[U-Boot] [PATCH 1/4] ARM: fix u-boot.lds for -ffunction-sections/-fdata-sections

[Stephen Warren]

On 10/18/2012 02:58 PM, Stephen Warren wrote:
> On 10/18/2012 02:36 PM, Albert ARIBAUD wrote:
>> Hi Stephen,
>>
>> On Wed, 17 Oct 2012 21:17:45 -0600, Stephen Warren
>> <swarren at wwwdotorg.org> wrote:
>>
>>> On 10/17/2012 05:58 PM, Simon Glass wrote:
>>>> Hi Stephen,
>>>>
>>>> On Tue, Oct 16, 2012 at 2:50 PM, Stephen Warren <swarren at wwwdotorg.org> wrote:
>>>>> From: Stephen Warren <swarren at nvidia.com>
>>>>>
>>>>> When -ffunction-sections or -fdata-section are used, symbols are placed
>>>>> into sections such as .data.eserial1_device and .bss.serial_current.
>>>>> Update the linker script to explicitly include these. Without this
>>>>> change (at least with my gcc-4.5.3 built using crosstool-ng), I see that
>>>>> the sections do end up being included, but __bss_end__ gets set to the
>>>>> same value as __bss_start.
>>>>>
>>>>> Signed-off-by: Stephen Warren <swarren at nvidia.com>
>>>>> ---
>>>>> This series fixes an SPL size overflow problem on Tegra. Tom Warren is
>>>>> out on vacation until Oct 25th, so he certainly won't be able to review
>>>>> this. Perhaps it could be applied directly to the ARM tree if enough
>>>>> Tegra people ack the series?
>>>>>
>>>>> Note that this series is not enough to make Tegra support work; either
>>>>> you must hack ./arch/arm/cpu/arm720t/tegra-common/spl.c to call
>>>>> serial_initialize() right before serial_init() in preloader_console_init()
>>>>> or wait for Allen Martin to rework Tegra's SPL support using the common
>>>>> SPL code.
>>>>
>>>> Are you going to submit a patch to enable function-sections, or is
>>>> that a separate discussion?
>>>
>>> For the SPL on Tegra, those flags were already on; this patch fixes a
>>> bug rather than prepares for new functionality.
>>>
>>>>> diff --git a/arch/arm/cpu/u-boot.lds b/arch/arm/cpu/u-boot.lds
>>>
>>>>> -       .u_boot_cmd : { *(.u_boot_cmd) }
>>>>> +       .u_boot_cmd : { *(.u_boot_cmd*) }
>>>>
>>>> I don't think this line is needed?
>>>>
>>> ...
>>>>> -               *(.dynsym)
>>>>> +               *(.dynsym*)
>>>>
>>>> Nor this one?
>>>
>>> Possibly. I changed all the section names to be future-proof. Perhaps a
>>> more targeted patch is warranted.
>>
>> Has this been (at least build-)tested on all boards which have
>> -ffunction-sections or -fdata-sections?
> 
> Yes; those flags both appear to be turned on when building the SPL for
> Tegra (although strangely, not when building the main U-Boot):
> 
> xxx-gcc -M -g  -Os   -fno-common -ffixed-r8 -msoft-float  -D__KERNEL__
> -ffunction-sections -fdata-sections -DCONFIG_SYS_TEXT_BASE=0x0010c000
> -DCONFIG_SPL_TEXT_BASE=0x00108000 -DCONFIG_SPL_BUILD
> -I/home/swarren/shared/git_wa/u-boot/include -fno-builtin -ffreestanding
> -nostdinc -isystem xxx -pipe -DCONFIG_ARM -D__ARM__ -marm
> -mno-thumb-interwork -mabi=aapcs-linux -march=armv4 -mtune=arm7tdmi -MQ
> xxx/spl/arch/arm/cpu/arm720t/interrupts.o interrupts.c >
> xxx/spl/arch/arm/cpu/arm720t/.depend.interrupts

I tried to test this more widely. I found that only
arch/arm/cpu/ixp/config.mk enables this on ARM, with the comment:

> # -fdata-sections triggers "section .bss overlaps section .rel.dyn" linker error
> PLATFORM_RELFLAGS += -ffunction-sections
> LDFLAGS_u-boot += --gc-sections

Indeed, if I edit arch/arm/cpu/armv7/config.mk to enable both
-ffunction-sections -fdata-sections, I do see that same section overlap
error building for Tegra.

Then, if I apply the .lds patch we're discussing, Tegra builds (and
runs) OK with those linker flags enabled. So, this patch seems to fix
two issues not just one - bonus:-)