src/devices/sensors/hal-interface.jd

page.title=HAL interface
@jd:body

<!--
    Copyright 2014 The Android Open Source Project

    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.
    You may obtain a copy of the License at

        http://www.apache.org/licenses/LICENSE-2.0

    Unless required by applicable law or agreed to in writing, software
    distributed under the License is distributed on an "AS IS" BASIS,
    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    See the License for the specific language governing permissions and
    limitations under the License.
-->
<div id="qv-wrapper">
  <div id="qv">
    <h2>In this document</h2>
    <ol id="auto-toc">
    </ol>
  </div>
</div>

<p>The HAL interface, declared in <a href="{@docRoot}devices/halref/sensors_8h.html">sensors.h</a>, represents the interface between the Android <a href="sensor-stack.html#framework">framework</a> and the hardware-specific software. A HAL implementation must define each
  function declared in sensors.h. The main functions are:</p>
<ul>
  <li><code>get_sensors_list</code> - Returns the list of all sensors. </li>
  <li><code>activate</code> - Starts or stops a sensor. </li>
  <li><code>batch</code> - Sets a sensor’s parameters such as sampling frequency and maximum
    reporting latency. </li>
  <li><code>setDelay</code> - Used only in HAL version 1.0. Sets the sampling frequency for a
    given sensor. </li>
  <li><code>flush</code> - Flushes the FIFO of the specified sensor and reports a flush complete
    event when this is done. </li>
  <li><code>poll</code> - Returns available sensor events. </li>
</ul>
<p>The implementation must be thread safe and allow these functions to be called
  from different threads.</p>
<p>The interface also defines several types used by those functions. The main
  types are:</p>
<ul>
  <li><code>sensors_module_t</code></li>
  <li><code>sensors_poll_device_t</code></li>
  <li><code>sensor_t</code></li>
  <li><code>sensors_event_t</code></li>
</ul>
<p>In addition to the sections below, see <a href="{@docRoot}devices/halref/sensors_8h.html">sensors.h</a> for more information on those types.</p>
<h2 id="get_sensors_list_list">get_sensors_list(list)</h2>
<pre>int (*get_sensors_list)(struct sensors_module_t* module, struct sensor_t
  const** list);</pre>
<p>Provides the list of sensors implemented by the HAL. See <a href="#sensor_t">sensor_t</a> for details on how the sensors are defined.</p>
<p>The order in which the sensors appear in the list is the order in which the
  sensors will be reported to the applications. Usually, the base sensors appear
  first, followed by the composite sensors.</p>
<p>If several sensors share the same sensor type and wake-up property, the first
  one in the list is called the “default” sensor. It is the one returned by
  <code>getDefaultSensor(int sensorType, bool wakeUp)</code>.</p>
<p>This function returns the number of sensors in the list.</p>
<h2 id="activate_sensor_true_false">activate(sensor, true/false)</h2>
<pre>int (*activate)(struct sensors_poll_device_t *dev, int sensor_handle, int
  enabled);</pre>
<p>Activates or deactivates a sensor.</p>
<p><code>sensor_handle</code> is the handle of the sensor to activate/deactivate. A sensor’s
  handle is defined by the <code>handle</code> field of its <a href="#sensor_t">sensor_t</a> structure.</p>
<p><code>enabled</code> is set to 1 to enable or 0 to disable the sensor.</p>
<p>One-shot sensors deactivate themselves automatically upon receiving an event,
  and they must still accept to be deactivated through a call to <code>activate(...,
  enabled=0)</code>.</p>
<p>Non-wake-up sensors never prevent the SoC from going into suspend mode; that
  is, the HAL shall not hold a partial wake-lock on behalf of applications.</p>
