| /* |
| * Copyright (C) 2013-2018 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. |
| */ |
| |
| #define LOG_TAG "Camera3-OutputStream" |
| #define ATRACE_TAG ATRACE_TAG_CAMERA |
| //#define LOG_NDEBUG 0 |
| |
| #include <algorithm> |
| #include <ctime> |
| #include <fstream> |
| |
| #include <aidl/android/hardware/camera/device/CameraBlob.h> |
| #include <aidl/android/hardware/camera/device/CameraBlobId.h> |
| #include "aidl/android/hardware/graphics/common/Dataspace.h" |
| |
| #include <android-base/unique_fd.h> |
| #include <cutils/properties.h> |
| #include <ui/GraphicBuffer.h> |
| #include <utils/Log.h> |
| #include <utils/Trace.h> |
| #include <camera/StringUtils.h> |
| |
| #include <common/CameraDeviceBase.h> |
| #include "api1/client2/JpegProcessor.h" |
| #include "Camera3OutputStream.h" |
| #include "utils/TraceHFR.h" |
| |
| #ifndef container_of |
| #define container_of(ptr, type, member) \ |
| (type *)((char*)(ptr) - offsetof(type, member)) |
| #endif |
| |
| namespace android { |
| |
| namespace camera3 { |
| |
| using aidl::android::hardware::camera::device::CameraBlob; |
| using aidl::android::hardware::camera::device::CameraBlobId; |
| |
| Camera3OutputStream::Camera3OutputStream(int id, |
| sp<Surface> consumer, |
| uint32_t width, uint32_t height, int format, |
| android_dataspace dataSpace, camera_stream_rotation_t rotation, |
| nsecs_t timestampOffset, const std::string& physicalCameraId, |
| const std::unordered_set<int32_t> &sensorPixelModesUsed, IPCTransport transport, |
| int setId, bool isMultiResolution, int64_t dynamicRangeProfile, |
| int64_t streamUseCase, bool deviceTimeBaseIsRealtime, int timestampBase, |
| int mirrorMode, int32_t colorSpace, bool useReadoutTimestamp) : |
| Camera3IOStreamBase(id, CAMERA_STREAM_OUTPUT, width, height, |
| /*maxSize*/0, format, dataSpace, rotation, |
| physicalCameraId, sensorPixelModesUsed, setId, isMultiResolution, |
| dynamicRangeProfile, streamUseCase, deviceTimeBaseIsRealtime, |
| timestampBase, colorSpace), |
| mConsumer(consumer), |
| mTransform(0), |
| mTraceFirstBuffer(true), |
| mUseBufferManager(false), |
| mTimestampOffset(timestampOffset), |
| mUseReadoutTime(useReadoutTimestamp), |
| mConsumerUsage(0), |
| mDropBuffers(false), |
| mMirrorMode(mirrorMode), |
| mDequeueBufferLatency(kDequeueLatencyBinSize), |
| mIPCTransport(transport) { |
| |
| if (mConsumer == NULL) { |
| ALOGE("%s: Consumer is NULL!", __FUNCTION__); |
| mState = STATE_ERROR; |
| } |
| |
| bool needsReleaseNotify = setId > CAMERA3_STREAM_SET_ID_INVALID; |
| mBufferProducerListener = new BufferProducerListener(this, needsReleaseNotify); |
| } |
| |
| Camera3OutputStream::Camera3OutputStream(int id, |
| sp<Surface> consumer, |
| uint32_t width, uint32_t height, size_t maxSize, int format, |
| android_dataspace dataSpace, camera_stream_rotation_t rotation, |
| nsecs_t timestampOffset, const std::string& physicalCameraId, |
| const std::unordered_set<int32_t> &sensorPixelModesUsed, IPCTransport transport, |
| int setId, bool isMultiResolution, int64_t dynamicRangeProfile, |
| int64_t streamUseCase, bool deviceTimeBaseIsRealtime, int timestampBase, |
| int mirrorMode, int32_t colorSpace, bool useReadoutTimestamp) : |
| Camera3IOStreamBase(id, CAMERA_STREAM_OUTPUT, width, height, maxSize, |
| format, dataSpace, rotation, physicalCameraId, sensorPixelModesUsed, |
| setId, isMultiResolution, dynamicRangeProfile, streamUseCase, |
| deviceTimeBaseIsRealtime, timestampBase, colorSpace), |
| mConsumer(consumer), |
| mTransform(0), |
| mTraceFirstBuffer(true), |
| mUseBufferManager(false), |
| mTimestampOffset(timestampOffset), |
| mUseReadoutTime(useReadoutTimestamp), |
| mConsumerUsage(0), |
| mDropBuffers(false), |
| mMirrorMode(mirrorMode), |
| mDequeueBufferLatency(kDequeueLatencyBinSize), |
| mIPCTransport(transport) { |
| |
| if (format != HAL_PIXEL_FORMAT_BLOB && format != HAL_PIXEL_FORMAT_RAW_OPAQUE) { |
| ALOGE("%s: Bad format for size-only stream: %d", __FUNCTION__, |
| format); |
| mState = STATE_ERROR; |
| } |
| |
| if (mConsumer == NULL) { |
| ALOGE("%s: Consumer is NULL!", __FUNCTION__); |
| mState = STATE_ERROR; |
| } |
| |
| bool needsReleaseNotify = setId > CAMERA3_STREAM_SET_ID_INVALID; |
| mBufferProducerListener = new BufferProducerListener(this, needsReleaseNotify); |
| } |
| |
| Camera3OutputStream::Camera3OutputStream(int id, |
| uint32_t width, uint32_t height, int format, |
| uint64_t consumerUsage, android_dataspace dataSpace, |
| camera_stream_rotation_t rotation, nsecs_t timestampOffset, |
| const std::string& physicalCameraId, |
| const std::unordered_set<int32_t> &sensorPixelModesUsed, IPCTransport transport, |
| int setId, bool isMultiResolution, int64_t dynamicRangeProfile, |
| int64_t streamUseCase, bool deviceTimeBaseIsRealtime, int timestampBase, |
| int mirrorMode, int32_t colorSpace, bool useReadoutTimestamp) : |
| Camera3IOStreamBase(id, CAMERA_STREAM_OUTPUT, width, height, |
| /*maxSize*/0, format, dataSpace, rotation, |
| physicalCameraId, sensorPixelModesUsed, setId, isMultiResolution, |
| dynamicRangeProfile, streamUseCase, deviceTimeBaseIsRealtime, |
| timestampBase, colorSpace), |
| mConsumer(nullptr), |
| mTransform(0), |
| mTraceFirstBuffer(true), |
| mUseBufferManager(false), |
| mTimestampOffset(timestampOffset), |
| mUseReadoutTime(useReadoutTimestamp), |
| mConsumerUsage(consumerUsage), |
| mDropBuffers(false), |
| mMirrorMode(mirrorMode), |
| mDequeueBufferLatency(kDequeueLatencyBinSize), |
| mIPCTransport(transport) { |
| // Deferred consumer only support preview surface format now. |
| if (format != HAL_PIXEL_FORMAT_IMPLEMENTATION_DEFINED) { |
| ALOGE("%s: Deferred consumer only supports IMPLEMENTATION_DEFINED format now!", |
| __FUNCTION__); |
| mState = STATE_ERROR; |
| } |
| |
| // Validation check for the consumer usage flag. |
| if ((consumerUsage & GraphicBuffer::USAGE_HW_TEXTURE) == 0 && |
| (consumerUsage & GraphicBuffer::USAGE_HW_COMPOSER) == 0) { |
| ALOGE("%s: Deferred consumer usage flag is illegal %" PRIu64 "!", |
| __FUNCTION__, consumerUsage); |
| mState = STATE_ERROR; |
| } |
| |
| mConsumerName = "Deferred"; |
| bool needsReleaseNotify = setId > CAMERA3_STREAM_SET_ID_INVALID; |
| mBufferProducerListener = new BufferProducerListener(this, needsReleaseNotify); |
| } |
| |
| Camera3OutputStream::Camera3OutputStream(int id, camera_stream_type_t type, |
| uint32_t width, uint32_t height, |
| int format, |
| android_dataspace dataSpace, |
| camera_stream_rotation_t rotation, |
| const std::string& physicalCameraId, |
| const std::unordered_set<int32_t> &sensorPixelModesUsed, |
| IPCTransport transport, |
| uint64_t consumerUsage, nsecs_t timestampOffset, |
| int setId, bool isMultiResolution, |
| int64_t dynamicRangeProfile, int64_t streamUseCase, |
| bool deviceTimeBaseIsRealtime, int timestampBase, |
| int mirrorMode, int32_t colorSpace, |
| bool useReadoutTimestamp) : |
| Camera3IOStreamBase(id, type, width, height, |
| /*maxSize*/0, |
| format, dataSpace, rotation, |
