If pm action fails within put_sync_locked, usage count should be
reset to reflect the state of the device.
Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
In the default PM policy, the function k_us_to_ticks_ceil32() is used
and does calculation using 64-bit integers which can be slow and
avoidable if the input is 0.
Signed-off-by: Bastien Beauchamp <bastien.beauchamp@silabs.com>
In a call to pm_device_runtime_get, if the power domain is taken but
action_cb fails, the power domain is not released.
Signed-off-by: Adrien Lessard <adrien.lessard@rbr-global.com>
More complex suspend and resume scheme might require exactly defined
location of this variable due to platform peculiar SW and HW requirement.
DTS zephyr,memory-region node with nodelabel `pm_s2ram` shall be used to
automatic definition of linker section for such objective.
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
Introduce pm_device_driver_deinit() which is required to fully
support device_deinit(). This function shall be called by drivers
when device_deinit() is called.
If PM_DEVICE is enabled, it will
simply check whether the device is either SUSPENDED or OFF, this
will prevent device_deinit() on a device which is currently in
use, and ensure the device is in the correct state if device_init()
is called later.
If PM_DEVICE is not enabled, it will invoke the SUSPEND action which
mirrors the pm_device_driver_init() behavior. Note that TURN_OFF
action is not called since the device is deinitialized after this
call.
Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
K_TICKS_FOREVER is defined as -1. Guard logic has been added for the case
when `ticks - exit_latency_ticks == -1` to prevent sys_clock_set_timeout()
from incorrectly setting a forever timeout.
Signed-off-by: Serhiy Katsyuba <serhiy.katsyuba@intel.com>
pm_device_runtime has inconsistent behavior regarding
getting and putting a device's domain. This commit
aligns it to get the domain only once, once device
is resumed, and put only once, once device is
suspended.
Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
Refactor this code so it has less nesting and is able to share some code
between functions, by making a function to find the device constraints
object which can be called before doing operations using it.
Signed-off-by: Declan Snyder <declan.snyder@nxp.com>
Async now uses its own work queue, which means it consumes more
resources. Since not all applications need the async API, we can make
it optional without any penalty for those applications.
Signed-off-by: Flavio Ceolin <flavio@hubblenetwork.com>
Device runtime is using the system workqueue to do operations
that are mostly blockers (suspend a device). This should not happen.
This commit adds an option to use dedicated queue for the device runtime
async operations.
The test for this API was assuming that the system workqueue
priority (which is cooperative) so we need to change the test priority
to be lower than the device runtime workqueue priority.
Signed-off-by: Flavio Ceolin <flavio@hubblenetwork.com>
Ensure `pm_state_force` returns `false` when it fails to retrieve or set
the desired power state. This change improves error handling and ensures
the function's behavior aligns with its intended purpose.
Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>
pm_state_get() is used to get a PM state that can be forced.
Disabled states (not included in automatic state selection) can
also be forced thus they should be included in search inside
pm_state_get().
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
pm_suspend is returning early if there are no states
available (due to locking or latency policy). However,
if state is forced it should not return but rather enter
forced power state.
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Update power state selection. Previously, it was iterating over states
starting from the last one so the most common short sleep periods were
taking the longest time to select. Order is now swapped so that short
sleeps will get power state as quick as possible.
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
When all states are locked or latency requirement cannot be met
by any power state it is important to be able to quickly exit
suspend procedure because that usually means that application
requires high performance. Add function for detecting if any
power state is available.
Additionally, add function pm_policy_state_is_available for
checking if given state is available which means that it is not
locked and fulfills current latency requirement.
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Use power state pointers instead of copies which improves performance.
Align power_mgmt_multicore test which was creating pm states in
runtime.
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
Set the value of `pm->state` as we move through the various stages of
`pm_device_driver_init`. This ensures hat if any of the code inside the
actions callbacks runs `pm_device_state_get` they get the correct state,
instead of always getting `PM_DEVICE_STATE_ACTIVE` (0, the value at
reset).
Signed-off-by: Jordan Yates <jordan@embeint.com>
When core is suspended and power mode has non zero exit_latency_us
a system timeout will be rescheduled to a point in time that is
earlier by exit_latency_us than request. It is to accommodate for
lengthy resuming procedure which would cause requested timeout to
be significantly late. However, setting additional wake up point
has cost, it is one more redundant core wake up and that impacts
performance and power consumption.
Add Kconfig option to chose what conversion method is used. It has
the biggest impact on small exit_latency_us where conversion may
result in 0 ticks (no pre-wake up) or 1 tick (wake up).
