[libcamera-devel] [PATCH v4 8/8] android: Implement flush() camera operation
Hirokazu Honda
hiroh at chromium.org
Fri Jun 4 10:48:45 CEST 2021
Hi Paul and Laurent,
On Wed, Jun 2, 2021 at 1:38 PM <paul.elder at ideasonboard.com> wrote:
> Hi Jacopo,
>
> On Wed, Jun 02, 2021 at 03:58:47AM +0300, Laurent Pinchart wrote:
> > Hi Jacopo,
> >
> > On Mon, May 31, 2021 at 03:59:53AM +0300, Laurent Pinchart wrote:
> > > On Fri, May 28, 2021 at 10:53:28AM +0300, Laurent Pinchart wrote:
> > > > On Fri, May 28, 2021 at 12:03:59AM +0200, Jacopo Mondi wrote:
> > > > > Implement the flush() camera operation in the CameraDevice class
> > > > > and make it available to the camera framework by implementing the
> > > > > operation wrapper in camera_ops.cpp.
> > > > >
> > > > > Introduce a new camera state State::Flushing to handle concurrent
> > > > > flush() and process_capture_request() calls.
> > > > >
> > > > > As flush() can race with processCaptureRequest() protect it
> > > > > by introducing a new State::Flushing state that
> > > > > processCaptureRequest() inspects before queuing the Request to the
> > > > > Camera. If flush() is in progress while processCaptureRequest() is
> > > > > called, return the current Request immediately in error state. If
> > > > > flush() has completed and a new call to processCaptureRequest() is
> > > > > made just after, start the camera again before queuing the request.
> > > > >
> > > > > Signed-off-by: Jacopo Mondi <jacopo at jmondi.org>
> > > > > ---
> > > > > src/android/camera_device.cpp | 74
> ++++++++++++++++++++++++++++++++++-
> > > > > src/android/camera_device.h | 3 ++
> > > > > src/android/camera_ops.cpp | 8 +++-
> > > > > 3 files changed, 83 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/src/android/camera_device.cpp
> b/src/android/camera_device.cpp
> > > > > index a20c3eaa0ff6..6a8d4d4d5f76 100644
> > > > > --- a/src/android/camera_device.cpp
> > > > > +++ b/src/android/camera_device.cpp
> > > > > @@ -797,6 +797,23 @@ void CameraDevice::close()
> > > > > camera_->release();
> > > > > }
> > > > >
> > > > > +void CameraDevice::flush()
> > > > > +{
> > > > > + {
> > > > > + MutexLocker stateLock(stateMutex_);
> > > > > + if (state_ != State::Running)
> > > > > + return;
> > > > > +
> > > > > + state_ = State::Flushing;
> > > > > + }
> > > > > +
> > > > > + worker_.stop();
> > > > > + camera_->stop();
> > > > > +
> > > > > + MutexLocker stateLock(stateMutex_);
> > > > > + state_ = State::Stopped;
> > > > > +}
> > > > > +
> > > > > void CameraDevice::stop()
> > > > > {
> > > > > MutexLocker stateLock(stateMutex_);
> > > > > @@ -1894,15 +1911,46 @@ int
> CameraDevice::processControls(Camera3RequestDescriptor *descriptor)
> > > > > return 0;
> > > > > }
> > > > >
> > > > > +void CameraDevice::abortRequest(camera3_capture_request_t
> *request)
> > > > > +{
> > > > > + notifyError(request->frame_number, nullptr,
> CAMERA3_MSG_ERROR_REQUEST);
> > > > > +
> > > > > + camera3_capture_result_t result = {};
> > > > > + result.num_output_buffers = request->num_output_buffers;
> > > > > + result.frame_number = request->frame_number;
> > > > > + result.partial_result = 0;
> > > > > +
> > > > > + std::vector<camera3_stream_buffer_t>
> resultBuffers(result.num_output_buffers);
> > > > > + for (auto [i, buffer] : utils::enumerate(resultBuffers)) {
> > > > > + buffer = request->output_buffers[i];
> > > > > + buffer.release_fence =
> request->output_buffers[i].acquire_fence;
> > > > > + buffer.acquire_fence = -1;
> > > > > + buffer.status = CAMERA3_BUFFER_STATUS_ERROR;
> > > > > + }
> > > > > + result.output_buffers = resultBuffers.data();
> > > > > +
> > > > > + callbacks_->process_capture_result(callbacks_, &result);
> > > > > +}
> > > > > +
> > > > > int CameraDevice::processCaptureRequest(camera3_capture_request_t
> *camera3Request)
> > > > > {
> > > > > if (!isValidRequest(camera3Request))
> > > > > return -EINVAL;
> > > > >
> > > > > {
> > > > > + /*
> > > > > + * Start the camera if that's the first request we
> handle after
> > > > > + * a configuration or after a flush.
