cpufreq: intel_pstate: Fix intel_pstate powersave min_perf_pct value
authorPrarit Bhargava <prarit@redhat.com>
Thu, 15 Oct 2015 11:34:15 +0000 (07:34 -0400)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Fri, 16 Oct 2015 20:35:11 +0000 (22:35 +0200)
commit51443fbf3d2cde16011b994252c8004ebcd66fb0
treeaaf73f92c85ce4b88ca04cf94cb34e340f2da3b3
parent7855e10294efd4925b351c56d2834dc9f7cff7a2
cpufreq: intel_pstate: Fix intel_pstate powersave min_perf_pct value

On systems that initialize the intel_pstate driver with the performance
governor, and then switch to the powersave governor will not transition to
lower cpu frequencies until /sys/devices/system/cpu/intel_pstate/min_perf_pct
is set to a low value.

The behavior of governor switching changed after commit a04759924e25
("[cpufreq] intel_pstate: honor user space min_perf_pct override on
 resume").  The commit introduced tracking of performance percentage
changes via sysfs in order to restore userspace changes during
suspend/resume.  The problem occurs because the global values of the newly
introduced max_sysfs_pct and min_sysfs_pct are not lowered on the governor
change and this causes the powersave governor to inherit the performance
governor's settings.

A simple change would have been to reset max_sysfs_pct to 100 and
min_sysfs_pct to 0 on a governor change, which fixes the problem with
governor switching.  However, since we cannot break userspace[1] the fix
is now to give each governor its own limits storage area so that governor
specific changes are tracked.

I successfully tested this by booting with both the performance governor
and the powersave governor by default, and switching between the two
governors (while monitoring /sys/devices/system/cpu/intel_pstate/ values,
and looking at the output of cpupower frequency-info).  Suspend/Resume
testing was performed by Doug Smythies.

[1] Systems which suspend/resume using the unmaintained pm-utils package
will always transition to the performance governor before the suspend and
after the resume.  This means a system using the powersave governor will
go from powersave to performance, then suspend/resume, performance to
powersave.  The simple change during governor changes would have been
overwritten when the governor changed before and after the suspend/resume.
I have submitted https://bugzilla.redhat.com/show_bug.cgi?id=1271225
against Fedora to remove the 94cpufreq file that causes the problem.  It
should be noted that pm-utils is obsoleted with newer versions of systemd.

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Acked-by: Kristen Carlson Accardi <kristen@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/cpufreq/intel_pstate.c