Hey Jerzy,

thanks for your help!

I already changed the adress of slot 0 in the bsp but this didn't resolve the problem. I tried using my old linker script, and now the problem counter is at least in the right range. Below is my *.map file:


There are no discarded input sections

Memory Configuration

Name             Origin             Length             Attributes
FLASH            0x08000000         0x000f8000         xr
ITCM             0x00000000         0x00004000         xr
DTCM             0x20000000         0x00010000         xrw
DATA_NO_CACHE    0x20010000         0x00000200         xrw
RAM              0x20010200         0x0003fe00         xrw
SDRAM            0xc0000000         0x00f80800         xrw
*default*        0x00000000         0xffffffff

Linker script and memory map

START GROUP
....
END GROUP
LOAD 
C:/msys64/mingw64/bin/../lib/gcc/arm-none-eabi/14.2.1/thumb/v7e-m/nofp\libgcc.a
START GROUP
LOAD 
C:/msys64/mingw64/bin/../lib/gcc/arm-none-eabi/14.2.1/thumb/v7e-m/nofp\libgcc.a
...
END GROUP
                0x00000000                        _imghdr_size = 0x0

.glue_7         0x08000000        0x0
 .glue_7        0x08000000        0x0 linker stubs

.glue_7t        0x08000000        0x0
 .glue_7t       0x08000000        0x0 linker stubs

.vfp11_veneer   0x08000000        0x0
 .vfp11_veneer  0x08000000        0x0 linker stubs

.v4_bx          0x08000000        0x0
 .v4_bx         0x08000000        0x0 linker stubs

.SDRAM          0xc0000000        0x0
                0xc0000000                        . = ALIGN (0x4)
 *(.SDRAM)

.DATA_NO_CACHE  0x20010000        0x0
                0x20010000                        . = ALIGN (0x4)
                0x20010000                        _s_ram_nocache = .
 *(.ram_nocache)
                0x20010000                        . = ALIGN (0x4)
                0x20010000                        _e_ram_nocache = .
OUTPUT(C:/Users/admin/git/proj/bin/targets/demo/app/apps/iptest/iptest.elf 
elf32-littlearm)
LOAD linker stubs

For testing I changed the flash adress to 0x08000000, in my old setup this worked and let me use the system without bootloader.

Do you have any idea what could be wrong? As always I'm very thankful for tips and hints, I really have to get this going fast..

Regards
Jan

Am 22.01.2025 um 17:52 schrieb Jerzy Kasenberg:
Hi Jan,

I checked some more.
BSP you mentioned has SLOT0 located at 0x08020000, you mentioned 0x08010000
as your starting point so you may check this.

I also checked that if you specify in your target syscfg.yml (or in
application or bsp)
INCLUDE_IMAGE_HEADER: 0
it should have same effect as _imghdr_size = 0x0;
basically it means that image header (32 bytes for MCUBOOT) is not to be
reserved. It can be verified in .map file look for something like that:

.image_header
  *(.image_header)

.interrupts     0x08040000      0x1f8
                 0x08040000                        . = ALIGN (0x4)
                 0x08040000                        __isr_vector = .
                 0x08040000                        __isr_vector_start = .

This was taken from STM32F767 which has slot0 located at 0x8040000
Following fragment is from build that has INCLUDE_IMAGE_HEADER: 1

.image_header   0x08040000       0x20
  *(.image_header)
  .image_header  0x08040000       0x20
H:\msys64\home\Jerzy\work\blinky\bin\targets\nucleo-f767zi-iptest\app\@apache-mynewt-core\mgmt\image_header\@apache-mynewt-core_mgmt_image_header.a(image_header.o)
                 0x08040000                image_header

.interrupts     0x08040020      0x1f8
                 0x08040020                        . = ALIGN (0x4)
                 0x08040020                        __isr_vector = .
                 0x08040020                        __isr_vector_start = .

So you can see that interrupt vectors are moved 32 bytes allowing for newt
to insert image header.

In any case if your custom bootloader jumps to 0x8010000 it should not work
before as well
_imghdr_size = 0 means that there is not space for image header but
interrupt vectors starting from initial SP are at the very beginning of you
build.
Unless you have something special at 0x8010000 your code should jump to
address store in 0x08010004

If your share some fragment of map file I can help you some more.

br
Jerzy

śr., 22 sty 2025 o 16:36 Jerzy Kasenberg <jerzy.kasenb...@codecoup.pl>
napisał(a):

Hi Jan,

Easiest way for you to proceed is to get old linker scripts that worked
for you before
Then instead of having bsp.yml file with line

  bsp.linkerscript: autogenerated

set bsp.linkerscript to what it was before

I'll check your case to see how to manage it using autogenerated scripts
and let you know later

br
Jerzy

śr., 22 sty 2025 o 13:44 Jan Clement <jan.clem...@cm-electronics.de>
napisał(a):

Hello Devlist,

during my process of upgrading apache mynewt from 1.9.0 to 1.13.0 I now
face the problem, that the binaries I generate are not running on my
system.

As a starting point I used the iptest application and the
stm32f7discovery bsp package. This worked with the old mynewt version. I
run my own bootloader implementation, after booting, it jumps to
0x08010000 and starts executing from there. For this to work in the old
implementation I had to set

  > _imghdr_size = 0x0;    /* size = 0 for direct flashing */
in the linker script.

I can't find any corresponding setting in the new version. The only
thing I found is:

  > INCLUDE_IMAGE_HEADER
which I set to 0.

For testing purposes I also tried to flash the image directly and use it
without the bootloader, this also didn't work.

When I debug my application I get following output:

  > xPSR: 00000000 pc: 00000000 msp: 0x96f3b83c

It seems that my program counter is not set correctly.

Any help is greatly appreciated, thanks!

Jan




Reply via email to