> > > > > + *
> > > > > + * If flush is in progress, return the pending
> request
> > > > > + * immediately in error state.
> > > > > + */
> > > > > MutexLocker stateLock(stateMutex_);
> > > > > + if (state_ == State::Flushing) {
> > > > > + abortRequest(camera3Request);
> > > > > + return 0;
> > > > > + }
> > > > >
> > > > > - /* Start the camera if that's the first request we
> handle. */
> > > > > if (state_ == State::Stopped) {
> > > > > worker_.start();
> > > > >
> > > > > @@ -2004,6 +2052,30 @@ int
> CameraDevice::processCaptureRequest(camera3_capture_request_t *camera3Reques
> > > > > if (ret)
> > > > > return ret;
> > > > >
> > > > > + /*
> > > > > + * Just before queuing the request, make sure flush() has
> not
> > > > > + * been called while this function was running. If flush
> is in progress
> > > > > + * abort the request. If flush has completed and has
> stopped the camera
> > > > > + * we have to re-start it to be able to process the
> request.
> > > > > + */
> > > > > + MutexLocker stateLock(stateMutex_);
> > > > > + if (state_ == State::Flushing) {
> > > > > + abortRequest(camera3Request);
> > > > > + return 0;
> > > > > + }
> > > >
> > > > It seems possibly overkill to do this check twice, but it shouldn't
> > > > hurt. I suspect we'll rework this code later, possibly by adding a
> > > > Camera::flush() in the libcamera native API, although I'm not
> entirely
> > > > sure what benefit it would bring compared to a stop/start. For now
> this
> > > > should be fine.
> > > >
> > > > > +
> > > > > + if (state_ == State::Stopped) {
> > > > > + worker_.start();
> > > > > +
> > > > > + ret = camera_->start();
> > > > > + if (ret) {
> > > > > + LOG(HAL, Error) << "Failed to start
> camera";
> > > > > + return ret;
> > > > > + }
> > > > > +
> > > > > + state_ = State::Running;
> > > > > + }
> > > >
> > > > This, however, bothers me a bit. Why do we need to start the camera
> in
> > > > two different locations ? Could we drop the first start above ? And
> if
> > > > we do so, given that preparing the request should be a short
> operation,
> > > > I wonder if we shouldn't also drop the first Flushing check at the
> top
> > > > of this function.
> > >
> > > This shouldn't be a blocker though, so I'll merge the series after
> > > running tests. We can address the issue on top.
> >
> > I'm afraid this series causes CTS regressions :-(
> >
> > I'm running the full camera tests with
> >
> > run cts-dev --skip-preconditions -a x86_64 -m CtsCameraTestCases
> >
> > With the current master branch, I get 22 or 23 failures (some are a bit
> > random), and with this series, it increases to 25. Here's the diff:
> >
> > @@ -3,14 +3,16 @@
> > android.hardware.camera2.cts.ImageReaderTest#testRawPrivate
> > android.hardware.camera2.cts.ImageReaderTest#testRepeatingRawPrivate
> > android.hardware.camera2.cts.RecordingTest#testSupportedVideoSizes
> > -android.hardware.camera2.cts.RecordingTest#testVideoSnapshot
> >
> android.hardware.camera2.cts.RobustnessTest#testMandatoryOutputCombinations
> > +android.hardware.camera2.cts.StillCaptureTest#testFocalLengths
> > +android.hardware.camera2.cts.StillCaptureTest#testJpegExif
> >
> android.hardware.camera2.cts.StillCaptureTest#testStillPreviewCombination
> > android.hardware.camera2.cts.SurfaceViewPreviewTest#testDeferredSurfaces
> > android.hardware.cts.CameraGLTest#testSetPreviewTextureBothCallbacks
> > android.hardware.cts.CameraGLTest#testSetPreviewTexturePreviewCallback
> > android.hardware.cts.CameraTest#testFocusDistances
> > android.hardware.cts.CameraTest#testImmediateZoom
> > +android.hardware.cts.CameraTest#testJpegExif
> > android.hardware.cts.CameraTest#testPreviewCallback
> > android.hardware.cts.CameraTest#testPreviewCallbackWithBuffer
> > android.hardware.cts.CameraTest#testPreviewCallbackWithPicture
>
> With the subplan, all of the existing ones are excluded. The
> RecordingTests are failing at random (I'll file a bug report for this).
>
> I've managed to reproduce the three extra failures though, and they
> consistently fail with a segfault.
>
>
I ran android.hardware.camera2.cts.StillCaptureTest and the failure tests
pass.
Could you tell me on top of what commit you applied the patch series?
-Hiro
>
> Paul
>
> >
> > There's some randomness in the RecordingTest, so that may not be
> > significant. The other three tests seem to pass consistently in master,
> > and fail consistently with the series applied. They also fail when run
> > in isolation.
> >
> > > > The rest of the patch looks good to me.
> > > >
> > > > > +
> > > > > worker_.queueRequest(descriptor.request_.get());
> > > > >
> > > > > {
> > > > > diff --git a/src/android/camera_device.h
> b/src/android/camera_device.h
> > > > > index c949fa509ca4..4aadb27c562c 100644
> > > > > --- a/src/android/camera_device.h
> > > > > +++ b/src/android/camera_device.h
> > > > > @@ -43,6 +43,7 @@ public:
> > > > >
> > > > > int open(const hw_module_t *hardwareModule);
> > > > > void close();
> > > > > + void flush();
> > > > >
> > > > > unsigned int id() const { return id_; }
> > > > > camera3_device_t *camera3Device() { return
> &camera3Device_; }
> > > > > @@ -92,6 +93,7 @@ private:
> > > > >
> > > > > enum class State {
> > > > > Stopped,
> > > > > + Flushing,
> > > > > Running,
> > > > > };
> > > > >
> > > > > @@ -106,6 +108,7 @@ private:
> > > > > getRawResolutions(const libcamera::PixelFormat
> &pixelFormat);
> > > > >
> > > > > libcamera::FrameBuffer *createFrameBuffer(const
> buffer_handle_t camera3buffer);
> > > > > + void abortRequest(camera3_capture_request_t *request);
> > > > > void notifyShutter(uint32_t frameNumber, uint64_t
> timestamp);
> > > > > void notifyError(uint32_t frameNumber, camera3_stream_t
> *stream,
> > > > > camera3_error_msg_code code);
> > > > > diff --git a/src/android/camera_ops.cpp
> b/src/android/camera_ops.cpp
> > > > > index 696e80436821..8a3cfa175ff5 100644
> > > > > --- a/src/android/camera_ops.cpp
> > > > > +++ b/src/android/camera_ops.cpp
> > > > > @@ -66,8 +66,14 @@ static void hal_dev_dump([[maybe_unused]] const
> struct camera3_device *dev,
> > > > > {
> > > > > }
> > > > >
> > > > > -static int hal_dev_flush([[maybe_unused]] const struct
> camera3_device *dev)
> > > > > +static int hal_dev_flush(const struct camera3_device *dev)
> > > > > {
> > > > > + if (!dev)
> > > > > + return -EINVAL;
> > > > > +
> > > > > + CameraDevice *camera = reinterpret_cast<CameraDevice
> *>(dev->priv);
> > > > > + camera->flush();
> > > > > +
> > > > > return 0;
> > > > > }
> > > > >
> >
> > --
> > Regards,
> >
> > Laurent Pinchart
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.libcamera.org/pipermail/libcamera-devel/attachments/20210604/75efff49/attachment-0001.htm>
More information about the libcamera-devel
mailing list