7

I have successfully upgraded from Ubuntu 15.04 to Ubuntu 15.10 using do-release-upgrade on my Acer Aspire E11 laptop. However the whole computer freezes (i.e. gets unresponsive) after some minutes of work. It's impossible to switch to text screens and there is nothing in the logs.

  • Windows is working well on the same machine, so perhaps there is a problem somewhere in Linux or its drivers.

  • The problem is not caused by the system upgrade. The problem is there also on a fresh installation of 15.10.

  • The problem is not caused by faulty RAM. The memtest86 was able to run successfully for 14 hours.

  • The problem doesn't seem to be related to WiFi because it occurs even if WiFi is deactivated in the menu, i.e. Enable Wifi is unchecked.

  • If Chromium browser is not running the system never freezes. How is it possible for the user-space application to block the whole system?

  • The system freezes even if Chromium browser is not used, but it doesn't happen so often in this case.

Honestly I don't know how to debug such a situation. How to obtain useful information from a frozen system? What shall I look for? What to enable/disable? What to try?

karel
  • 114,770
dlask
  • 175
  • Try running with only 1 RAM, IF YOU HAVE JUST 1 RAM, it may be faulty. – Star OS Oct 25 '15 at 09:00
  • RAM is OK, see the updated question. – dlask Oct 27 '15 at 06:36
  • 1
    The same problem on ASUS X551 – Extender Nov 02 '15 at 03:08
  • 1
    Same problem exists on Acer E3-111. It seems to happen prominently while watching videos (VLC and Chrome/Youtube) and playing Ren'Py games – matega Nov 03 '15 at 08:46
  • Do you know what SSH is and how to use it? If so, try SSHing into your system when it's freezed, and see if you can. If you can it could simply be a graphical freeze, not a whole system + kernel freeze. – Léo Lam Jan 06 '16 at 11:43
  • Also, I was having this issue on an Acer laptop before, and the issue was Wi-Fi. Even disabled the driver would crash and since it's a kernel module, the whole kernel would panic and crash. – Léo Lam Jan 06 '16 at 11:45
  • I have to install the SSH server first but it is not a problem. However, there is a possibly interesting hint: It happened once that the system got frozen but a piece of sound kept playing, like - say - 250 ms fragment played in a loop. Could it mean anything? – dlask Jan 06 '16 at 11:47
  • I have the same problem on the same machine, glad I found you! I'm also glad that the HDD should be ruled out since I swapped it out with an SSD (more glad that it means I can keep the SSD) – matega Jan 31 '16 at 18:46
  • May I know your kernel version please. That may solve the problem. – sangharsh Feb 12 '16 at 12:55
  • I use the kernel provided by Ubuntu 15.10, currently 4.2.0-27-generic x86_64. Is there any relationship between the kernel version and the reported problem? – dlask Feb 12 '16 at 17:36
  • 1
    Seems to be a bug in upstream kernel https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1509723 https://bugzilla.kernel.org/show_bug.cgi?id=109051 – Peter Gibson May 23 '16 at 01:19

2 Answers2

2

I have been suffering freezes on my Acer E5-511 ever since 15.10, and upgraded to 16.04 2 weeks ago in the hope it would go away, but the problems continued, freeze every 10-30 min max ( and usually when moving the mouse strangely!).

I finally tried this 3 days ago in /etc/default/grub

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash acpi_osi=Linux i8042.nopnp intel_idle.max_cstate=0 processor.max_cstate=1"

and have so far been crash free

Updated by OP

The single required option is intel_idle.max_cstate=2 so the necessary modification of /etc/default/grub is:

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_idle.max_cstate=2"

Perhaps the Acer Aspire E11 hardware is buggy, unable to wake up CPU from idle states higher than C2.

dlask
  • 175
David Walker
  • 315
  • I am going to try it. More precisely, I would try anything because I am really desperate. What's the meaning of these parameters? Provided that they really solve the problem, does it mean there is a bug in Linux kernel? – dlask Jan 29 '16 at 08:28
  • It's something to do with power management - see http://www.breakage.org/2012/11/14/processor-max_cstate-intel_idle-max_cstate-and-devcpu_dma_latency/ – David Walker Jan 29 '16 at 10:16
  • You suggest these parameters: acpi_osi=Linux, i8042.nopnp, intel_idle.max_cstate=0, and processor.max_cstate=1. Are all of them necessary to fix the problem? – dlask Jan 30 '16 at 16:06
  • The ones in italics are relevant - the " i8042.nopnp" is an attempt to solve a touchpad problem which seems to plague all ACER Exx laptops . I am now 5 days without a freeze, so it looks like this is the cure. – David Walker Jan 31 '16 at 17:09
  • Having studied many things about the *Intel C states", I am experimenting with the intel_idle.max_cstate parameter alone. The value 3 doesn't solve the problem so I am using the value 2 now. The test proceeds. – dlask Jan 31 '16 at 17:15
  • don't forget "sudo update-grub" after the CMDLINE change, otherwise the changes aren't applied. Just in case – David Walker Jan 31 '16 at 17:17
  • Don't worry, I will accept your answer because it seems to really address the problem. Just (if you don't mind) I would extend it with the minimal set of necessary parameters together with their precise values after my test is complete. However, a question persists: Is it a hardware or software problem? – dlask Jan 31 '16 at 17:25
  • 1
    I think it's hardware - it seems to be only ACER E1 / E5 that are affected. I have had 2 Acers before this one (E5-511) which is a pile of ****. The MB blew up 3 weeks after purchase and Acer took 4 months to repair it under guarantee – David Walker Jan 31 '16 at 17:33
  • @David Walker Its not just about ACER because I faced the same problem on Lenovo notebook. It is probably about kernel. – sangharsh Feb 12 '16 at 12:57
2

Intel recently issued an erratum regarding all processors based on the Bay Trail architecture. Because of a design mistake the C6 power state cannot be used normally and using this power state can lead to a software failure (see VLP52 in this document).

A script written by Wolgang M. Reimer disables the problematic C6 states. I copied this script into /etc/rc.local. This fix can lead to greater power consumptions but prevented Kubuntu from freezing on my Acer Aspire V3.

Your Intel processor is affected if it is based on the Bay Trail architecture. A list of Bay Trail based products is available on ark.intel.com. Currently the following products are affected:

Intel Atom E3805 E3815 E3825 E3826 E3827 E3845
Intel Atom Z3735D Z3735E Z3735F Z3735G Z3736F Z3736G Z3740 Z3740D Z3745 Z3745D Z3770 Z3770D Z3775 Z3775D Z3785 Z3795
Intel Celeron J1750 J1800 J1850 J1900
Intel Celeron N2805 N2806 N2807 N2808 N2810 N2815 N2820 N2830 N2840 N2910 N2920 N2930 N2940
Intel Pentium A1020
Intel Pentium J2850 J2900
Intel Pentium N3510 N3520 N3530 N3540

This information was published on heise.de.

Stéphane Tréboux
  • 2,979