I got PANIC at PC : 0x2ffec98b - Why?

[DMårt]

I made a custom board with the processor STM32MP151AAC3 and own firmwire https://github.com/DanielMartensson/STM32-Computer/tree/main/STM32CubeIDE/STM32-Computer-Firmware . I have compiled the TF-A by using STM32CubeIDE. It was some bugs with the TF-A configuration, but it seems I fixed them in this thread: 

Solved: BUG: Label or path X_pins_y not found - Compiling ... - STMicroelectronics Community

PANIC at PC : 0x2ffec98b

Exception mode=0x00000016 at: 0x2ffec98b

I have tried to flash it with firmware of STM32MP157 with another I2C4 pin configuration. And it failed. 

NOTICE:  CPU: STM32MP151AAC Rev.Z
NOTICE:  Model: STMicroelectronics STM32MP157A eval daughter on eval mother
ERROR:   I2C device not ready
PANIC at PC : 0x2fff101b

Exception mode=0x00000016 at: 0x2fff101b

Trying the Flashout Delete All DK1 STM32MP157a.
Exception mode=0x00000016 at: 0x2fff101b

NOTICE:  CPU: STM32MP151AAC Rev.Z
NOTICE:  Model: STMicroelectronics STM32MP157A-DK1 Discovery Board
ERROR:   I2C device not ready
PANIC at PC : 0x2fff101b

That means the following:

1. The HSE clock is working
2. The HSE clock is soldered correctly
3. The I2C is working, even if it was not ready for communication (wrong pins)

Question:

This new error must lead to an answer. What can it be?

[PatrickF]

Hi,

please clarify how did you get this message ? Boot from flash or download from CubeProgrammer ?

Few tips (maybe already covered):

- check if HSE setting (digital clock (bypass) or crystal oscillator) is aligned between HW (e.g. OSC_OUT) and TF-A Device Tree.
   e.g. if you are using a crystal, the st,digbypass should not be present in the DT
https://github.com/STMicroelectronics/arm-trusted-firmware/blob/v2.8-stm32mp/docs/devicetree/bindings/clock/st,stm32mp13-rcc.txt#L409

- Check (with an oscilloscope) if there is I2C correct transactions on the STPMIC1 pins

- Check what function could be related around address 0x2ffec98b in your compiled code.

- add more debugging messages

See also https://wiki.st.com/stm32mpu/wiki/How_to_debug_TF-A_BL2

Regards,

[DMårt]

Hi @PatrickF 

>> please clarify how did you get this message ? Boot from flash or download from CubeProgrammer ?

I got the message from UART via Arduino Monitor. I downloaded TF-A from CubeProgrammer.

>> - check if HSE setting (digital clock (bypass) or crystal oscillator) is aligned between HW (e.g. OSC_OUT) and TF-A Device Tree.
   e.g. if you are using a crystal, the st,digbypass should not be present in the DT

Yes. I have chosen Digital Pass and "st,digbypass" is present due to my 24 MHz Oscillator (not crystal).

STM32-Computer/STM32CubeIDE/STM32-Computer-Firmware/CA7/DeviceTree/STM32-Computer-Firmware/tf-a/stm32mp151a-stm32-computer-firmware-mx.dts at main · DanielMartensson/STM32-Computer (github.com)

>> - Check (with an oscilloscope) if there is I2C correct transactions on the STPMIC1 pins

It must be. I mean, it seems to be soldered correctly.

>> - Check what function could be related around address 0x2ffec98b in your compiled code.

How can I do that? 

Update:

I re-configure all the clocks to make sure that I did not miss anything.

