kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

positron · 2011-07-20 20:15:57

kokoko3k wrote:

@positron: Did you had the wakeup issue after resuming from ram/disk ?
What's your hardware?
Thanks

Sorry for the lag....

No, i had random heavy loads of kworker, even just after booting up. But unfortunately I am still facing problems. My problem is back.
It is totaly random.

My graphics card is VGA compatible controller: Intel Corporation Mobile 4 Series Chipset Integrated Graphics Controller

and i get with powertop interrupts form i915 with 2.6.39 kernel and xf86-intel-git driver.

matiit · 2011-07-20 22:00:23

I'm having the same problem, Even I have the same Grahics card (intel)

Markus00000 · 2011-08-09 09:40:24

It seems as if this issue completely disappeared for me sometime within the last week. Now using 3.0-ARCH on x86_64.

Zucca · 2011-08-16 12:49:38

Markus00000 wrote:

It seems as if this issue completely disappeared for me sometime within the last week. Now using 3.0-ARCH on x86_64.

... But here's one who's having this problem even with 3.0-ARCH.

In my case kworker starts to eat CPU when I plug USB device in while my laptop is on battery power. After I plug power source in, the usb device gets recognized and kworker stops running so bad.

There have been few cases when plugging memory card in does not cause this, but it's extremely rare. If it happens to work, after ejecting and and putting it back, kworker gets mad again. Basically it might work out by unplugging and plugging device many times until kworker "eases down". Only truly working solution I have is to plug power source in. I have some doubts that this is a bug in some power saving component in kernel... Maybe usbautosuspend? I have added 'usbcore.autosuspend=-1' to my kernel boot line, but it has no effect.

Last edited by Zucca (2011-08-16 12:50:30)

Markus00000 · 2011-08-16 13:12:08

Sorry to hear that you still have this annoying issue. I haven't seen it since.

Zucca wrote:

Only truly working solution I have is to plug power source in.

Just in case you missed this:

# rmmod ehci_hcd
# modprobe ehci_hcd

For me this stopped kworker instantly while USB devices were plugged in.

EDIT: The issue is back... But at least the above commands still fix it.

Last edited by Markus00000 (2011-08-17 07:22:47)

Zucca · 2011-08-19 09:45:52

Markus00000 wrote:

Just in case you missed this:
# rmmod ehci_hcd
# modprobe ehci_hcd
For me this stopped kworker instantly while USB devices were plugged in.
EDIT: The issue is back... But at least the above commands still fix it.

Works, but it's so d*mn dirty.

Zucca · 2011-08-19 10:17:48

Oh WAIT...
It only works partially. The device gets recognized, but kworker still likes the taste of my CPU. (OM-NOM <3) ... (I'm really losing my mind on this...)

If someone finds this information useful:

$ inxi -xxSMCf
System:    Host: Mieu Kernel: 3.0-ARCH x86_64 (64 bit, gcc: 4.6.1) Desktop LXDE (Openbox 3.5.0) Distro: Arch Linux
Machine:   System: Acer (portable) product: Aspire 1810TZ version: v1.3314
           Mobo: Acer model: JM11-MS Bios: INSYDE version: v1.3314 date: 08/31/2010
CPU:       Dual core Intel CPU U4100 (-MCP-) cache: 2048 KB bmips: 5188.82 
           Clock Speeds: 1: 1300.00 MHz 2: 1200.00 MHz
           CPU Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts 
           acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx lm constant_tsc arch_perfmon pebs bts rep_good nopl 
           aperfmperf pni dtes64 monitor ds_cpl est tm2 ssse3 cx16 xtpr pdcm xsave lahf_lm dts

kpiche · 2011-09-06 17:44:03

On my laptop I found I had to reload the uhci_hcd module as well as the ehci_hcd module.

modprobe -r uhci_hcd
modprobe -r ehci_hcd
modprobe ehci_hcd
modprobe uhci_hcd

bbhl · 2011-09-09 15:27:38

Had the same issue with kworker on a new 3.0 kernel installation...
This did help seems: http://souriguha.wordpress.com/2011/03/ … 2-35-2-36/

Zucca · 2011-09-09 16:00:47

Strangely this bug seems to be very random.
With some people simple mopdule reloading works, with the rest it doesn't.
And at least in my case it occurs randomly, but still quite often. Only working remedy is to plug power source in.
I need to test those last two ways in last two post...

joe of loath · 2011-09-12 13:43:48

I've been having this problem since ~2.6.36, still no fix. I used the LTS kernel for a while, but the last LTS upgrade broke my graphics, and I can only run in 1024x768 with no hardware acceleration.

