View previous topic :: View next topic |
Author |
Message |
Voltago Advocate
Joined: 02 Sep 2003 Posts: 2593 Location: userland
|
Posted: Thu Jan 12, 2012 8:17 pm Post subject: Laggy GUI applications with kernel 3.2.0 |
|
|
So after using linux-2.6.39 for ages, I recently upgraded to linux-3.2.0-gentoo-r1 and ran into a problem that is rather hard to diagnose: After an uptime of two, three hours, the GUI of some (not all!) applications starts to get unresponsive and laggy. First, firefox and thunderbird are affected, and KDE's plasma desktop a little later. The applications just freeze up for 10-20s, then behave normally for a short time, then freeze up again. There doesn't seem to be any uncommonly high CPU usage going on at the time, and no unusual memory usage.
Has anybody witnessed similar behaviour in recent kernels? |
|
Back to top |
|
|
avx Advocate
Joined: 21 Jun 2004 Posts: 2152
|
Posted: Thu Jan 12, 2012 9:01 pm Post subject: |
|
|
For me, Opera sometimes has these short-time freezes, too. Don't know if it's related to the kernel or an Opera-bug, though. _________________ ++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
Back to top |
|
|
Voltago Advocate
Joined: 02 Sep 2003 Posts: 2593 Location: userland
|
Posted: Thu Jan 12, 2012 10:37 pm Post subject: |
|
|
Actually in my case it's rather short-time-unfreeze. I'm positive that what I'm seeing is somehow kernel-related (it might be a userland bug somewhere that shows up with the new kernel). Does reverting to an older kernel make your freezes go away? |
|
Back to top |
|
|
avx Advocate
Joined: 21 Jun 2004 Posts: 2152
|
Posted: Fri Jan 13, 2012 4:46 am Post subject: |
|
|
Hard to tell, I've got no way to forcefully reproduce this, yet. Besides, here Opera is the only GUI-application on this box, everything else runs fine, sorry.
Now that I've finally got WLAN on this machine - only took a year, damn Broadcom - I'd rather not downgrade and lay out all the cables again
BTW, this is my notebook I'm complaining about, my desktop doesn't have this problem(as of now). _________________ ++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
Back to top |
|
|
toralf Developer
Joined: 01 Feb 2004 Posts: 3922 Location: Hamburg
|
Posted: Fri Jan 13, 2012 12:31 pm Post subject: |
|
|
Hhm, I observed a similar behaviour with newer kernel - and dunno how to reproduce it. But sometimes it is related to my wlan which suddenly stops working w/o any messages ... |
|
Back to top |
|
|
Voltago Advocate
Joined: 02 Sep 2003 Posts: 2593 Location: userland
|
Posted: Sat Feb 11, 2012 2:47 pm Post subject: |
|
|
For the record, my troubles were apparently caused by one of the kernel boot parameters
Code: | i915.i915_enable_fbc=1 i915.lvds_downclock=1 |
which are supposed to lower power consumption for Intel integrated graphics. |
|
Back to top |
|
|
toralf Developer
Joined: 01 Feb 2004 Posts: 3922 Location: Hamburg
|
Posted: Sat Feb 11, 2012 4:07 pm Post subject: |
|
|
Voltago wrote: | For the record, my troubles were apparently caused by one of the kernel boot parameters
Code: | i915.i915_enable_fbc=1 i915.lvds_downclock=1 |
which are supposed to lower power consumption for Intel integrated graphics. | For me there were only few power consumptions tips useful.
One is to use the patch of kernel 3.2.5 (or 3.0.20) - that saves ~3 W. Beside that these helped me : Code: | $ sudo cat /etc/local.d/power.start
#!/bin/sh
#
#
# power saving
#
F=/sys/devices/system/cpu/cpufreq/ondemand/ignore_nice_load
if [[ ! -f $F ]]; then
echo "module acpi_cpufreq not loaded ?!"
modprobe acpi_cpufreq
fi
echo 1 > $F || echo "$F doesn't exist !"
/usr/sbin/ethtool -s eth0 wol d
/usr/sbin/hciconfig hci0 down
/sbin/iwconfig wlan0 power on
echo 3000 > /proc/sys/vm/dirty_writeback_centisecs
|
|
|
Back to top |
|
|
Voltago Advocate
Joined: 02 Sep 2003 Posts: 2593 Location: userland
|
Posted: Sat Feb 11, 2012 6:00 pm Post subject: |
|
|
Thanks, Toralf. By the way, I noticed that by default my system has
Code: | linux ~ # cat /proc/sys/vm/dirty_writeback_centisecs
60000
|
600 seconds for dirty_writeback seems a bit high. Should I be alarmed? |
|
Back to top |
|
|
nikulinpi n00b
Joined: 12 Jan 2007 Posts: 37
|
Posted: Sat Mar 17, 2012 4:34 pm Post subject: |
|
|
Same problem! I bet it should be related to some kind of cgroups nastiness |
|
Back to top |
|
|
toralf Developer
Joined: 01 Feb 2004 Posts: 3922 Location: Hamburg
|
Posted: Sat Mar 17, 2012 5:46 pm Post subject: |
|
|
Voltago wrote: | Thanks, Toralf. By the way, I noticed that by default my system has
Code: | linux ~ # cat /proc/sys/vm/dirty_writeback_centisecs
60000
|
600 seconds for dirty_writeback seems a bit high. Should I be alarmed? | Did you installed app-laptop/laptop-mode-tools ? |
|
Back to top |
|
|
nikulinpi n00b
Joined: 12 Jan 2007 Posts: 37
|
|
Back to top |
|
|
nikulinpi n00b
Joined: 12 Jan 2007 Posts: 37
|
|
Back to top |
|
|
|