AM625: boot ATF error

Hello,

By ATF error do you mean the booting procedure doesn't go beyond the above shared logs?

mars lin said:

it have ATF error happen occasionally.

Please make sure you are building ATF compliant with the SDK version as mentioned in: https://software-dl.ti.com/processor-sdk-linux/esd/AM62X/09_01_00_08/exports/docs/devices/AM62X/linux/Release_Specific_Release_Notes.html#build-information

Regards,
Aparna

Aparna, 

This is TI digital FAE, customer is doing power cycle test (power on/off cycle) and found device will stop booting occasionally.

Re-power cycle can make it back to work.  

Certain boards have higher possibility to fail rate (one out of 100 times) and others also have low changes to encounter. (zero to one out of 4000 times)

Customer system use eMMC as boot device for all boot files. 

We currently suspect eMMC stability. 

Is there a way to enable debug log to investigate further after ATF? 

_____                    _____           _         _

|  _  |___ ___ ___ ___   |  _  |___ ___  |_|___ ___| |_

|     |  _| .'| . | . |  |   __|  _| . | | | -_|  _|  _|

|__|__|_| |__,|_  |___|  |__|  |_| |___|_| |___|___|_|

              |___|                    |___|

Arago Project am62xx-evm -

Arago 2023.10 am62xx-evm -

am62xx-evm login: [   14.839986] fpga_manager fpga0: writing CNC2_FC_A7_50T_CSG324_M3_BUS3.bin to Xilinx Slave Serial FPGA Manager

U-Boot SPL 2023.04-ga37da23008 (Apr 02 2024 - 05:22:32 +0000)

SYSFW ABI: 3.1 (firmware rev 0x0009 '9.1.8--v09.01.08 (Kool Koala)')

SPL initial stack usage: 13384 bytes

Trying to boot from MMC1

Authentication passed

Authentication passed

Authentication passed

Authentication passed

Authentication passed

Starting ATF on ARM64 core...

NOTICE:  BL31: v2.9(r

NOTICE:  BL31: Built : 09:34:15, Aug 24 2023  -->It stop here.

BR, Rich

Hello,

I am routing this query to the eMMC expert for comments.

Regards,
Aparna

Aparna, 

For normal boot, the log should be 

U-Boot SPL 2023.04-g576b2c98c0 (Mar 11 2024 - 07:42:03 +0000)
SYSFW ABI: 3.1 (firmware rev 0x0009 '9.1.8--v09.01.08 (Kool Koala)')
SPL initial stack usage: 13384 bytes
Trying to boot from MMC1
Authentication passed
Authentication passed
Authentication passed
Authentication passed
Authentication passed
Starting ATF on ARM64 core...

NOTICE: BL31: v2.9(release):v2.9.0-614-gd7a7135d32-dirty
NOTICE: BL31: Built : 09:34:15, Aug 24 2023

However, now the fail cases 1

NOTICE:  BL31: v2.9(r

NOTICE:  BL31: Built : 09:34:15, Aug 24 2023  -->It stop here.

Fail case 2 has more wired message

NOTICE:  BL31: v2.9(rea{e):~:.;/8-6=6-wewa7575v=r-vmbt{AzENqlseqtorts Scbcz?d-on-nm. Please set build flag CTX_INCLUDE_AARCH32_REGS = 0

TIBoot3.bin is actually created by TI and customer doesn't change it. 

What possibility to see such debug message? 

BR, Rich

Hi Apama::

This kind of log never happens here.

The customer is doing the cycling power on + off testing.

They never touch the SPL + ATF code.

I guess the Warm - Reset too soon for causing this syndrome?

May we have your comment here?

Thanks.

BR Rio

Rich Chen said:

We currently suspect eMMC stability. 

The images are being read correctly as indicated by the logs "Authentication passed" so the issue is not related to the eMMC. Instead, it seems to be originating from the ATF as the logs stop there.

Once the boot stops, you may want to connect to the A53 core via the CCS & note down the address at which the core is suspended. This should give an idea if the control in really stuck in the ATF or something else.

Prashant, 

Two points, 

1. Why the NOTICE:BL31 output will be different  and it looks contain some debug message from code. 

2. The reason why suspect eMMC is because the boot stop. Itcould be fail to access eMMC or DDR.

CCS connection for PC (program counter) can be a check point for investigation. 

BR, Rich 

1. Probably something is going wrong during the ATF execution.

2. There is no eMMC access during ATF. If anything, this may very well be a DDR instability issue.

Let us know the results from CCS investigation.