On my laptop, plugging in the power doesn't seem to make a difference, either, but if I rmmod ehci_hcd, then CPU usage goes back to normal, but it's such a pain.

hramat · 2011-10-07 18:51:14

Andevellicus wrote:

I'm having the same issue. Kernel 2.6.38, on my ThinkPad T61.
I found that for whatever reason the problem started once I ran the WICD daemon. I uninstalled WICD and went with networkmanager, but as soon as I started the networkmanager daemon, kworker nuked my CPU again. Stopping WICD daemon, stopping networkmanager daemon, killing wicd, etc - none of that worked.
For the moment I uninstalled WICD and networkmanager and configure my wireless manually (which is a pain, but it works). The weird thing is that it just happened all of a sudden.
EDIT:
Went back to kernel 2.6.37 and I'm not having any more problems.

I just got to the same trouble (after kernel update) as Andevellicus earlier, also having ThinkPadT61.
Setting pcie_ports=compat or native did not help

Had to also uninstall the WICD and to configure wireless manually (and later via rc.conf)

gdirty · 2011-10-13 02:58:08

Yeah I'm having the same deal on a Lenovo T400, it's obnoxious. I had huge problems with udevd for a long time and now this. The behavior has basically no rhyme or reason.
I notice I get a lot of interrupts from ps2 when I run powertop.

Is there anything better to figure out what is actually going on? My home computer is basically dead. It's extremely depressing.

For the alternative of running an older kernel, how do you get that going?

hramat · 2011-10-13 10:06:32

gdirty,
It is not entirely clear if the problem is the same as I have reported. Did you problem also appeared after kernel update? Do you run networkmanager or wicd.

In my case my computer works very nice without these network daemons. However I had to change the network setting in arch for manual, which is actually fine.

gdirty · 2011-10-13 15:47:36

well I had problems with udevd in earlier kernels, then kworker since .36 maybe? I use 'netcfg' for my wired connection, and wicd for wireless. Doesn't seem to matter what I do though. I have not yet tried just running various programs without an internet connection.

kokoko3k · 2011-10-13 17:23:03

the interrupt issue from ps2 is normal if you keep your finger on the touchpad, it is an hw issue and there's nothing the kernel can do.

hramat · 2011-10-13 18:10:16

gdirty

what is the reason for using netcfg for wired connection? netcfg wiki: For systems connecting to a single network, the network daemon may be more appropriate

coolaman · 2011-10-14 05:53:23

Same behavior .

But after removing the line "echo 0 > /proc/sys/kernel/nmi_watchdog" like powertop say, it works normally.

gdirty · 2011-10-18 06:39:21

After booting a live cd and observing the same problem in another distro I did another round of googling and ended up with this as the very first result:
https://bugzilla.redhat.com/show_bug.cgi?id=655694

Adding:
drm_kms_helper.poll=0
to my boot options seems to have ended the problem. I found instructions suggesting to add

options drm_kms_helper poll=N

to /etc/modprobe.d/modprobe.conf but the problem remained until changing the actual boot options.

Basically the reason I use wmii/ netcfg is that I am lazy and also in the 21st century I want my computer to do most of its own thinking

piotroxp · 2011-10-18 22:11:09

Hiya,
Arch 3.0, Acer S3820TG, fresh install as of yesterday. Powertop states that kworker is munching around 50% of computing power. Also, I have 2 kworker threads (one for each i3 core?) . All the suggestions above are not working , this is how my boot cmd line looks like:

kernel /boot/vmlinuz-linux root=/dev/sda5 ro pcie_aspm=force acpi_osi=Linux pcie_ports=native usbcore.autosuspend=-1 drm_kms_helper.poll=0

The SOB is eating up my batter alive. Any suggestions? Also, have laptop tools and others installed. Dropbox is causing the fuss maybe ??

BR

Zucca · 2011-10-26 18:20:34

Hi. I solved the problem here. I hope this works for you as well. I actually solved the problem a while ago, but it took some time to remember what I actually did, because at first I didn't even noticed that everything was ok.

I did some searching using find and grep. Laptop mode tools caused this by forcing usb autosuspend on.

You could try to list all the files that have something to do with "autosuspend". Searching /etc/laptop-mode is enough I think.
Run:

find /etc/laptop-mode/ -type f | xargs grep -il 'autosuspend'

Then check the files. If you find any 'CONTROL_USB_AUTOSUSPEND' set them to '0'

CONTROL_USB_AUTOSUSPEND="0"

If this does not work then edit /etc/laptop-mode/laptop-mode.conf and disable module auto-loading

ENABLE_AUTO_MODULES=0

and edit module config files (in /etc/laptop-mode/conf.d/) and enable or disable them by hand.