<p>Wake-up sensors, when delivering events continuously, can prevent the SoC from
  going into suspend mode, but if no event needs to be delivered, the partial
  wake-lock must be released.</p>
<p>If <code>enabled</code> is 1 and the sensor is already activated, this function is a no-op
  and succeeds.</p>
<p>If <code>enabled</code> is 0 and the sensor is already deactivated, this function is a no-op
  and succeeds.</p>
<p>This function returns 0 on success and a negative error number otherwise.</p>
<h2 id="batch_sensor_flags_sampling_period_maximum_report_latency">batch(sensor, flags, sampling period, maximum report latency)</h2>
<pre>
int (*batch)(
     struct sensors_poll_device_1* dev,
     int sensor_handle,
     int flags,
     int64_t sampling_period_ns,
     int64_t max_report_latency_ns);
</pre>
<p>Sets a sensor’s parameters, including <a href="#sampling_period_ns">sampling frequency</a> and <a href="#max_report_latency_ns">maximum report latency</a>. This function can be called while the sensor is activated, in which case it
  must not cause any sensor measurements to be lost: Transitioning from one
  sampling rate to the other cannot cause lost events, nor can transitioning from
  a high maximum report latency to a low maximum report latency.</p>
<p><code>sensor_handle</code> is the handle of the sensor to configure.</p>
<p><code>flags</code> is currently unused.</p>
<p><code>sampling_period_ns</code> is the sampling period at which the sensor should run, in
  nanoseconds. See <a href="#sampling_period_ns">sampling_period_ns</a> for more details.</p>
<p><code>max_report_latency_ns</code> is the maximum time by which events can be delayed before
  being reported through the HAL, in nanoseconds. See the <a href="#max_report_latency_ns">max_report_latency_ns</a> paragraph for more details.</p>
<p>This function returns 0 on success and a negative error number otherwise.</p>
<h3 id="sampling_period_ns">sampling_period_ns</h3>
<p>What the <code>sampling_period_ns</code> parameter means depends on the specified sensor's
  reporting mode:</p>
<ul>
  <li> Continuous: <code>sampling_period_ns</code> is the sampling rate, meaning the rate at which
    events are generated. </li>
  <li> On-change: <code>sampling_period_ns</code> limits the sampling rate of events, meaning
    events are generated no faster than every <code>sampling_period_ns</code> nanoseconds. There
    might be periods longer than <code>sampling_period_ns</code> where no event is generated if
    the measured values do not change for long periods. See <a
    href="report-modes.html#on-change">on-change</a> reporting mode for more
    details. </li>
  <li> One-shot: <code>sampling_period_ns</code> is ignored. It has no effect. </li>
  <li> Special: See the specific <a href="sensor-types.html">sensor type
  descriptions</a> for details on how <code>sampling_period_ns</code> is used
  for special sensors. </li>
</ul>
<p>See <a href="report-modes.html">Reporting modes</a> for more information
  about the impact of <code>sampling_period_ns</code> in the different modes.</p>
<p>For continuous and on-change sensors,</p>
<ul>
  <li> if <code>sampling_period_ns</code> is less than
    <code>sensor_t.minDelay</code>, then the HAL implementation must silently
    clamp it to <code>max(sensor_t.minDelay, 1ms)</code>. Android
    does not support the generation of events at more than 1000Hz. </li>
  <li> if <code>sampling_period_ns</code> is greater than
    <code>sensor_t.maxDelay</code>, then the HAL
    implementation must silently truncate it to <code>sensor_t.maxDelay</code>. </li>
</ul>
<p>Physical sensors sometimes have limitations on the rates at which they can run
  and the accuracy of their clocks. To account for this, we allow the actual
  sampling frequency to differ from the requested frequency, as long as it
