On 03/06/2014 01:06 PM, David C. Rankin wrote: > On 03/06/2014 12:55 PM, David C. Rankin wrote: >> On 03/06/2014 12:29 PM, Timothy Pearson wrote: >>>> On 03/06/2014 10:52 AM, Dr. Nikolaus Klepp wrote: >>>>> Am Donnerstag, 6. März 2014 schrieb David C. Rankin: >>>>>> All, >>>>>> >>>>>> Testing the new soft-freeze packages has disclosed a horrible problem >>>>>> with >>>>>> tdepowersave taking nearly 100% of the CPU. My laptop was nearly on >>>>>> fire this >>>>>> morning: >>>>>> >>>>>> [screenshot] >>>>>> http://www.3111skyline.com/dl/dt/trinity/ss/tdeowersave-97percent.jpg >>>>>> >>>>>> Earlier in development I saw tdepowesave in the 20-40% range, but >>>>>> never at >>>>>> near 100% of CPU. What to try? >>>>>> >>>>> >>>>> What kernel do you use? I had the same problem on wheezy with 3.5.13.2. >>>>> tdepowersave took 100% of one core. It occured after tdepowersave put >>>>> the x61 to powersave the second time. It looks like the problem went >>>>> away after changing to kernel 3.12-0.bpo.1. >>>>> >>>>> nik >>>>> >>>>> >>>> >>>> Kernel is 3.13-5.1 (just a few days old), So this is happening with >>>> current kernels. >>>> >>>> -- >>>> David C. Rankin, J.D.,P.E. >>> >>> Try attaching gdb to the runaway tdepowersave process--that should at >>> least show you where it is stuck. >>> >>> Tim >>> >> >> Will do - and report back. >> >> > > Darn, > > On attaching the debugger to the launch, you get some interesting output, but > then tdepowersave goes back to taking its normal 5-7% of CPU instead of 97% of > CPU. Maybe this just happens on initial startup? Is its failure to register with > systemd the problem? Here is the output: > > [Thread debugging using libthread_db enabled] > Using host libthread_db library "/usr/lib/libthread_db.so.1". > [FIXME] UNCLASSIFIED DEVICE name: broadcast type: (null) subsystem: clockevents > driver: (null) [Node Path: (null)] [Syspath: > /sys/devices/system/clockevents/broadcast] [(null):(null)] > [FIXME] UNCLASSIFIED DEVICE name: clockevent0 type: (null) subsystem: > clockevents driver: (null) [Node Path: (null)] [Syspath: > /sys/devices/system/clockevents/clockevent0] [(null):(null)] > TQSocketNotifier: Multiple socket notifiers for same socket 12 and type read > tdepowersave: WARNING: The session is not registered with systemd > [Inferior 1 (process 31672) exited normally] > > Is there a way I can configure it to start with the debugger attached on tde > start? > Tim, tdepowersave starts at it's normal 5-7% and then steadily climbs in CPU use until it will take 100% on my box. Attaching the pid to gdb yielded the following: sudo gdb attach 674 ... 0xb7795424 in __kernel_vsyscall () (gdb) continue Continuing the tdepowersave process I watch the CPU use steadily climbing over the next few minutes. I captured a screenshot at 14%, now it is well over 26% of the CPU: http://www.3111skyline.com/dl/dt/trinity/ss/tdeowersave-14percent.jpg Interrupting tdepowesave in gdb output the following: ^C Program received signal SIGINT, Interrupt. 0xb5ac1b1f in ?? () from /usr/lib/libdbus-1.so.3 I don't know if that is where it is stuck or if it is stuck at the __kernel_vsyscall ()? What would you like me to try? -- David C. Rankin, J.D.,P.E.
