Add templates to mojo to generate code for the IPC mechanism. These
templates generate:
- module header
- module serializer
- IPA proxy cpp, header, and worker
Given an input data definition mojom file for a pipeline.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The ControlSerializer saves all ControlInfoMaps that it has already
(de)serialized, in order to (de)serialize ControlLists that contain the
ControlInfoMaps. Leverage this to cache ControlInfoMaps, such that the
ControlSerializer will not re-(de)serialize a ControlInfoMap that it has
previously (de)serialized.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Currently the call to the configure method of rkisp1 IPA
is called upon the 'start' of the pipeline. This should be done
in the 'configure' method instead.
Signed-off-by: Dafna Hirschfeld <dafna.hirschfeld@collabora.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
MappedBuffers have a custom move constructor and assignment operator to
ensure that memory is not unmapped during object manipulation.
Because of the user-defined move constructor, the implicitly-declared
copy-constructor will already be deleted, however delete it explicitly
to help readability of the code, and make it clear that the object can
not be copied.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Replace existing use cases where the copy constructor and copy
assignment operator are deleted with the LIBCAMERA_DISABLE_COPY
statement
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The ControlId and Control classes disable the copy constructor and
assignment operator, but they should also prevent move construction and
assignment.
Utilise LIBCAMERA_DISABLE_COPY_AND_MOVE to fully disable these
functions.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Convert MediaLink, MediaPad, and MediaEntity to declare
LIBCAMERA_DISABLE_COPY_AND_MOVE. These classes already deleted their
copy constructor but not the assignment operator.
They should also not be movable, so expand to fully disable both copying
and moving.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Convert the existing copy, move and assignment operators to use the new
macros.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
It can be difficult to correctly parse the syntax for copy/move and the
associated assignment operators.
Provide helpers as syntactic sugar to facilitate disabling either the
copy constructor, and copy assignment operator, and the move constructor
and move assignment operator in a way which is explicit and clear.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The Extensible concept is a generic Class helper. To prepare for
further class helper additions, move the specific extensible
implementation and header to a more generic class header and source.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Applying the manual_r_ and manual_b_ values is entirely removed from
the asynchronous thread where their use constituted a race hazard. The
main thread now deals with them entirely, involving the following
changes.
1. SetManualGains() applies the new values directly to the
"sync_results", meaning that Prepare() will jump to the new values
immediately (which is a better behaviour).
2. Process() does not restart the asynchronous thread when manual
gains are in force.
3. The asynchronous thread might be running when manual gains are set,
so we ignore the results produced in this case.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Naushir Patuck <naush@raspberrypi.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The meson custom_target() function accepts a 'depend_files' argument to
list additional dependency files. Files already listed in the 'command'
argument are handled automatically, and don't need to be added manually.
Drop them.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The meson.build files mix array declarations with and without a space
after the opening and before the closing square bracket. The vast
majority of cases don't use spaces, so standardize on that.
While it it, fix indentation in a few places.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The variable frame_count2_ is not needed as Prepare() and Process()
always run in lock step one after the other.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Naushir Patuck <naush@raspberrypi.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The application provided noise reduction mode gets passed into the
denoise controller. The denoise controller in turn returns the mode to
the IPA which now sets up the colour denoise processing appropriately.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
This denoise algorithm class will be used to pass in the user requested
denoise operating mode to the controller. The existing Denoise
controller will derive from this new DenoiseAlgorithm class.
Add a denoise mode field in the denoise status metadata object for the
IPA to use when configuring the ISP.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
This change is in anticipation of the addition of a DenoiseAlgorithm
base class which the SDN class will derive from. We want to match the
metadata object name with the base class algorithm name.
This renames:
- SdnStatus metadata object to DenoiseStatus
- "sdn.status" metadata string key to "denoise.status"
- sdn_status.h header file to denoise_status.h
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Update the bcm2835-isp.h file with the latest version available in the
downstream Raspberry Pi linux repo at commit:
68878170d8a98afd6d519a3a2c909080c19de4ce
This change adds support for colour denoise processing, and the
following downstream kernel changes must be available:
https://github.com/raspberrypi/linux/pull/4069https://github.com/raspberrypi/linux/pull/4083
The Raspberry Pi image must also be running the latest firmware,
obtained by running rpi-update.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
In preparation for fast colour denoise, set the low resolution ISP
output stream (Output1) to a 1/4 resolution of the application requested
stream (Output0). This only happens if the application has not requested
an additional YUV or RGB stream.