  satisfies the requirements in the table below.</p>
<table>
  <tr>
    <th><p>If the requested frequency is</p></th>
    <th><p>Then the actual frequency must be</p></th>
  </tr>
  <tr>
    <td><p>below min frequency (&lt;1/maxDelay)</p></td>
    <td><p>between 90% and 110% of the min frequency</p></td>
  </tr>
  <tr>
    <td><p>between min and max frequency</p></td>
    <td><p>between 90% and 220% of the requested frequency</p></td>
  </tr>
  <tr>
    <td><p>above max frequency (&gt;1/minDelay)</p></td>
    <td><p>between 90% and 110% of the max frequency</p>
      <p>and below 1100Hz</p></td>
  </tr>
</table>
<p>Note that this contract is valid only at the HAL level, where there is always a
  single client. At the SDK level, applications might get different rates, due to
  the multiplexing happening in the Framework. See <a
  href="sensor-stack.html#framework">Framework</a> for more details.</p>
<h3 id="max_report_latency_ns">max_report_latency_ns</h3>
<p><code>max_report_latency_ns</code> sets the maximum time in nanoseconds, by which events can
  be delayed and stored in the hardware FIFO before being reported through the
  HAL while the SoC is awake.</p>
<p>A value of zero signifies that the events must be reported as soon as they are
  measured, either skipping the FIFO altogether, or emptying the FIFO as soon as
  one event from this sensor is present in it.</p>
<p>For example, an accelerometer activated at 50Hz with <code>max_report_latency_ns=0</code>
  will trigger interrupts 50 times per second when the SoC is awake.</p>
<p>When <code>max_report_latency_ns&gt;0</code>, sensor events do not need to be reported as soon
  as they are detected. They can be temporarily stored in the hardware FIFO and
  reported in batches, as long as no event is delayed by more than
  max_report_latency_ns nanoseconds. That is, all events since the previous batch
  are recorded and returned at once. This reduces the amount of interrupts sent
  to the SoC and allows the SoC to switch to a lower power mode (idle) while the
  sensor is capturing and batching data.</p>
<p>Each event has a timestamp associated with it. Delaying the time at which an
  event is reported does not impact the event timestamp. The timestamp must be
  accurate and correspond to the time at which the event physically happened, not
  the time it is being reported. </p>
<p>Allowing sensor events to be stored temporarily in the hardware FIFO does not
  modify the behavior of <code>poll</code>: events from different sensors can be interleaved,
  and as usual, all events from the same sensor are time-ordered.</p>
<p>See <a href="batching.html">Batching</a> for more details on sensor
batching, including behaviors in suspend mode and out of suspend mode.</p>
<h2 id="setdelay_sensor_sampling_period">setDelay(sensor, sampling period)</h2>
<pre>
int (*setDelay)(
     struct sensors_poll_device_t *dev,
     int sensor_handle,
     int64_t sampling_period_ns);
</pre>
<p>After HAL version 1.0, this function is deprecated and is never called.
  Instead, the <code>batch</code> function is called to set the
  <code>sampling_period_ns</code> parameter.</p>
<p>In HAL version 1.0, setDelay was used instead of batch to set <a href="#sampling_period_ns">sampling_period_ns</a>.</p>
<h2 id="flush_sensor">flush(sensor)</h2>
<pre>int (*flush)(struct sensors_poll_device_1* dev, int sensor_handle);</pre>
<p>Add a <a href="#metadata_flush_complete_events">flush complete event</a> to the end of the hardware FIFO for the specified sensor and flushes the FIFO;
  those events are delivered as usual (i.e.: as if the maximum reporting latency
  had expired) and removed from the FIFO.</p>
<p>The flush happens asynchronously (i.e.: this function must return immediately).
