On Sun, Sep 6, 2009 at 11:01 AM, Sebastien Roy<Sebastien.Roy at sun.com> wrote: > On Sun, 2009-09-06 at 09:52 +0800, Li, Aubrey wrote: >> PowerTOP should not be the case that causes 100% highest p-state >> residency in idle. We have this bug >> >> "6818514 Event based CPUPM too easily shifts into high gear" >> >> fixed and putback since build 117 by rev#. 9802. It looks like it does not >> fix the problem. >> >> But if you always didn't see the lowest frequency residency when idle, >> there is probably another bug here. ?It would be great if Seb can post the >> powertop report here. > > Sure, here it is: > > ? ? ? ? ? ? ? ? ? ? ? ?OpenSolaris PowerTOP version 1.2 > > C-states (idle power) ? Avg ? ? Residency ? ? ? P-states (frequencies) > C0 (cpu running) ? ? ? ? ? ? ? ?(0.0%) ? ? ? ? ? 800 Mhz ? ? ? ?0.0% > C1 ? ? ? ? ? ? ? ? ? ? ?3.2ms ? (100.0%) ? ? ? ?1060 Mhz ? ? ? ?0.0% > ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?1330 Mhz ? ? ? ?100.0% > > Wakeups-from-idle per second: 632.7 ? ? interval: 5.0s > Power usage (ACPI estimate): 0.000W (running on AC power, fully charged) > > Top causes for wakeups: > 30.0% (189.8) ? ? ? ? ? ? ? <kernel> : ?genunix`cv_wakeup > 18.3% (115.6) ? ? ? ? ? ? ? ? ?sched : ?<xcalls> unix`dtrace_xcall_func > 15.8% (100.0) ? ? ? ? ? ? ? <kernel> : ?genunix`clock > ?4.8% ( 30.1) ? ? ? ? ? ?<interrupt> : ?iwk#0 > ?2.4% ( 15.2) ? ? ? ? ? ? ? <kernel> : > ?uhci`uhci_handle_root_hub_status_change > ?1.6% ( 10.2) ? ? ? ? ? ? ? <kernel> : ?genunix`delay_wakeup > ?0.6% ( ?4.0) ? ? ? ? ? ? ? <kernel> : ?genunix`schedpaging > ?0.6% ( ?4.0) ? ? ? ? ? ? ? <kernel> : ?uhci`uhci_cmd_timeout_hdlr > ?0.6% ( ?3.8) ? ? ? ? ? ? ? <kernel> : > ?ehci`ehci_handle_root_hub_status_change > > -Seb
okay, thanks but need your another help. I did a quick investigation and found that there is a xorg related kthread causing the scheduler ping-pong between "idle" thread and this xorg related thread. Could you please disable X by "pfexec svcadm disable gdm" and try powertop again? Please make sure you are still using event mode, not poll-mode. Thanks, -Aubrey