| physicalCameraId, sensorPixelModesUsed, setId, isMultiResolution, |
| dynamicRangeProfile, streamUseCase, deviceTimeBaseIsRealtime, |
| timestampBase, colorSpace), |
| mTransform(0), |
| mTraceFirstBuffer(true), |
| mUseBufferManager(false), |
| mTimestampOffset(timestampOffset), |
| mUseReadoutTime(useReadoutTimestamp), |
| mConsumerUsage(consumerUsage), |
| mDropBuffers(false), |
| mMirrorMode(mirrorMode), |
| mDequeueBufferLatency(kDequeueLatencyBinSize), |
| mIPCTransport(transport) { |
| |
| bool needsReleaseNotify = setId > CAMERA3_STREAM_SET_ID_INVALID; |
| mBufferProducerListener = new BufferProducerListener(this, needsReleaseNotify); |
| |
| // Subclasses expected to initialize mConsumer themselves |
| } |
| |
| |
| Camera3OutputStream::~Camera3OutputStream() { |
| disconnectLocked(); |
| } |
| |
| status_t Camera3OutputStream::getBufferLocked(camera_stream_buffer *buffer, |
| const std::vector<size_t>&) { |
| ATRACE_HFR_CALL(); |
| |
| ANativeWindowBuffer* anb; |
| int fenceFd = -1; |
| |
| status_t res; |
| res = getBufferLockedCommon(&anb, &fenceFd); |
| if (res != OK) { |
| return res; |
| } |
| |
| /** |
| * FenceFD now owned by HAL except in case of error, |
| * in which case we reassign it to acquire_fence |
| */ |
| handoutBufferLocked(*buffer, &(anb->handle), /*acquireFence*/fenceFd, |
| /*releaseFence*/-1, CAMERA_BUFFER_STATUS_OK, /*output*/true); |
| |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::queueBufferToConsumer(sp<ANativeWindow>& consumer, |
| ANativeWindowBuffer* buffer, int anwReleaseFence, |
| const std::vector<size_t>&) { |
| return consumer->queueBuffer(consumer.get(), buffer, anwReleaseFence); |
| } |
| |
| status_t Camera3OutputStream::returnBufferLocked( |
| const camera_stream_buffer &buffer, |
| nsecs_t timestamp, nsecs_t readoutTimestamp, |
| int32_t transform, const std::vector<size_t>& surface_ids) { |
| ATRACE_HFR_CALL(); |
| |
| if (mHandoutTotalBufferCount == 1) { |
| returnPrefetchedBuffersLocked(); |
| } |
| |
| status_t res = returnAnyBufferLocked(buffer, timestamp, readoutTimestamp, |
| /*output*/true, transform, surface_ids); |
| |
| if (res != OK) { |
| return res; |
| } |
| |
| mLastTimestamp = timestamp; |
| mFrameCount++; |
| |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::fixUpHidlJpegBlobHeader(ANativeWindowBuffer* anwBuffer, int fence) { |
| // Lock the JPEG buffer for CPU read |
| sp<GraphicBuffer> graphicBuffer = GraphicBuffer::from(anwBuffer); |
| void* mapped = nullptr; |
| base::unique_fd fenceFd(dup(fence)); |
| // Use USAGE_SW_WRITE_RARELY since we're going to re-write the CameraBlob |
| // header. |
| GraphicBufferLocker gbLocker(graphicBuffer); |
| status_t res = |
| gbLocker.lockAsync( |
| GraphicBuffer::USAGE_SW_READ_OFTEN | GraphicBuffer::USAGE_SW_WRITE_RARELY, |
| &mapped, fenceFd.release()); |
| if (res != OK) { |
| ALOGE("%s: Failed to lock the buffer: %s (%d)", __FUNCTION__, strerror(-res), res); |
| return res; |
| } |
| |
| uint8_t *hidlHeaderStart = |
| static_cast<uint8_t*>(mapped) + graphicBuffer->getWidth() - sizeof(camera_jpeg_blob_t); |
| // Check that the jpeg buffer is big enough to contain HIDL camera blob |
| if (hidlHeaderStart < static_cast<uint8_t *>(mapped)) { |
| ALOGE("%s, jpeg buffer not large enough to fit HIDL camera blob %" PRIu32, __FUNCTION__, |
| graphicBuffer->getWidth()); |
| return BAD_VALUE; |
| } |
| camera_jpeg_blob_t *hidlBlobHeader = reinterpret_cast<camera_jpeg_blob_t *>(hidlHeaderStart); |
| |
| // Check that the blob is indeed the jpeg blob id. |
| if (hidlBlobHeader->jpeg_blob_id != CAMERA_JPEG_BLOB_ID) { |
| ALOGE("%s, jpeg blob id %d is not correct", __FUNCTION__, hidlBlobHeader->jpeg_blob_id); |
| return BAD_VALUE; |
| } |
| |
| // Retrieve id and blob size |
| CameraBlobId blobId = static_cast<CameraBlobId>(hidlBlobHeader->jpeg_blob_id); |
| uint32_t blobSizeBytes = hidlBlobHeader->jpeg_size; |
| |
| if (blobSizeBytes > (graphicBuffer->getWidth() - sizeof(camera_jpeg_blob_t))) { |
| ALOGE("%s, blobSize in HIDL jpeg blob : %d is corrupt, buffer size %" PRIu32, __FUNCTION__, |
| blobSizeBytes, graphicBuffer->getWidth()); |
| } |
| |
| uint8_t *aidlHeaderStart = |
| static_cast<uint8_t*>(mapped) + graphicBuffer->getWidth() - sizeof(CameraBlob); |
| |
| // Check that the jpeg buffer is big enough to contain AIDL camera blob |
| if (aidlHeaderStart < static_cast<uint8_t *>(mapped)) { |
| ALOGE("%s, jpeg buffer not large enough to fit AIDL camera blob %" PRIu32, __FUNCTION__, |
| graphicBuffer->getWidth()); |
| return BAD_VALUE; |
| } |
| |
| if (static_cast<uint8_t*>(mapped) + blobSizeBytes > aidlHeaderStart) { |
| ALOGE("%s, jpeg blob with size %d , buffer size %" PRIu32 " not large enough to fit" |
| " AIDL camera blob without corrupting jpeg", __FUNCTION__, blobSizeBytes, |
| graphicBuffer->getWidth()); |
| return BAD_VALUE; |
| } |
| |
| // Fill in JPEG header |
| CameraBlob aidlHeader = { |
| .blobId = blobId, |
| .blobSizeBytes = static_cast<int32_t>(blobSizeBytes) |
| }; |
| memcpy(aidlHeaderStart, &aidlHeader, sizeof(CameraBlob)); |
| graphicBuffer->unlock(); |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::returnBufferCheckedLocked( |
| const camera_stream_buffer &buffer, |
| nsecs_t timestamp, |
| nsecs_t readoutTimestamp, |
| [[maybe_unused]] bool output, |
| int32_t transform, |
| const std::vector<size_t>& surface_ids, |
| /*out*/ |
| sp<Fence> *releaseFenceOut) { |
| |
| ALOG_ASSERT(output, "Expected output to be true"); |
| |
| status_t res; |
| |
| // Fence management - always honor release fence from HAL |
| sp<Fence> releaseFence = new Fence(buffer.release_fence); |
| int anwReleaseFence = releaseFence->dup(); |
| |
| /** |
| * Release the lock briefly to avoid deadlock with |
| * StreamingProcessor::startStream -> Camera3Stream::isConfiguring (this |
| * thread will go into StreamingProcessor::onFrameAvailable) during |
| * queueBuffer |
| */ |
| sp<ANativeWindow> currentConsumer = mConsumer; |
| StreamState state = mState; |
| mLock.unlock(); |
| |
| ANativeWindowBuffer *anwBuffer = container_of(buffer.buffer, ANativeWindowBuffer, handle); |
| bool bufferDeferred = false; |
| /** |
| * Return buffer back to ANativeWindow |
| */ |
| if (buffer.status == CAMERA_BUFFER_STATUS_ERROR || mDropBuffers || timestamp == 0) { |
| // Cancel buffer |
| if (mDropBuffers) { |
| ALOGV("%s: Dropping a frame for stream %d.", __FUNCTION__, mId); |
| } else if (buffer.status == CAMERA_BUFFER_STATUS_ERROR) { |
| ALOGV("%s: A frame is dropped for stream %d due to buffer error.", __FUNCTION__, mId); |
| } else { |
| ALOGE("%s: Stream %d: timestamp shouldn't be 0", __FUNCTION__, mId); |
| } |
| |
| res = currentConsumer->cancelBuffer(currentConsumer.get(), |
| anwBuffer, |
| anwReleaseFence); |
| if (shouldLogError(res, state)) { |
| ALOGE("%s: Stream %d: Error cancelling buffer to native window:" |
| " %s (%d)", __FUNCTION__, mId, strerror(-res), res); |
| } |
| |
| notifyBufferReleased(anwBuffer); |
| if (mUseBufferManager) { |
| // Return this buffer back to buffer manager. |
