[Bug 955287] Re: Ubuntu should handle "hot" CPUs by taking preemptive action and warning users

[Steve Langasek]

This probably should not be handled in the acpi-support package, since
acpi-support is considered deprecated and we're trying (without much
success) to phase it out.  I think notifications would be better handled
through upowerd.

However, even without desktop notifications, the kernel is supposed to
have built-in support for stepping down the CPU automatically when the
system gets too hot.  Perhaps the thresholds are wrong in your ACPI
tables?  For instance, my ACPI is buggy and has a "passive" threshold
that's at a higher temperature than the "critical" threshold:

$ cat /sys/class/thermal/thermal_zone0/trip_point_*
100000
critical
127500
passive
$

My understanding is that when these are correct, the kernel will take
care of things automatically; and when they're wrong, as above, I'm not
sure there's much point to working around it in userspace.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to acpi-support in Ubuntu.
https://bugs.launchpad.net/bugs/955287

Title:
  Ubuntu should handle "hot" CPUs by taking preemptive action and
  warning users

Status in “acpi-support” package in Ubuntu:
  New

Bug description:
  If the kernel detects your CPU(s) is/are too hot currently (see bug
  751689), the kernel calls /sbin/poweroff. This will provide a
  "graceful" system shutdown. If /sbin/poweroff fails, the kernel just
  forcibly shuts the system down. However, both strategies are last
  resorts and are called when the system temperature has reached a
  critical level.

  However, the kernel seems to emit an ACPI event when it detects the
  CPU(s) are merely "hot". I suggest we consider adding an acpi hook to
  attempt to avoid a "critical" scenario.

  Currently, the user experience when "critical" gets hit is not good -
  the system just shuts down with no warning whatsoever. This is
  alarming in the extreme to users.

  Ideas:

  - proactively attempt to kill off power hogging processes (use powertop?)
  - ramp fans to maximum and present the user with a warning window explaining what is happening.
  - present the user with a window of high-power processes and ask *them* to select the processes they'd like to kill off in an effort
    to avoid a system shutdown.

  Problems:

  - it is unclear (to me atleast) how close (in terms of degrees
  centigrade) "hot" and "critical" are (is it different for all CPUs ?)
  As such, it is unclear how long (time) it might take for a system that
  is hot to go critical and just shutdown.

  ProblemType: Bug
  DistroRelease: Ubuntu 12.04
  Package: acpi-support 0.140
  ProcVersionSignature: Ubuntu 3.2.0-18.29-generic-pae 3.2.9
  Uname: Linux 3.2.0-18-generic-pae i686
  NonfreeKernelModules: nvidia
  ApportVersion: 1.94.1-0ubuntu2
  Architecture: i386
  Date: Wed Mar 14 17:22:09 2012
  InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release i386 (20101007)
  ProcEnviron:
   TERM=xterm
   PATH=(custom, user)
   LANG=fr_CA.UTF8
   SHELL=/bin/bash
  SourcePackage: acpi-support
  UpgradeStatus: Upgraded to precise on 2012-01-12 (62 days ago)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/acpi-support/+bug/955287/+subscriptions