Re: [PATCH] add big endian version of ld_/st_ IO access macros and convert main 8xx code to use it

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

> e is shared by all architectures. Doing something like
> 
> /*
>  * 8, 16 and 32 bit, big and little endian I/O operations, with barrier.
>  */
> extern inline int in_8(volatile unsigned char __iomem *addr)
> {
>         int ret;
> 
>         __asm__ __volatile__(
>                 "lbz%U1%X1 %0,%1;\n"
> #ifndef CONFIG_8xx
>                 "twi 0,%0,0;\n"
>                 "isync" : "=r" (ret) : "m" (*addr));
> #else
>                 : "=r" (ret) : "m" (*addr));
> #endif
>         return ret;
> }
> 
> Seems somewhat ugly?

Yes. You can work around that by macro-generating the functions with
different rules for different CPUs, or having the twi/isync be a
separate asm block that is itself in a macro or inline that gets defined
differently on 8xx

Ben.