Re: [PATCH 2/2] locking/xchg/alpha: Add leading smp_mb() to xchg(), cmpxchg()

[Paul E. McKenney]

On Thu, Feb 22, 2018 at 10:24:48AM +0100, Andrea Parri wrote:
> Successful RMW operations are supposed to be fully ordered, but
> Alpha's xchg() and cmpxchg() do not align to this requirement.
> 
> Will reported that:
> 
> > So MP using xchg:
> >
> > WRITE_ONCE(x, 1)
> > xchg(y, 1)
> >
> > smp_load_acquire(y) == 1
> > READ_ONCE(x) == 0
> >
> > would be allowed.
> 
> (thus violating the above requirement).  Amend this by adding a
> leading smp_mb() to the implementations of xchg(), cmpxchg().
> 
> Reported-by: Will Deacon <will.deacon@xxxxxxx>
> Signed-off-by: Andrea Parri <parri.andrea@xxxxxxxxx>

Acked-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>

> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
> Cc: Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>
> Cc: Ivan Kokshaysky <ink@xxxxxxxxxxxxxxxxxxxx>
> Cc: Matt Turner <mattst88@xxxxxxxxx>
> Cc: Richard Henderson <rth@xxxxxxxxxxx>
> Cc: linux-alpha@xxxxxxxxxxxxxxx
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> ---
>  arch/alpha/include/asm/xchg.h | 21 ++++++++++++++++++---
>  1 file changed, 18 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/alpha/include/asm/xchg.h b/arch/alpha/include/asm/xchg.h
> index e1facf6fc2446..e2b59fac5257d 100644
> --- a/arch/alpha/include/asm/xchg.h
> +++ b/arch/alpha/include/asm/xchg.h
> @@ -12,6 +12,10 @@
>   * Atomic exchange.
>   * Since it can be used to implement critical sections
>   * it must clobber "memory" (also for interrupts in UP).
> + *
> + * The leading and the trailing memory barriers guarantee that these
> + * operations are fully ordered.
> + *
>   */
> 
>  static inline unsigned long
> @@ -19,6 +23,7 @@ ____xchg(_u8, volatile char *m, unsigned long val)
>  {
>  	unsigned long ret, tmp, addr64;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"	andnot	%4,7,%3\n"
>  	"	insbl	%1,%4,%1\n"
> @@ -43,6 +48,7 @@ ____xchg(_u16, volatile short *m, unsigned long val)
>  {
>  	unsigned long ret, tmp, addr64;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"	andnot	%4,7,%3\n"
>  	"	inswl	%1,%4,%1\n"
> @@ -67,6 +73,7 @@ ____xchg(_u32, volatile int *m, unsigned long val)
>  {
>  	unsigned long dummy;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"1:	ldl_l %0,%4\n"
>  	"	bis $31,%3,%1\n"
> @@ -87,6 +94,7 @@ ____xchg(_u64, volatile long *m, unsigned long val)
>  {
>  	unsigned long dummy;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"1:	ldq_l %0,%4\n"
>  	"	bis $31,%3,%1\n"
> @@ -128,9 +136,12 @@ ____xchg(, volatile void *ptr, unsigned long x, int size)
>   * store NEW in MEM.  Return the initial value in MEM.  Success is
>   * indicated by comparing RETURN with OLD.
>   *
> - * The memory barrier is placed in SMP unconditionally, in order to
> - * guarantee that dependency ordering is preserved when a dependency
> - * is headed by an unsuccessful operation.
> + * The leading and the trailing memory barriers guarantee that these
> + * operations are fully ordered.
> + *
> + * The trailing memory barrier is placed in SMP unconditionally, in
> + * order to guarantee that dependency ordering is preserved when a
> + * dependency is headed by an unsuccessful operation.
>   */
> 
>  static inline unsigned long
> @@ -138,6 +149,7 @@ ____cmpxchg(_u8, volatile char *m, unsigned char old, unsigned char new)
>  {
>  	unsigned long prev, tmp, cmp, addr64;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"	andnot	%5,7,%4\n"
>  	"	insbl	%1,%5,%1\n"
> @@ -165,6 +177,7 @@ ____cmpxchg(_u16, volatile short *m, unsigned short old, unsigned short new)
>  {
>  	unsigned long prev, tmp, cmp, addr64;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"	andnot	%5,7,%4\n"
>  	"	inswl	%1,%5,%1\n"
> @@ -192,6 +205,7 @@ ____cmpxchg(_u32, volatile int *m, int old, int new)
>  {
>  	unsigned long prev, cmp;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"1:	ldl_l %0,%5\n"
>  	"	cmpeq %0,%3,%1\n"
> @@ -215,6 +229,7 @@ ____cmpxchg(_u64, volatile long *m, unsigned long old, unsigned long new)
>  {
>  	unsigned long prev, cmp;
> 
> +	smp_mb();
>  	__asm__ __volatile__(
>  	"1:	ldq_l %0,%5\n"
>  	"	cmpeq %0,%3,%1\n"
> -- 
> 2.7.4
>