Commit a9ceb78bc75c (cpuidle,menu: use interactivity_req to disable
polling) changed the behavior of the fallback state selection part
of menu_select() so it looks at interactivity_req instead of
data->next_timer_us when it makes its decision.  That effectively
caused polling to be used more often as fallback idle which led to
significant increases of energy consumption in some cases.

Commit e132b9b3bc7f (cpuidle: menu: use high confidence factors
only when considering polling) changed that logic again to be more
predictable, but that didn't help with the increased energy
consumption problem.

For this reason, go back to making decisions on which state to fall
back to based on data->next_timer_us which is the time we know for
sure something will happen rather than a prediction (which may be
inaccurate and turns out to be so often enough to be problematic).
However, take the target residency of the first proper idle state
(C1) into account, so that state is not used as the fallback one
if its target residency is greater than data->next_timer_us.

Fixes: a9ceb78bc75c (cpuidle,menu: use interactivity_req to disable polling)
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reported-and-tested-by: Doug Smythies <dsmythies@telus.net>

The menu governor uses five different factors to pick the
idle state:
 - the user configured latency_req
 - the time until the next timer (next_timer_us)
 - the typical sleep interval, as measured recently
 - an estimate of sleep time by dividing next_timer_us by an observed factor
 - a load corrected version of the above, divided again by load

Only the first three items are known with enough confidence that
we can use them to consider polling, instead of an actual CPU
idle state, because the cost of being wrong about polling can be
excessive power use.

The latter two are used in the menu governor's main selection
loop, and can result in choosing a shallower idle state when
the system is expected to be busy again soon.

This pushes a busy system in the "performance" direction of
the performance<>power tradeoff, when choosing between idle
states, but stays more strictly on the "power" state when
deciding between polling and C1.

Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

We know that the avg variable actually ends up holding a 32 bit
quantity, since it's an average of such numbers. It is only a u64
because it is temporarily used to hold the sum. Making it an actual
u32 allows gcc to generate slightly better code, e.g. when computing
the square, it can do a 32x32->64 multiply.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Computing the integer square root is a rather expensive operation, at
least compared to doing a 64x64 -> 64 multiply (avg*avg) and, on 64
bit platforms, doing an extra comparison to a constant (variance <=
U64_MAX/36).

On 64 bit platforms, this does mean that we add a restriction on the
range of the variance where we end up using the estimate (since
previously the stddev <= ULONG_MAX was a tautology), but on the other
hand, we extend the range quite substantially on 32 bit platforms - in
both cases, we now allow standard deviations up to 715 seconds, which
is for example guaranteed if all observations are less than 1430
seconds.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

* pm-cpuidle:
  cpuidle: coupled: remove unused define cpuidle_coupled_lock
  cpuidle: fix fallback mechanism for suspend to idle in absence of enter_freeze

* pm-cpufreq:
  cpufreq: cpufreq-dt: avoid uninitialized variable warnings:
  cpufreq: pxa2xx: fix pxa_cpufreq_change_voltage prototype
  cpufreq: Use list_is_last() to check last entry of the policy list
  cpufreq: Fix NULL reference crash while accessing policy->governor_data

* pm-domains:
  PM / Domains: Fix typo in comment
  PM / Domains: Fix potential deadlock while adding/removing subdomains
  PM / domains: fix lockdep issue for all subdomains

* pm-sleep:
  PM: APM_EMULATION does not depend on PM

This was found with the -RT patch enabled, but the fix should apply to
non-RT also.

Used multi_v7_defconfig+PREEMPT_RT_FULL=y and this caused a compilation
warning without this fix:
../drivers/cpuidle/coupled.c:122:21: warning: 'cpuidle_coupled_lock'
defined but not used [-Wunused-variable]

Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Commit 51164251f5c3 "sched / idle: Drop default_idle_call() fallback
from call_cpuidle()" made find_deepest_state() return non-negative
value and check all the states with index > 0.  Also as a result,
find_deepest_state() returns 0 even when enter_freeze callbacks are not
implemented and enter_freeze_proper() is called which ends up crashing
the kernel.

This patch updates the check for index > 0 in cpuidle_enter_freeze and
cpuidle_idle_call(when idle_should_freeze is true) to restore the
suspend-to-idle functionality in absence of enter_freeze callback.

Fixes: 51164251f5c3 "sched / idle: Drop default_idle_call() fallback from call_cpuidle()"
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Pull more power management and ACPI updates from Rafael Wysocki:
 "This includes fixes on top of the previous batch of PM+ACPI updates
  and some new material as well.

