uboot jump to kernel affects llce_lin output

Masson

Hi, NXP

We use bootloader to boot M core + A core, M core configured with llce_lin output;
S32g u-boot jumps to kernel, llce_lin output is broken, query the specific code and found;
u-boot code: bootm.c->announce_and_cleanup()-> cleanup_before_linux(void)->board_cleanup_before_linux();->scmi_reset_agent();
we found that because of the call to the scmi_reset_agent function leads to, the M core configure llce_lin to stop sending;
due to scmi ( System Control and Management Interface The system control and management interface (SCMI) is a lower level function in u-boot and kernel, controlling clock, power, io, reset, etc., which involves the underlying control logic of s32g;
How to modify or configure to solve the problem of initialization of A-core which leads to the stoppage of llce_lin?

Best regards，

Norajohnson

The transition from U-Boot to the Linux kernel can indeed affect the output of peripherals, including those related to LLCE-LIN (Low Latency Communication Engine - Local Interconnect Network), especially in complex embedded systems like the NXP S32G. This is primarily due to the different ways U-Boot and the monkey type Linux kernel manage and initialize hardware resources. We use bootloader to boot M core + A core, M core configured with llce_lin output; S32g u-boot jumps to kernel, llce_lin output is broken, query ...However, they serve different purposes and are managed quite differently. ->U-boot device tree is used to configure and initialize the hardware components needed during the boot process, while the Linux kernel used the device tree to describe the hardware configuration once the kernel is booted.

Masson

hi,nxp

I am continuing to trace the code in u-boot and found the following:
call chain: scmi_reset_agent(void) ->
devm_scmi_process_msg(dev, &scmi_msg) ->
scmi_smccc_process_msg(struct udevice *dev, struct scmi_msg *msg)->
arm_smccc_smc(chan->func_id, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, &res);->
SMCCC smc
The final call to the assembly is made, and I don't know the relevant details, so what does this part of the code do? How does it affect the M core operation and how can I skip this part of the code execution.

BR.

chenyin_h

Hello, @Masson

Thanks for the reply.

From my understanding, the U-boot will not set the clock, the tfa will set it during the boot.

For current issue, would you mind testing it by commenting out the scmi_reset_agent from the source, to check if there is still issue existed in your test?

BR

Chenyin

Masson

Hi，chenyin

I now have some questions about the clock in the S32G, basically similar to the questions in the picture;
1. won't tf-a exit after tf-a initializes the clock?
2. how does u-boot and kernel initialize or use the clock，Need to communicate with tf-a?
3. can you explain in detail how clk is used in tf-a, u-boot and kernel?

https://community.nxp.com/t5/NXP-Designs-Knowledge-Base/S32G-Bootloader-Customzition/tac-p/1771379/e...

BR.

Masson

Hi，chenyin

Thanks for the reply
I tested the commented scmi_reset_agent function and confirmed that the llce_lin output is normal during the u-boot->kernel jump.
However, I found that there are anomalies after kernel boot, such as eth0 ping can not be reached, because I am not sure if there are other problems, so I did not follow this line of investigation;
Do you suggest me to do this kind of investigation?

BR.

chenyin_h

Hello, @Masson

Thanks for your reply.

Commenting out the corresponding line is only a kind of test, it is not suggested way to fix the issue, since it may impact several parts, which may cause issues like you had found.

I think further analysis and test are needed, for more convenient discussion and resource sharing, let me help to create a support case to continue the support, I will directly reply you there via mails.

Sorry for your inconvenience.

BR

Chenyin

chenyin_h

Hello, @Masson

Thanks for your reply.

Sorry that temporally not, I suggest firstly trying checking the UART clock and its sources in TFA.

BR

Chenyin

Masson

Hi,chenyin

Please advise a question, we are now booting the flow as follows
bootloader(M7)->tf-a(A53)->u-boot(A53)->kernel(A53)
We have configured the uart clock in the bootloader, the uart clock is not configured in the tf-a, and no modification is made in the u-boot;

Will the uart module and clock be reconfigured in u-boot?

BR.

chenyin_h

Hello, @Masson

Thanks for your reply.

Since only ddr, xbar_2x clocks are kept in your tf-a, u-boot, do you mean that the clock for UART and Peripheral PLL3 are also disabled and set in bootloader instead of in TF-A?(the TFA/uboot only inherit the clock settings from bootloader to drive the UART using in TFA/uboot?)

BR

Chenyin

Masson

Hi，chenyi

I understand that's how it runs now.

Do you have any other findings?

BR.

chenyin_h

Hello, @Masson

Thanks for your post.

1. May I know if the LIN clock is configured in your bootloader or the LIN application running on the M7 core?

2. Which version BSP is used in your test setup? any modifications for the clock settings? would you mind sharing the clock dump from your u-boot?

BR

Chenyin

Masson

Hi,chenyi

The clk dump in u-boot is as follows：

Masson

Hi，chenyin

1. The Lin clock is configured in the bootloader;

2. we use bsp42, only ddr, xbar_2x clocks are kept in tf-a, u-boot, all other configurations are switched off, relying on the bootloader configuration;

BR.