Hello On a X201 core i7 (westmere), I am seeing an important number of wakeups per second. The system is very minimal (openbox to display lxterminal) yet I get : - about 100 wakeups per second with only the terminal open - about 300 wakeups per second with firefox Currently, with firefox opened: Summary: 283.0 wakeups/second, 2.0 GPU ops/seconds, 0.0 VFS ops/sec and 2.3% CPU us Usage Events/s Category Description 1973 rpm Device Laptop fan 1.5 ms/s 109.7 Process /opt/google/talkplugin/Google 1.1 ms/s 37.2 Timer hrtimer_wakeup 5.3 ms/s 35.3 Process /usr/bin/firefox 188.7 µs/s 19.6 Process [xfsaild/dm-0] 218.8 µs/s 14.7 Timer tick_sched_timer 0.8 ms/s 11.8 Interrupt [6] tasklet(softirq) 1.1 ms/s 8.8 Process lxterminal --geometry=79x29 426.9 µs/s 9.8 Process /opt/firefox/plugin-container 166.7 µs/s 8.8 kWork ieee80211_iface_work 126.0 µs/s 8.8 Process [rcu_sched] 590.5 µs/s 4.9 Process /usr/bin/X vt12 8.8 ms/s 0.00 Process powertop 1.0 ms/s 2.9 Interrupt [16] ath9k 55.9 µs/s 2.0 Process [ips-monitor] 426.9 µs/s 1.0 Interrupt [9] RCU(softirq) 139.3 µs/s 1.0 Process [ksoftirqd/3] 41.7 µs/s 1.0 Process /usr/sbin/acpid -f -d -c /etc 26.4 µs/s 1.0 Interrupt [43] i915 6.4 µs/s 1.0 Interrupt [3] net_rx(softirq) 5.2 µs/s 1.0 kWork flush_to_ldisc 4.5 µs/s 1.0 kWork console_callback 3.4 µs/s 1.0 kWork ath_tx_complete_poll_work I tried to look for more information about which kernel configure options could be causing that, but I could not find anything. When I stop everything and monitor with eventstat, I still generally get about 100 wakeups per second: Event/s PID Task Init Function Callback 27.00 0 [swapper/0] hrtimer_start tick_sched_timer 19.00 194 [xfsaild/dm-0] xfsaild process_timeout 8.00 0 [swapper/3] hrtimer_start tick_sched_timer 5.00 0 [swapper/0] hrtimer_start_range_ns tick_sched_timer 5.00 0 [swapper/2] hrtimer_start tick_sched_timer 4.00 7 [rcu_sched] rcu_gp_kthread process_timeout 4.00 459 Xorg hrtimer_start_range_ns hrtimer_wakeup 4.00 525 lxterminal hrtimer_start_range_ns hrtimer_wakeup 3.00 0 [swapper/0] add_unmap flush_unmaps_timeout 3.00 39 [kworker/3:1] hrtimer_start_range_ns hrtimer_wakeup 3.00 0 [swapper/1] hrtimer_start tick_sched_timer 2.00 0 [swapper/1] clocksource_watchdog clocksource_watchdog 1.00 393 gpm hrtimer_start_range_ns hrtimer_wakeup 1.00 426 systemd-network hrtimer_start timerfd_tmrproc 1.00 143 [kworker/3:2] e1000_watchdog_task e1000_watchdog 1.00 0 [swapper/0] tpt_trig_timer tpt_trig_timer 1.00 746 python hrtimer_start_range_ns hrtimer_wakeup 1.00 0 [swapper/2] br_hello_timer_expired br_hello_timer_expired 1.00 225 [kworker/2:2] queue_delayed_work_on delayed_work_timer_fn 1.00 135 [kworker/u8:6] queue_delayed_work_on delayed_work_timer_fn 1.00 36 [kworker/0:1] queue_delayed_work_on delayed_work_timer_fn 97 Total events, 97.00 events/sec (kernel: 85.00, userspace: 12.00) I believe the swapper is to be ignored (https://lkml.org/lkml/2009/6/25/166) but that still leaves xfsaild and rcu_sched My suspects at the moment are: - the hrtimer issue reported on https://bugzilla.kernel.org/show_bug.cgi?id=14424 - the xfs issues reported before ( http://marc.info/?l=linux-xfs&m=134088105208917&w=2) but the patch has been included in 3.14.25. I might move everything into a ext4 filesystem if that can help. - some westmere related issues (http://lwn.net/Articles/479814/: Extend mwait idle to optimize away IPIs when possible) but there is no more idle=mwait I'm running a 3.14.25 kernel, whose cmdline is: BOOT_IMAGE=/kernel/bzImage-test root=/dev/sda4 swap=/dev/sda3 plymouth.enable=0 libahci.ignore_sss=1 raid=noautodetect i915.lvds_downclock=1 i915.i915_enable_rc6=1 i915.fastboot=1 i915.semaphores=1 nohz=on thinkpad_acpi.brightness_enable=0 thinkpad_acpi.force_load=1 thinkpad_acpi.fan_control=1 snd-hda-intel.index=0 snd_hda_intel.power_save=10 snd-pcsp.index=1 btusb.reset=1 ath9k.ps_enable=1 ath9k.btcoex_enable=1 pcie=aspm pcie_aspm.policy=powersave tpm_tis.force=1 tpm_tis.interrupts=0 tpm_tis.hid=PNP0C31 crashkernel=512M-2G:64M,2G-:128M nmi_watchdog=0 quiet i915.i915_enable_fbc=1 drm.vblankoffdelay=1 rcutree.rcu_idle_gp_delay=6 iommu=pt systemd.unit=graphical.target (The rcutree.rcu_idle_gp_delay was just a recent attempt - didn't help) The kernel .config is attached. Powertop is happy with all my settings except SATA link power management, but I need link_power_management_policy set to medium_power or I do have weird disconnect with my SSD (Buffer I/O error on device sda1, logical block 0, etc) Is is the best list to discuss that? (or is it only for powertop development) Is so, could I please ask for some suggestions to reduce the number of wakeups per second? Thanks