[U-Boot] Zynq bug and fix by generic board

[Masahiro Yamada]

Hi Michal and Albert,


On Thu, 05 Jun 2014 12:12:37 +0200
Michal Simek <monstr at monstr.eu> wrote:

> Hi,
> 
> On 06/05/2014 10:59 AM, Masahiro Yamada wrote:
> > Hi Albert and Michal,
> > 
> > 
> > On Wed, 4 Jun 2014 12:54:23 +0200
> > Michal Simek <michal.simek at xilinx.com> wrote:
> > 
> >> Hi Albert and Masahiro,
> >>
> >> surprisingly I have tested zynq in Tom's repo and
> >> I am getting stuck in full u-boot.
> >> I have also tried SPL with MMC and it is hanging in the same location
> >>
> >> U-Boot SPL 2014.07-rc2-00063-g3fe1a8545b55 (Jun 04 2014 - 11:57:12)
> >> mmc boot
> >> reading system.dtb
> >> spl_load_image_fat_os: error reading image system.dtb, err - -1
> >> reading u-boot-dtb.img
> >> reading u-boot-dtb.img
> >>
> >>
> >> U-Boot 2014.07-rc2-00063-g3fe1a8545b55 (Jun 04 2014 - 11:57:12)
> >>
> >> I2C:   ready
> >> Memory: ECC disabled
> >> DRAM:  1 GiB
> >>
> >> Problem is with relocation and it is not related to OF_CONTROL.
> >>
> >> I have added 3 pending patches on the top of current master next
> >> and "ARM: zynq: Enable generic board for Xilinx Zynq" fix the problem.
> >> That's why I think that is easier for just apply it to mainline repo
> >> because we need to apply this patch anyway.
> >>
> >> Masahiro: Can you please check if the latest u-boot is working for you?
> >> Albert: If Masahiro confirms the same issue, can I send you one more pull
> >> request with these 3 fixes? (I have also run buildman on them too)
> >> (Enabling generic board, Fix message which is caused by enabling generic board,
> >> + one fix from Masahiro regarding compilation out of tree files for SPL).
> > 
> > 
> > I found how to fix the problem. 
> > 
> > Add one line to arch/arm/cpu/armv7/zynq/u-boot.lds as follows:
> > 
> > 
> >   --- a/arch/arm/cpu/armv7/zynq/u-boot.lds
> >  +++ b/arch/arm/cpu/armv7/zynq/u-boot.lds
> >  @@ -18,6 +18,7 @@ SECTIONS
> >          .text :
> >          {
> >                  *(.__image_copy_start)
> >  +               *(.vectors)
> >                  CPUDIR/start.o (.text*)
> >                  *(.text*)
> >          }
> > 
> > 
> > It  worked on my ZC706 board.
> > 
> 
> Interesting that it is not needed for GENERIC_BOARD but anyway
> this fix is also works for me.

I think it is working with GENERIC_BOARD by coincidence.

"_start" symbol should be 0x04000000 (=CONFIG_SYS_TEXT_BASE)
But it is set to 0  without fixing the linker script.



> Masahiro: Can you please send proper patch which Albert or Tom
> can directly take? Link to origin commit sha will be also useful
> to point where the problem happened.
> 
> Please add this to the patch.
> Tested-by: Michal Simek <monstr at monstr.eu>

Done.
http://patchwork.ozlabs.org/patch/356300/


Albert,
This is a fatal bug. Please pick it up asap.

And please apply the followings as well:
http://patchwork.ozlabs.org/patch/352484/
http://patchwork.ozlabs.org/patch/352672/
They also fix the bugs you added by commit 41623c91.


Best Regards
Masahiro Yamada