Commit da781936e7 ("thermal: gov_power_allocator: Allow binding
without trip points") allowed the governor to bind even when trip_max
is NULL. This allows a NULL pointer dereference to happen in the manage
callback.
Add an early return to prevent it, since the governor is expected to not do
anything in this case.
Fixes: da781936e7 ("thermal: gov_power_allocator: Allow binding without trip points")
Signed-off-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>
Link: https://patch.msgid.link/20240702-power-allocator-null-trip-max-v1-1-47a60dc55414@collabora.com
Cc: All applicable <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Passive polling is enabled by setting the 'passive' field in
struct thermal_zone_device to a positive value so long as the
'passive_delay_jiffies' field is greater than zero. It causes
the thermal core to actively check the thermal zone temperature
periodically which in theory should be done after crossing a
passive trip point on the way up in order to allow governors to
react more rapidly to temperature changes and adjust mitigation
more precisely.
However, the 'passive' field in struct thermal_zone_device is currently
managed by governors which is quite problematic. First of all, only
two governors, Step-Wise and Power Allocator, update that field at
all, so the other governors do not benefit from passive polling,
although in principle they should. Moreover, if the zone governor is
changed from, say, Step-Wise to Fair-Share after 'passive' has been
incremented by the former, it is not going to be reset back to zero by
the latter even if the zone temperature falls down below all passive
trip points.
For this reason, make handle_thermal_trip() increment 'passive'
to enable passive polling for the given thermal zone whenever a
passive trip point is crossed on the way up and decrement it
whenever a passive trip point is crossed on the way down. Also
remove the 'passive' field updates from governors and additionally
clear it in thermal_zone_device_init() to prevent passive polling
from being enabled after a system resume just beacuse it was enabled
before suspending the system.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Tested-by: Lukasz Luba <lukasz.luba@arm.com>
Notice that the passive field in struct thermal_zone_device is not
used by the Power Allocator governor itself and so the ordering of
its updates with respect to allow_maximum_power() or allocate_power()
does not matter.
Accordingly, make power_allocator_manage() update that field right
before returning, which allows the current value of it to be passed
directly to allow_maximum_power() without using the additional update
variable that can be dropped.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
The Power Allocator governor really only wants to be called once per
thermal zone update and it does a special check to skip the extra,
from its perspective, invocations of the .throttle() callback.
Make it use .manage() instead of .throttle().
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
The threshold field in struct thermal_trip is only used internally by
the thermal core and it is better to prevent drivers from misusing it.
It also takes some space unnecessarily in the trip tables passed by
drivers to the core during thermal zone registration.
For this reason, introduce struct thermal_trip_desc as a wrapper around
struct thermal_trip, move the threshold field directly into it and make
the thermal core store struct thermal_trip_desc objects in the internal
thermal zone trip tables. Adjust all of the code using trip tables in
the thermal core accordingly.
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
IPA probe function was recently refactored to perform extra error checks
and make sure the thermal zone has trip points necessary for the IPA
operation. With this change, if a thermal zone is probed such that it
has no trip points that IPA can use, IPA will fail and the TZ won't be
created. This is the case if a platform defines a TZ without cooling
devices and only with "hot"/"critical" trip points, often found on some
Qualcomm devices [1].
Documentation across IPA code (notably get_governor_trips() kerneldoc)
suggests that IPA is supposed to handle such TZ even if it won't
actually do anything.
This commit partially reverts the previous change to allow IPA to bind
to such "empty" thermal zones.
Fixes: e83747c2f8 ("thermal: gov_power_allocator: Set up trip points earlier")
Link: arch/arm64/boot/dts/qcom/sc7180.dtsi#n4776 # [1]
Signed-off-by: Nikita Travkin <nikita@trvn.ru>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
IPA was recently refactored to split out memory allocation into a
separate funciton. That funciton was made to return -EINVAL if there is
zero power_actors and thus no memory to allocate. This causes IPA to
fail probing when the thermal zone has no attached cooling devices.
Since cooling devices can attach after the thermal zone is created and
the governer is attached to it, failing probe due to the lack of cooling
devices is incorrect.
Change the allocate_actors_buffer() to return success when there is no
cooling devices present.
