[U-Boot] [PATCH] arm: Handle .gnu.hash section in ldscripts

["Andreas Färber" <afaerber@suse.de>]

Hi,

Am 09.03.2014 21:34, schrieb Albert ARIBAUD:
> On Sun, 09 Mar 2014 21:22:34 +0100, Andreas F?rber <afaerber@suse.de>
> wrote:
>> Am 09.03.2014 21:05, schrieb Albert ARIBAUD:
>>> On Mon, 3 Mar 2014 08:00:52 +0100, Albert ARIBAUD
>>> <albert.u.boot@aribaud.net> wrote:
>>>> On Sun, 02 Mar 2014 23:00:19 +0100, Andreas F?rber
>>>> <afaerber@suse.de> wrote:
>>>>> Am 02.03.2014 21:57, schrieb Andreas F?rber:
>>>>>> Am 02.03.2014 20:15, schrieb Albert ARIBAUD:
>>>>>>> On Sun, 02 Mar 2014 18:57:14 +0100, Andreas F?rber <afaerber@suse.de>
>>>>>>> wrote:
>>>>>>>> Am 22.02.2014 14:27, schrieb Albert ARIBAUD:
>>>>>>>>> On Thu, 13 Feb 2014 12:39:07 +0100, Albert ARIBAUD
>>>>>>>>> <albert.u.boot@aribaud.net> wrote:
>>>>>>>>>
>>>>>>>>>> Can you point me to the toolchain you are using, and if you are not
>>>>>>>>>> building mainline code, can you make this code available to me somehow?
>>>>>> [...]
>>>>>>>> Our latest GCC 4.8.2 can be found here:
>>>>>>>> https://build.opensuse.org/package/show/openSUSE:Factory:ARM/gcc48
>>>>>>>>
>>>>>>>> Matching binutils 2.24 are here:
>>>>>>>> https://build.opensuse.org/package/show/openSUSE:Factory:ARM/binutils
>>>>>>>>
>>>>>>>> Our latest U-Boot tarball and patches:
>>>>>>>> https://build.opensuse.org/package/show/Base:System/u-boot
>>>>>>>> Build logs can be found in the u-boot-* packages in the same project,
>>>>>>>> with underscores stripped, e.g.:
>>>>>>>> https://build.opensuse.org/package/show/Base:System/u-boot-am335xevm
>>>>>>>
>>>>>>> Thanks for your reply. Apparently, the problem does not appear for the
>>>>>>> boards in mainline U-Boot which I listed above with the source code
>>>>>>> from our v2014.01 tag and using gcc 4.8.1 from Linaro. I can install
>>>>>>> 4.8.2, but meanwhile, can you try to build v2014.01 for these boards
>>>>>>> with your setup?
>>>>>>>
>>>>>>> am335x_evm colibri_t20_iris highbank mx53loco omap3_beagle
>>>>>>> omap4_panda paz00 rpi_b
>>>>>>
>>>>>> I have commented out the patch in:
>>>>>> https://build.opensuse.org/project/show/home:a_faerber:branches:Base:System
>>>>>>
>>>>>> While there are apparently frequent rebuilds scheduled ATM, it seemed
>>>>>> like the colibri_t20_iris config intermittently built OK without it.
>>>>>> Maybe some gcc 4.8 stable branch update [1] resolved it ... will keep an
>>>>>> eye on the build status to confirm. If everything builds, I'm happy.
>>>>>
>>>>> Nah, still fails without the patch:
>>>>>
>>>>> [  178s] gcc -E -g  -Os   -ffunction-sections -fdata-sections
>>>>> -fno-common -ffixed-r9 -msoft-float  -D__KERNEL__
>>>>> -DCONFIG_SYS_TEXT_BASE=0x0010E000
>>>>> -I/home/abuild/rpmbuild/BUILD/u-boot-2014.01/include
>>>>> -I/home/abuild/rpmbuild/BUILD/u-boot-2014.01/arch/arm/include
>>>>> -fno-builtin -ffreestanding -nostdinc -isystem
>>>>> /usr/lib/gcc/armv7hl-suse-linux-gnueabi/4.8/include -pipe  -DCONFIG_ARM
>>>>> -D__ARM__ -marm -mno-thumb-interwork -mabi=aapcs-linux
>>>>> -mword-relocations -march=armv7-a -include
>>>>> /home/abuild/rpmbuild/BUILD/u-boot-2014.01/include/u-boot/u-boot.lds.h
>>>>> -DCPUDIR=arch/arm/cpu/armv7  -ansi -D__ASSEMBLY__ -P -
>>>>> </home/abuild/rpmbuild/BUILD/u-boot-2014.01/arch/arm/cpu/u-boot.lds
>>>>>> u-boot.lds
>>>>> [  178s] cd /home/abuild/rpmbuild/BUILD/u-boot-2014.01 && ld.bfd  -pie
>>>>> -T u-boot.lds --gc-sections -Bstatic -Ttext 0x0010E000
>>>>> arch/arm/cpu/armv7/start.o --start-group arch/arm/cpu/armv7/built-in.o
>>>>> arch/arm/cpu/armv7/tegra20/built-in.o arch/arm/cpu/built-in.o
>>>>> arch/arm/lib/built-in.o board/toradex/colibri_t20_iris/built-in.o
>>>>> common/built-in.o disk/built-in.o drivers/built-in.o
>>>>> drivers/dma/built-in.o drivers/gpio/built-in.o drivers/i2c/built-in.o
>>>>> drivers/input/built-in.o drivers/mmc/built-in.o drivers/mtd/built-in.o
>>>>> drivers/mtd/nand/built-in.o drivers/mtd/onenand/built-in.o
>>>>> drivers/mtd/spi/built-in.o drivers/net/built-in.o
>>>>> drivers/net/phy/built-in.o drivers/pci/built-in.o
>>>>> drivers/power/battery/built-in.o drivers/power/built-in.o
>>>>> drivers/power/fuel_gauge/built-in.o drivers/power/mfd/built-in.o
>>>>> drivers/power/pmic/built-in.o drivers/serial/built-in.o
>>>>> drivers/spi/built-in.o drivers/usb/eth/built-in.o
>>>>> drivers/usb/gadget/built-in.o drivers/usb/host/built-in.o
>>>>> drivers/usb/musb-new/built-in.o drivers/usb/musb/built-in.o
>>>>> drivers/usb/phy/built-in.o drivers/usb/ulpi/built-in.o fs/built-in.o
>>>>> lib/built-in.o lib/libfdt/built-in.o net/built-in.o test/built-in.o
>>>>> --end-group
>>>>> /home/abuild/rpmbuild/BUILD/u-boot-2014.01/arch/arm/lib/eabi_compat.o
>>>>> -L /usr/lib/gcc/armv7hl-suse-linux-gnueabi/4.8 -lgcc -Map u-boot.map -o
>>>>> u-boot
>>>>> [  178s] ld.bfd: u-boot: could not find output section .gnu.hash
>>>>> [  178s] ld.bfd: final link failed: Nonrepresentable section on output
>>>>> [  178s] Makefile:556: recipe for target 'u-boot' failed
>>>>> [  178s] make: *** [u-boot] Error 1
>>>>
>>>> You are trying this with your codebase, right? If so, can you try and
>>>> build vanilla v2014.01 with the exact same build environment?
>>
>> Commented out all our patches at:
>> https://build.opensuse.org/project/show/home:a_faerber:branches:Base:System
>>
>> Failure of any sunxi-added targets is to be expected, of course.
> 
> Not sure I'm getting this: if you build vanilla v2014.01, there
> shouldn't be any sunxi target built at all, should it?

