locking/atomic/x86: Introduce the read64_nonatomic macro to x86_32 with cx8

As described in commit:

e73c4e34a0e9 ("locking/atomic/x86: Introduce arch_atomic64_read_nonatomic() to x86_32")

the value preload before the CMPXCHG loop does not need to be atomic.

Introduce the read64_nonatomic assembly macro to load the value from a
atomic_t location in a faster non-atomic way and use it in
atomic64_cx8_32.S.

Signed-off-by: Uros Bizjak <[e-mail]>
Signed-off-by: Ingo Molnar <[e-mail]>
Cc: Linus Torvalds <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

arch/x86/lib/atomic64_cx8_32.S

···
 	cmpxchg8b (\reg)
 .endm
 
+.macro read64_nonatomic reg
+	movl (\reg), %eax
+	movl 4(\reg), %edx
+.endm
+
 SYM_FUNC_START(atomic64_read_cx8)
 	read64 %ecx
 	RET
···
 	movl %edx, %edi
 	movl %ecx, %ebp
 
-	read64 %ecx
+	read64_nonatomic %ecx
 1:
 	movl %eax, %ebx
 	movl %edx, %ecx
···
 SYM_FUNC_START(atomic64_\func\()_return_cx8)
 	pushl %ebx
 
-	read64 %esi
+	read64_nonatomic %esi
 1:
 	movl %eax, %ebx
 	movl %edx, %ecx