Fixes: 912e97c67c ("thermal: gov_power_allocator: Move memory allocation out of throttle()")
Signed-off-by: Nikita Travkin <nikita@trvn.ru>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When the PID coefficients k_* are set via sysfs before the IPA
algorithm is triggered then the coefficients would be overwritten after
IPA throttle() is called. The old configuration values might be
different than the new values estimated by the IPA internal algorithm.
There might be a time delay when this overwriting happens. It
depends on the thermal zone temperature value. The temperature value
needs to cross the first trip point value then IPA algorithms start
operating. Although, the PID coefficients setup time should not be
affected or linked to any later operating phase and values must not be
overwritten.
This patch initializes params->sustainable_power when the governor
binds to thermal zone to avoid overwriting k_*. The basic function won't
be affected, as the k_* still can be estimated if the sustainable_power
is modified.
Signed-off-by: Di Shen <di.shen@unisoc.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 0952177f2a ("thermal/core/power_allocator: Update once
cooling devices when temp is low") adds an update flag to avoid
triggering a thermal event when there is no need, and the thermal
cdev is updated once when the temperature is low.
But when the trips are writable, and switch_on_temp is set to be a
higher value, the cooling device state may not be reset to 0,
because last_temperature is smaller than switch_on_temp.
For example:
First:
switch_on_temp=70 control_temp=85;
Then userspace change the trip_temp:
switch_on_temp=45 control_temp=55 cur_temp=54
Then userspace reset the trip_temp:
switch_on_temp=70 control_temp=85 cur_temp=57 last_temp=54
At this time, the cooling device state should be reset to 0.
However, because cur_temp(57) < switch_on_temp(70)
last_temp(54) < switch_on_temp(70) ----> update = false,
update is false, the cooling device state can not be reset.
Using the observation that tz->passive can also be regarded as the
temperature status, set the update flag to the tz->passive value.
When the temperature drops below switch_on for the first time, the
states of cooling devices can be reset once, and tz->passive is updated
to 0. In the next round, because tz->passive is 0, cdev->state will not
be updated.
By using the tz->passive value as the "update" flag, the issue above
can be solved, and the cooling devices can be updated only once when the
temperature is low.
Fixes: 0952177f2a ("thermal/core/power_allocator: Update once cooling devices when temp is low")
Cc: 5.13+ <stable@vger.kernel.org> # 5.13+
Suggested-by: Wei Wang <wvw@google.com>
Signed-off-by: Di Shen <di.shen@unisoc.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When the thermal instance's weight is updated from the sysfs the governor
update_tz() callback is triggered. Implement proper reaction to this
event in the IPA, which would save CPU cycles spent in throttle().
This will speed-up the main throttle() IPA function and clean it up
a bit.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There is a need to check if the cooling device in the thermal zone
supports IPA callback and is set for control trip point.
Refactor the code which validates the power actor capabilities and
make it more consistent in all places.
No intentional functional impact.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The new thermal callback allows to react to the change of cooling
instances in the thermal zone. Move the memory allocation to that new
callback and save CPU cycles in the throttle() code path.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Change trace event trace_thermal_power_allocator() to not use dynamic
array for requested power and granted power for all power actors.
Instead, simplify the trace event and print other simple values.
Add new trace event to print power actor information of requested power
and granted power. That trace event would be called in a loop for each
power actor. The trace data would be easier to parse comparing to the
dynamic array implementation.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Simplify the code and remove one extra 'if' block.
No intentional functional impact.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In preparation for a subsequent change, rearrange check_power_actors().
No intentional functional impact.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Rearrange the initialization of local variables in allocate_power() so
as to improve code clarity and the visibility of the initial values.
This change is not expected to alter the general functionality.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Local variable 'ret' in allocate_power() is only used in the return
statement, so drop it.
Local variable 'trip_max' in allocate_power() is only used for caching
the params->trip_max value which may as well be accessed directly as
needed, so drop it either.
This change is not expected to alter the general functionality.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The 'cdev' pointer in allow_maximum_power() is valid, so there is no
need to use 'instance->cdev' instead of it.
This change is not expected to alter the general functionality.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Rearrange the order of local variable definitions in multiple functions
so as to follow the kernel coding style in that respect.
Also, move local variable definitions located in nested code blocks to
the beginning of each function to improve the visibility of all local
variables in use.