| mBufferProducerListener->onBufferReleased(); |
| } |
| } else { |
| if (mTraceFirstBuffer && (stream_type == CAMERA_STREAM_OUTPUT)) { |
| { |
| char traceLog[48]; |
| snprintf(traceLog, sizeof(traceLog), "Stream %d: first full buffer\n", mId); |
| ATRACE_NAME(traceLog); |
| } |
| mTraceFirstBuffer = false; |
| } |
| // Fix CameraBlob id type discrepancy between HIDL and AIDL, details : http://b/229688810 |
| if (getFormat() == HAL_PIXEL_FORMAT_BLOB && (getDataSpace() == HAL_DATASPACE_V0_JFIF || |
| (getDataSpace() == |
| static_cast<android_dataspace_t>( |
| aidl::android::hardware::graphics::common::Dataspace::JPEG_R)))) { |
| if (mIPCTransport == IPCTransport::HIDL) { |
| fixUpHidlJpegBlobHeader(anwBuffer, anwReleaseFence); |
| } |
| // If this is a JPEG output, and image dump mask is set, save image to |
| // disk. |
| if (mImageDumpMask) { |
| dumpImageToDisk(timestamp, anwBuffer, anwReleaseFence); |
| } |
| } |
| |
| nsecs_t captureTime = ((mUseReadoutTime || mSyncToDisplay) && readoutTimestamp != 0 ? |
| readoutTimestamp : timestamp) - mTimestampOffset; |
| if (mPreviewFrameSpacer != nullptr) { |
| nsecs_t readoutTime = (readoutTimestamp != 0 ? readoutTimestamp : timestamp) |
| - mTimestampOffset; |
| res = mPreviewFrameSpacer->queuePreviewBuffer(captureTime, readoutTime, |
| transform, anwBuffer, anwReleaseFence); |
| if (res != OK) { |
| ALOGE("%s: Stream %d: Error queuing buffer to preview buffer spacer: %s (%d)", |
| __FUNCTION__, mId, strerror(-res), res); |
| return res; |
| } |
| bufferDeferred = true; |
| } else { |
| nsecs_t presentTime = mSyncToDisplay ? |
| syncTimestampToDisplayLocked(captureTime, releaseFence) : captureTime; |
| |
| setTransform(transform, true/*mayChangeMirror*/); |
| res = native_window_set_buffers_timestamp(mConsumer.get(), presentTime); |
| if (res != OK) { |
| ALOGE("%s: Stream %d: Error setting timestamp: %s (%d)", |
| __FUNCTION__, mId, strerror(-res), res); |
| return res; |
| } |
| |
| queueHDRMetadata(anwBuffer->handle, currentConsumer, dynamic_range_profile); |
| |
| res = queueBufferToConsumer(currentConsumer, anwBuffer, anwReleaseFence, surface_ids); |
| if (shouldLogError(res, state)) { |
| ALOGE("%s: Stream %d: Error queueing buffer to native window:" |
| " %s (%d)", __FUNCTION__, mId, strerror(-res), res); |
| } |
| } |
| } |
| mLock.lock(); |
| |
| if (bufferDeferred) { |
| mCachedOutputBufferCount++; |
| } |
| |
| // Once a valid buffer has been returned to the queue, can no longer |
| // dequeue all buffers for preallocation. |
| if (buffer.status != CAMERA_BUFFER_STATUS_ERROR) { |
| mStreamUnpreparable = true; |
| } |
| |
| *releaseFenceOut = releaseFence; |
| |
| return res; |
| } |
| |
| void Camera3OutputStream::dump(int fd, [[maybe_unused]] const Vector<String16> &args) const { |
| std::string lines; |
| lines += fmt::sprintf(" Stream[%d]: Output\n", mId); |
| lines += fmt::sprintf(" Consumer name: %s\n", mConsumerName); |
| write(fd, lines.c_str(), lines.size()); |
| |
| Camera3IOStreamBase::dump(fd, args); |
| |
| mDequeueBufferLatency.dump(fd, |
| " DequeueBuffer latency histogram:"); |
| } |
| |
| status_t Camera3OutputStream::setTransform(int transform, bool mayChangeMirror) { |
| ATRACE_CALL(); |
| Mutex::Autolock l(mLock); |
| if (mMirrorMode != OutputConfiguration::MIRROR_MODE_AUTO && mayChangeMirror) { |
| // If the mirroring mode is not AUTO, do not allow transform update |
| // which may change mirror. |
| return OK; |
| } |
| |
| return setTransformLocked(transform); |
| } |
| |
| status_t Camera3OutputStream::setTransformLocked(int transform) { |
| status_t res = OK; |
| |
| if (transform == -1) return res; |
| |
| if (mState == STATE_ERROR) { |
| ALOGE("%s: Stream in error state", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| |
| mTransform = transform; |
| if (mState == STATE_CONFIGURED) { |
| res = native_window_set_buffers_transform(mConsumer.get(), |
| transform); |
| if (res != OK) { |
| ALOGE("%s: Unable to configure stream transform to %x: %s (%d)", |
| __FUNCTION__, transform, strerror(-res), res); |
| } |
| } |
| return res; |
| } |
| |
| status_t Camera3OutputStream::configureQueueLocked() { |
| status_t res; |
| |
| mTraceFirstBuffer = true; |
| if ((res = Camera3IOStreamBase::configureQueueLocked()) != OK) { |
| return res; |
| } |
| |
| if ((res = configureConsumerQueueLocked(true /*allowPreviewRespace*/)) != OK) { |
| return res; |
| } |
| |
| // Set dequeueBuffer/attachBuffer timeout if the consumer is not hw composer or hw texture. |
| // We need skip these cases as timeout will disable the non-blocking (async) mode. |
| if (!(isConsumedByHWComposer() || isConsumedByHWTexture())) { |
| if (mUseBufferManager) { |
| // When buffer manager is handling the buffer, we should have available buffers in |
| // buffer queue before we calls into dequeueBuffer because buffer manager is tracking |
| // free buffers. |
| // There are however some consumer side feature (ImageReader::discardFreeBuffers) that |
| // can discard free buffers without notifying buffer manager. We want the timeout to |
| // happen immediately here so buffer manager can try to update its internal state and |
| // try to allocate a buffer instead of waiting. |
| mConsumer->setDequeueTimeout(0); |
| } else { |
| mConsumer->setDequeueTimeout(kDequeueBufferTimeout); |
| } |
| } |
| |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::configureConsumerQueueLocked(bool allowPreviewRespace) { |
| status_t res; |
| |
| mTraceFirstBuffer = true; |
| |
| ALOG_ASSERT(mConsumer != 0, "mConsumer should never be NULL"); |
| |
| // Configure consumer-side ANativeWindow interface. The listener may be used |
| // to notify buffer manager (if it is used) of the returned buffers. |
| res = mConsumer->connect(NATIVE_WINDOW_API_CAMERA, |
| /*reportBufferRemoval*/true, |
| /*listener*/mBufferProducerListener); |
| if (res != OK) { |
| ALOGE("%s: Unable to connect to native window for stream %d", |
| __FUNCTION__, mId); |
| return res; |
| } |
| |
| mConsumerName = mConsumer->getConsumerName(); |
| |
| res = native_window_set_usage(mConsumer.get(), mUsage); |
| if (res != OK) { |
| ALOGE("%s: Unable to configure usage %" PRIu64 " for stream %d", |
| __FUNCTION__, mUsage, mId); |
| return res; |
| } |
| |
| res = native_window_set_scaling_mode(mConsumer.get(), |
| NATIVE_WINDOW_SCALING_MODE_SCALE_TO_WINDOW); |
| if (res != OK) { |
| ALOGE("%s: Unable to configure stream scaling: %s (%d)", |
| __FUNCTION__, strerror(-res), res); |
| return res; |
| } |
| |
| if (mMaxSize == 0) { |
| // For buffers of known size |
| res = native_window_set_buffers_dimensions(mConsumer.get(), |
| camera_stream::width, camera_stream::height); |
| } else { |
| // For buffers with bounded size |
| res = native_window_set_buffers_dimensions(mConsumer.get(), |
| mMaxSize, 1); |
| } |
| if (res != OK) { |
| ALOGE("%s: Unable to configure stream buffer dimensions" |
| " %d x %d (maxSize %zu) for stream %d", |
| __FUNCTION__, camera_stream::width, camera_stream::height, |
