ivybridge system gets completely unusable after upgrade to vivid thanks to defaulting to immature intel_pstate driver

[Oliver Grawert]

I finally got around to upgrade my work laptop (first gen dell XPS sputnik system (ivy bridge)) yesterday, to find myself with a completely unusable and sluggish desktop, constantly spinning CPU fans and a hot enough laptop case that you can burn your fingers after the first reboot after install ... the device didn't stay on for more than 10-15min in a row before the BIOS temperature monitor initiated emergency shutdowns.

i also noticed a constant load average above 30.00 in htop

after a few hours of researching i found that we apparently switched to the intel_pstate driver by default in vivid ...

running a very simple script like:
while true; do
    clear
    cat /proc/cpuinfo |grep MHz
    sleep 1
done

...revealed that the cores unconditionally switch to something like 133MHz even though there is currently a lot going on in the system (firefox trying to recover 150 tabs after boot, evolution syncing with an IMAP server etc). it looks to me like the intel_pstate driver (at least on ivybridge hardware) is far from being ready for end users and if i wouldn't have known how to do a detailed research on these issues i would most likely have switched to another distro or to a different known working ubuntu release. could we turn this default back to the ondemand cpufreq governor in an SRU until the driver has matured enough, so we do not lose our ivybridge users and do not generate bad press ?

ProblemType: Bug
DistroRelease: Ubuntu 15.04
Package: linux-image-3.19.0-18-generic 3.19.0-18.18
ProcVersionSignature: Ubuntu 3.19.0-18.18-generic 3.19.6
Uname: Linux 3.19.0-18-generic x86_64
ApportVersion: 2.17.2-0ubuntu1.1
Architecture: amd64
AudioDevicesInUse:
 USER PID ACCESS COMMAND
 /dev/snd/controlC0: ogra 2120 F.... pulseaudio
CurrentDesktop: Unity
Date: Sun May 31 11:53:35 2015
DistributionChannelDescriptor:
 # This is a distribution channel descriptor
 # For more information see http://wiki.ubuntu.com/DistributionChannelDescriptor
 canonical-oem-somerville-precise-amd64-20120703-2
HibernationDevice: RESUME=UUID=145593f3-8085-4308-99b1-e1ff9c80f273
InstallationDate: Installed on 2013-11-05 (572 days ago)
InstallationMedia: Ubuntu 12.04 "Precise" - Build amd64 LIVE Binary 20120703-15:08
MachineType: Dell Inc. XPS L322X
ProcFB: 0 inteldrmfb
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-3.19.0-18-generic root=UUID=d0a6e4d2-9705-4ce3-97be-8ec33e06f2cb ro quiet splash intel_pstate=disable crashkernel=384M-:128M vt.handoff=7
RelatedPackageVersions:
 linux-restricted-modules-3.19.0-18-generic N/A
 linux-backports-modules-3.19.0-18-generic N/A
 linux-firmware 1.143.1
SourcePackage: linux
SystemImageInfo:
 current build number: 0
 device name:
 channel: daily
 last update: Unknown
UpgradeStatus: Upgraded to vivid on 2015-05-30 (0 days ago)
dmi.bios.date: 08/28/2013
dmi.bios.vendor: Dell Inc.
dmi.bios.version: A10
dmi.board.name: 0PJHXN
dmi.board.vendor: Dell Inc.
dmi.board.version: A00
dmi.chassis.type: 8
dmi.chassis.vendor: Dell Inc.
dmi.chassis.version: 0.1
dmi.modalias: dmi:bvnDellInc.:bvrA10:bd08/28/2013:svnDellInc.:pnXPSL322X:pvr:rvnDellInc.:rn0PJHXN:rvrA00:cvnDellInc.:ct8:cvr0.1:
dmi.product.name: XPS L322X
dmi.sys.vendor: Dell Inc.

[Oliver Grawert (ogra)]

just for completeness (in case someone looks for a wrokaround), adding intel_pstate=disable to GRUB_CMDLINE_LINUX_DEFAULT in /etc/default/grub and running sudo update-grub got me the ondemand driver back and got the system back to behave as stable as in 14.10.

[Brad Figg (brad-figg)]

This change was made by a bot.

[Alberto Salvia Novella (es20490446e)]

Also affects Sandy Bridge, so probably it is affecting several processors.

[Alberto Salvia Novella (es20490446e)]

In Sandy bridge the system is not rendered completely unusable. But it still becomes hot, with fans constantly turning on and off, and the computer powering off from time to time.

And this did not happen in Ubuntu 14.10.

[Colin Ian King (colin-king)]

I suggest trying thermald to see if this helps

[Oliver Grawert (ogra)]

one thing i noticed is that after dropping intel_pstate=disable from teh kernel commandline, thermald does not seem to star/run at all .. starting it manually i captured http://paste.ubuntu.com/11541063/
note that the behavior gets minimally better with thermald running, but switching workspaces is still as slow as about 20-30 sec in slow motion ...
http://imgur.com/Ukuv34t shows one terminal with the script above and one terminal with htop running ... this is with thermald active ... the CPU normally switches between 800MHz and 2.8GHz with the ondemand governor in use ...
with thermald i se it between 133MHz and at very short spikes up to 3.5GHz (but even when the cores run at that sped the system only gets responsive for a few seconds)

turning intel_pstate back off gets me a relatively usable system back ...

[Oliver Grawert (ogra)]

oops, totally forgot this is still open, the problem turned out to be a completely stuck fan wrapped in filth ... cleaning it ave the system proper values so thermald does the right thing now.