This change is not expected to alter the general functionality.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The throttling logic only cares about the last passive trip point and
the cooling devices attached to it.
Therefore, there is no need to bail out if other trip points have
cooling devices which are not a supported by the IPA.
Check the cooling devices only for 'trip_max' during the binding.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Set up the trip points at the beginning of the binding function.
This simplifies the code a bit and allows for further cleanups.
Also add a check to fail the binding if the last passive trip point is
not found.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Refactor the code and rename the last passive trip point field.
There is a comment describing the field properly. Use shorter field name
so as to allow to clarify the code.
This change is not expected to alter the general functionality.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Modify the governor .throttle() callback definition so that it takes a
trip pointer instead of a trip index as its second argument, adjust the
governors accordingly and update the core code invoking .throttle().
This causes the governors to become independent of the representation
of the list of trips in the thermal zone structure.
This change is not expected to alter the general functionality.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Modify the power allocator thermal governor to use trip pointers instead
of trip indices everywhere except for the power_allocator_throttle()
second argument that will be changed subsequently along with the
definition of the .throttle() governor callback.
The general functionality is not expected to be changed.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Tested-by: Lukasz Luba <lukasz.luba@arm.com>
Replace the integer trip number stored in struct thermal_instance with
a pointer to the relevant trip and adjust the code using the structure
in question accordingly.
The main reason for making this change is to allow the trip point to
cooling device binding code more straightforward, as illustrated by
subsequent modifications of the ACPI thermal driver, but it also helps
to clarify the overall design and allows the governor code overhead to
be reduced (through subsequent modifications).
The only case in which it adds complexity is trip_point_show() that
needs to walk the trips[] table to find the index of the given trip
point, but this is not a critical path and the interface that
trip_point_show() belongs to is problematic anyway (for instance, it
doesn't cover the case when the same cooling devices is associated
with multiple trip points).
This is a preliminary change and the affected code will be refined by
a series of subsequent modifications of thermal governors, the core and
the ACPI thermal driver.
The general functionality is not expected to be affected by this change.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The traces are exported but only local to the thermal core code. On
the other side, the traces take the thermal zone device structure as
argument, thus they have to rely on the exported thermal.h header
file. As we want to move the structure to the private thermal core
header, first we have to relocate those traces to the same place as
many drivers do.
Cc: Steven Rostedt <rostedt@goodmis.org>
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20230307133735.90772-2-daniel.lezcano@linaro.org
The governors are using the ops->get_trip_* functions, Replace these
calls with thermal_zone_get_trip().
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com> # IPA
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/20221003092602.1323944-5-daniel.lezcano@linaro.org
All the governors throttling ops are taking/releasing the lock at the
beginning and the end of the function.
We can move the mutex to the throttling call site instead.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20220805153834.2510142-4-daniel.lezcano@linaro.org
The thermal zone lock is taken in the different places in the
throttling path.
At the first glance it does not hurt to move them at the beginning and
the end of the 'throttle' function. That will allow a consolidation of
the lock in the next following changes.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20220805153834.2510142-3-daniel.lezcano@linaro.org
In order to use thermal trips defined in the thermal structure, rename
the 'trips' field to 'num_trips' to have the 'trips' field containing the
thermal trip points.
Cc: Alexandre Bailon <abailon@baylibre.com>
Cc: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linexp.org>
Link: https://lore.kernel.org/r/20220722200007.1839356-8-daniel.lezcano@linexp.org
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Use the new helper function and avoid unnecessery second lock/unlock,
which was present in old approach with thermal_cdev_update().
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20210422153624.6074-4-lukasz.luba@arm.com
The cooling device state change generates an event, also when there is no
need, because temperature is low and device is not throttled. Avoid to
unnecessary update the cooling device which means also not sending event.
The cooling device state has not changed because the temperature is still
below the first activation trip point value, so we can do this.
Add a tracking mechanism to make sure it updates cooling devices only
once - when the temperature dropps below first trip point.
