You are not logged in.

#1 2016-01-17 15:51:35

ewaller
Administrator
From: Pasadena, CA
Registered: 2009-07-13
Posts: 19,744

Kernel 4.4 ACPI Blocking

I use an HP Envy 15 touch with an  Intel(R) Core(TM) i7-4720HQ CPU @ 2.60GHz

This system has problems with reading from the /sys file system for ACPI generated "files" when the system is busy.  I do not not if the trigger is temperature or CPU activity, but when the processor is under extended load (compiling a kernel, doing extreme calculations), reads from the sys file system as to processor temperature will block until the processor returns to lower levels.  I have seen it stay blocked for an hour while compiling Android.  During this time, i3bar stops updating, conky will stop updating, emacs (which displays processor load) stops updating.  In the mean time, i7z continues happily, as does htop, chromium, and most other user space software.  Builds continue just fine.  Journalctl reports CPU package over temperature, throttling, and MCEs being thrown.

The situation first appeared with the change to 4.x kernels, but has improved over the last half year to where it almost never happened in the last 4.3 kernel.  I also never had any MCEs or over temperature conditions.  Throttling was present, but I would expect that as a normal part of cooling.

With 4.4, the problem is back with a vengeance -- as bad as it has ever been with a 4.x kernel.

For now, I have backed down to the 4.3 kernel.  This is likely an upstream problem, but I thought I would post this prior to the 4.4 kernel going mainline.

Last edited by ewaller (2016-01-18 14:50:31)


Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way

Offline

#2 2016-01-19 05:39:00

ewaller
Administrator
From: Pasadena, CA
Registered: 2009-07-13
Posts: 19,744

Re: Kernel 4.4 ACPI Blocking

I've been delving into this and found that changing the policy from step_wise to fair_share for my thermal zone of type x86_pkg_temp works perfectly.  So, on this laptop with its default ACPI tables the problem seems to lie in the step_wise policy.

ewaller@turing ~ 1059 %cd /sys/class/thermal 
ewaller@turing /sys/class/thermal 1060 %ls
cooling_device0  cooling_device3  cooling_device6  thermal_zone0
cooling_device1  cooling_device4  cooling_device7  thermal_zone1
cooling_device2  cooling_device5  cooling_device8
ewaller@turing /sys/class/thermal 1061 %cd thermal_zone1 
ewaller@turing /sys/class/thermal/thermal_zone1 1062 %ls
available_policies  k_pu    subsystem          trip_point_1_temp
integral_cutoff     offset  sustainable_power  trip_point_1_type
k_d                 policy  temp               type
k_i                 power   trip_point_0_temp  uevent
k_po                slope   trip_point_0_type
ewaller@turing /sys/class/thermal/thermal_zone1 1063 %sudo bash -c "echo fair_share > policy"                                                                   
[sudo] password for ewaller:                               
ewaller@turing /sys/class/thermal/thermal_zone1 1064 %cat type 
x86_pkg_temp
ewaller@turing /sys/class/thermal/thermal_zone1 1065 %cat available_policies 
user_space bang_bang fair_share step_wise 
ewaller@turing /sys/class/thermal/thermal_zone1 1066 %

Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way

Offline

Board footer

Powered by FluxBB