[PATCH] RADEONCopySwap: Fix RADEON_HOST_DATA_SWAP_16BIT case.

[Benjamin Herrenschmidt]

On Mon, 2012-04-16 at 15:39 +0200, Michel Dänzer wrote:
> From: Michel Dänzer <michel.daenzer at amd.com>
> 
> It was the same code as for RADEON_HOST_DATA_SWAP_32BIT. This caused bus errors
> on FreeBSD/PPC, but I'm not sure how it could not cause problems anywhere...

What kind of bus errors ? Alignment faults ? I think we go lucky the
kernel fixed them up on Linux.... at a performance penalty :-)

Cheers,
Ben.

> Reported-by: Andreas Tobler <andreast at fgznet.ch>
> Signed-off-by: Michel Dänzer <michel.daenzer at amd.com>
> ---
>  src/radeon_accel.c |    5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/src/radeon_accel.c b/src/radeon_accel.c
> index 1cebcf6..784502d 100644
> --- a/src/radeon_accel.c
> +++ b/src/radeon_accel.c
> @@ -982,10 +982,9 @@ void RADEONCopySwap(uint8_t *dst, uint8_t *src, unsigned int size, int swap)
>  
>  	    for (; nwords > 0; --nwords, ++d, ++s)
>  #ifdef __powerpc__
> -		asm volatile("stwbrx %0,0,%1" : : "r" (*s), "r" (d));
> +		asm volatile("sthbrx %0,0,%1" : : "r" (*s), "r" (d));
>  #else
> -	        *d = ((*s >> 24) & 0xff) | ((*s >> 8) & 0xff00)
> -			| ((*s & 0xff00) << 8) | ((*s & 0xff) << 24);
> +	        *d = (*s >> 8) | (*s << 8);
>  #endif
>  	    return;
>  	}