Reported-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20210422114308.29684-4-lukasz.luba@arm.com
When the temperature is below the first activation trip point the cooling
devices are not checked, so they cannot maintain fresh statistics. It
leads into the situation, when temperature crosses first trip point, the
statistics are stale and show state for very long period. This has impact
on IPA algorithm calculation and wrong decisions. Thus, check the cooling
devices even when the temperature is low, to refresh these statistics.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20210422114308.29684-3-lukasz.luba@arm.com
The division is used directly in re-divvying up power, the decimal part will
be discarded, devices will get less than the extra_actor_power - 1.
if using round the division to make the calculation more accurate.
For example:
actor0 received more than its max_power, it has the extra_power 759
actor1 received less than its max_power, it require extra_actor_power 395
actor2 received less than its max_power, it require extra_actor_power 365
actor1 and actor2 require the total capped_extra_power 760
using division in re-divvying up power
actor1 would actually get the extra_actor_power 394
actor2 would actually get the extra_actor_power 364
if using round the division in re-divvying up power
actor1 would actually get the extra_actor_power 394
actor2 would actually get the extra_actor_power 365
Signed-off-by: Jeson Gao <jeson.gao@unisoc.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/1615796737-4688-1-git-send-email-gao.yunxiao6@gmail.com
The thermal zone can have cooling devices which are missing power actor
API. This could be due to missing Energy Model for devfreq or cpufreq
cooling device. In this case it is safe to fail the binding rather than
trying to workaround and control the temperature in such thermal zone.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20210119114126.19480-1-lukasz.luba@arm.com
The code does no longer use the ms unit based fields to set the
delays as they are replaced by the jiffies.
Remove them and replace their user to use the jiffies version instead.
Cc: Thara Gopinath <thara.gopinath@linaro.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Reviewed-by: Peter Kästle <peter@piie.net>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20201216220337.839878-3-daniel.lezcano@linaro.org
The PID coefficients should be estimated again when there was a change to
sustainable power value made by user. This change removes unused argument
'force' and makes the function ready for such updates.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Reviewed-by: Ionela Voinescu <ionela.voinescu@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20201124161025.27694-4-lukasz.luba@arm.com
The sustainable power value might come from the Device Tree or can be
estimated in run time. The sustainable power might be updated by the user
via sysfs interface, which should trigger new estimation of PID
coefficients. There is no need to estimate it every time when the
governor is called and temperature is high. Instead, store the estimated
value and make it available via standard sysfs interface, so it can be
checked from the user-space.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Reviewed-by: Ionela Voinescu <ionela.voinescu@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20201124161025.27694-3-lukasz.luba@arm.com
Intelligent Power Allocation (IPA) is built around the PID controller
concept. The initialization code tries to setup the environment based on
the information available in DT or estimate the value based on minimum
power reported by each of the cooling device. The estimation will have an
impact on the PID controller behaviour via the related 'k_po', 'k_pu',
'k_i' coefficients and also on the power budget calculation.
This change prevents the situation when 'k_i' is relatively big compared
to 'k_po' and 'k_pu' values. This might happen when the estimation for
'sustainable_power' returned small value, thus 'k_po' and 'k_pu' are
small.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Reviewed-by: Ionela Voinescu <ionela.voinescu@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20201124161025.27694-2-lukasz.luba@arm.com
Since the power actor section has one function power_actor_set_power()
move it into Intelligent Power Allocation (IPA). There is no other user
of that helper function. It would also allow to remove the check of
cdev_is_power_actor() because the code which calls it in IPA already does
the needed check. Make the function static since only IPA use it.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20201015112441.4056-5-lukasz.luba@arm.com
The thermal cooling device specified in DT might be instantiated for
a thermal zone trip point with a limited set of OPPs to operate on. This
configuration should be supported by Intelligent Power Allocation (IPA),
since it is a standard for other governors. Change the code and allow IPA
to get power value of lower and upper bound set for a given cooling
device.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20201015112441.4056-3-lukasz.luba@arm.com
1. devfreq_cooling.c: The variable *tz is not used in
devfreq_cooling_get_requested_power(), devfreq_cooling_state2power()
and devfreq_cooling_power2state().
2. cpufreq_cooling.c: After 84fe2cab48, the variable *tz is not used
anymore in cpufreq_get_requested_power(), cpufreq_state2power() and
cpufreq_power2state().
Remove the variable *tz.
Signed-off-by: zhuguangqing <zhuguangqing@xiaomi.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20200914071101.13575-1-zhuguangqing83@gmail.com
