[PATCH rtems 3/3] bsps/fdt: Make sure data is cache aligned.

Gedare Bloom gedare at rtems.org
Fri Jul 17 15:53:29 UTC 2020


The other 2 BSP-specific patches look fine. I didn't look too closely
at the imx-specific part.

On Thu, Jul 16, 2020 at 11:55 PM Christian Mauderer
<christian.mauderer at embedded-brains.de> wrote:
>
> The cache of the fdt blob is flushed after copy. Therefore it should be
> aligned.
> ---
>  bsps/shared/start/bsp-fdt.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/bsps/shared/start/bsp-fdt.c b/bsps/shared/start/bsp-fdt.c
> index 7e8d8922a8..f55273e4ca 100644
> --- a/bsps/shared/start/bsp-fdt.c
> +++ b/bsps/shared/start/bsp-fdt.c
> @@ -28,10 +28,10 @@
>  #endif
>
>  #ifdef BSP_FDT_BLOB_READ_ONLY
> -static const uint32_t
> +static const uint32_t CPU_STRUCTURE_ALIGNMENT

Typically seen the alignment after the variable decl. Maybe it is
flexible, but we might consider a rule for consistency if both
placements work.

>  bsp_fdt_blob[BSP_FDT_BLOB_SIZE_MAX / sizeof(uint32_t)] = { 0xdeadbeef };
>  #else
> -static uint32_t
> +static uint32_t CPU_STRUCTURE_ALIGNMENT
>  bsp_fdt_blob[BSP_FDT_BLOB_SIZE_MAX / sizeof(uint32_t)];
>  #endif
>
> @@ -48,6 +48,7 @@ void bsp_fdt_copy(const void *src)
>
>    if (s != d) {
>      size_t m = MIN(sizeof(bsp_fdt_blob), fdt_totalsize(src));
> +    size_t aligned_size = ((m-1) | (CPU_CACHE_LINE_BYTES-1)) + 1;

This is a little bit of magic, I guess it works with cache line bytes
a power of 2. Do we have a macro for this?

>      size_t n = (m + sizeof(*d) - 1) / sizeof(*d);
>      size_t i;
>
> @@ -55,7 +56,7 @@ void bsp_fdt_copy(const void *src)
>        d[i] = s[i];
>      }
>
> -    rtems_cache_flush_multiple_data_lines(d, m);
> +    rtems_cache_flush_multiple_data_lines(d, aligned_size);
>    }
>  }
>
> --
> 2.26.2
>
> _______________________________________________
> devel mailing list
> devel at rtems.org
> http://lists.rtems.org/mailman/listinfo/devel


More information about the devel mailing list