[U-Boot] [PATCH 06/42] Use separate options for TPL support

[Simon Glass]

Hi Masahiro,

On 24 August 2016 at 20:20, Masahiro Yamada
<yamada.masahiro at socionext.com> wrote:
> 2016-08-25 1:51 GMT+09:00 Simon Glass <sjg at chromium.org>:
>> At present TPL uses the same options as SPL support. In a few cases the board
>> config enables or disables the SPL options depending on whether
>> CONFIG_TPL_BUILD is defined.
>>
>> With the move to Kconfig, options are determined for the whole build and
>> (without a hack like an #undef in a header file) cannot be controlled in this
>> way.
>>
>> Create new TPL options for these and update users. This will allow Kconfig
>> conversion to proceed for these boards.
>>
>> Signed-off-by: Simon Glass <sjg at chromium.org>
>> ---
>>
>>  common/Makefile                | 12 +++++++++++-
>>  drivers/Makefile               | 16 +++++++++++++++-
>>  include/common.h               | 15 +++++++++------
>>  include/configs/C29XPCIE.h     | 16 ++++++++--------
>>  include/configs/P1010RDB.h     | 14 +++++++-------
>>  include/configs/P1022DS.h      | 14 +++++++-------
>>  include/configs/p1_p2_rdb_pc.h | 14 +++++++-------
>>  lib/Makefile                   |  9 +++++++--
>>  scripts/Makefile.spl           | 10 +++++++++-
>>  9 files changed, 80 insertions(+), 40 deletions(-)
>>
>> diff --git a/common/Makefile b/common/Makefile
>> index 21619b3..9a9a065 100644
>> --- a/common/Makefile
>> +++ b/common/Makefile
>> @@ -99,10 +99,16 @@ obj-$(CONFIG_SPL_USB_SUPPORT) += usb.o usb_hub.o
>>  obj-$(CONFIG_USB_STORAGE) += usb_storage.o
>>  endif
>>  # environment
>> -ifdef CONFIG_SPL_ENV_SUPPORT
>> +ifdef CONFIG_TPL_BUILD
>> +obj-$(CONFIG_TPL_ENV_SUPPORT) += env_attr.o
>> +obj-$(CONFIG_TPL_ENV_SUPPORT) += env_flags.o
>> +obj-$(CONFIG_TPL_ENV_SUPPORT) += env_callback.o
>> +else
>>  obj-$(CONFIG_SPL_ENV_SUPPORT) += env_attr.o
>>  obj-$(CONFIG_SPL_ENV_SUPPORT) += env_flags.o
>>  obj-$(CONFIG_SPL_ENV_SUPPORT) += env_callback.o
>> +endif
>> +ifneq ($(CONFIG_TPL_ENV_SUPPORT)$(CONFIG_SPL_ENV_SUPPORT),)
>>  obj-$(CONFIG_ENV_IS_NOWHERE) += env_nowhere.o
>>  obj-$(CONFIG_ENV_IS_IN_MMC) += env_mmc.o
>>  obj-$(CONFIG_ENV_IS_IN_FAT) += env_fat.o
>> @@ -123,7 +129,11 @@ obj-$(CONFIG_SPD_EEPROM) += ddr_spd.o
>>  obj-$(CONFIG_HWCONFIG) += hwconfig.o
>>  obj-$(CONFIG_BOUNCE_BUFFER) += bouncebuf.o
>>  ifdef CONFIG_SPL_BUILD
>> +ifdef CONFIG_TPL_BUILD
>> +obj-$(CONFIG_TPL_SERIAL_SUPPORT) += console.o
>> +else
>>  obj-$(CONFIG_SPL_SERIAL_SUPPORT) += console.o
>> +endif
>>  else
>>  obj-y += console.o
>>  endif
>> diff --git a/drivers/Makefile b/drivers/Makefile
>> index 7861d34..ca98273 100644
>> --- a/drivers/Makefile
>> +++ b/drivers/Makefile
>> @@ -40,8 +40,22 @@ obj-$(CONFIG_OMAP_USB_PHY) += usb/phy/
>>  obj-$(CONFIG_SPL_SATA_SUPPORT) += block/
>>  obj-$(CONFIG_SPL_USB_HOST_SUPPORT) += block/
>>  obj-$(CONFIG_SPL_MMC_SUPPORT) += block/
>> +endif
>> +
>> +ifdef CONFIG_TPL_BUILD
>> +
>> +obj-$(CONFIG_TPL_I2C_SUPPORT) += i2c/
>> +obj-$(CONFIG_TPL_DRIVERS_MISC_SUPPORT) += misc/ sysreset/
>> +obj-$(CONFIG_TPL_MMC_SUPPORT) += mmc/
>> +obj-$(CONFIG_TPL_MPC8XXX_INIT_DDR_SUPPORT) += ddr/fsl/
>> +obj-$(CONFIG_TPL_NAND_SUPPORT) += mtd/nand/
>> +obj-$(CONFIG_TPL_SERIAL_SUPPORT) += serial/
>> +obj-$(CONFIG_TPL_SPI_FLASH_SUPPORT) += mtd/spi/
>> +obj-$(CONFIG_TPL_SPI_SUPPORT) += spi/
>> +
>> +endif
>>
>> -else
>> +ifeq ($(CONFIG_SPL_BUILD)$(CONFIG_TPL_BUILD),)
>>
>>  obj-y += adc/
>>  obj-$(CONFIG_DM_DEMO) += demo/
>> diff --git a/include/common.h b/include/common.h
>
>
>
> I'd say a strong "no" to this patch.
>
> The root cause of this pita is in the booting order.
>
> For CONFIG_SPL=y && CONFIG_TPL=n,
>     Boot ROM ->  SPL  ->  U-Boot proper
>
>
> For CONFIG_SPL=y && CONFIG_TPL=y,
>     Boot ROM ->  SPL  ->  TPL  -> U-Boot proper
>
>
> For the latter case, TPL can be big enough to use
> environments and/or device drivers.
> While, SPL is a really small program
> that is only possible in ad-hoc implementation,
> so that probably enable none of CONFIG_SPL_ options.
>
>
> How about this idea:
> Let's exploit the fact that TPL exists for SoCs
> that load a really small program (4KB or less).
> Swap the order of SPL and TPL like follows:
>
>   Boot ROM ->  TPL  ->  SPL   ->  U-Boot proper
>
> With this change,  TPL is too small to enable luxury features.
> No need to add CONFIG_TPL_ stuff in Kconfig.
>
> Let's think TPL is a short of "Tiny Program Loader" instead of
> "Tertiary Program Loader".
> (SPL is no longer "Secondary" in this case, so a bit funny, though.)

At present SPL is full-featured for many board, but for things that
use TPL, SPL is very small. That is a mismatch. It makes sense I think
that SPL is the full-featured thing in all cases. Thus TPL is the
small thing, if it is needed at all.

That would involve refactoring the boards that use TPL. I think it
could be done though, and the idea has merit.

However, this is certainly not feasible in the same patch that moves
these things to Kconfig. We need to move things forward in stages. If
people want to bisect to find problems they really can't deal with
landing on a patch that moves things to Kconfig and changes the
meaning of the options at the same time.

Also, I notice that even TPL does enable some options (e.g. serial,
NAND), so we can't get away without the TPL options. They exist for a
reason.

Regards,
Simon