Re: [PATCHv2 04/10] x86/spinlock: Replace ACCESS_ONCE with READ_ONCE/ASSIGN_ONCE

[Christian Borntraeger <borntraeger@de.ibm.com>]

Am 25.11.2014 um 13:38 schrieb Christian Borntraeger:
> ACCESS_ONCE does not work reliably on non-scalar types. For
> example gcc 4.6 and 4.7 might remove the volatile tag for such
> accesses during the SRA (scalar replacement of aggregates) step
> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58145)
> 
> Change the spinlock code to replace ACCESS_ONCE with READ_ONCE
> and ASSIGN_ONCE.
> 
> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> ---
>  arch/x86/include/asm/spinlock.h | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/x86/include/asm/spinlock.h b/arch/x86/include/asm/spinlock.h
> index 9295016..af6e673 100644
> --- a/arch/x86/include/asm/spinlock.h
> +++ b/arch/x86/include/asm/spinlock.h
> @@ -92,7 +92,7 @@ static __always_inline void arch_spin_lock(arch_spinlock_t *lock)
>  		unsigned count = SPIN_THRESHOLD;
> 
>  		do {
> -			if (ACCESS_ONCE(lock->tickets.head) == inc.tail)
> +			if (ASSIGN_ONCE(inc.tail, lock->tickets.head))

As Mike pointed out: this should be 
			if (READ_ONCE(lock->tickets.head) == inc.tail)
of course.