| mMaxSize, mId); |
| return res; |
| } |
| res = native_window_set_buffers_format(mConsumer.get(), |
| camera_stream::format); |
| if (res != OK) { |
| ALOGE("%s: Unable to configure stream buffer format %#x for stream %d", |
| __FUNCTION__, camera_stream::format, mId); |
| return res; |
| } |
| |
| res = native_window_set_buffers_data_space(mConsumer.get(), |
| camera_stream::data_space); |
| if (res != OK) { |
| ALOGE("%s: Unable to configure stream dataspace %#x for stream %d", |
| __FUNCTION__, camera_stream::data_space, mId); |
| return res; |
| } |
| |
| int maxConsumerBuffers; |
| res = static_cast<ANativeWindow*>(mConsumer.get())->query( |
| mConsumer.get(), |
| NATIVE_WINDOW_MIN_UNDEQUEUED_BUFFERS, &maxConsumerBuffers); |
| if (res != OK) { |
| ALOGE("%s: Unable to query consumer undequeued" |
| " buffer count for stream %d", __FUNCTION__, mId); |
| return res; |
| } |
| |
| ALOGV("%s: Consumer wants %d buffers, HAL wants %d", __FUNCTION__, |
| maxConsumerBuffers, camera_stream::max_buffers); |
| if (camera_stream::max_buffers == 0) { |
| ALOGE("%s: Camera HAL requested max_buffer count: %d, requires at least 1", |
| __FUNCTION__, camera_stream::max_buffers); |
| return INVALID_OPERATION; |
| } |
| |
| mTotalBufferCount = maxConsumerBuffers + camera_stream::max_buffers; |
| |
| int timestampBase = getTimestampBase(); |
| bool isDefaultTimeBase = (timestampBase == |
| OutputConfiguration::TIMESTAMP_BASE_DEFAULT); |
| if (allowPreviewRespace) { |
| bool forceChoreographer = (timestampBase == |
| OutputConfiguration::TIMESTAMP_BASE_CHOREOGRAPHER_SYNCED); |
| bool defaultToChoreographer = (isDefaultTimeBase && |
| isConsumedByHWComposer()); |
| bool defaultToSpacer = (isDefaultTimeBase && |
| isConsumedByHWTexture() && |
| !isConsumedByCPU() && |
| !isVideoStream()); |
| if (forceChoreographer || defaultToChoreographer) { |
| mSyncToDisplay = true; |
| // For choreographer synced stream, extra buffers aren't kept by |
| // camera service. So no need to update mMaxCachedBufferCount. |
| mTotalBufferCount += kDisplaySyncExtraBuffer; |
| } else if (defaultToSpacer) { |
| mPreviewFrameSpacer = new PreviewFrameSpacer(this, mConsumer); |
| // For preview frame spacer, the extra buffer is kept by camera |
| // service. So update mMaxCachedBufferCount. |
| mMaxCachedBufferCount = 1; |
| mTotalBufferCount += mMaxCachedBufferCount; |
| res = mPreviewFrameSpacer->run((std::string("PreviewSpacer-") |
| + std::to_string(mId)).c_str()); |
| if (res != OK) { |
| ALOGE("%s: Unable to start preview spacer: %s (%d)", __FUNCTION__, |
| strerror(-res), res); |
| return res; |
| } |
| } |
| } |
| mHandoutTotalBufferCount = 0; |
| mFrameCount = 0; |
| mLastTimestamp = 0; |
| |
| if (isDeviceTimeBaseRealtime()) { |
| if (isDefaultTimeBase && !isConsumedByHWComposer() && !isVideoStream()) { |
| // Default time base, but not hardware composer or video encoder |
| mTimestampOffset = 0; |
| } else if (timestampBase == OutputConfiguration::TIMESTAMP_BASE_REALTIME || |
| timestampBase == OutputConfiguration::TIMESTAMP_BASE_SENSOR) { |
| mTimestampOffset = 0; |
| } |
| // If timestampBase is CHOREOGRAPHER SYNCED or MONOTONIC, leave |
| // timestamp offset as bootTime - monotonicTime. |
| } else { |
| if (timestampBase == OutputConfiguration::TIMESTAMP_BASE_REALTIME) { |
| // Reverse offset for monotonicTime -> bootTime |
| mTimestampOffset = -mTimestampOffset; |
| } else { |
| // If timestampBase is DEFAULT, MONOTONIC, SENSOR or |
| // CHOREOGRAPHER_SYNCED, timestamp offset is 0. |
| mTimestampOffset = 0; |
| } |
| } |
| |
| res = native_window_set_buffer_count(mConsumer.get(), |
| mTotalBufferCount); |
| if (res != OK) { |
| ALOGE("%s: Unable to set buffer count for stream %d", |
| __FUNCTION__, mId); |
| return res; |
| } |
| |
| res = native_window_set_buffers_transform(mConsumer.get(), |
| mTransform); |
| if (res != OK) { |
| ALOGE("%s: Unable to configure stream transform to %x: %s (%d)", |
| __FUNCTION__, mTransform, strerror(-res), res); |
| return res; |
| } |
| |
| /** |
| * Camera3 Buffer manager is only supported by HAL3.3 onwards, as the older HALs requires |
| * buffers to be statically allocated for internal static buffer registration, while the |
| * buffers provided by buffer manager are really dynamically allocated. Camera3Device only |
| * sets the mBufferManager if device version is > HAL3.2, which guarantees that the buffer |
| * manager setup is skipped in below code. Note that HAL3.2 is also excluded here, as some |
| * HAL3.2 devices may not support the dynamic buffer registeration. |
| * Also Camera3BufferManager does not support display/texture streams as they have its own |
| * buffer management logic. |
| */ |
| if (mBufferManager != 0 && mSetId > CAMERA3_STREAM_SET_ID_INVALID && |
| !(isConsumedByHWComposer() || isConsumedByHWTexture())) { |
| uint64_t consumerUsage = 0; |
| getEndpointUsage(&consumerUsage); |
| uint32_t width = (mMaxSize == 0) ? getWidth() : mMaxSize; |
| uint32_t height = (mMaxSize == 0) ? getHeight() : 1; |
| StreamInfo streamInfo( |
| getId(), getStreamSetId(), width, height, getFormat(), getDataSpace(), |
| mUsage | consumerUsage, mTotalBufferCount, |
| /*isConfigured*/true, isMultiResolution()); |
| wp<Camera3OutputStream> weakThis(this); |
| res = mBufferManager->registerStream(weakThis, |
| streamInfo); |
| if (res == OK) { |
| // Disable buffer allocation for this BufferQueue, buffer manager will take over |
| // the buffer allocation responsibility. |
| mConsumer->getIGraphicBufferProducer()->allowAllocation(false); |
| mUseBufferManager = true; |
| } else { |
| ALOGE("%s: Unable to register stream %d to camera3 buffer manager, " |
| "(error %d %s), fall back to BufferQueue for buffer management!", |
| __FUNCTION__, mId, res, strerror(-res)); |
| } |
| } |
| |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::getBufferLockedCommon(ANativeWindowBuffer** anb, int* fenceFd) { |
| ATRACE_HFR_CALL(); |
| status_t res; |
| |
| if ((res = getBufferPreconditionCheckLocked()) != OK) { |
| return res; |
| } |
| |
| bool gotBufferFromManager = false; |
| |
| if (mUseBufferManager) { |
| sp<GraphicBuffer> gb; |
| res = mBufferManager->getBufferForStream(getId(), getStreamSetId(), |
| isMultiResolution(), &gb, fenceFd); |
| if (res == OK) { |
| // Attach this buffer to the bufferQueue: the buffer will be in dequeue state after a |
| // successful return. |
| *anb = gb.get(); |
| res = mConsumer->attachBuffer(*anb); |
| if (shouldLogError(res, mState)) { |
| ALOGE("%s: Stream %d: Can't attach the output buffer to this surface: %s (%d)", |
| __FUNCTION__, mId, strerror(-res), res); |
| } |
| if (res != OK) { |
| checkRetAndSetAbandonedLocked(res); |
| return res; |
| } |
| gotBufferFromManager = true; |
| ALOGV("Stream %d: Attached new buffer", getId()); |
| } else if (res == ALREADY_EXISTS) { |
| // Have sufficient free buffers already attached, can just |
| // dequeue from buffer queue |
| ALOGV("Stream %d: Reusing attached buffer", getId()); |
| gotBufferFromManager = false; |