  From the new material perspective the most significant are the driver
  core changes that should allow USB devices to stay suspended over
  system suspend/resume cycles if they have been runtime-suspended
  already beforehand.  Apart from that, ACPICA is updated to upstream
  revision 20160108 (cosmetic mostly, but including one fixup on top of
  the previous ACPICA update) and there are some devfreq updates the
  didn't make it before (due to timing).

  A few recent regressions are fixed, most importantly in the cpuidle
  menu governor and in the ACPI backlight driver and some x86 platform
  drivers depending on it.

  Some more bugs are fixed and cleanups are made on top of that.

  Specifics:

   - Modify the driver core and the USB subsystem to allow USB devices
     to stay suspended over system suspend/resume cycles if they have
     been runtime-suspended already beforehand and fix some bugs on top
     of these changes (Tomeu Vizoso, Rafael Wysocki).

   - Update ACPICA to upstream revision 20160108, including updates of
     the ACPICA's copyright notices, a code fixup resulting from a
     regression fix that was necessary in the upstream code only (the
     regression fixed by it has never been present in Linux) and a
     compiler warning fix (Bob Moore, Lv Zheng).

   - Fix a recent regression in the cpuidle menu governor that broke it
     on practically all architectures other than x86 and make a couple
     of optimizations on top of that fix (Rafael Wysocki).

   - Clean up the selection of cpuidle governors depending on whether or
     not the kernel is configured for tickless systems (Jean Delvare).

   - Revert a recent commit that introduced a regression in the ACPI
     backlight driver, address the problem it attempted to fix in a
     different way and revert one more cosmetic change depending on the
     problematic commit (Hans de Goede).

   - Add two more ACPI backlight quirks (Hans de Goede).

   - Fix a few minor problems in the core devfreq code, clean it up a
     bit and update the MAINTAINERS information related to it (Chanwoo
     Choi, MyungJoo Ham).

   - Improve an error message in the ACPI fan driver (Andy Lutomirski).

   - Fix a recent build regression in the cpupower tool (Shreyas
     Prabhu)"

* tag 'pm+acpi-4.5-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (32 commits)
  cpuidle: menu: Avoid pointless checks in menu_select()
  sched / idle: Drop default_idle_call() fallback from call_cpuidle()
  cpupower: Fix build error in cpufreq-info
  cpuidle: Don't enable all governors by default
  cpuidle: Default to ladder governor on ticking systems
  time: nohz: Expose tick_nohz_enabled
  ACPICA: Update version to 20160108
  ACPICA: Silence a -Wbad-function-cast warning when acpi_uintptr_t is 'uintptr_t'
  ACPICA: Additional 2016 copyright changes
  ACPICA: Reduce regression fix divergence from upstream ACPICA
  ACPI / video: Add disable_backlight_sysfs_if quirk for the Toshiba Satellite R830
  ACPI / video: Revert "thinkpad_acpi: Use acpi_video_handles_brightness_key_presses()"
  ACPI / video: Document acpi_video_handles_brightness_key_presses() a bit
  ACPI / video: Fix using an uninitialized mutex / list_head in acpi_video_handles_brightness_key_presses()
  ACPI / video: Revert "ACPI / video: driver must be registered before checking for keypresses"
  ACPI / fan: Improve acpi_device_update_power error message
  ACPI / video: Add disable_backlight_sysfs_if quirk for the Toshiba Portege R700
  cpuidle: menu: Fix menu_select() for CPUIDLE_DRIVER_STATE_START == 0
  MAINTAINERS: Add devfreq-event entry
  MAINTAINERS: Add missing git repository and directory for devfreq
  ...

If menu_select() cannot find a suitable state to return, it will
return the state index stored in data->last_state_idx.  This
means that it is pointless to look at the states whose indices
are less than or equal to data->last_state_idx in the main loop,
so don't do that.

Given that those checks are done on every idle state selection, this
change can save quite a bit of completely unnecessary overhead.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Sudeep Holla <sudeep.holla@arm.com>

After commit 9c4b2867ed7c (cpuidle: menu: Fix menu_select() for
CPUIDLE_DRIVER_STATE_START == 0) it is clear that menu_select()
cannot return negative values.  Moreover, ladder_select_state()
will never return a negative value too, so make find_deepest_state()
return non-negative values too and drop the default_idle_call()
fallback from call_cpuidle().

This eliminates one branch from the idle loop and makes the governors
and find_deepest_state() handle the case when all states have been
disabled from sysfs consistently.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Sudeep Holla <sudeep.holla@arm.com>