2

I am customising the linux kernel.

In normal Linux, the BH thread priority is set to a constant priority (MAX_USER_RT_PRIO / 2) [1], but I am changing the BH thread priority from a constant priority (MAX_USER_RT_PRIO / 2) to another value in the interrupt handler [2]. Specifically, this priority is changed in __irq_wake_thread().

#code1

action->thread->prio = prio_input

or
#code2

sched_setscheduler_nocheck(action->thread, SCHED_FIFO, & param_input);

But with such a customised kernel, the hardware crashes and reboots two seconds later.

I debugged it and identified #code1, #code2 as the cause, but I don't understand what's wrong.

what was the error when it crashed?

It simply downed without any pop-ups like error messages. dmesg has no clues.

output of command "last"

reboot system boot 4.9.201-rt134 Sun Jul 10 11:24 still running 
him :1 :1 Sun Jul 10 11:21 - crash (00:02) 
reboot system boot 4.9.201-rt134 Fri Dec 31 20:00 still running

Are priority changes ever not allowed in Interrupt?

Environment
Hardware: Jetson nano
Power mode: 10 W
Power: 5V4A
Kernel: linux kernel-4.9
How to build: https://forums.developer.nvidia.com/t/applying-a-preempt-rt-patch-to-jetpack-4-5-on-jetson-nano/168428/4

[1] In setup_irq_thread()
[2] In __irq_wake_thread()

Reply for other comments:

changing that could cause some critical BHs to not run in time and break things.

I change only the priority of a few interrupts(BH), not all interrupts. It is set with care.

Tsum
  • 21
  • 2
  • what was the error when it crashed? – stark Jul 10 '22 at 13:41
  • I don't know if priority changes are allowed in interrupt context, but if BHs have a fixed priority set to a certain value there is most likely a reason for it, and changing that could cause some critical BHs to not run in time and break things. – Marco Bonelli Jul 11 '22 at 14:05

0 Answers0