| } else if (res != OK) { |
| ALOGE("%s: Stream %d: Can't get next output buffer from buffer manager: %s (%d)", |
| __FUNCTION__, mId, strerror(-res), res); |
| return res; |
| } |
| } |
| if (!gotBufferFromManager) { |
| /** |
| * Release the lock briefly to avoid deadlock for below scenario: |
| * Thread 1: StreamingProcessor::startStream -> Camera3Stream::isConfiguring(). |
| * This thread acquired StreamingProcessor lock and try to lock Camera3Stream lock. |
| * Thread 2: Camera3Stream::returnBuffer->StreamingProcessor::onFrameAvailable(). |
| * This thread acquired Camera3Stream lock and bufferQueue lock, and try to lock |
| * StreamingProcessor lock. |
| * Thread 3: Camera3Stream::getBuffer(). This thread acquired Camera3Stream lock |
| * and try to lock bufferQueue lock. |
| * Then there is circular locking dependency. |
| */ |
| sp<Surface> consumer = mConsumer; |
| size_t remainingBuffers = (mState == STATE_PREPARING ? mTotalBufferCount : |
| camera_stream::max_buffers) - mHandoutTotalBufferCount; |
| mLock.unlock(); |
| |
| nsecs_t dequeueStart = systemTime(SYSTEM_TIME_MONOTONIC); |
| |
| size_t batchSize = mBatchSize.load(); |
| if (batchSize == 1) { |
| sp<ANativeWindow> anw = consumer; |
| res = anw->dequeueBuffer(anw.get(), anb, fenceFd); |
| } else { |
| std::unique_lock<std::mutex> batchLock(mBatchLock); |
| res = OK; |
| if (mBatchedBuffers.size() == 0) { |
| if (remainingBuffers == 0) { |
| ALOGE("%s: cannot get buffer while all buffers are handed out", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| if (batchSize > remainingBuffers) { |
| batchSize = remainingBuffers; |
| } |
| batchLock.unlock(); |
| // Refill batched buffers |
| std::vector<Surface::BatchBuffer> batchedBuffers; |
| batchedBuffers.resize(batchSize); |
| res = consumer->dequeueBuffers(&batchedBuffers); |
| batchLock.lock(); |
| if (res != OK) { |
| ALOGE("%s: batch dequeueBuffers call failed! %s (%d)", |
| __FUNCTION__, strerror(-res), res); |
| } else { |
| mBatchedBuffers = std::move(batchedBuffers); |
| } |
| } |
| |
| if (res == OK) { |
| // Dispatch batch buffers |
| *anb = mBatchedBuffers.back().buffer; |
| *fenceFd = mBatchedBuffers.back().fenceFd; |
| mBatchedBuffers.pop_back(); |
| } |
| } |
| |
| nsecs_t dequeueEnd = systemTime(SYSTEM_TIME_MONOTONIC); |
| mDequeueBufferLatency.add(dequeueStart, dequeueEnd); |
| |
| mLock.lock(); |
| |
| if (mUseBufferManager && res == TIMED_OUT) { |
| checkRemovedBuffersLocked(); |
| |
| sp<GraphicBuffer> gb; |
| res = mBufferManager->getBufferForStream( |
| getId(), getStreamSetId(), isMultiResolution(), |
| &gb, fenceFd, /*noFreeBuffer*/true); |
| |
| if (res == OK) { |
| // Attach this buffer to the bufferQueue: the buffer will be in dequeue state after |
| // a successful return. |
| *anb = gb.get(); |
| res = mConsumer->attachBuffer(*anb); |
| gotBufferFromManager = true; |
| ALOGV("Stream %d: Attached new buffer", getId()); |
| |
| if (res != OK) { |
| if (shouldLogError(res, mState)) { |
| ALOGE("%s: Stream %d: Can't attach the output buffer to this surface:" |
| " %s (%d)", __FUNCTION__, mId, strerror(-res), res); |
| } |
| checkRetAndSetAbandonedLocked(res); |
| return res; |
| } |
| } else { |
| ALOGE("%s: Stream %d: Can't get next output buffer from buffer manager:" |
| " %s (%d)", __FUNCTION__, mId, strerror(-res), res); |
| return res; |
| } |
| } else if (res != OK) { |
| if (shouldLogError(res, mState)) { |
| ALOGE("%s: Stream %d: Can't dequeue next output buffer: %s (%d)", |
| __FUNCTION__, mId, strerror(-res), res); |
| } |
| checkRetAndSetAbandonedLocked(res); |
| return res; |
| } |
| } |
| |
| if (res == OK) { |
| checkRemovedBuffersLocked(); |
| } |
| |
| return res; |
| } |
| |
| void Camera3OutputStream::checkRemovedBuffersLocked(bool notifyBufferManager) { |
| std::vector<sp<GraphicBuffer>> removedBuffers; |
| status_t res = mConsumer->getAndFlushRemovedBuffers(&removedBuffers); |
| if (res == OK) { |
| onBuffersRemovedLocked(removedBuffers); |
| |
| if (notifyBufferManager && mUseBufferManager && removedBuffers.size() > 0) { |
| mBufferManager->onBuffersRemoved(getId(), getStreamSetId(), isMultiResolution(), |
| removedBuffers.size()); |
| } |
| } |
| } |
| |
| void Camera3OutputStream::checkRetAndSetAbandonedLocked(status_t res) { |
| // Only transition to STATE_ABANDONED from STATE_CONFIGURED. (If it is |
| // STATE_PREPARING, let prepareNextBuffer handle the error.) |
| if ((res == NO_INIT || res == DEAD_OBJECT) && mState == STATE_CONFIGURED) { |
| mState = STATE_ABANDONED; |
| } |
| } |
| |
| bool Camera3OutputStream::shouldLogError(status_t res, StreamState state) { |
| if (res == OK) { |
| return false; |
| } |
| if ((res == DEAD_OBJECT || res == NO_INIT) && state == STATE_ABANDONED) { |
| return false; |
| } |
| return true; |
| } |
| |
| void Camera3OutputStream::onCachedBufferQueued() { |
| Mutex::Autolock l(mLock); |
| mCachedOutputBufferCount--; |
| // Signal whoever is waiting for the buffer to be returned to the buffer |
| // queue. |
| mOutputBufferReturnedSignal.signal(); |
| } |
| |
| status_t Camera3OutputStream::disconnectLocked() { |
| status_t res; |
| |
| if ((res = Camera3IOStreamBase::disconnectLocked()) != OK) { |
| return res; |
| } |
| |
| // Stream configuration was not finished (can only be in STATE_IN_CONFIG or STATE_CONSTRUCTED |
| // state), don't need change the stream state, return OK. |
| if (mConsumer == nullptr) { |
| return OK; |
| } |
| |
| returnPrefetchedBuffersLocked(); |
| |
| if (mPreviewFrameSpacer != nullptr) { |
| mPreviewFrameSpacer->requestExit(); |
| } |
| |
| ALOGV("%s: disconnecting stream %d from native window", __FUNCTION__, getId()); |
| |
| res = native_window_api_disconnect(mConsumer.get(), |
| NATIVE_WINDOW_API_CAMERA); |
| /** |
| * This is not an error. if client calling process dies, the window will |
| * also die and all calls to it will return DEAD_OBJECT, thus it's already |
| * "disconnected" |
| */ |
| if (res == DEAD_OBJECT) { |
| ALOGW("%s: While disconnecting stream %d from native window, the" |
| " native window died from under us", __FUNCTION__, mId); |
| } |
| else if (res != OK) { |
| ALOGE("%s: Unable to disconnect stream %d from native window " |
| "(error %d %s)", |
| __FUNCTION__, mId, res, strerror(-res)); |
| mState = STATE_ERROR; |
| return res; |
| } |
| |
| // Since device is already idle, there is no getBuffer call to buffer manager, unregister the |
| // stream at this point should be safe. |
| if (mUseBufferManager) { |
| res = mBufferManager->unregisterStream(getId(), getStreamSetId(), isMultiResolution()); |
| if (res != OK) { |
| ALOGE("%s: Unable to unregister stream %d from buffer manager " |
| "(error %d %s)", __FUNCTION__, mId, res, strerror(-res)); |
| mState = STATE_ERROR; |
| return res; |
| } |
| // Note that, to make prepare/teardown case work, we must not mBufferManager.clear(), as |
| // the stream is still in usable state after this call. |
| mUseBufferManager = false; |
| } |
| |
| mState = (mState == STATE_IN_RECONFIG) ? STATE_IN_CONFIG |
