Re: [PATCH] arch/powerpc/include/asm/barrier.h: redefine rmb and wmb to lwsync

[Kautuk Consul]

Hi All,

On Wed, Feb 22, 2023 at 11:31:07AM +0530, Kautuk Consul wrote:
>  /* The sub-arch has lwsync */
>  #if defined(CONFIG_PPC64) || defined(CONFIG_PPC_E500MC)
> -#    define SMPWMB      LWSYNC
> +#undef rmb
> +#undef wmb
> +/* Redefine rmb() to lwsync. */
> +#define rmb()	({__asm__ __volatile__ ("lwsync" : : : "memory"); })
> +/* Redefine wmb() to lwsync. */
> +#define wmb()	({__asm__ __volatile__ ("lwsync" : : : "memory"); })
> +#define SMPWMB      LWSYNC
>  #elif defined(CONFIG_BOOKE)
> -#    define SMPWMB      mbar
> +#define SMPWMB      mbar
>  #else
> -#    define SMPWMB      eieio
> +#define SMPWMB      eieio
>  #endif

I think I am conceptually right about this patch but I lack the
resources currently to tets this out on PowerPC 64 bit servers.

I request IBM/Non-IBM employees to test this patch out for:
a) functionality breaking. This patch is no good if this breaks current
   kernel functionality.
b) performance impact. If functionality doesn't break, can anyone do
   some reliable kernel load testing on ppc64 servers to see if there
   is any significant performance gain ?

Thanks a lot!