AM62P: Watchdog Reset not Working From Linux

One more observation to add,

On TI AM62P SK board (running tisdk-default-image-am62pxx-evm-10.01.10.04.rootfs.wic.xz), I am unable to reset the device using the watchdog from Linux doing the same simple test.

Logs: /cfs-file/__key/communityserver-discussions-components-files/791/AM62P_5F00_SK_5F00_Board_5F00_Watchdog_5F00_Linux_5F00_Reboot_5F00_Not_5F00_Working_5F00_logs.txt

Hello Parth,

Apologies for this feature not working out-of-the-box in SDK 10.1. I can replicate your observations on my end with the default SDK 10.1 image:

root@am62pxx-evm:~# uname -a
Linux am62pxx-evm 6.6.58-ti-01497-ga7758da17c28-dirty #1 SMP PREEMPT Wed Nov 27 13:23:15 UTC 2024 aarch64 GNU/L
inux
root@am62pxx-evm:~# echo 1 > /dev/watchdog
[265732.523329] watchdog: watchdog0: nowayout prevents watchdog being stopped!
[265732.530313] watchdog: watchdog0: watchdog did not stop!
root@am62pxx-evm:~#
// board did not time out after 60 seconds

We wanted to add watchdog support for AM62Px starting in SDK 10.0, but we did not properly set the ESM interrupt sources in u-boot. The ESM module is the thing that receives the interrupt when the watchdog counts down to zero, so if the ESM inputs are wrong, then nothing happens when the watchdog signal triggers.

You can find the modifications to make to both Linux and U-boot here:
https://software-dl.ti.com/processor-sdk-linux/esd/AM62PX/10_01_10_04/exports/docs/linux/Foundational_Components/Kernel/Kernel_Drivers/Watchdog.html#fixes-for-10-0-sdk

I assumed this would be fixed for SDK 10.1, but it does not look like the patches made their way into SDK 10.1 either. The latest version of the ti-u-boot-2024.04 branch still has the old incorrect code:
https://git.ti.com/cgit/ti-u-boot/ti-u-boot/tree/arch/arm/dts/k3-am62p-main.dtsi?h=ti-u-boot-2024.04#n269

Please try modifying your uboot code and tell me if it starts working for you. In the meantime, I will followup on what happened for ti-u-boot-2024.04, and make sure that this will be fixed for the upcoming SDK 11.0.

Regards,

Nick

Hello Parth,

Ok, so we HAVE fixed the uboot code in ti-u-boot-2025.01.

Starting in this version of u-boot, the devicetree files are in a different spot. It looks like we are just mirroring the upstream Linux devicetree files. So the AM62Px devicetree file is here:
https://git.ti.com/cgit/ti-u-boot/ti-u-boot/tree/dts/upstream/src/arm64/ti/k3-am62p-j722s-common-main.dtsi?h=ti-u-boot-2025.01-next#n286

and then this devicetree entry tells the processor when during the boot flow to probe the devicetree node:
bootph-pre-ram;

Regards,

Nick