I removed the patches from the spec file in our build service. That
doesn't drop the packages for those no longer patched-in targets, so
failures of those targets are to be expected. I am not building locally.

Here's a vanilla "highbank" target failing:
https://build.opensuse.org/package/live_build_log/home:a_faerber:branches:Base:System/u-boot-highbank/openSUSE_Factory_ARM/armv7l

>>> If it
>>>> builds fine, then can you tell me which changes your patch(es) made to
>>>> linker scripts in your tree wrt 2014.01?
>>>
>>> Ping.
>>
>> I had previously pointed to our patches and explained that they do not
>> touch on any existing linker scripts. They merely change CONFIG_
>> options, tweak environment settings and add the sunxi targets.
> 
> I am not asking about changing linker scripts; I am asking about
> trying your build environment on the pristine v2014.01 codebase, e.g.
> build all mainline ARM targets with your toolchains.

You did ask above "which changes" our patches "made to linker scripts".
And the answer is: None, apart from the already commented-out
gnuhash.patch, which corresponds to this very patch under discussion.

>> As for the other open question, OBS does not package build
>> intermediaries of failed builds, so I'll need to reproduce it locally.
>> ATM I'm lacking time because QEMU is about to enter Hard Freeze for 2.0.
> 
> Could you just run this in a git clone of U-Boot with v2014.01 checked
> out?
> 
> 	ARCH=arm CROSS_COMPILE=your-toolchain-prefix- ./MAKEALL -a arm
> 
> It would build all targets unattended using your toolchain (assuming it
> is in your PATH of course). 

My local ARM machines are all 13.1, not Factory, but I can try running
that command there later.

Amicalement,
Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 N?rnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imend?rffer; HRB 16746 AG N?rnberg