Hi there,
I am trying to debug random system freezes when running our application software on a preempt-rt patched kernel. At the moment of the system freeze, there is nothing to see in the logs. No response to keyboard input, no messages on the serial console, no replies to pings.
We are running kernel 4.14.93-rt on a quad-core x86_64 Intel Atom SMP system. As this is an Intel Baytrail system, we limited the max cstate to 1 as deeper cstates are known to cause system freezes.
When reviewing the dmesg output, I noticed we are getting a large number of "NOHZ: local_softirq_pending 80" messages but only when running our software.
These messages also occur when running all applications using a normal scheduling policy (SCHED_OTHER) with default priority / niceness values.
Note that our application software only uses up to 50% on average of our CPU resources.
I also tried running the system using nohz=off, but this did not solve our system freezes.
Does anyone have an idea whether these messages signal a potentially dangerous issue? Could this be related to our system freezes? Any thoughts on how to debug these messages and solve the underlying issue?
Best regards,
Tom Putzeys
