This bug was fixed in the package ubuntu-drivers-common - 1:

ubuntu-drivers-common (1: xenial-proposed; urgency=medium

  [ Alberto Milone ]
  * gpu-manager.c:
    - Fix memory leak in get_alternatives(). Make sure to always free
      alternatives->current_core and alternatives->current_egl.
    - Partially drop the drm code, and get information about connectors
      from sysf. This fixes a long standing performance issue (LP: #1586933).
    - Disable persistence mode before powering down the dGPU.
      If persistence mode is not disabled, gpu-manager will not be able to
      unload the nvidia module (LP: #1619306).
    - Make sure to load and unload nvidia-drm. This fixes a problem that
      prevents the dGPU from being powered off in hybrid systems with
      recent nvidia drivers (LP: #1619306).

  [ Ying-Chun Liu (PaulLiu) ]
  * Quirks/
    - Catch and ignore UnicodeDecodeErrors in dmi info, as some machines
      may report board_name as ffffffffffff0a to mean empty, and this
      causes an error in python 3 (LP: #1320868).

 -- Alberto Milone <>  Mon, 06 Jun 2016
12:49:19 +0200

** Changed in: ubuntu-drivers-common (Ubuntu Xenial)
       Status: Fix Committed => Fix Released

You received this bug notification because you are a member of नेपाली
भाषा समायोजकहरुको समूह, which is subscribed to Xenial.
Matching subscriptions: Ubuntu 16.04 Bugs

  gpu-manager's drm output detection affects boot time

Status in ubuntu-drivers-common package in Ubuntu:
  Fix Released
Status in ubuntu-drivers-common source package in Xenial:
  Fix Released

Bug description:
  SRU request:

  The current drm code to probe the connected outputs is too resource 
intensive, and slow down the boot process.

  [Test Case]
  1) Boot Ubuntu and run the following command, looking for gpu-manager:
  systemd-analize blame

  This will show how long gpu-manager took. Write down the number

  2) Install the package from -proposed, and run systemd-analize blame.
  See if the number has become smaller.

  [Regression Potential]
  Low, all the changes are already included in Yakkety.

  Rather than calling drm directly, gpu-manager should simply check the outputs 
connector status by using the sysfs filesystem (e.g. 
/sys/class/drm/card0-HDMI-A-1/status, etc.), thus relying on data that is 
already available.

  This should make gpu-manager a bit less resource intensive, especially
  on boot.

  The attached perf shows the problem in gpu-manager.

To manage notifications about this bug go to:

Mailing list:
Post to     :
Unsubscribe :
More help   :

Reply via email to