std::mutex and std::unique_lock are used in v4l2 directory,
mixing Mutex and MutexLocker. This consolidates them to Mutex
and MutexLocker.
This also fixes vidioc_dqbuf argument to take Mutex instead of
MutexLocker.
Signed-off-by: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Set the compressed flag in ENUM_FMT if the format is MJPEG. As the only
compressed format that libcamera currently supports is MJPEG, this
should be sufficient.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jean-Michel Hautbois <jeanmichel.hautbois@ideasonboard.com>
Now that libcamera has the V4L2 format names, retrieve and report those
names in ENUM_FMT. While at it, refactor the code slightly with
PixelFormatInfo.
This fixes the test failures on v4l2-compliance with the v4l2
compatilibity layer that were observed in version v4l2-compliance
version 1.21.0-4838.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jean-Michel Hautbois <jeanmichel.hautbois@ideasonboard.com>
The V4L2 compatibility layer supports the single-planar API only, and
thus exposes a single V4L2 buffer plane to applications, regardless of
the number of planes in the FrameBuffer. For multi-planar frame buffers,
the bytesused value isn't correct as it only takes the first plane into
account. Fix it by summing the bytesused values for all FrameBuffer
planes.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Jean-Michel Hautbois <jeanmichel.hautbois@ideasonboard.com>
The number of metadata planes should always match the number of frame
buffer planes. Enforce this by making the vector private and providing
accessor functions.
As this changes the public API, update all in-tree users accordingly.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Jean-Michel Hautbois <jeanmichel.hautbois@ideasonboard.com>
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Replace manual searches for V4L2 pixel format in the PixelFormatInfo
with the V4L2PixelFormat::fromPixelFormat() helper function. This
prepares for multi-planar support that will modify how V4L2 pixel
formats are stored in PixelFormatInfo.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jean-Michel Hautbois <jeanmichel.hautbois@ideasonboard.com>
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Move the functionality for the following components to the new
base support library:
- BoundMethod
- EventDispatcher
- EventDispatcherPoll
- Log
- Message
- Object
- Signal
- Semaphore
- Thread
- Timer
While it would be preferable to see these split to move one component
per commit, these components are all interdependent upon each other,
which leaves us with one big change performing the move for all of them.
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Rename the internal member variables to use the post-fixed '_' member
variable naming style. This in turn ensures variable shadowing does not
occur in the member initializer list of the constructor.
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The V4L2CameraProxy always calls V4L2Camera::getStreamConfig() right
after V4L2Camera::open(), and never afterwards. Simplify the code by
returning the initial configuration from V4L2Camera::open() and removing
V4L2Camera::getStreamConfig().
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
The V4L2 compatibility layer only uses the fmt.pix field of
curV4L2Format_. There's no need to cache the full v4l2_format, store
v4l2_pix_format only and rename the member variable from curV4L2Format_
to v4l2PixFormat_. While at it, group the V4L2-related member variables
together in the V4L2CameraProxy class.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
The setFmtFromConfig() method doesn't modify its parameter, make it
const.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@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>
For handling try_fmt, the values should be filled in by validating the
stream configuration, and not by recalculating them or manually checking
against the cached list of formats and sizes. Add a new
V4L2Camera::validateConfiguration() function to validate a configuration
and use it to obtain size, format, stride, and frameSize values.
If the format negotiation fails, return error from try_fmt and s_fmt.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The stride and frameSize should be obtained through StreamConfiguration
rather than PixelFormatInfo, as pipeline handlers might have different
values (eg. for alignment). Get the stride and frameSize values from
StreamConfiguration instead of from PixelFormatInfo.
This removes the need for V4L2CameraProxy's calculation helper functions
(bplMultiplier, imageSize, v4l2ToDrm, drmToV4L2, calculateSizeImage) and
formats, so remove them.
This also removes the need for V4L2CameraProxy::calculateSizeImage, so
remove it,.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Make the V4L2 compatibility layer thread-safe by serializing accesses to
the V4L2CameraProxy with a lock. Release the lock when blocking for
dqbuf.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Fix buffer flags related to queueing and dequeueing:
- don't allow a buffer with the same index that is already in the queue
to be enqueued again
- don't clear the done flag upon qbuf
- do clear the done flag upon dqbuf
- set the flags in V4L2CameraProxy's internal buffers, and not just in
the buffers returned from qbuf
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
In VIDIOC_REQBUFS with count = 0, if the buffers are still mmaped, they
should not be allowed to be freed. Add a check for this.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Make VIDIOC_STREAMON return -EINVAL if no buffers have been allocated
with reqbufs.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
There were no bounds checks for the index argument for VIDIOC_QUERYBUF,
VIDIOC_QBUF, and VIDIOC_DQBUF. Add them.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
In V4L2, a blocked dqbuf should not not also block a streamoff. This
means that on streamoff, the blocked dqbuf must return (with error). We
cannot do this with the libcamera semaphore, so pull out the necessary
components of a semaphore, and put them into V4L2Camera, so that dqbuf
from V4L2CameraProxy can wait on a disjunct condition of the
availability of the semaphore or the stopping of the stream.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
If VIDIOC_REQBUFS with count = 0 is called when the stream is not on,
clear the proxy's internal vector of buffer. If the stream is on when
reqbufs 0 is called, return -EBUSY.