We also constrain this 1/4 resolution to at most 1200 pixels in the
largest dimension to avoid being too taxing on memory usage and system
bandwidth.
Also switch the default StreamRole::VideoRecording to YUV420 to allow
fast colour denoise to run.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Refactor the high/low resolution stream format and output selection
routine. This change is in preparation of adding 1/4 resolution output
for fast colour denoise.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Buffers dequeued from the ImgU may not be associated with a Request as
they are internal (statistics, parameters and some RAW buffers). Fetch
the request they are used for from the frame information instead.
Fixes: 9708f49fec ("libcamera: ipu3: Share parameter and statistic buffers with IPA")
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Tested-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Pause() and Resume() are only called synchronously so paused_ does not
need to be atomic.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Calculate the JPEG maximum size using the maximum preview format size
multiplied by a 1.5 factor.
The same multiplication factor is used in the existing HAL
implementations in ChromeOS.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Supply the missing SetCurrentAperture() method (even though no one is
calling it). If we did call it, it would be called synchronously so
the atomic variable is not required.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
mode_factor_ does not need to be atomic - it is set by SwitchMode()
which runs synchronously.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
SetBrightness() and SetContrast() are only called synchronously so
there is no need for brightness_ and contrast_ to be atomic.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
SetSaturation() gets called synchronously so there is no need for
saturation_ to be atomic.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Fix a couple of places where notify_one() was called with the lock
held. Also restartAsync doesn't need the lock for its entire duration.
This change exactly matches commit db552b0b92 ("libcamera: ipa:
raspberrypi: ALSC: Improve locking in a few places") where we do the
same for ALSC (the asynchronous thread arrangement there is identical).
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
AWB settings get updated synchronously with the main thread, so the
settings_mutex_ and associated locking can be removed.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Commit 7208e70211 ("libcamera: ipu3: Always use sensor full frame size")
changed the CIO2 configuration procedure to always select the full
sensor resolution to work around an ImgU configuration issue, as
reported in the todo item introduced in said commit.
When capturing a RAW stream only and the ImgU is not involved, the CIO2
has to be configured with the requested stream size to avoid adjusting
the stream to the sensor resolution.
As an example, capturing a raw frame smaller than the sensor resolution
with this patch applied results in the stream configuration to be
correctly assigned.
$ cam -c1 -swidth=1056,height=784,role=raw
ipu3.cpp:207 CIO2 configuration: 1056x784-SGRBG10_IPU3
ipu3.cpp:222 Validating stream: 1056x784-SGRBG10_IPU3
ipu3.cpp:233 Assigned 1056x784-SGRBG10_IPU3 to the raw stream
Without this patch the same operation results in the stream resolution
to be adjusted to the sensor resolution.
$ cam -c1 -swidth=1056,height=784,role=raw
ipu3.cpp:201 CIO2 configuration: 4224x3136-SGRBG10_IPU3
ipu3.cpp:216 Validating stream: 1056x784-SGRBG10_IPU3
ipu3.cpp:227 Assigned 4224x3136-SGRBG10_IPU3 to the raw stream
ipu3.cpp:297 Stream 0 configuration adjusted to 4224x3136-SGRBG10_IPU3
Camera configuration adjusted
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The CaptureResultTest#testCameraCaptureResultAllKeys enforces the
presence of the ANDROID_STATISTICS_HOT_PIXEL_MAP_MODE result
metadata.
Report ANDROID_STATISTICS_HOT_PIXEL_MAP_MODE_OFF as the Camera HAL
does not support the generation of an hot pixel correction map.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The CaptureResultTest#testCameraCaptureResultAllKeys enforces the
presence of the ANDROID_SENSOR_TEST_PATTERN_MODE result metadata.
Report the only supported value of ANDROID_SENSOR_TEST_PATTERN_MODE_OFF.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Report the crop region for every completed request.
The crop region is initialized as the sensor's analogue crop
rectangle and updated when a Request with a new region completes.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The CaptureResultTest#testCameraCaptureResultAllKeys CTS test
enforces the presence of the auto focus mode in the preview
request template.
Add the only supported mode AF_MODE_OFF in the generated template.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The camera HAL does currently list AF_MODE_OFF as the only supported
auto-focus mode in the ANDROID_CONTROL_AF_AVAILABLE_MODES static
metadata. For this reason reporting AF_MODE_AUTO as scene override
is wrong.
At the moment scene override is disabled in the Camera HAL, but this
is worth a fix nonetheless.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The CaptureResultTest#testCameraCaptureResultAllKeys enforces the
presence of the ANDROID_CONTROL_AE_EXPOSURE_COMPENSATION result
metadata.
Report 0 as the current value for
ANDROID_CONTROL_AE_EXPOSURE_COMPENSATION in the result metadata
pack.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The CaptureResultTest#testCameraCaptureResultAllKeys enforces the
presence of the ANDROID_COLOR_CORRECTION_ABERRATION_MODE result
metadata.
Report the only supported value of
ANDROID_COLOR_CORRECTION_ABERRATION_MODE_OFF.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The cam option --capture=N is suppose to only capture N requests. But if
the processing done for each request is large (such as writing it to a
slow disk) the current implementation could queue more than N requests
before the exit condition is detected and capturing stopped.
Solve this by only queueing N requests while still waiting for N
requests to complete before exiting.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cam uses libevent to deal with threads and idle loop while still
implementing its own event queue. This creates issues when the event
loop is terminated as it might get stuck in the idle loop if exit() is
called while the thread is busy with dispatchCalls().
Solve this by removing the custom event execution and instead injecting
the calls as events to the base event loop.
Reported-by: Sebastian Fricke <sebastian.fricke@posteo.net>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The 'event' variable name is usually used for events added to the base
event loop, not the loop itself. Rename the struct event_base member to
base_ as a preparation for future work adding events to the loop.
There is no functional change.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Use the IPU3Frames helper to share parameters and statistics buffers
with the IPA. Track which parameter and statistic buffer is used for
which request and make sure the parameter buffers is filled in by the
IPA before it's needed and that the statistic buffer is consumed and
meta data generated before completing the request.
With this change the IPU3 pipeline is prepared to fully operate with an
IPA component.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add a helper class to aid in associating a parameter and statistic
buffer with each request queued to the pipeline. The helper helps with
tracking the state of the extra buffers and in completing the request
once all extra processing is done.
This change only adds the helper more work is needed to integrate it
with the pipeline and an IPA.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Map and unmap the parameters and statistic buffers in the IPA when the
pipeline handler allocates and frees the buffers.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Instead of preparing for buffer importing allocate buffers that can be
used by an IPA.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Attach to the IPA and allow it to push V4L2 controls that apply to the
camera sensor. The IPA is not fully integrated in the pipeline as
statistics and parameters buffers are not yet allocated, processed by
the IPA nor queued to the hardware.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add an empty IPA skeleton for the IPU3 pipeline. The skeleton IPA
handles the flow of parameter and statistic buffers but does not read or
write anything in the buffers. It also allows the IPA to set sensor
controls but does not implement any logic to set optimal values and
instead sets the V4L2 exposure and gain controls to max and keeps them
at that setting.
This IPA is meant as a base to allow the pipeline handler to be wired up
to an IPA. The image algorithms can then later be added to the IPA
independently from also having to add plumbing to the pipeline handler.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
When adding IPA plumbing to the IPU3 pipeline handler it will be needed
to track which raw buffer was queued to the CIO2 for a specific request.
Currently if using an internally allocated raw buffer the FrameBuffer is
not exposed outside the CIO2Device as it was not needed. Prepare for the
IPA by exposing which internal raw buffer is picked for each request,
later changes will associate this with a Request.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add the parameters video device to the data structure of the ImgUDevice.
Even if the video device is configured, prepared to import buffers and
started no buffers are ever queued to it so it does not yet effect the
capture. Nor is does this change hinder the current capture mode to
function.
This is done in preparation to attach an IPA to the IPU3 pipeline.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
There is no reason to expose and call a separate configureStat() when
the statistics video device can be configured with the exact same
parameters as part of configure(). Move the configuration internally to
the ImgUDevice simplifying the interface, there is no functional change.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
