Description:

When I am browsing the web with chromium (with lots of tabs/images/etc.) or running virtual machines with kvm, occasionally (at least a few times a day) my system will suddenly become unresponsive for several minutes, until the OOM killer is invoked and the system returns to normal. However, I'm not convinced that it is simply an out-of-memory problem, because:

- OOM killer is always invoked when there is plenty of free RAM (at least 4-700MB out of 4GB, I have no swap), see dmesg output below which shows usage of current processes that does not go over the limit
- the majority of all processes instantly become stuck in the D state out of nowhere when the hanging occurs
- killing e.g. qemu or chromium from the console always returns the system to normal (again, when there is still even free RAM)

After installing ulatencyd from AUR I was able to have enough system responsiveness to run a few commands from the console during the hang:

About the D processes, here is output of ps axl | awk '$10 ~ /D/':

1 0 36 2 20 0 0 0 conges D ? 0:25 [kswapd0]
4 0 203 187 20 0 598860 297336 conges Dsl+ tty1 31:44 /usr/sbin/X :0 -auth /run/lightdm/root/:0 -nolisten tcp vt1 -novtswitch
4 0 14486 1 20 0 146056 21264 conges DLsl ? 11:48 /usr/bin/ulatencyd
0 1000 16630 566 20 0 976948 171812 conges Dl ? 1:46 /usr/lib/chromium/chromium --incognito --password-store=kwallet
1 1000 17740 16639 25 5 1054504 70300 conges DNl ? 0:11 /usr/lib/chromium/chromium --type=renderer --disable-databases --lang=en-US --force-fieldtrials=DeferBackgroundExtensionCreation/Deferred/Prefetch/ContentPrefetchPrefetchOn/Prerender/PrerenderControl/PrerenderFromOmnibox/OmniboxPrerenderEnabled/UMA-New-Install-Uniformity-Trial/Control/UMA-Session-Randomized-Uniformity-Trial-5-Percent/group_12/UMA-Uniformity-Trial-1-Percent/group_68/UMA-Uniformity-Trial-10-Percent/group_09/UMA-Uniformity-Trial-100-Percent/group_01/UMA-Uniformity-Trial-20-Percent/group_02/UMA-Uniformity-Trial-5-Percent/group_12/UMA-Uniformity-Trial-50-Percent/group_01/ --enable-deadline-scheduling --disable-client-side-phishing-detection --disable-gl-multisampling --disable-accelerated-2d-canvas --disable-accelerated-video-decode --channel=16630.30.1536067666
0 1000 25026 1 20 0 179156 840 conges D ? 0:00 journalctl -rn3
0 1000 25046 25044 20 0 8 4 conges D+ pts/3 0:00 [awk]

Not sure what function could be "conges" here, I'm assuming it's waiting on the hard drive or some driver bug perhaps?

And a picture of iotop during the hang, showing very high IO of most processes for seemingly no reason: http://i.imgur.com/Ws0mE7S.jpg

The problem has been occurring at least since the last few updates to the linux package, but I cannot remember how long this has been happening. I have not noticed it with any other applications besides qemu and chromium, but I also don't do a lot of intensive computing outside of those programs.

Another interesting tidbit is that my USB mouse always appears to disconnect/re-connect itself sometime during the hang, but no other USB device exhibits this problem.

[477439.679672] usb 3-2.2: USB disconnect, device number 13
[477441.146367] usb 3-2.2: new low-speed USB device number 14 using xhci_hcd
[477441.164535] usb 3-2.2: ep 0x81 - rounding interval to 64 microframes, ep desc says 80 microframes
[477441.239936] input: Logitech USB Optical Mouse as /devices/pci0000:00/0000:00:14.0/usb3/3-2/3-2.2/3-2.2:1.0/input/input23
[477441.240216] hid-generic 0003:046D:C05A.000C: input,hidraw1: USB HID v1.11 Mouse [Logitech USB Optical Mouse] on usb-0000:00:14.0-2.2/input0

The hardware is Acer W700 tablet with i5-3337U CPU, 256GB SSD, 4GB ram. I have another laptop (Panasonic CF-SX1 with i5-M) with the same software setup that does not have this problem.

Additional info:

Kernel version: 3.12.8-1-ARCH
Devices in system: http://dpaste.com/1572964/
dmesg: http://dpaste.com/1572965/