On Wednesday 04 May 2016 20:16:19 Horia Geantă wrote:
> @@ -625,6 +645,16 @@ static inline u32 ioread32be(const volatile void __iomem *addr)
> }
> #endif
>
> +#ifdef CONFIG_64BIT
> +#ifndef ioread64be
> +#define ioread64be ioread64be
> +static inline u64 ioread64be(const volatile void __iomem *addr)
> +{
> + return __be64_to_cpu(__raw_readq(addr));
> +}
> +#endif
> +#endif /* CONFIG_64BIT */
> +
> #ifndef iowrite16be
> #define iowrite16be iowrite16be
> static inline void iowrite16be(u16 value, void volatile __iomem *addr)
> @@ -641,6 +671,16 @@ static inline void iowrite32be(u32 value, volatile void __iomem *addr)
> }
> #endif
>
> +#ifdef CONFIG_64BIT
> +#ifndef iowrite64be
> +#define iowrite64be iowrite64be
> +static inline void iowrite64be(u64 value, volatile void __iomem *addr)
> +{
> + __raw_writeq(__cpu_to_be64(value), addr);
> +}
> +#endif
> +#endif /* CONFIG_64BIT */
> +
>
I just noticed that these two are both a bit wrong, but they copy the
mistake that already exists in the 16 and 32 bit versions: If an
architecture overrides readq/writeq to have barriers but does not override
ioread64be/iowrite64be, this will lack the barriers and behave differently
from the little-endian version. I think the only affected architecture
is ARC, since ARM and ARM64 both override the big-endian accessors to
have the correct barriers, and all others don't use barriers at all.
Maybe you can add a patch before this one to replace the 16/32-bit accessors
with ones that do a
static inline void iowrite32be(u32 value, volatile void __iomem *addr)
{
writel(swab32(value), addr);
}
This will lead to a double-swap on architectures that don't override it,
but it will work correctly on all architectures without them having
to override the big-endian accessors.
Aside from that, the patch looks fine.
Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