| : STATE_CONSTRUCTED; |
| |
| mDequeueBufferLatency.log("Stream %d dequeueBuffer latency histogram", mId); |
| mDequeueBufferLatency.reset(); |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::getEndpointUsage(uint64_t *usage) const { |
| |
| status_t res; |
| |
| if (mConsumer == nullptr) { |
| // mConsumerUsage was sanitized before the Camera3OutputStream was constructed. |
| *usage = mConsumerUsage; |
| return OK; |
| } |
| |
| res = getEndpointUsageForSurface(usage, mConsumer); |
| |
| return res; |
| } |
| |
| void Camera3OutputStream::applyZSLUsageQuirk(int format, uint64_t *consumerUsage /*inout*/) { |
| if (consumerUsage == nullptr) { |
| return; |
| } |
| |
| // If an opaque output stream's endpoint is ImageReader, add |
| // GRALLOC_USAGE_HW_CAMERA_ZSL to the usage so HAL knows it will be used |
| // for the ZSL use case. |
| // Assume it's for ImageReader if the consumer usage doesn't have any of these bits set: |
| // 1. GRALLOC_USAGE_HW_TEXTURE |
| // 2. GRALLOC_USAGE_HW_RENDER |
| // 3. GRALLOC_USAGE_HW_COMPOSER |
| // 4. GRALLOC_USAGE_HW_VIDEO_ENCODER |
| if (format == HAL_PIXEL_FORMAT_IMPLEMENTATION_DEFINED && |
| (*consumerUsage & (GRALLOC_USAGE_HW_TEXTURE | GRALLOC_USAGE_HW_RENDER | |
| GRALLOC_USAGE_HW_COMPOSER | GRALLOC_USAGE_HW_VIDEO_ENCODER)) == 0) { |
| *consumerUsage |= GRALLOC_USAGE_HW_CAMERA_ZSL; |
| } |
| } |
| |
| status_t Camera3OutputStream::getEndpointUsageForSurface(uint64_t *usage, |
| const sp<Surface>& surface) const { |
| status_t res; |
| uint64_t u = 0; |
| |
| res = native_window_get_consumer_usage(static_cast<ANativeWindow*>(surface.get()), &u); |
| applyZSLUsageQuirk(camera_stream::format, &u); |
| *usage = u; |
| return res; |
| } |
| |
| bool Camera3OutputStream::isVideoStream() const { |
| uint64_t usage = 0; |
| status_t res = getEndpointUsage(&usage); |
| if (res != OK) { |
| ALOGE("%s: getting end point usage failed: %s (%d).", __FUNCTION__, strerror(-res), res); |
| return false; |
| } |
| |
| return (usage & GRALLOC_USAGE_HW_VIDEO_ENCODER) != 0; |
| } |
| |
| status_t Camera3OutputStream::setBufferManager(sp<Camera3BufferManager> bufferManager) { |
| Mutex::Autolock l(mLock); |
| if (mState != STATE_CONSTRUCTED) { |
| ALOGE("%s: this method can only be called when stream in CONSTRUCTED state.", |
| __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| mBufferManager = bufferManager; |
| |
| return OK; |
| } |
| |
| status_t Camera3OutputStream::updateStream(const std::vector<sp<Surface>> &/*outputSurfaces*/, |
| const std::vector<OutputStreamInfo> &/*outputInfo*/, |
| const std::vector<size_t> &/*removedSurfaceIds*/, |
| KeyedVector<sp<Surface>, size_t> * /*outputMapo*/) { |
| ALOGE("%s: this method is not supported!", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| |
| void Camera3OutputStream::BufferProducerListener::onBufferReleased() { |
| sp<Camera3OutputStream> stream = mParent.promote(); |
| if (stream == nullptr) { |
| ALOGV("%s: Parent camera3 output stream was destroyed", __FUNCTION__); |
| return; |
| } |
| |
| Mutex::Autolock l(stream->mLock); |
| if (!(stream->mUseBufferManager)) { |
| return; |
| } |
| |
| ALOGV("Stream %d: Buffer released", stream->getId()); |
| bool shouldFreeBuffer = false; |
| status_t res = stream->mBufferManager->onBufferReleased( |
| stream->getId(), stream->getStreamSetId(), stream->isMultiResolution(), |
| &shouldFreeBuffer); |
| if (res != OK) { |
| ALOGE("%s: signaling buffer release to buffer manager failed: %s (%d).", __FUNCTION__, |
| strerror(-res), res); |
| stream->mState = STATE_ERROR; |
| } |
| |
| if (shouldFreeBuffer) { |
| sp<GraphicBuffer> buffer; |
| // Detach and free a buffer (when buffer goes out of scope) |
| stream->detachBufferLocked(&buffer, /*fenceFd*/ nullptr); |
| if (buffer.get() != nullptr) { |
| stream->mBufferManager->notifyBufferRemoved( |
| stream->getId(), stream->getStreamSetId(), stream->isMultiResolution()); |
| } |
| } |
| } |
| |
| void Camera3OutputStream::BufferProducerListener::onBuffersDiscarded( |
| const std::vector<sp<GraphicBuffer>>& buffers) { |
| sp<Camera3OutputStream> stream = mParent.promote(); |
| if (stream == nullptr) { |
| ALOGV("%s: Parent camera3 output stream was destroyed", __FUNCTION__); |
| return; |
| } |
| |
| if (buffers.size() > 0) { |
| Mutex::Autolock l(stream->mLock); |
| stream->onBuffersRemovedLocked(buffers); |
| if (stream->mUseBufferManager) { |
| stream->mBufferManager->onBuffersRemoved(stream->getId(), |
| stream->getStreamSetId(), stream->isMultiResolution(), buffers.size()); |
| } |
| ALOGV("Stream %d: %zu Buffers discarded.", stream->getId(), buffers.size()); |
| } |
| } |
| |
| void Camera3OutputStream::onBuffersRemovedLocked( |
| const std::vector<sp<GraphicBuffer>>& removedBuffers) { |
| sp<Camera3StreamBufferFreedListener> callback = mBufferFreedListener.promote(); |
| if (callback != nullptr) { |
| for (const auto& gb : removedBuffers) { |
| callback->onBufferFreed(mId, gb->handle); |
| } |
| } |
| } |
| |
| status_t Camera3OutputStream::detachBuffer(sp<GraphicBuffer>* buffer, int* fenceFd) { |
| Mutex::Autolock l(mLock); |
| return detachBufferLocked(buffer, fenceFd); |
| } |
| |
| status_t Camera3OutputStream::detachBufferLocked(sp<GraphicBuffer>* buffer, int* fenceFd) { |
| ALOGV("Stream %d: detachBuffer", getId()); |
| if (buffer == nullptr) { |
| return BAD_VALUE; |
| } |
| |
| sp<Fence> fence; |
| status_t res = mConsumer->detachNextBuffer(buffer, &fence); |
| if (res == NO_MEMORY) { |
| // This may rarely happen, which indicates that the released buffer was freed by other |
| // call (e.g., attachBuffer, dequeueBuffer etc.) before reaching here. We should notify the |
| // buffer manager that this buffer has been freed. It's not fatal, but should be avoided, |
| // therefore log a warning. |
| *buffer = 0; |
| ALOGW("%s: the released buffer has already been freed by the buffer queue!", __FUNCTION__); |
| } else if (res != OK) { |
| // Treat other errors as abandonment |
| if (shouldLogError(res, mState)) { |
| ALOGE("%s: detach next buffer failed: %s (%d).", __FUNCTION__, strerror(-res), res); |
| } |
| mState = STATE_ABANDONED; |
| return res; |
| } |
| |
| if (fenceFd != nullptr) { |
| if (fence!= 0 && fence->isValid()) { |
| *fenceFd = fence->dup(); |
| } else { |
| *fenceFd = -1; |
| } |
| } |
| |
| // Here we assume detachBuffer is called by buffer manager so it doesn't need to be notified |
| checkRemovedBuffersLocked(/*notifyBufferManager*/false); |
| return res; |
| } |
| |
| status_t Camera3OutputStream::dropBuffers(bool dropping) { |
| Mutex::Autolock l(mLock); |
| mDropBuffers = dropping; |
| return OK; |
| } |
| |
| const std::string& Camera3OutputStream::getPhysicalCameraId() const { |
| Mutex::Autolock l(mLock); |
| return physicalCameraId(); |
| } |
| |
| status_t Camera3OutputStream::notifyBufferReleased(ANativeWindowBuffer* /*anwBuffer*/) { |
| return OK; |
| } |
| |
| bool Camera3OutputStream::isConsumerConfigurationDeferred(size_t surface_id) const { |
| Mutex::Autolock l(mLock); |
