[U-Boot] [PATCH 1/4] spl, common, serial: build SPL without serial support

[Heiko Schocher <hs@denx.de>]

Hello Tom,

Am 15.06.2015 17:32, schrieb Tom Rini:
> On Mon, Jun 15, 2015 at 02:20:59PM +0200, Heiko Schocher wrote:
>
>> This patch enables building SPL without
>> CONFIG_SPL_SERIAL_SUPPORT support.
>>
>> Signed-off-by: Heiko Schocher <hs@denx.de>
>
> I like that we're starting down this path.  But can you explain why we
> need each of these changes:

I try it ...

The determining change is in "include/common.h", which twist the
putc, puts, printf and vprintf functions to /dev/null ... so the
compiler/linker can drop all them ... all other changes are only
fixing compiler warnings, as for example vars no longer used ...

>>   arch/arm/lib/interrupts.c          | 4 ++--
>>   arch/arm/mach-at91/arm926ejs/cpu.c | 2 +-
>>   common/cli_readline.c              | 2 +-
>>   common/cli_simple.c                | 2 +-
>>   lib/display_options.c              | 4 ++--
>
> Having gone down the path of disabling output in SPL before due to size
> constraints I didn't need to modify those files (but it was on an older
> version)

Do you have a patch?

>>   common/image.c                     | 8 ++++++--
>
> In this case I just said we only need those functions (and their
> strings!) when we have SPL_SERIAL available.  It may be worth
> re-checking what parts of this file we really need available in SPL
> even.

Hmm... my change is only to get rid of compiler warning "var defined
but not used" with marking this vars with "__maybe_unused" ... so it
is the less intrusive way ...

If we have a header before the kernel image, we need functions from
here in SPL, and maybe FIT is sometime supported/used in SPL ...
maybe someone wants to print header info in SPL before booting it ...
so I do not know if this would be an easy job ...

>>   common/spl/spl.c                   | 9 +++++++++
>
> We should be able to be fine without a preloader_console_init() func I
> think.

I just defined a version for the non serial case, at this place
there is already this function ...

I only set:

         gd->bd = &bdata;
         gd->have_console = 0;

which seems not to bad to me ... ok, "gd->have_console" should be 0.
And why is in preloader_console_init() "gd->bd = &bdata;" ... this
seems not the right place to me, as it has nothing to do with
"console init" ... so maybe we really can drop it, but it gets called
from:

$ grep -lr preloader_console_init .
./arch/microblaze/cpu/spl.c
./arch/powerpc/cpu/ppc4xx/spl_boot.c
./arch/powerpc/cpu/mpc5xxx/spl_boot.c
./arch/arm/mach-davinci/spl.c
./arch/arm/mach-uniphier/spl.c
./arch/arm/mach-mvebu/spl.c
./arch/arm/mach-at91/spl_at91.c
./arch/arm/mach-at91/spl_atmel.c
./arch/arm/mach-tegra/spl.c
./arch/arm/mach-zynq/spl.c
./arch/arm/cpu/armv7/omap-common/boot-common.c
./arch/arm/cpu/armv7/omap-common/boot-common.c.orig
./arch/arm/cpu/armv7/sunxi/board.c
./arch/arm/cpu/armv7/omap3/board.c
./arch/arm/mach-socfpga/spl.c
./20150615/0004-spl-common-serial-build-SPL-without-serial-support.patch
./board/compulab/cm_fx6/spl.c
./board/gateworks/gw_ventana/gw_ventana_spl.c
./board/solidrun/mx6cuboxi/mx6cuboxi.c
./board/barco/platinum/spl_titanium.c
./board/barco/platinum/spl_picon.c
./board/kosagi/novena/novena_spl.c
./board/ti/ks2_evm/board.c
./board/LaCie/edminiv2/edminiv2.c
./board/wandboard/spl.c
./board/freescale/mx6sxsabresd/mx6sxsabresd.c
./board/freescale/ls1021aqds/ls1021aqds.c
./board/freescale/ls1021atwr/ls1021atwr.c
./board/freescale/mx6sabresd/mx6sabresd.c
./board/bachmann/ot1200/ot1200_spl.c
./board/work-microwave/work_92105/work_92105_spl.c
./board/woodburn/woodburn.c

so ... we need a dummy function ... or?

Thanks for your comments!

bye,
Heiko
-- 
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany