The cpufreq timing statistics are all zeros:
$ sort --unique /sys/devices/system/cpu/cpu*/cpufreq/stats/time_in_state | column
1600000 0 2100000 0 2600000 0 3100000 0
1700000 0 2200000 0 2800000 0 3300000 0
1900000 0 2400000 0 2900000 0 3400000 0
2000000 0 2500000 0 3000000 0 3401000 0
How is this possible? Is there really no time spent in any state?
Details
/proc/cpuinfo
The CPUs are clearly transitioning between frequency states:
$ grep --perl-regexp --only-matching '(?<=cpu MHz\t\t: )[\d.]+' /proc/cpuinfo | column 1600.000 1600.000 3401.000 1600.000 2900.000 1600.000 2100.000 1600.000 $ while sleep 10; do column /sys/devices/system/cpu/cpu*/cpufreq/stats/total_trans; done 17839 17278 17497 17336 12332 14358 14054 17647 17885 17313 17545 17384 12359 14394 14082 17691 17888 17323 17556 17391 12359 14399 14085 17693 17904 17336 17576 17417 12374 14402 14091 17696
Someone else asked about this on a forum but never received a reply.
There is a problem regarding this in the Linux kernel 3.11. The bug was fixed with the following patch:
http://permalink.gmane.org/gmane.linux.kernel.cpufreq/11672%3E
https://lkml.org/lkml/2013/9/7/126
Upgrading your kernel (to 3.12-rc1 or higher) should fix the problem.