| |
| if (surface_id != 0) { |
| ALOGE("%s: surface_id %zu for Camera3OutputStream should be 0!", __FUNCTION__, surface_id); |
| } |
| return mConsumer == nullptr; |
| } |
| |
| status_t Camera3OutputStream::setConsumers(const std::vector<sp<Surface>>& consumers) { |
| Mutex::Autolock l(mLock); |
| if (consumers.size() != 1) { |
| ALOGE("%s: it's illegal to set %zu consumer surfaces!", |
| __FUNCTION__, consumers.size()); |
| return INVALID_OPERATION; |
| } |
| if (consumers[0] == nullptr) { |
| ALOGE("%s: it's illegal to set null consumer surface!", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| |
| if (mConsumer != nullptr) { |
| ALOGE("%s: consumer surface was already set!", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| |
| mConsumer = consumers[0]; |
| return OK; |
| } |
| |
| bool Camera3OutputStream::isConsumedByHWComposer() const { |
| uint64_t usage = 0; |
| status_t res = getEndpointUsage(&usage); |
| if (res != OK) { |
| ALOGE("%s: getting end point usage failed: %s (%d).", __FUNCTION__, strerror(-res), res); |
| return false; |
| } |
| |
| return (usage & GRALLOC_USAGE_HW_COMPOSER) != 0; |
| } |
| |
| bool Camera3OutputStream::isConsumedByHWTexture() const { |
| uint64_t usage = 0; |
| status_t res = getEndpointUsage(&usage); |
| if (res != OK) { |
| ALOGE("%s: getting end point usage failed: %s (%d).", __FUNCTION__, strerror(-res), res); |
| return false; |
| } |
| |
| return (usage & GRALLOC_USAGE_HW_TEXTURE) != 0; |
| } |
| |
| bool Camera3OutputStream::isConsumedByCPU() const { |
| uint64_t usage = 0; |
| status_t res = getEndpointUsage(&usage); |
| if (res != OK) { |
| ALOGE("%s: getting end point usage failed: %s (%d).", __FUNCTION__, strerror(-res), res); |
| return false; |
| } |
| |
| return (usage & GRALLOC_USAGE_SW_READ_MASK) != 0; |
| } |
| |
| void Camera3OutputStream::dumpImageToDisk(nsecs_t timestamp, |
| ANativeWindowBuffer* anwBuffer, int fence) { |
| // Deriver output file name |
| std::string fileExtension = "jpg"; |
| char imageFileName[64]; |
| time_t now = time(0); |
| tm *localTime = localtime(&now); |
| snprintf(imageFileName, sizeof(imageFileName), "IMG_%4d%02d%02d_%02d%02d%02d_%" PRId64 ".%s", |
| 1900 + localTime->tm_year, localTime->tm_mon + 1, localTime->tm_mday, |
| localTime->tm_hour, localTime->tm_min, localTime->tm_sec, |
| timestamp, fileExtension.c_str()); |
| |
| // Lock the image for CPU read |
| sp<GraphicBuffer> graphicBuffer = GraphicBuffer::from(anwBuffer); |
| void* mapped = nullptr; |
| base::unique_fd fenceFd(dup(fence)); |
| status_t res = graphicBuffer->lockAsync(GraphicBuffer::USAGE_SW_READ_OFTEN, &mapped, |
| fenceFd.release()); |
| if (res != OK) { |
| ALOGE("%s: Failed to lock the buffer: %s (%d)", __FUNCTION__, strerror(-res), res); |
| return; |
| } |
| |
| // Figure out actual file size |
| auto actualJpegSize = android::camera2::JpegProcessor::findJpegSize((uint8_t*)mapped, mMaxSize); |
| if (actualJpegSize == 0) { |
| actualJpegSize = mMaxSize; |
| } |
| |
| // Output image data to file |
| std::string filePath = "/data/misc/cameraserver/"; |
| filePath += imageFileName; |
| std::ofstream imageFile(filePath, std::ofstream::binary); |
| if (!imageFile.is_open()) { |
| ALOGE("%s: Unable to create file %s", __FUNCTION__, filePath.c_str()); |
| graphicBuffer->unlock(); |
| return; |
| } |
| imageFile.write((const char*)mapped, actualJpegSize); |
| |
| graphicBuffer->unlock(); |
| } |
| |
| status_t Camera3OutputStream::setBatchSize(size_t batchSize) { |
| Mutex::Autolock l(mLock); |
| if (batchSize == 0) { |
| ALOGE("%s: invalid batch size 0", __FUNCTION__); |
| return BAD_VALUE; |
| } |
| |
| if (mUseBufferManager) { |
| ALOGE("%s: batch operation is not supported with buffer manager", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| |
| if (!isVideoStream()) { |
| ALOGE("%s: batch operation is not supported with non-video stream", __FUNCTION__); |
| return INVALID_OPERATION; |
| } |
| |
| if (camera_stream::max_buffers < batchSize) { |
| ALOGW("%s: batch size is capped by max_buffers %d", __FUNCTION__, |
| camera_stream::max_buffers); |
| batchSize = camera_stream::max_buffers; |
| } |
| |
| size_t defaultBatchSize = 1; |
| if (!mBatchSize.compare_exchange_strong(defaultBatchSize, batchSize)) { |
| ALOGE("%s: change batch size from %zu to %zu dynamically is not supported", |
| __FUNCTION__, defaultBatchSize, batchSize); |
| return INVALID_OPERATION; |
| } |
| |
| return OK; |
| } |
| |
| void Camera3OutputStream::onMinDurationChanged(nsecs_t duration, bool fixedFps) { |
| Mutex::Autolock l(mLock); |
| mMinExpectedDuration = duration; |
| mFixedFps = fixedFps; |
| } |
| |
| void Camera3OutputStream::setStreamUseCase(int64_t streamUseCase) { |
| Mutex::Autolock l(mLock); |
| camera_stream::use_case = streamUseCase; |
| } |
| |
| void Camera3OutputStream::returnPrefetchedBuffersLocked() { |
| std::vector<Surface::BatchBuffer> batchedBuffers; |
| |
| { |
| std::lock_guard<std::mutex> batchLock(mBatchLock); |
| if (mBatchedBuffers.size() != 0) { |
| ALOGW("%s: %zu extra prefetched buffers detected. Returning", |
| __FUNCTION__, mBatchedBuffers.size()); |
| batchedBuffers = std::move(mBatchedBuffers); |
| } |
| } |
| |
| if (batchedBuffers.size() > 0) { |
| mConsumer->cancelBuffers(batchedBuffers); |
| } |
| } |
| |
| nsecs_t Camera3OutputStream::syncTimestampToDisplayLocked(nsecs_t t, sp<Fence> releaseFence) { |
| nsecs_t currentTime = systemTime(); |
| if (!mFixedFps) { |
| mLastCaptureTime = t; |
| mLastPresentTime = currentTime; |
| return t; |
| } |
| |
| ParcelableVsyncEventData parcelableVsyncEventData; |
| auto res = mDisplayEventReceiver.getLatestVsyncEventData(&parcelableVsyncEventData); |
| if (res != OK) { |
| ALOGE("%s: Stream %d: Error getting latest vsync event data: %s (%d)", |
| __FUNCTION__, mId, strerror(-res), res); |
| mLastCaptureTime = t; |
| mLastPresentTime = currentTime; |
| return t; |
| } |
| |
| const VsyncEventData& vsyncEventData = parcelableVsyncEventData.vsync; |
| nsecs_t minPresentT = mLastPresentTime + vsyncEventData.frameInterval / 2; |
| |
| // Find the best presentation time without worrying about previous frame's |
| // presentation time if capture interval is more than kSpacingResetIntervalNs. |
| // |
| // When frame interval is more than 50 ms apart (3 vsyncs for 60hz refresh rate), |
| // there is little risk in starting over and finding the earliest vsync to latch onto. |
| // - Update captureToPresentTime offset to be used for later frames. |
| // - Example use cases: |
| // - when frame rate drops down to below 20 fps, or |
| // - A new streaming session starts (stopPreview followed by |
| // startPreview) |
| // |
| nsecs_t captureInterval = t - mLastCaptureTime; |
| if (captureInterval > kSpacingResetIntervalNs) { |
| for (size_t i = 0; i < vsyncEventData.frameTimelinesLength; i++) { |
| const auto& timeline = vsyncEventData.frameTimelines[i]; |
| if (timeline.deadlineTimestamp >= currentTime && |
| timeline.expectedPresentationTime > minPresentT) { |
