[U-Boot] [PATCH] spl: add a new stub spl_early_board_init() for early SoC-specific setup

[Simon Glass]

Hi Masahiro,

On 8 March 2016 at 04:37, Masahiro Yamada <yamada.masahiro at socionext.com> wrote:
> We are generally supposed to implement SoC/board-specific SPL init
> code in spl_board_init(), but it is called after spl_init() where the
> FDT is setup and devices are bound.
>
> This new stub spl_early_board_init() would be useful to put something
> really SoC-specific, for example, debug_uart_init().
>
> In fact, I was hit by some problems on FDT setup when I was tackling
> on a completely new platform.  I wished I could use the debug UART
> earlier in situations like that.
>
> Signed-off-by: Masahiro Yamada <yamada.masahiro at socionext.com>
> ---
>
>  common/spl/spl.c | 6 ++++++
>  include/spl.h    | 1 +
>  2 files changed, 7 insertions(+)
>
> diff --git a/common/spl/spl.c b/common/spl/spl.c
> index e5167bf..df85b09 100644
> --- a/common/spl/spl.c
> +++ b/common/spl/spl.c
> @@ -150,6 +150,10 @@ static int spl_ram_load_image(void)
>  }
>  #endif
>
> +void __weak spl_early_board_init(void)
> +{
> +}
> +
>  int spl_init(void)
>  {
>         int ret;
> @@ -344,6 +348,8 @@ void board_init_r(gd_t *dummy1, ulong dummy2)
>  {
>         int i;
>
> +       spl_early_board_init();

Why not put this in board_init_f()? That is a little earlier.

In fact you can replace that function with your own version.

(although it would be better if we had a single board_init_f() and
called out to board-specific code from there.)

> +
>         debug(">>spl:board_init_r()\n");
>

[snip]

Regards,
Simon