Still got the same error code:

  19:27:23 : STM32CubeProgrammer API v2.17.0 | Windows-64Bits 
  19:28:25 : Read TSV File: C:\Users\danie\Downloads\en.FLASH-stm32mp1-openstlinux-6.1-yocto-mickledore-mpu-v24.06.26\stm32mp1-openstlinux-6.1-yocto-mickledore-mpu-v24.06.26\images\stm32mp1\flashlayout_st-image-weston\optee\FlashLyout_STM32_Computer-Firmware.tsv
  19:28:25 : Number of partitions: 13
  19:29:20 : UR connection mode is defined with the HWrst reset mode
  19:29:21 : USB speed   : High Speed (480MBit/s)
  19:29:21 : Manuf. ID   : STMicroelectronics
  19:29:21 : Product ID  : DFU in HS Mode @Device ID /0x500, @Revision ID /0x0000
  19:29:21 : SN          : 002600373431510433383938
  19:29:21 : DFU protocol: 1.1
  19:29:21 : Board       : --
  19:29:21 : Device ID   : 0x0500
  19:29:27 : Start Embedded Flashing service
  19:29:27 : Memory Programming ...
  19:29:27 : Opening and parsing file: tf-a-stm32mp151a-stm32-computer-firmware-mx-optee-emmc.stm32
  19:29:27 :   File          : tf-a-stm32mp151a-stm32-computer-firmware-mx-optee-emmc.stm32
  19:29:27 :   Size          : 224.44 KB 
  19:29:27 :   Partition ID  : 0x01 
  19:29:27 : Download in Progress:
  19:29:30 : File download complete
  19:29:30 : Time elapsed during download operation: 00:00:02.536
  19:29:30 : RUNNING Program ... 
  19:29:30 :   PartID:      :0x01 

----------------------------
PANIC at PC : 0x2ffec98b

Exception mode=0x00000016 at: 0x2ffec98b

[DMårt]

Hi @PatrickF 

I tried to change the clocks, without success. 

I tried to change the clocks so they will be exactly the same as these ones: STM32MP15 clock tree - stm32mpu

Then I generated the TF-A meta with stm32mp-stm32-computer-mx.stm32 file. I ONLY flashing the stm32mp-stm32-computer-mx.stm32 and not the fip/bin file.

I get the exactly the same issue:

PANIC at PC : 0x2ffec98b
Exception mode=0x00000016 at: 0x2ffec98b

But then I tried to use a pre-made FlashLayout FlashLayout_emmc_stm32mp157c-ed1-optee.tsv with CubeProgrammer. Then I got the classical issue that my I2C4 have wrong pinouts.

NOTICE:  CPU: STM32MP151AAC Rev.Z
NOTICE:  Model: STMicroelectronics STM32MP157C eval daughter on eval mother
ERROR:   I2C device not ready
PANIC at PC : 0x2fff101b

Exception mode=0x00000016 at: 0x2fff101b

 So that means:

• The clocks is working
• The soldering is working
• No physical damage on the board

So my next question is: If I get the error 0x2ffec98b....and it's not the clocks. Perhaps you could send me the .ioc configuration file for STM32MP157x so I can copy over the most necessary parts to STM32MP151? I mean, it quite obvious that I have forgot something in the .ioc configuration. 

Update:

Do you think this has caused the problem: Solved: BUG: Label or path X_pins_y not found - Compiling ... - STMicroelectronics Community

I just changed the stm32mp15-bl2.dtsi file because the pins could not be found.

[PatrickF]

Hi @DMårt 

ST boards .ioc configuration could obtained by just loading an STM32MP157x board (EV1 or DK1/2) in CubeMX. But it might not bring so much.

For ST boards .dts, please have a look to this folder https://github.com/STMicroelectronics/arm-trusted-firmware/tree/v2.8-stm32mp/fdts (see  also https://wiki.st.com/stm32mpu/wiki/STM32_MPU_device_tree#STM32MP1_series)

Regards.

[DMårt]

Hi @PatrickF 

I have looked at stm32mp15-bl2.dtsi and it seems that there are some errors there.

If I comment the /omit-if-no-ref/, then the error disapear. But I might think that will cause the issue that my clocks wont start.

If I don't comment the /omit-if-no-ref/ then I will get the error "Label or path not found", because my pins are named e.g "i2c2_pins_mx" and not "i2c2_pins_a".

I think the ending "_mx" is the issue here. What do you think?

[PatrickF]

Hi,

the I2C naming of pins reference should be consistant with pinmux definitions elsewhere in the device tree.

