Allow for both the main and self path streams to be configured. This
change adds the self path as an internal stream to the pipeline handler.
It is not exposed as a Camera stream so it can not yet be used.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Create RkISP1Frames from camera data instead of picking information out
from it. This is done to prepare for multi stream support where more
information from the camera data will be needed.
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 buffer ready handlers are designed for a single application facing
stream from the main path. To prepare for multiple application facing
streams from main and/or self path the handlers need to be prepared.
The data keeping track of the frame number and advancing the timeline
can be moved from the application facing buffer ready handler to the
statistics handler. For each request processed there will always be a
statistic buffer and as the ISP is inline and is the source of both
main, self and statistic paths there is no change in behavior.
The application facing handler no longer needs a special case for
cancelled frames and can be made simpler. With this change the handlers
are ready to deal with any combinations of application facing streams.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
In preparation of supporting both the main and self path configure all
the media graph links as a part of the configuration step. Before this
change the link between ISP and DMA engine was setup at match time as
the only supported path was the main path and only the link between
sensor and ISP was updated at part of the configuration step.
The main path is still the only path between ISP and DMA engine that is
possible to enable.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
There is no need to check if Request contains a buffer belonging the
RkISP1 Camera as this is already done in Camera::queueRequest(), remove
the redundant check.
Reported-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
The RkISP1 pipeline originally only supported NV formats which have 2
planes. When support for YUV formats was added the plane count on the
output format was not made to reflect this. Instead of hard coding the
plane count to 2 fetch the number of planes from the format information.
Reported-by: Jacopo Mondi <jacopo@jmondi.org>
Fixes: 2b1a908b52 ("libcamera: camera: Add a validation API to the CameraConfiguration class")
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Handle the case where a FrameBuffer that has been externally allocated
(i.e. not through the v4l2 video device) is passed into a Request.
We must store the buffer pointer in the stream internal buffer list to
identify when used.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
By using a set container, we can easily insert/remove buffer ids that have
been mmaped by the IPA. This will be required to track buffers allocated
externally and passed to the pipeline handler through a Request.
Move the IPA buffer mapping code into a function to remove duplicated
code.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
By using a map container, we can easily insert/remove buffers from the
buffer list. This will be required to track buffers allocated externally
and passed to the pipeline handler through a Request.
Replace the buffer index tracking with an id generated internally by the
stream object.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Add further queueing into the RPiStream object to ensure that we always
follow the buffer ordering (be it internal or external) given by
incoming Requests.
This is essential, otherwise we risk dropping frames that are meant to
be part of a Request, and can cause the pipeline to stall indefinitely.
This also prevents any possibility of mismatched frame buffers going
through the pipeline and out to the application.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Tested-by: David Plowman <david.plowman@raspberrypi.com>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Stop using v4l2_videodevice::allocateBuffer() for internal buffers and
instead export/import all buffers. This allows the pipeline to return
any stream buffer requested by the application as zero-copy.
Advertise the Unicam Image stream as the RAW capture stream now.
The RPiStream object now maintains a new list of buffers that are
available to queue into a device. This is needed to distinguish between
FrameBuffers allocated for internal use vs externally provided buffers.
When a Request comes in, if a buffer is not provided for an exported
stream, we re-use a buffer from this list.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Tested-by: David Plowman <david.plowman@raspberrypi.com>
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
When calculating the pipeline configuration for the IPU3 platform,
libcamera tries to be smart and select the smallest sensor frame
resolution large enough to accommodate the stream sizes
requested by the application.
While this makes a lot of sense, in practice optimizing the
selected sensor resolution makes the pipeline configuration calculation
process fail in multiple occasions, or results in stalls during capture.
As a trivial example, capturing with cam with the following command
line results in a stall:
$ cam -swidth=1280,height=720 -swidth=640,height=480 -c1 -C
Likewise, the Android HAL supported format enumeration fails in
reporting smaller resolutions as supported when used with the OV5670
sensor.
320x240:
DEBUG IPU3 ipu3.cpp:192 CIO2 configuration: 648x486-SGRBG10_IPU3
ERROR IPU3 imgu.cpp:408 Failed to calculate pipe configuration
ERROR IPU3 ipu3.cpp:299 Failed to calculate pipe configuration: unsupported resolutions.
640x480:
DEBUG IPU3 ipu3.cpp:192 CIO2 configuration: 320x240-SGRBG10_IPU3
ERROR IPU3 imgu.cpp:408 Failed to calculate pipe configuration
ERROR IPU3 ipu3.cpp:299 Failed to calculate pipe configuration: unsupported resolutions.
Furthermore the reference xml files used for the IPU3 camera
configuration on the ChromeOS platform restricts the number of sensor
resolution to be used for the OV5670 sensor to 2 from the 6 supported by
the driver [1].
The selection criteria of the correct CIO2 mode are not specified, and
for the time being, as a workaround, always use the sensor maximum
resolution at the expense of frame rate and bus bandwidth to allow the
pipeline to successfully support smaller modes for the OV5670 sensor and
solve pipeline stalls when capturing with both sensors.
[1] See the <sensor_modes> enumeration in:
https://chromium.googlesource.com/chromiumos/overlays/board-overlays/+/master/baseboard-poppy/media-libs/cros-camera-hal-configs-poppy/files/gcss/graph_settings_ov5670.xml
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Improve the device opening at class construction time by testing all
the available device paths and printout the appropriate error message.
