Bug 11091 - (int-184396) Bug in cpufreq module - exposed in /proc/timer_stats
(int-184396)
: Bug in cpufreq module - exposed in /proc/timer_stats
Status: UNCONFIRMED
Product: Core
Kernel
: 5.0:(10.2010.19-1)
: All Maemo
: Unspecified minor (vote)
: ---
Assigned To: unassigned
: linux-kernel-bugs
:
:
:
:
  Show dependency tree
 
Reported: 2010-08-08 00:32 UTC by Ian Stirling
Modified: 2010-08-10 00:17 UTC (History)
1 user (show)

See Also:


Attachments


Note

You need to log in before you can comment on or make changes to this bug.


Description Ian Stirling (reporter) 2010-08-08 00:32:02 UTC
SOFTWARE VERSION:
(Settings > General > About product)
PR1.2

EXACT STEPS LEADING TO PROBLEM: 
(Explain in detail what you do (e.g. tap on OK) and what you see (e.g. message
Connection Failed appears))
1. 
echo 1 >/proc/timer_stats
cat /proc/timer_stats
cat /proc/timer_stats 
Timer Stats Version: v0.2
Sample period: 25.818 s
  136,     0 swapper          tick_nohz_restart_sched_tick (tick_sched_timer)
   65,   470 wl12xx           queue_delayed_work (delayed_work_timer_fn)
   30,   470 wl12xx           schedule_timeout (process_timeout)
   37,    10 omap2_mcspi      sk_reset_timer (tcp_delack_timer)
  121,     0 swapper          hrtimer_start (tick_sched_timer)
   27,  1208 wlancond         ieee80211_ioctl_siwpower
(ieee80211_dynamic_ps_timer)
   27,   470 wl12xx           schedule_timeout (process_timeout)
   27,   470 wl12xx           schedule_timeout (process_timeout)
    1,   342 pdflush          ubifs_wbuf_write_nolock
(wbuf_timer_callback_nolock)
    1,   342 pdflush          ubifs_wbuf_write_nolock
(wbuf_timer_callback_nolock)

  34D,    37 awk              cpufreq_governor_dbs (delayed_work_timer_fn)
...

555 total events, 21.496 events/sec


EXPECTED OUTCOME:
Something like the line above shown being 
0D,     37 cpufreq           cpufreq_governor_dbs...
ACTUAL OUTCOME:
The above
REPRODUCIBILITY:
always

The cpufreq_governor_dbs timer is as I understand it a timer that's used by the
cpufreq code, to see if it needs to change frequency.

There is no awk process running - I'm confused why it manages to pick a
plausible name - perhaps it's just set on boot, when awk is the first process
always run after enabling the governor?

However. The process ID always is seemingly random.

I tracked this down to the cpufreq governor patches from nokia, but was unable
to work out how the info in /proc/timer_stats is actually set - which part of
the timer callback sets it.