  If the implementation uses a single FIFO for several sensors, that FIFO is
  flushed and the flush complete event is added only for the specified sensor.</p>
<p>If the specified sensor has no FIFO (no buffering possible), or if the FIFO,
  was empty at the time of the call, <code>flush</code> must still succeed and send a flush
  complete event for that sensor. This applies to all sensors other than one-shot
  sensors.</p>
<p>When <code>flush</code> is called, even if a flush event is already in the FIFO for that
  sensor, an additional one must be created and added to the end of the FIFO, and
  the FIFO must be flushed. The number of <code>flush</code> calls must be
  equal to the number of flush complete events created.</p>
<p><code>flush</code> does not apply to <a href="report-modes.html#one-shot">one-shot</a>
  sensors; if <code>sensor_handle</code> refers to a one-shot sensor,
  <code>flush</code> must return <code>-EINVAL</code> and not generate any
  flush complete metadata event.</p>
<p>This function returns 0 on success, <code>-EINVAL</code> if the specified sensor is a
  one-shot sensor or wasn’t enabled, and a negative error number otherwise.</p>
<h2 id="poll">poll()</h2>
<pre>int (*poll)(struct sensors_poll_device_t *dev, sensors_event_t* data, int
  count);</pre>
<p>Returns an array of sensor data by filling the <code>data</code> argument. This function
  must block until events are available. It will return the number of events read
  on success, or a negative error number in case of an error.</p>
<p>The number of events returned in <code>data</code> must be less or equal to
  the <code>count</code> argument. This function shall never return 0 (no event).</p>
<h2 id="sequence_of_calls">Sequence of calls</h2>
<p>When the device boots, <code>get_sensors_list</code> is called.</p>
<p>When a sensor gets activated, the <code>batch</code> function will be called with the
  requested parameters, followed by <code>activate(..., enable=1)</code>.</p>
<p>Note that in HAL version 1_0, the order was the opposite: <code>activate</code> was called
  first, followed by <code>set_delay</code>.</p>
<p>When the requested characteristics of a sensor are changing while it is
  activated, the <code>batch</code> function is called.</p>
<p><code>flush</code> can be called at any time, even on non-activated sensors (in which case
  it must return <code>-EINVAL</code>)</p>
<p>When a sensor gets deactivated, <code>activate(..., enable=0)</code> will be called.</p>
<p>In parallel to those calls, the <code>poll</code> function will be called repeatedly to
  request data. <code>poll</code> can be called even when no sensors are activated.</p>
<h2 id="sensors_module_t">sensors_module_t</h2>
<p><code>sensors_module_t</code> is the type used to create the Android hardware module for the
  sensors. The implementation of the HAL must define an object
  <code>HAL_MODULE_INFO_SYM</code> of this type to expose the <a
  href="#get_sensors_list_list">get_sensors_list</a> function. See the definition
  of <code>sensors_module_t</code> in <a
  href="{@docRoot}devices/halref/sensors_8h.html">sensors.h</a> and the
  definition of <code>hw_module_t</code> for more information.</p>
<h2 id="sensors_poll_device_t_sensors_poll_device_1_t">sensors_poll_device_t / sensors_poll_device_1_t</h2>
<p><code>sensors_poll_device_1_t</code> contains the rest of the methods defined above:
  <code>activate</code>, <code>batch</code>, <code>flush</code> and
  <code>poll</code>. Its <code>common</code> field (of type <a
  href="{@docRoot}devices/halref/structhw__device__t.html">hw_device_t</a>)
  defines the version number of the HAL.</p>
<h2 id="sensor_t">sensor_t</h2>
<p><code>sensor_t</code> represents an <a href="index.html">Android sensor</a>. Here are some of its important fields:</p>
<p><strong>name:</strong> A user-visible string that represents the sensor. This string often
  contains the part name of the underlying sensor, the type of the sensor, and
  whether it is a wake-up sensor. For example, “LIS2HH12 Accelerometer”,
  “MAX21000 Uncalibrated Gyroscope”, “BMP280 Wake-up Barometer”, “MPU6515 Game
  Rotation Vector”</p>
<p><strong>handle:</strong> The integer used to refer to the sensor when registering to it or
  generating events from it.</p>
<p><strong>type:</strong> The type of the sensor. See the explanation of sensor
type in <a href="index.html">What are Android sensors?</a> for more details, and see <a
href="sensor-types.html">Sensor types</a> for official sensor types. For
non-official sensor types, <code>type</code> must start with <code>SENSOR_TYPE_DEVICE_PRIVATE_BASE</code></p>
<p><strong>stringType:</strong> The type of the sensor as a string. When the sensor has an official
  type, set to <code>SENSOR_STRING_TYPE_*</code>. When the sensor has a manufacturer specific
  type, <code>stringType</code> must start with the manufacturer reverse domain name. For
  example, a sensor (say a unicorn detector) defined by the
  <em>Cool-product</em> team at Fictional-Company could use
  <code>stringType=”com.fictional_company.cool_product.unicorn_detector”</code>.
