3

Problem : Userland process hogging CPU. System Conf: Linux 2.6.34

Executed strace with different command line options and below are the few for the queries:

strace -f -rtT -p 12661

 0.000058 restart_syscall(<... resuming interrupted call ...>) = -487663192 <0.000017>
 0.000056 restart_syscall(<... resuming interrupted call ...>) = -487663192 <0.000017>
 0.000056 restart_syscall(<... resuming interrupted call ...>) = -487663192 <0.000017>
 0.000057 restart_syscall(<... resuming interrupted call ...>) = -487663192 <0.000017>
 0.000057 restart_syscall(<... resuming interrupted call ...>) = -487663192 <0.000018>
 0.000057 restart_syscall(<... resuming interrupted call ...>) = -487663192 <0.000017>

strace -c -p 12661

% time seconds usecs/call calls errors syscall
100.00 0.001191 0 23216 restart_syscall
0.00 0.000000 0 188 sigreturn
0.00 0.000000 0 1 rt_sigprocmask
0.00 0.000000 0 21 stat64
100.00 0.001191 23426 total

I'm confused with strace first column "time" which is saying 100 % does this directly refer to CPU occupancy ?

And also, restart_syscall occupancy (100%) is due to the strace ??

Objective of userland process is to emulate the not supported instructions/memory and also have one clone process to dequeue the packets from Eth (using netfilter).

Could someone help/clarify the above points ?

Thanks in Advance

Aaron Digulla
  • 321,842
  • 108
  • 597
  • 820
Snake
  • 63
  • 1
  • 9
  • Wrong place to ask. Perhaps your process is getting a lot of signals. – Basile Starynkevitch Nov 10 '14 at 15:15
  • @BasileStarynkevitch. Could you please help me to choose the right place. I posted here because found similar queries in this platform .http://stackoverflow.com/questions/4053142/how-to-track-child-process-using-strace?rq=1 – Snake Nov 10 '14 at 15:32

0 Answers0