmpxchg(local_t *l, long old, long new)
+{
+	return cmpxchg_local(&l->a.counter, old, new);
+}
+
+static __inline__ bool local_try_cmpxchg(local_t *l, long *old, long new)
+{
+	typeof(l->a.counter) *__old = (typeof(l->a.counter) *) old;
+	return try_cmpxchg_local(&l->a.counter, __old, new);
+}
+
 #define local_xchg(l, n) (atomic_long_xchg((&(l)->a), (n)))
 
 /**

arch/openrisc/include/asm/cmpxchg.h

@@ -147,8 +147,8 @@ static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old,
 extern unsigned long __xchg_called_with_bad_pointer(void)
 	__compiletime_error("Bad argument size for xchg");
 
-static inline unsigned long __xchg(volatile void *ptr, unsigned long with,
-		int size)
+static inline unsigned long
+__arch_xchg(volatile void *ptr, unsigned long with, int size)
 {
 	switch (size) {
 	case 1:
@@ -163,9 +163,9 @@ static inline unsigned long __xchg(volatile void *ptr, unsigned long with,
 
 #define arch_xchg(ptr, with) 						\
 	({								\
-		(__typeof__(*(ptr))) __xchg((ptr),			\
-					    (unsigned long)(with),	\
-					    sizeof(*(ptr)));		\
+		(__typeof__(*(ptr))) __arch_xchg((ptr),			\
+						 (unsigned long)(with),	\
+						 sizeof(*(ptr)));	\
 	})
 
 #endif /* __ASM_OPENRISC_CMPXCHG_H */

arch/parisc/include/asm/cmpxchg.h

@@ -22,7 +22,7 @@ extern unsigned long __xchg64(unsigned long, volatile unsigned long *);
 
 /* optimizer better get rid of switch since size is a constant */
 static inline unsigned long
-__xchg(unsigned long x, volatile void *ptr, int size)
+__arch_xchg(unsigned long x, volatile void *ptr, int size)
 {
 	switch (size) {
 #ifdef CONFIG_64BIT
@@ -49,7 +49,7 @@ __xchg(unsigned long x, volatile void *ptr, int size)
 	__typeof__(*(ptr)) __ret;					\
 	__typeof__(*(ptr)) _x_ = (x);					\
 	__ret = (__typeof__(*(ptr)))					\
-		__xchg((unsigned long)_x_, (ptr), sizeof(*(ptr)));	\
+		__arch_xchg((unsigned long)_x_, (ptr), sizeof(*(ptr)));	\
 	__ret;								\
 })
 

arch/powerpc/include/asm/cmpxchg.h

@@ -229,7 +229,7 @@ __xchg_local(void *ptr, unsigned long x, unsigned int size)
 		return __xchg_u64_local(ptr, x);
 #endif
 	}
-	BUILD_BUG_ON_MSG(1, "Unsupported size for __xchg");
+	BUILD_BUG_ON_MSG(1, "Unsupported size for __xchg_local");
 	return x;
 }
 
@@ -248,7 +248,7 @@ __xchg_relaxed(void *ptr, unsigned long x, unsigned int size)
 		return __xchg_u64_relaxed(ptr, x);
 #endif
 	}
-	BUILD_BUG_ON_MSG(1, "Unsupported size for __xchg_local");
+	BUILD_BUG_ON_MSG(1, "Unsupported size for __xchg_relaxed");
 	return x;
 }
 #define arch_xchg_local(ptr,x)						     \

arch/powerpc/include/asm/local.h

@@ -90,6 +90,17 @@ static __inline__ long local_cmpxchg(local_t *l, long o, long n)
 	return t;
 }
 
+static __inline__ bool local_try_cmpxchg(local_t *l, long *po, long n)
+{
+	long o = *po, r;
+
+	r = local_cmpxchg(l, o, n);
+	if (unlikely(r != o))
+		*po = r;
+
+	return likely(r == o);
+}
+
 static __inline__ long local_xchg(local_t *l, long n)
 {
 	long t;

arch/riscv/include/asm/atomic.h

@@ -261,7 +261,7 @@ c_t arch_atomic##prefix##_xchg_release(atomic##prefix##_t *v, c_t n)	\
 static __always_inline							\
 c_t arch_atomic##prefix##_xchg(atomic##prefix##_t *v, c_t n)		\
 {									\
-	return __xchg(&(v->counter), n, size);				\
+	return __arch_xchg(&(v->counter), n, size);			\
 }									\
 static __always_inline							\
 c_t arch_atomic##prefix##_cmpxchg_relaxed(atomic##prefix##_t *v,	\

arch/riscv/include/asm/cmpxchg.h

@@ -114,7 +114,7 @@
 					    _x_, sizeof(*(ptr)));	\
 })
 
-#define __xchg(ptr, new, size)						\
+#define __arch_xchg(ptr, new, size)					\
 ({									\
 	__typeof__(ptr) __ptr = (ptr);					\
 	__typeof__(new) __new = (new);					\
@@ -143,7 +143,7 @@
 #define arch_xchg(ptr, x)						\
 ({									\
 	__typeof__(*(ptr)) _x_ = (x);					\
-	(__typeof__(*(ptr))) __xchg((ptr), _x_, sizeof(*(ptr)));	\
+	(__typeof__(*(ptr))) __arch_xchg((ptr), _x_, sizeof(*(ptr)));	\
 })
 
 #define xchg32(ptr, x)							\