  The <code>stringType</code> is used to uniquely identify non-official sensors types. See <a
  href="{@docRoot}devices/halref/sensors_8h.html">sensors.h</a> for more
  information on types and string types.</p>
<p><strong>requiredPermission:</strong> A string representing the permission that applications must
  possess to see the sensor, register to it and receive its data. An empty string
  means applications do not require any permission to access this sensor. Some
  sensor types like the <a href="sensor-types.html#heart_rate">heart rate
  monitor</a> have a mandatory <code>requiredPermission</code>. All sensors
  providing sensitive user information (such as the heart rate) must be protected by a permission.</p>
<p><strong>flags:</strong> Flags for this sensor, defining the sensor’s reporting mode and whether
  the sensor is a wake-up sensor or not. For example, a one-shot wake-up sensor
  will have <code>flags = SENSOR_FLAG_ONE_SHOT_MODE | SENSOR_FLAG_WAKE_UP</code>. The bits of
  the flag that are not used in the current HAL version must be left equal to 0.</p>
<p><strong>maxRange:</strong> The maximum value the sensor can report, in the same unit as the
  reported values. The sensor must be able to report values without saturating
  within <code>[-maxRange; maxRange]</code>. Note that this means the total range of the
  sensor in the generic sense is <code>2*maxRange</code>. When the sensor reports values over
  several axes, the range applies to each axis. For example, a “+/- 2g”
  accelerometer will report <code>maxRange = 2*9.81 = 2g</code>.</p>
<p><strong>resolution:</strong> The smallest difference in value that the sensor can measure.
  Usually computed based on <code>maxRange</code> and the number of bits in the measurement.</p>
<p><strong>power:</strong> The power cost of enabling the sensor, in milliAmps. This is nearly
  always more that the power consumption reported in the datasheet of the
  underlying sensor. See <a
href="sensor-types.html#base_sensors_=_not_equal_to_physical_sensors">Base
sensors != physical sensors</a> for more details and see <a
href="power-use.html#power_measurement_process">Power measurement process</a> for details on
how to measure the power consumption of a sensor. If the
  sensor’s power consumption depends on whether the device is moving, the power
  consumption while moving is the one reported in the <code>power</code> field.</p>
<p><strong>minDelay:</strong> For continuous sensors, the sampling period, in microseconds,
  corresponding to the fastest rate the sensor supports. See <a href="#sampling_period_ns">sampling_period_ns</a> for details on how this value is used. Beware that <code>minDelay</code> is expressed in
  microseconds while <code>sampling_period_ns</code> is in nanoseconds. For on-change and
  special reporting mode sensors, unless otherwise specified, <code>minDelay</code> must be 0.
