Possible Bug: pixhawk hanged and sensors failed with v1.11.1

[RamDG]

Hi,
What am I doing: I am using jetson nano board with pixhawk cube black (fmuv_3) for object tracking. I am using the latest firmware v1.11.1 stable for my Hexacopter. i have created some parameters for object selection in a new module i have created. When i touch on video that particular point in the video will be selected as target point and the same is tracked by the algo running on jetson . When i click on the object only once, that xyz location is written on the parameters i have created in the QGC parameter list.

Possible Bug: when i touch on the video multiple times. CPU load goes to 100% and i start getting errors like accel_0 accel_1 failed or mag failed. Yesterday i flew my Hexacopter. It was in the position hold mode. When i touched the video to select the object for tracking mag failed and then the entire gps unit failed. The vehicle log says it tried for fail-safe but it crashed. Here is a link for the log file:
https://logs.px4.io/plot_app?log=b78baa94-c48a-4f98-b032-5cd031f53cf6

After the crash i checked the same issue on v1.10.1 and ardupilot. The issue was not there, (Didn't flew just tested on ground)

Along with that this issue is alwyas there on v1.11.1
#16035

[dagar]

Can you comment on what the custom gimbal_control module is doing on these changes?

[RamDG]

@dagar
I have created a simple 1 axis gimbal which has a laser pointer.... gimbal is free to move in pitch direction only.
I am using aux pins to connect 2 servos on aux1 and aux2 to control the gimbal movement. One more servo on aux3 is being used to turn on and off the laser when i am tracking the object. Also, all the tracking related parameters are being accessed through this module.

FYI : just for testing purpose i stopped this module and tried to write on some other available parameter continuously through a loop.... then also i observed the same . CPU load goes to 100% and sensors start failing one by one.
I used baudrate= 57600 and 115200 to communicate between jetson and pixhawk

[dagar]

write on some other available parameter continuously through a loop

When you write a parameter all modules in the system are notified of the update and then they refresh their local copy. At a bare minimum you can use param_set_no_notification(), but really you should be using uORB sharing data between modules.

[RamDG]

Suggestion: I believe px4 must not get hanged in any case. At least when system is ARM. Therefore, a software level protection is required to keep the sensors and all other essential tasks running to land the vehicle safe!!!!!!!!!

[dagar]

Suggestion: I believe px4 must not get hanged in any case. At least when system is ARM. Therefore, a software level protection is required !!!!!!!!!

High rate parameter updates are a bit of an abuse of the system, but we can add protections. #16189

[RamDG]

@dagar I tried param_set_no_notification() for the parameter that i am using for tracking which is GOD_S1.
Still every-time when i select an object to track , cpu load goes to 100%. please let me know how it can be resolved.

https://logs.px4.io/plot_app?log=ca8d6da5-af38-4057-ba6b-2fdf78966b2a

Also in the watchdog i am seeing FD for gimbal_control is 10 and for other apps it is quite low. Can you please tell what is it representing. and what is a nominal value for the same.