While at it, initialize dmaHeapHandle_ to -1 and check for its value
at destruction time.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The default sensor orientation must be set early on in match() to ensure
generateConfiguration() and configure() return out the correct Bayer
ordering to the application. This is particularly important for RAW
capture streams.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
We build libcamera with -Wno-unused-parameter and this doesn't cause
much issue internally. However, it prevents catching unused parameters
in inline functions defined in public headers. This can lead to
compilation warnings for applications compiled without
-Wno-unused-parameter.
To catch those issues, remove -Wno-unused-parameter and fix all the
related warnings with [[maybe_unused]].
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
The Stream pointer just acts as a key in the Request object. There is no
good use-case to modify a stream from a pointer retrieved from the
Request, make it const. This allows pipeline handlers to better express
that the Stream pointer is retrieved in a Request should just be treated
as a key.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
When switching from V4L2 to DRM pixel formats V4L2_PIX_FMT_GREY was
dropped form the list of supported formats but the arrays size was never
decreased, fix this.
Fixes: 448716d8f7 ("libcamera: Switch PixelFormat to DRM FourCC values")
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Rename Camera::name() to camera::id() to better describe what it
represents, a unique and stable ID for the camera. While at it improve
the documentation for the camera ID to describe it needs to be stable
for a camera between resets of the system.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Generate camera names that are unique and persistent between system
resets. The name is constructed from the USB device information as well
as the USB controller on the host.
Before this change example of camera names:
Venus USB2.0 Camera: Venus USB2
Logitech Webcam C930e
After this change the same cameras are:
\_SB_.PCI0.RP05.PXSX-2.1.1:1.0-0ac8:3420
\_SB_.PCI0.RP05.PXSX-2.4:1.0-046d:0843
On OF-based system:
/base/soc/usb@7e980000/usb-port@1-1.2:1.0-0ac8:3420
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Use the CameraSensor ID as the camera name in pipelines that uses a
CameraSensors, this is done in preparation of turning the camera name
into an ID. The CameraSensor ID meets the requirements that will be put
on camera ID.
Before this change example of camera names:
* OF based systems
ov5695 7-0036
ov2685 7-003c
* ACPI based systems
ov13858 8-0010
ov5670 10-0036
* VIMC
VIMC Sensor B
After this change the same cameras are:
* OF based systems
/base/i2c@ff160000/camera@36
/base/i2c@ff160000/camera@36
* ACPI based systems
\_SB_.PCI0.I2C2.CAM0
\_SB_.PCI0.I2C4.CAM1
* VIMC
platform/vimc.0 Sensor B
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
GCC5 does not provide prototypes for the math library functions defined
in the math.h header for the std:: namespace.
Include the C++ <cmath> header in place of <math.h> as it defines
overloads for the std::abs and std::fmod function.
This goes intentionally against the libcamera coding guidelines, and
is reported as warning by checkpatch.py.
Fixes: 968ab9bad0 ("libcamera: ipu3: imgu: Calculate ImgU pipe configuration")
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The ImgUDevice::configureInput() function does not only configure the
input format but applies rectangles to the IF, BDS and GDC components.
Rename it to ImgUDevice::configure().
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Instrument the ImgUDevice::configureInput() function to use the provided
pipe configuration parameters to configure the IF, BDS and GDC
rectangles.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Collect the desired ImgU pipe configuration while assigning
streams in the pipeline handler validate() function and ask the
ImgUDevice class to calculate the pipe configuration parameters.
If the requested pipe configuration results in a non-valid
configuration, return an error from the validate() function.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Instrument the ImgU component to dynamically calculate the image
manipulation pipeline intermediate sizes.
To correctly configure the ImgU it is necessary to program the IF, BDS
and GDC sizes, which are currently fixed to the input frame size.
The procedure used to calculate the intermediate sizes has been ported
from the pipe_config.py python script, available at:
https://github.com/intel/intel-ipu3-pipecfg
at revision:
61e83f2f7606 ("Add more information into README")
Define two structures (ImgUDevice::Pipe and ImgUDevice::PipeConfig)
to allow the pipeline handler to supply and retrieve configuration
parameters from the ImgU.
Finally, add a new operation to the ImgUDevice that calculates
the pipe configuration parameters based on the requested input and
output sizes.
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
The IPU3CameraConfiguration::streams_ field was used to keep an
association between the StreamConfiguration and the assigned streams
before CameraConfiguration::setStream() was called at configure() time.
The stream assignment was based on the order in which elements were
inserted in the vector, implementing a fragile association between
streams and their intended configurations.
As it is now possible to assign streams at validation time, there is no
need to keep that association in place, and the streams_ vector is now
unused.
Remove it and the associated accessor method from the
IPU3CameraConfiguration class.
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Remove the adjustStream() and assignStream() methods, and perform stream
adjustment and assignment while iterating the StreamConfiguration
items.
The adjustStream() implementation had some arbitrary assumption, like
the main output having to be as large as the sensor resolution, and did
not take into account the different alignment requirements between the
main output and the viewfinder output.
The assignStream() implementation also assumes only full-size streams
can be produced by the main output, and having it as a separate function
prevents adjusting streams according to which output they are assigned.
Blend the two implementation in a single loop and perform the required
stream adjustment and assignment in one go.
As streams are now assigned at validate() time, remove the same
operation from the configure() function.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Add a const version of the CIO2Device::sensor() method, that retrieves
a const pointer to the sensor_ class member, to be called by users
which only own a const reference to a CIO2Device class instance.
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
