1

On multi-core RHEL6, there is a FIFO realtime process in a deadloop, so it occupies all the CPU resources on the core (process is bound to the core). However, the other cores are pretty fine.

At this time, pstack for the process in deadloop failed to print anything. At the same time, pstack for any other processes works well.

Could anyone please help to let me know why?

Thanks!

Steven Ding
  • 74
  • 1
  • 9

1 Answers1

1

try in this way, put the debuger command in a free core with taskset -c corenumber /path/command and after this try change the scheduling class in fifo of your debug process using chrt command

I think you found this problem, because the fifo process cannot be stoped by normal process

c4f4t0r
  • 1,563
  • 15
  • 24
  • Just tried as you said, but it seems the debugger is still not able to catch the stack trace. My "test" program is FIFO priority 1. My debugger is FIFO priority 99. The result seemed the same. However, when I tried to lock the shell to the free core, it seems to be working - even without raise the debugger's priority. :-) Now the question is why a "locked" core will block a floating process? Really strange because the free core has 0% CPU used. Thanks! – Steven Ding Dec 26 '13 at 13:00