kernel-packages team mailing list archive

[Will Deacon <1512153@xxxxxxxxxxxxxxxxxx>]

I tried using the v4.4-rc5 mainline build from :

http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.4-rc5-wily/

and everything is now working nicely.

Consequently, I'm assigning this to the linux (kernel) package, since it
looks like the issue is there. Unfortunately, I don't specifically what
needs to be backported.

** Package changed: ubuntu => linux (Ubuntu)

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1512153

Title:
  Secondary CPUs switch to "performance" cpufreq following
  suspend/resume cycle

Status in linux package in Ubuntu:
  Confirmed

Bug description:
  I have a Dell XPS 13 2015 laptop (Broadwell i7) using the intel_pstate
  driver for CPU power management. After a cold boot, the CPUs are all
  using the "powersave" cpufreq governor:

  $ cat /sys/devices/system/cpu/cpu{0..3}/cpufreq/scaling_governor
  powersave
  powersave
  powersave
  powersave

  However, following a suspend/resume cycle (e.g. by closing and opening the lid), the secondary CPUs end up in
  "performance" state:

  $ cat /sys/devices/system/cpu/cpu{0..3}/cpufreq/scaling_governor
  powersave
  performance
  performance
  performance

  This has the undesirable effect of forcing /sys/devices/system/cpu/intel_pstate/min_perf_pct to 100, pegging all of the CPUs
  at ~3.2Ghz and hosing the battery life of the system. I would expect the governor setting to persist across the suspend on all cores. Manually setting the secondary cores back to the "powersave" governor resolves the issue.

  There does appear to be a known kernel issue in this area (although it
  doesn't sound exactly the same):

    https://www.mail-archive.com/linux-
  kernel@xxxxxxxxxxxxxxx/msg996192.html

  I backported the supposed fix and the issue remained (but note that
  the backport is a PITA and I ended up pulling in most of the
  intel_pstate changes queued in linux-next). I'm currently working
  around the issue by compiling out the "performance" governor entirely.

  I didn't have this problem with 15.04, but reverting to the vivid 3.19-based kernel breaks resume entirely with wily userspace, so
  it's hard to tell exactly what is responsible. pm-utils has some scripts to poke around at the cpufreq governor, but I commented
  all that out and didn't see any change in behaviour.

  Any ideas?

  --->8

  $ lsb_release -rd
  Description:	Ubuntu 15.10
  Release:	15.10

  $ apt-cache policy linux-image-4.2.0-16-generic 
  linux-image-4.2.0-16-generic:
    Installed: 4.2.0-16.19
    Candidate: 4.2.0-16.19
    Version table:
   *** 4.2.0-16.19 0
          500 http://gb.archive.ubuntu.com/ubuntu/ wily/main amd64 Packages
          100 /var/lib/dpkg/status

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1512153/+subscriptions