Finally, restart laptop-mode

sudo rc.d restart laptop-mode

akurei · 2011-10-26 19:10:45

I am Not using laptop-mode-tools. Still I experience those wakeups.

Zucca · 2011-10-27 22:13:30

This:

options usbcore autosuspend=0

in /etc/modprobe.d/modprobe.conf too?

piotroxp · 2011-10-31 20:40:42

Okay, so I've done a lot of things mentioned in this topic, combining, changing etc. etc. Nothing changes. But let me post some observations. I am using ARCH-3.0 on a AS3820TG ACER laptop. Updated on the daily.

- When the OS starts, the fans start work as if the load on the CPU is changing. One can make out a periodical change in the fan speeds, which pretty much resembles how the kworker threads work. I can't see, however, during the startup of which module the symptom described above starts taking place. Don't want to test it by disabling all modules and running one at a time, or daemons etc... I don't have the time.
- powertop is showing kworker0:0, kworker0:1 and Load balancing tick to be constantly using over 50% of the processor time. It's retarded.
- I've tried disabling the usb autosuspend via the kernel cmd line, via the modprobe method, loading/unloading uhci/ehci and by configuring laptop-mode, but to no avail. Since there is someone who ruled out laptop-tools as the responsible software, I'm not removing it.
- kworker 0:1 is almost always over 30%.
- I've gone all the way with powertop, abiding to whatever it wanted (disabling, enabling etc etc). Kworker persists.
- powertop says that enabling CONFIG_NOTIFY in the kernel options would help (I believe that it's something to reduce load on the HDD) - thats where powertop stops being a smartass. and kworker still using up power like crazy
- the battry is being drained. seriously. when on windows, the battery is used up by a ratio of almost 0,3 times slower. (6 hrs w7 vs 2,5 hrs arch). It's most certainly because kworker is .... kworking.
- when I stop typing (for instance) the kworker load rises to more than 50% almost instantaneously. wth?
- unplugging/plugging usb devices - nothing changes . devices work properly though.
- unplugging/plugging DC - the kworker threads eat up almost 70% of the processor time.
- strange - upon pluggin in an usb mouse the load in powertop was lesser than 0 although -60% wakeup from kworker is still 60%...

So long post short - there is something that kworker 1 and 0 do, that requires a lot of load balancing. wat do?

--little update:
I've tried to disable all the load balancers that are supposed to save power when using laptop-mode, but that didn't help either.

posting my kernel cmd line

kernel /boot/vmlinuz-linux root=/dev/sda5 ro pcie_aspm=force acpi_osi=Linux pcie_ports=native drm_kms_helper.poll=0

When I wasn't using the pcie_ports=native I couldn't use the fn key + brightness change on my laptop. That occured as very strange to me...

Last edited by piotroxp (2011-11-01 08:06:29)

jason32835 · 2011-11-02 01:07:24

Wow, great post piotroxp.. this is exactly the same situation I'm experiencing to a tee.

I did go ahead and disable every module and daemon I could and kworker kept chugging away. I may install a fresh system to see if this persists. This is a show stopper for me as the battery life is total crap.

Has anyone experienced this on a fresh install? Is it even worth checking?

Arch Linux

#26 2011-07-20 20:15:57

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#27 2011-07-20 22:00:23

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#28 2011-08-09 09:40:24

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#29 2011-08-16 12:49:38

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#30 2011-08-16 13:12:08

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#31 2011-08-19 09:45:52

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#32 2011-08-19 10:17:48

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#33 2011-09-06 17:44:03

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#34 2011-09-09 15:27:38

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#35 2011-09-09 16:00:47

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#36 2011-09-12 13:43:48

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#37 2011-10-07 18:51:14

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#38 2011-10-13 02:58:08

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#39 2011-10-13 10:06:32

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#40 2011-10-13 15:47:36

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#41 2011-10-13 17:23:03

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#42 2011-10-13 18:10:16

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#43 2011-10-14 05:53:23

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#44 2011-10-18 06:39:21

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#45 2011-10-18 22:11:09

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#46 2011-10-26 18:20:34

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#47 2011-10-26 19:10:45

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#48 2011-10-27 22:13:30

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#49 2011-10-31 20:40:42

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

#50 2011-11-02 01:07:24

Re: kworker constantly using 15/20% of CPU since upgrade to 2.6.36-ARCH

Board footer