AM62P: Panic and black screen occur upon startup

Hi,

Can you confirm this is happening when making use of DPI w/ HDMI bridge? Also, please confirm this HMI interface test is being done by making use of the AM62P SKEVM itself. From the log, it seems like Linux is being botted up from eMMC. Please confirm. Have you tried the latest SDK version to see if the same behavior is experienced? Can you provide more details on the voltage drop test that's being done? Is the display the load that the 13.5V-4.5V voltage is being supplied to in this case? 

Best regards,

Luis Parga

Hi,

Sorry, I don't quite understand. The current project screen only simply outputs lvds signals and does not use the bridge mode.  

The latest SDK version has not been used for testing, and there are no resources available for the verification of the new SDK version for the time being.  

Test steps: The power-on voltage is 13.5v, maintained for 100ms, then reduced to 4.5v and maintained for 500ms. It rises again to 13.5v and waits for about 60 seconds.  Keep repeating the above steps for the test.  

When the panic problem occurs, the voltage supply is 13.5v.  

Are there any further analytical steps? 

Thanks

Hi Shawn,

Thanks for the additional details. Can you help to let us know what the 13.5V supply is connected to exactly  in this HMI system with LVDS display? Also, please confirm you are using AM62P SKEVM for this testing? In the meantime, we are internally reviewing the Kernel panic message to understand the root cause of the GPU driver issue.

Best regards,

Luis Parga

Hi Luis,

I will reply to you after confirming the power supply pins and their positions. The machine currently being tested does not use an EVM board.  

Let me know at any time if there is any progress regarding Kernel panic.

Thanks.

Supplement relevant information:

Hi Shawn,

It seems like your testing somehow killed the GPU and it's not being responsive at all. The error above just means that GPU is unreachable. As an experiment, you can try to disable the APM as listed here: https://texasinstruments.github.io/processor-sdk-doc/processor-sdk-linux-AM62PX/esd/docs/master/linux/Foundational_Components/Graphics/Rogue/Rogue_Power_Management_Info.html. Try again with your testing and let me know if you see the issue again.

Regards,
Krunal

Hi Krunal,

Thank you for your information.  

According to the Rogue_Power_Management_Info.html document, can the interface return value be modified to "on" to disable APM?  

Currently, the return value of the /sys/devices/path_to_device/power/control interface is "auto", and I execute the instruction:  echo on > /sys/devices/platform/bus@f0000/fd00000.gpu/power/control, has been disabled?

Also, why don't the modifications I made in the device tree take effect?  The interface return value remains "auto" after compilation.  If the device tree is not modified through echo while it is running, how can it be modified to take effect? 

be modified as follows:

#power-domain-cells = <0>;
clocks = <&k3_clks 187 0>;
ti,power-domain-always-on;

gpu: gpu@fd00000 {
		compatible = "ti,am62-gpu", "img,img-axe";
		reg = <0x00 0x0fd00000 0x00 0x20000>;
		clocks = <&k3_clks 187 0>;
		clock-names = "core";
		interrupts = <GIC_SPI 86 IRQ_TYPE_LEVEL_HIGH>;
		power-domains = <&k3_pds 187 TI_SCI_PD_EXCLUSIVE>;
		/*Add the following definitions */
		#power-domain-cells = <0>;
		clocks = <&k3_clks 187 0>;
		ti,power-domain-always-on;
		
	};

Thanks

Hi,

I am not sure I follow your DT comment but you just need to set the value to "on" to disable the APM. This is a runtime change only and after reboot, the value will reset back to auto. 

Regards,
Krunal

OK，I get it.

If I don't want to set it to "on" at every runtime, I want to set this value to the permanent value, and it is still "on" after restarting, what should I do?

Thanks

Hi,

Before we go into that, can you confirm is disabling the APM indeeds resolves your issue?

Regards,
Krunal

Hi Krunal,

We are currently conducting stress tests using software with APM disabled. We have tested 400 times so far and have not yet reproduced the same issue.  The test is still ongoing. If the problem occurs again, I will let you know.  

So I would like to consult at the same time on how to modify the APM value to the permanent value "ON". 

Thanks

Hi Krunal,

Let me confirm with you again whether changing ".benableactivepm = IMG_TRUE "to FALSE means disabling APM? Is there anything else that needs to be modified?

/* Setup RGX specific timing data */
static RGX_TIMING_INFORMATION gsRGXTimingInfo = {
	.ui32CoreClockSpeed = 800000000,
	// .bEnableActivePM = IMG_TRUE, //Modify to FALSE 
	.bEnableActivePM = IMG_FALSE,
	.ui32ActivePMLatencyms = SYS_RGX_ACTIVE_POWER_LATENCY_MS,
	.bEnableRDPowIsland = IMG_FALSE,
};

Also, how does this sysconfig.c compile into the kernel?  Under the path of extra-drivers in the SDK, how can it be linked and compiled into the kernel? 

board-support/extra-drivers/ti-img-rogue-driver-24.1.6554834/services/system/rogue/bxs_am62p/sysconfig.c

Thanks

Hi,

I am not sure I understand your question but the GPU driver is standalone and it doesn't get compiled into Kernel.

Regards,
Krunal