Note that this is contrary to what the V4L2 docs say (reqbufs 0 when
streaming should also streamoff), but it is how the V4L2 implementation
works. v4l2-compliance doesn't seem to care either way, however, so we
cater to the implementation, and no longer call streamoff on reqbufs 0.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
If VIDIOC_STREMAON is called when the stream is already on, do a noop.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Make VIDIOC_DQBUF return -EINVAL if the stream is not turned on.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement VIDIOC_ENUM_FRAMESIZES in the V4L2 compatibility layer.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement VIDIOC_ENUMINPUT, VIDIOC_G_INPUT, and VIDIOC_S_INPUT. Only the
zeroth input device is supported, and the info returned by enuminput is
hardcoded and basic. This is sufficient to pass v4l2-compliance.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement VIDIOC_G_PRIORITY and VIDIOC_S_PRIORITY. The behaviour
documented in the V4L2 specification doesn't match the implementation in
the Linux kernel, implement the latter.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Fix the following v4l2-compliance error:
fail: v4l2-compliance.cpp(652): !(caps & V4L2_CAP_EXT_PIX_FORMAT)
Simply add V4L2_CAP_EXT_PIX_FORMAT to capabilities in querycap.
In addition, populate the necessary fields in struct v4l2_pix_format to
support extended pixel formats in try_fmt and g/s_fmt, and clear the
reserved field for enum_fmt.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The ioctl handlers currently don't check if arg is null, so if it ever
is, it will cause a segfault. Check that arg is null and return -EFAULT
in the main vidioc ioctl handler.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Set buffer flag V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC at VIDIOC_REQBUFS
after the buffers have been allocated.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Clear the reserved field in arg struct v4l2_reqbuffers of
VIDIOC_REQBUFS.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Free buffers, if any were previously allocated, at VIDIOC_REQBUFS with
count > 0.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Previously, since we acquired the libcamera camera upon open(), it was
impossible to support multiple open, as any subsequent opens would
return error because the camera would already be acquired.
To fix this, we first initialize the camera in the first call to
V4L2CameraProxy::open(), just to heat up the stream format cache. We
then add ownership by a V4L2CameraFile of a V4L2Camera via the
V4L2CameraProxy. All vidioc ioctls prior to reqbufs > 0 (except for
s_fmt) are able to access the camera without ownership. A call to
reqbufs > 0 (and s_fmt) will take ownership, and the ownership will be
released at reqbufs = 0. While ownership is assigned, the eventfd that
should be signaled (and cleared) by V4L2Camera and V4L2CameraProxy is
set to the V4L2CameraFile that has ownership, and is cleared when the
ownership is released. In case close() is called without a
reqbufs = 0 first, the ownership is also released on close().
We also use the V4L2CameraFile to contain all the information specific
to an open instance of the file. This removes the need to keep track of
such information within V4L2CameraProxy via multiple maps from int fd to
info.
Since V4L2 does not expect reqbufs 0 to ever return error, make
V4L2CameraProxy::freeBuffers() return void.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Prepare for using the V4L2CameraFile as a container for file-specific
information in the V4L2 compatibility layer by making it a required
argument for all V4L2CameraProxy calls that are directed from
V4L2CompatManager, which are intercepted via LD_PRELOAD. Change
V4L2CameraFile accordingly.
Also change V4L2CompatManager accordingly. Instead of keeping a map of
file descriptors to V4L2CameraProxy instances, we keep a map of
V4L2CameraFile instances to V4L2CameraProxy instances. When the
proxy methods are called, feed the file as a parameter.
The dup function is also modified, in that it is removed from
V4L2CameraProxy, and is handled completely in V4L2CompatManager, as a
map from file descriptors to V4L2CameraFile instances.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
These formats can be helpful when downstream applications or libraries
support them natively (avoiding a costly conversion).
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Tested-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Some applications (eg. Firefox, Google Chrome, Skype) use open64,
openat64, and mmap64 instead of their non-64 versions that we currently
intercept. Intercept these calls as well. _LARGEFILE64_SOURCE needs to
be set so that the 64-bit symbols are available and not synonymous to
the non-64-bit versions on 64-bit systems.
Also, since we set _FILE_OFFSET_BITS to 32 to force the various open and
mmap symbols that we export to not be the 64-bit versions, our dlsym to
get the original open and mmap calls will not automatically be converted
to their 64-bit versions. Since we intercept both 32-bit and 64-bit
versions of open and mmap, we should be using the 64-bit version to
service both. Fetch the 64-bit versions of openat and mmap directly.
musl defines the 64-bit symbols as macros that are equivalent to the
non-64-bit symbols, so we put compile guards that check if the 64-bit
symbols are defined.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Tested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> # Compile with musl
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Use the new pixel format constants to replace usage of macros from
drm_fourcc.h.