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
When PM_POLICY_DEVICE_CONSTRAINTS is enabled, we create a list of
devices with the property “zephyr,disabling-power-states” and their
respective states to lock. However, if a device is declared but not
built (e.g., due to a Kconfig option), we have an undefined reference.
This PR changes the code to use the device name instead of the
reference. This increases lookup time (which could be cached at
runtime), and we still have an entry for a non-existent device.
Signed-off-by: Flavio Ceolin <flavio@hubblenetwork.com>
Use shell_device_get_binding() instead of device_get_binding() so that
we get the device based on its name and in addition by its label.
Signed-off-by: Yishai Jaffe <yishai1999@gmail.com>
Update events to use uptime ticks, which is a monotonic clock which
in the same res as kernel ticks. This makes comparisons simple and
removes the complexity of dealing with wrapping counter values.
The wrapping is particularly problematic for events since this makes
it quite complex to track if an event has occured in the past, or
will occur in the future. This info is needed to know if an event
has actually been handled or not.
Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
A warning is giving for missing initalizer for field `exit_latency_us`
of `struct pm_state_info`. This adds the additional init fields.
Signed-off-by: Ryan McClelland <ryanmcclelland@meta.com>
There is no point in using lock or semaphore to read current
usage counter as it may change after unlocking or giving
back the semaphore. Value can only be trusted in the controlled
environment (e.g. test).
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
PM_DEVICE_FLAG_ISR_SAFE is an enum and it must be converted to
a bit mask before masking with flags.
Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>
The default policy currently directly references the private
variable next_event from policy_events.c to then convert the cycle
of said event (if exists) to a kernel tick in the future, something
policy_events.c already implements and exposes through
pm_policy_next_event_ticks().
Additionally, the implementation of pm_policy_next_state() in
policy_default.c already gets the nearest kernel tick, wherein
the next event has already been accounted for in, see
implementation of pm_system_suspend().
This commit removes the redundant and layer violating computation
if the tick of the next event from policy_default.c and updates
the test test_pm_policy_events to not use default policy to
determine if pm_policy_next_event_ticks() is correct.
Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
The pm_policy_event_register() API takes absolute cycles as the
second arg, like pm_policy_event_update(), but the arg is renamed
time_us and treated as a relative time in us rather than abs
cycles.
Fix implementation of pm_policy_event_register() to treat cycles
like pm_policy_event_update() and API docs suggest.
Signed-off-by: Bjarki Arge Andreasen <bjarki.andreasen@nordicsemi.no>
This option allows using the pm_policy_latency* APIs to gather latency
requirements on systems that do not support PM (e.g. systems whithout
CPU idle states). Because the API has a subscription mechanism, it can
be useful to perform system-level adjustments based on latency
requirements gathered from multiple system components.
Signed-off-by: Gerard Marull-Paretas <gerard@teslabs.com>
policy.c has grown organically, it contained many independent pieces of
code. This patch splits each logical unit into its own C file, making it
easier to browse the code.
Signed-off-by: Gerard Marull-Paretas <gerard@teslabs.com>
Some devices, e.g. SoC level devices like I2C peripheral, can never be
powerd off as they are always energized. Such devices can only go from an
active state or to a low power state (suspended). Allow them to simply
return -ENOTSUP when called with TURN_ON (or TURN_OFF).
Signed-off-by: Gerard Marull-Paretas <gerard@teslabs.com>
Some events needs to be handled with a very low latency constraint.
If the system is in deep sleep, exit latency from this low level state
exceeds sometimes the maximum latency constraint of these events.
Before suspending the system, select which events is happening sooner,
kernel events or normal events.
CPU will be up just before the next event occurs taking into account the
exit latency of the current power state
Change also the policy event API to take as argument absolute time in HW
cycles instead of time in us
Signed-off-by: Riadh Ghaddab <rghaddab@baylibre.com>
Refactor pm_device_driver_init code to keep the normal execution path
inline and the early exit branches at a single indentation, this is
commonly done throughout the code base.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
PM_DEVICE_RUNTIME_EXCLUSIVE was deprecated and its behavior
is achived with PM_DEVICE_SYSTEM_MANAGED=n.
Fixes#76037
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Move information about device power state constraints from device
to policy.
It slows down the constraints lookup since we now have to find the
constraints for a device in a global array, but it saves resources
because we don't need to add a reference to constraints in all devices
instances.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Add a symbol to enable device power state constraints this
saves resources when this feature is not needed.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Declare power state constraints for a device in devicetree.
It allows a map between device instances and power states that disable
their power. This information is used by a new API
(pm_policy_device_power_lock_put/get) that automically set/release
pm state constraints.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
