[U-Boot] [PATCH] arm: spl: Allow board_init_r() to run with a larger stack

[Simon Glass]

Hi Masahiro,

On 22 January 2015 at 00:33, Masahiro Yamada <yamada.m at jp.panasonic.com> wrote:
> Simon, Heiko,
>
>
> On Thu, 22 Jan 2015 07:28:37 +0100
> Heiko Schocher <hs at denx.de> wrote:
>
>> > But for your particular case, I certainly would like the UART to be
>> > available early in SPL. We are in the very early days of driver model
>> > for SPL. Once we get it merged (thanks to Tom's work this should be
>> > soon) we can look at how to get the UART available early. One hacky
>> > idea is to do the DM scan in board_init_f() in SPL, for serial only.
>> > But it can be done.
>> >
>> > Yes generic board for SPL is where I'd like to head (but hopefully
>> > using driver model for all init). This patch is a step along the way.
>>
>> I know, I am geting annoying, but we should not forget boards/SoCs
>> which have minimal sram ... like the sam9260 with 4k only! So we must
>> have an option to get SPL running there ... to set BSS and stack
>> into RAM is a great option, so we have no problems with that anymore,
>> but there is currently no place to get an UART running on it... so I
>> feel DM is also to big ...
>
>
> Simon,
> Have you ever estimated how much the DM increase the binary image size?
> (for ARM)

I did originally (around 2.2KB for the core) but it has grown since.
When I revisit SPL (soon) I suppose we will get another data point.

One problem is that the new bus functionality is likely not needed in
SPL for things like AT91. I have a SAM9260 board so can try it out.
However, that chip is very old. Are there chips in common use still
which are so limited in SRAM?

Also Masahiro you should badger your hardware guys to add e.g. 128KB
of SRAM to their next chip!

>
> Heiko,
> In your case, perhaps, you can disable the DM for SPL.
> If you do not use UART on SPL, the DM core is not needed either.

Regards,
Simon