_mx is when it has been generated from CubeMX. _a is for our Starter Package DT (not generated from CubeMX) . This is because in starter package DT, we have to ensure multiple board support with same portions of the DTs (so we sometimes use _b when there is pin differences).

https://github.com/STMicroelectronics/arm-trusted-firmware/blob/v2.8-stm32mp/fdts/stm32mp15-pinctrl.dtsi#L34

https://github.com/STMicroelectronics/arm-trusted-firmware/blob/v2.8-stm32mp/fdts/stm32mp157a-ed1.dts#L58

So, you should align your i2c node.

Regards.

[DMårt]

@PatrickF 

But when I compile with the bulild "meta stm32mp-stm32-computer-firmware-mx" option, then I get that error.

Try it your self.

Generate an empty STM32MP151AAC3 project, setup the SDK, add the TF-A. Then go to build artifacts and build the TF-A with the "meta stm32-<your project>-mx" command.

Then you will get the errors.

I have avoid these errors by comment the "/omit-if-not-ref/". But I think that will cause another future problem. I don't know. 

[DMårt]

@PatrickF 

Hi Patrick! Do you think it's the SYSRAM that causing this code?

I'm sure that it's only a configuration. My I2C peferial did start when I used another .tsv file.

I'm pretty sure that the cause of the error 0x2ffec98b

So what should I do next?

By the way, I found the new solution to the build artifact without removing any code lines. Yes, it's a BUG, but very simple to solve.

Watch: Solved: BUG: Label or path X_pins_y not found - Compiling ... - STMicroelectronics Community

[DMårt]

@PatrickF 

I have made one change. First, I did not use tf-a-usb.stm32 for fsbl-boot. Instead I was using tf-a-stm32mp151a-<projet name>-mx-optee-emmc.stm32 or tf-a-stm32mp151a-<project name>-mx-optee-sdcard.stm32. Which gave the same error code: 0x2ffec98b

But now I'm using tf-a-stm32mp151a-<project name>-mx-usb.stm32 for fsbl-boot and I got this error code now:

PANIC at PC : 0x2ffead5d

Exception mode=0x00000016 at: 0x2ffead5d

The memory map of ARM Trusted Firmware: 6.34. STMicroelectronics STM32MP1 — Trusted Firmware-A documentation

So that means that It's somthing with Bootloader 2 (BL2). 

0x00000000 +-----------------+
           |                 |   ROM
0x00020000 +-----------------+
           |                 |
           |       ...       |
           |                 |
0x2FFC0000 +-----------------+ \
           |     BL32 DTB    | |
0x2FFC5000 +-----------------+ |
           |       BL32      | |
0x2FFDF000 +-----------------+ |
           |       ...       | |
0x2FFE3000 +-----------------+ |
           |     BL2 DTB     | | Embedded SRAM
0x2FFEA000 +-----------------+ |
           |       BL2       | |
0x2FFFF000 +-----------------+ |
           |  SCMI mailbox   | |
0x30000000 +-----------------+ /
           |                 |
           |       ...       |
           |                 |
0x40000000 +-----------------+
           |                 |
           |                 |   Devices
           |                 |
0xC0000000 +-----------------+ \
           |                 | |
0xC0100000 +-----------------+ |
           |       BL33      | | Non-secure RAM (DDR)
           |       ...       | |
           |                 | |
0xFFFFFFFF +-----------------+ /

What I have done is to change the build artifact for tf-a sources. I have replaced ..=optee-sdcard with ..=optee-usb.

When I'm trying to build the stm32 meta <project name>-mx build artifact. I'm not sure if that's correct.

I have also change the clocks so they are 100% the same as the default initialization of the system clocks of STM32MP157F-DK1.

Notice:

In STM32CubeIDE, for U-boot and Linux, you need to manually configure the clocks by pasting them into the .dts file. Yes, it's a bug and I hope the dev team of STM32CubeIDE will solve that in next version 1.17.0.

Question:

Which build artifact for tf-a in order to implement it into the SYSRAM by changing BL2?

https://github.com/DanielMartensson/STM32-Computer/tree/main/Firmware/STM32-Computer-Firmware