  For one-shot sensors, it must be -1.</p>
<p><strong>maxDelay:</strong> For continuous and on-change sensors, the sampling period, in
  microseconds, corresponding to the slowest rate the sensor supports. See <a href="#sampling_period_ns">sampling_period_ns</a> for details on how this value is used. Beware that <code>maxDelay</code> is expressed in
  microseconds while <code>sampling_period_ns</code> is in nanoseconds. For special and
  one-shot sensors, <code>maxDelay</code> must be 0.</p>
<p><strong>fifoReservedEventCount:</strong> The number of events reserved for this sensor in the
  hardware FIFO. If there is a dedicated FIFO for this sensor, then
  <code>fifoReservedEventCount</code> is the size of this dedicated FIFO. If the FIFO is
  shared with other sensors, <code>fifoReservedEventCount</code> is the size of the part of
  the FIFO that is reserved for that sensor. On most shared-FIFO systems, and on
  systems that do not have a hardware FIFO this value is 0.</p>
<p><strong>fifoMaxEventCount:</strong> The maximum number of events that could be stored in the
  FIFOs for this sensor. This is always greater or equal to
  <code>fifoReservedEventCount</code>. This value is used to estimate how quickly the FIFO
  will get full when registering to the sensor at a specific rate, supposing no
  other sensors are activated. On systems that do not have a hardware FIFO,
  <code>fifoMaxEventCount</code> is 0. See <a href="batching.html">Batching</a> for more details.</p>
<p>For sensors with an official sensor type, some of the fields are overwritten by
  the framework. For example, <a
  href="sensor-types.html#accelerometer">accelerometer</a> sensors are forced to
  have a continuous reporting mode, and <a
  href="sensor-types.html#heart_rate">heart rate</a> monitors are forced to be
  protected by the <code>SENSOR_PERMISSION_BODY_SENSORS</code> permission.</p>
<h2 id="sensors_event_t">sensors_event_t</h2>
<p>Sensor events generated by Android sensors and reported through the <a
href="#poll">poll</a> function are of <code>type sensors_event_t</code>. Here are some
important fields of <code>sensors_event_t</code>:</p>
<p><strong>version:</strong> Must be <code>sizeof(struct sensors_event_t)</code></p>
<p><strong>sensor:</strong> The handle of the sensor that generated the event, as defined by
  <code>sensor_t.handle</code>.</p>
<p><strong>type:</strong> The sensor type of the sensor that generated the event, as defined by
  <code>sensor_t.type</code>.</p>
<p><strong>timestamp:</strong> The timestamp of the event in nanoseconds. This is the time the
  event happened (a step was taken, or an accelerometer measurement was made),
  not the time the event was reported. <code>timestamp</code> must be synchronized with the
  <code>elapsedRealtimeNano</code> clock, and in the case of continuous sensors, the jitter
  must be small. Timestamp filtering is sometimes necessary to satisfy the CDD
  requirements, as using only the SoC interrupt time to set the timestamps
  causes too high jitter, and using only the sensor chip time to set the
  timestamps can cause de-synchronization from the
  <code>elapsedRealtimeNano</code> clock, as the sensor clock drifts.</p>
<p><strong>data and overlapping fields:</strong> The values measured by the sensor. The meaning and
  units of those fields are specific to each sensor type. See <a
  href="{@docRoot}devices/halref/sensors_8h.html">sensors.h</a> and the
  definition of the different <a href="sensor-types.html">Sensor types</a> for a
  description of the data fields. For some sensors, the accuracy of the
  readings is also reported as part of the data, through a <code>status</code> field. This
  field is only piped through for those select sensor types, appearing at the SDK
  layer as an accuracy value. For those sensors, the fact that the status field
  must be set is mentioned in their <a href="sensor-types.html">sensor type</a> definition.</p>
<h3 id="metadata_flush_complete_events">Metadata flush complete events</h3>
<p>Metadata events have the same type as normal sensor events:
  <code>sensors_event_meta_data_t = sensors_event_t</code>. They are returned together with
  other sensor events through poll. They possess the following fields:</p>
<p><strong>version:</strong> Must be <code>META_DATA_VERSION</code></p>
<p><strong>type:</strong> Must be <code>SENSOR_TYPE_META_DATA</code></p>
<p><strong>sensor, reserved, and timestamp</strong>: Must be 0</p>
<p><strong>meta_data.what:</strong> Contains the metadata type for this event. There is currently a
  single valid metadata type: <code>META_DATA_FLUSH_COMPLETE</code>.</p>
<p><code>META_DATA_FLUSH_COMPLETE</code> events represent the completion of the flush of a
  sensor FIFO. When <code>meta_data.what=META_DATA_FLUSH_COMPLETE</code>, <code>meta_data.sensor</code>
  must be set to the handle of the sensor that has been flushed. They are
  generated when and only when <code>flush</code> is called on a sensor. See the section on
  the <a href="#flush_sensor">flush</a> function for more information.</p>