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 return value of the write to the eventfd (to signal POLLIN) from
V4L2Camera and the read from the eventfd (to clear POLLIN) from
V4L2CameraProxy was ignored. Check the return value, and print an error
message.
Reported-by: Coverity CID=290743
Reported-by: Coverity CID=290744
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Add an entry for MJPEG in V4L2CameraProxy's PixelFormatInfo list to
allow proper calculation of sizeimage for MJPEG, such that the
parameters to mmap can align properly instead of failing. This allows
MJPEG to be used in the V4L2 compatibility layer.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The V4L2 compatibility layer keeps track of the index of the next buffer
to be dequeued, to handle VIDIOC_DQBUF. This index is set to 0 on
startup and incremented (modulo #frames), and is otherwise never reset.
This means that if the last handled frame index is not #frames-1, and
the stream is restarted without restarting libcamera and the V4L2
compatilibity layer, the buffer index number won't match with the
libcamera buffer index number, causing a segfault. Fix this by resetting
the current buffer index to zero on VIDIOC_STREAMON.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
To support polling, we need to be able to signal when data is
available to be read (POLLIN), as well as events (POLLPRI). Add the
necessary calls to eventfd to allow signaling POLLIN. We signal POLLIN
by writing writing to the eventfd, and clear it by reading from the
eventfd, upon VIDIOC_DQBUF.
Note that eventfd does not support signaling POLLPRI, so we don't yet
support V4L2 events.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
If VIDIOC_REQBUFS returns -EINVAL, it signals to the application that
the requested buffer or memory type is not supported. If we return
-EINVAL due to a zero sizeimage, then the application will think that we
don't support a memory type that we actually do. We cannot error on a
zero sizeimage, because reqbufs could be called merely to probe what IO
methods we support; qv4l2, for example, called reqbufs once with userptr
and once more with mmap, both times with count=1.
On the other hand, sizeimage will be zero for formats whose size we
don't know how to calculate, such as MJPEG. If we try to stream such
formats anyway, we will get a floating point exception and crash. Issue
a warning for now, and don't return -EINVAL, so that we can continue
operation.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
We use a semaphore to atomically keep track of how many buffers are
available for dequeueing. The check for how to acquire the semaphore was
incorrect, leading to a double acquire upon a successful nonblocking
acquire. Fix this.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
VIDIOC_ENUM_FMT is meant to return -EINVAL if the requested index is out
of bounds. This bounds is obtained from the libcamera Camera's list of
formats. The bounds check for this list was incorrect; fix it.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
The V4L2 compatibility layer is licensed under the GPL. It is compiled
as a binary separate from libcamera.so, and is loaded into the address
space of processes through LD_PRELOAD to intercept calls to the C
library.
It is our understanding and intent that the GPL license doesn't
propagate to the binaries whose calls are intercepted, considering those
binaries are not derivative work of the V4L2 compatibility layer and are
not designed to be linked to the V4L2 compatibility layer. There is
however a possibly grey area if binaries are packaged with a shell
script wrapper that loads the V4L2 compatibility layer. This could lead
to license-related issues if such packaging is performed by Linux
distributions or system integrators.
To clarify the intent and lift the doubts, relicense the V4L2
compatibility layer under the LGPL. The V4L2 compatibility layer code
itself still benefits from the license protection, while its usage with
third-party binaries is clearly allowed as intended.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Acked-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Acked-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
The libcamera internal headers are located in src/libcamera/include/.
The directory is added to the compiler headers search path with a meson
include_directories() directive, and internal headers are included with
(e.g. for the internal semaphore.h header)
#include "semaphore.h"
All was well, until libcxx decided to implement the C++20
synchronization library. The __threading_support header gained a
#include <semaphore.h>
to include the pthread's semaphore support. As include_directories()
adds src/libcamera/include/ to the compiler search path with -I, the
internal semaphore.h is included instead of the pthread version.
Needless to say, the compiler isn't happy.
Three options have been considered to fix this issue:
- Use -iquote instead of -I. The -iquote option instructs gcc to only
consider the header search path for headers included with the ""
version. Meson unfortunately doesn't support this option.
- Rename the internal semaphore.h header. This was deemed to be the
beginning of a long whack-a-mole game, where namespace clashes with
system libraries would appear over time (possibly dependent on
particular system configurations) and would need to be constantly
fixed.
- Move the internal headers to another directory to create a unique
namespace through path components. This causes lots of churn in all
the existing source files through the all project.
The first option would be best, but isn't available to us due to missing
support in meson. Even if -iquote support was added, we would need to
fix the problem before a new version of meson containing the required
support would be released.
The third option is thus the only practical solution available. Bite the
bullet, and do it, moving headers to include/libcamera/internal/.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
PixelFormat was previously an alias for unsigned int but is now a
class. Make all functions taking PixelFormat do so as a const reference.
Signed-off-by: Niklas Söderlund <niklas.soderlund@ragnatech.se>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
