locking/atomic/x86: Introduce arch_try_cmpxchg64

Introduce arch_try_cmpxchg64 for 64-bit and 32-bit targets to improve
code using cmpxchg64.  On 64-bit targets, the generated assembly improves
from:

  ab:	89 c8                	mov    %ecx,%eax
  ad:	48 89 4c 24 60       	mov    %rcx,0x60(%rsp)
  b2:	83 e0 fd             	and    $0xfffffffd,%eax
  b5:	89 54 24 64          	mov    %edx,0x64(%rsp)
  b9:	88 44 24 60          	mov    %al,0x60(%rsp)
  bd:	48 89 c8             	mov    %rcx,%rax
  c0:	c6 44 24 62 f2       	movb   $0xf2,0x62(%rsp)
  c5:	48 8b 74 24 60       	mov    0x60(%rsp),%rsi
  ca:	f0 49 0f b1 34 24    	lock cmpxchg %rsi,(%r12)
  d0:	48 39 c1             	cmp    %rax,%rcx
  d3:	75 cf                	jne    a4 <t+0xa4>

to:

  b3:	89 c2                	mov    %eax,%edx
  b5:	48 89 44 24 60       	mov    %rax,0x60(%rsp)
  ba:	83 e2 fd             	and    $0xfffffffd,%edx
  bd:	89 4c 24 64          	mov    %ecx,0x64(%rsp)
  c1:	88 54 24 60          	mov    %dl,0x60(%rsp)
  c5:	c6 44 24 62 f2       	movb   $0xf2,0x62(%rsp)
  ca:	48 8b 54 24 60       	mov    0x60(%rsp),%rdx
  cf:	f0 48 0f b1 13       	lock cmpxchg %rdx,(%rbx)
  d4:	75 d5                	jne    ab <t+0xab>

where a move and a compare after cmpxchg is saved.  The improvements
for 32-bit targets are even more noticeable, because dual-word compare
after cmpxchg8b gets eliminated.

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220515184205.103089-3-ubizjak@gmail.com

Author: ubizjak

arch/x86/include/asm/cmpxchg_32.h

@@ -42,6 +42,9 @@ static inline void set_64bit(volatile u64 *ptr, u64 value)
 #define arch_cmpxchg64_local(ptr, o, n)					\
 	((__typeof__(*(ptr)))__cmpxchg64_local((ptr), (unsigned long long)(o), \
 					       (unsigned long long)(n)))
+#define arch_try_cmpxchg64(ptr, po, n)					\
+	__try_cmpxchg64((ptr), (unsigned long long *)(po), \
+			(unsigned long long)(n))
 #endif
 
 static inline u64 __cmpxchg64(volatile u64 *ptr, u64 old, u64 new)
@@ -70,6 +73,24 @@ static inline u64 __cmpxchg64_local(volatile u64 *ptr, u64 old, u64 new)
 	return prev;
 }
 
+static inline bool __try_cmpxchg64(volatile u64 *ptr, u64 *pold, u64 new)
+{
+	bool success;
+	u64 old = *pold;
+	asm volatile(LOCK_PREFIX "cmpxchg8b %[ptr]"
+		     CC_SET(z)
+		     : CC_OUT(z) (success),
+		       [ptr] "+m" (*ptr),
+		       "+A" (old)
+		     : "b" ((u32)new),
+		       "c" ((u32)(new >> 32))
+		     : "memory");
+
+	if (unlikely(!success))
+		*pold = old;
+	return success;
+}
+
 #ifndef CONFIG_X86_CMPXCHG64
 /*
  * Building a kernel capable running on 80386 and 80486. It may be necessary

arch/x86/include/asm/cmpxchg_64.h

@@ -19,6 +19,12 @@ static inline void set_64bit(volatile u64 *ptr, u64 val)
 	arch_cmpxchg_local((ptr), (o), (n));				\
 })
 
+#define arch_try_cmpxchg64(ptr, po, n)					\
+({									\
+	BUILD_BUG_ON(sizeof(*(ptr)) != 8);				\
+	arch_try_cmpxchg((ptr), (po), (n));				\
+})
+
 #define system_has_cmpxchg_double() boot_cpu_has(X86_FEATURE_CX16)
 
 #endif /* _ASM_X86_CMPXCHG_64_H */