| nsecs_t presentT = vsyncEventData.frameTimelines[i].expectedPresentationTime; |
| mCaptureToPresentOffset = presentT - t; |
| mLastCaptureTime = t; |
| mLastPresentTime = presentT; |
| |
| // If releaseFence is available, store the fence to check signal |
| // time later. |
| mRefVsyncData = vsyncEventData; |
| mReferenceCaptureTime = t; |
| mReferenceArrivalTime = currentTime; |
| if (releaseFence->isValid()) { |
| mReferenceFrameFence = new Fence(releaseFence->dup()); |
| } else { |
| mFenceSignalOffset = 0; |
| } |
| |
| // Move the expected presentation time back by 1/3 of frame interval to |
| // mitigate the time drift. Due to time drift, if we directly use the |
| // expected presentation time, often times 2 expected presentation time |
| // falls into the same VSYNC interval. |
| return presentT - vsyncEventData.frameInterval/3; |
| } |
| } |
| } |
| |
| // If there is a reference frame release fence, get the signal time and |
| // update the captureToPresentOffset. |
| if (mReferenceFrameFence != nullptr) { |
| mFenceSignalOffset = 0; |
| nsecs_t signalTime = mReferenceFrameFence->getSignalTime(); |
| // Now that the fence has signaled, recalculate the offsets based on |
| // the timeline which was actually latched |
| if (signalTime != INT64_MAX) { |
| for (size_t i = 0; i < mRefVsyncData.frameTimelinesLength; i++) { |
| const auto& timeline = mRefVsyncData.frameTimelines[i]; |
| if (timeline.deadlineTimestamp >= signalTime) { |
| nsecs_t originalOffset = mCaptureToPresentOffset; |
| mCaptureToPresentOffset = timeline.expectedPresentationTime |
| - mReferenceCaptureTime; |
| mLastPresentTime = timeline.expectedPresentationTime; |
| mFenceSignalOffset = signalTime > mReferenceArrivalTime ? |
| signalTime - mReferenceArrivalTime : 0; |
| |
| ALOGV("%s: Last deadline %" PRId64 " signalTime %" PRId64 |
| " original offset %" PRId64 " new offset %" PRId64 |
| " fencesignal offset %" PRId64, __FUNCTION__, |
| timeline.deadlineTimestamp, signalTime, originalOffset, |
| mCaptureToPresentOffset, mFenceSignalOffset); |
| break; |
| } |
| } |
| mReferenceFrameFence.clear(); |
| } |
| } |
| |
| nsecs_t idealPresentT = t + mCaptureToPresentOffset; |
| nsecs_t expectedPresentT = mLastPresentTime; |
| nsecs_t minDiff = INT64_MAX; |
| |
| // In fixed FPS case, when frame durations are close to multiples of display refresh |
| // rate, derive minimum intervals between presentation times based on minimal |
| // expected duration. The minimum number of Vsyncs is: |
| // - 0 if minFrameDuration in (0, 1.5] * vSyncInterval, |
| // - 1 if minFrameDuration in (1.5, 2.5] * vSyncInterval, |
| // - and so on. |
| // |
| // This spaces out the displaying of the frames so that the frame |
| // presentations are roughly in sync with frame captures. |
| int minVsyncs = (mMinExpectedDuration - vsyncEventData.frameInterval / 2) / |
| vsyncEventData.frameInterval; |
| if (minVsyncs < 0) minVsyncs = 0; |
| nsecs_t minInterval = minVsyncs * vsyncEventData.frameInterval; |
| |
| // In fixed FPS case, if the frame duration deviates from multiples of |
| // display refresh rate, find the closest Vsync without requiring a minimum |
| // number of Vsync. |
| // |
| // Example: (24fps camera, 60hz refresh): |
| // capture readout: | t1 | t1 | .. | t1 | .. | t1 | .. | t1 | |
| // display VSYNC: | t2 | t2 | ... | t2 | ... | t2 | ... | t2 | |
| // | : 1 frame |
| // t1 : 41.67ms |
| // t2 : 16.67ms |
| // t1/t2 = 2.5 |
| // |
| // 24fps is a commonly used video frame rate. Because the capture |
| // interval is 2.5 times of display refresh interval, the minVsyncs |
| // calculation will directly fall at the boundary condition. In this case, |
| // we should fall back to the basic logic of finding closest vsync |
| // timestamp without worrying about minVsyncs. |
| float captureToVsyncIntervalRatio = 1.0f * mMinExpectedDuration / vsyncEventData.frameInterval; |
| float ratioDeviation = std::fabs( |
| captureToVsyncIntervalRatio - std::roundf(captureToVsyncIntervalRatio)); |
| bool captureDeviateFromVsync = ratioDeviation >= kMaxIntervalRatioDeviation; |
| bool cameraDisplayInSync = (mFixedFps && !captureDeviateFromVsync); |
| |
| // Find best timestamp in the vsync timelines: |
| // - Only use at most kMaxTimelines timelines to avoid long latency |
| // - Add an extra timeline if display fence is used |
| // - closest to the ideal presentation time, |
| // - deadline timestamp is greater than the current time, and |
| // - For fixed FPS, if the capture interval doesn't deviate too much from refresh interval, |
| // the candidate presentation time is at least minInterval in the future compared to last |
| // presentation time. |
| // - For variable FPS, or if the capture interval deviates from refresh |
| // interval for more than 5%, find a presentation time closest to the |
| // (lastPresentationTime + captureToPresentOffset) instead. |
| int fenceAdjustment = (mFenceSignalOffset > 0) ? 1 : 0; |
| int maxTimelines = std::min(kMaxTimelines + fenceAdjustment, |
| (int)vsyncEventData.frameTimelinesLength); |
| float biasForShortDelay = 1.0f; |
| for (int i = 0; i < maxTimelines; i ++) { |
| const auto& vsyncTime = vsyncEventData.frameTimelines[i]; |
| if (minVsyncs > 0) { |
| // Bias towards using smaller timeline index: |
| // i = 0: bias = 1 |
| // i = maxTimelines-1: bias = -1 |
| biasForShortDelay = 1.0 - 2.0 * i / (maxTimelines - 1); |
| } |
| if (std::abs(vsyncTime.expectedPresentationTime - idealPresentT) < minDiff && |
| vsyncTime.deadlineTimestamp >= currentTime + mFenceSignalOffset && |
| ((!cameraDisplayInSync && vsyncTime.expectedPresentationTime > minPresentT) || |
| (cameraDisplayInSync && vsyncTime.expectedPresentationTime > |
| mLastPresentTime + minInterval + |
| static_cast<nsecs_t>(biasForShortDelay * kTimelineThresholdNs)))) { |
| expectedPresentT = vsyncTime.expectedPresentationTime; |
| minDiff = std::abs(vsyncTime.expectedPresentationTime - idealPresentT); |
| } |
| } |
| |
| if (expectedPresentT == mLastPresentTime && expectedPresentT < |
| vsyncEventData.frameTimelines[maxTimelines-1].expectedPresentationTime) { |
| // Couldn't find a reasonable presentation time. Using last frame's |
| // presentation time would cause a frame drop. The best option now |
| // is to use the next VSync as long as the last presentation time |
| // doesn't already has the maximum latency, in which case dropping the |
| // buffer is more desired than increasing latency. |
| // |
| // Example: (60fps camera, 59.9hz refresh): |
| // capture readout: | t1 | t1 | .. | t1 | .. | t1 | .. | t1 | |
| // \ \ \ \ \ \ \ \ \ |
| // queue to BQ: | | | | | | | | | |
| // \ \ \ \ \ \ \ \ \ |
| // display VSYNC: | t2 | t2 | ... | t2 | ... | t2 | ... | t2 | |
| // |
| // |: 1 frame |
| // t1 : 16.67ms |
| // t2 : 16.69ms |
| // |
| // It takes 833 frames for capture readout count and display VSYNC count to be off |
| // by 1. |
| // - At frames [0, 832], presentationTime is set to timeline[0] |
| // - At frames [833, 833*2-1], presentationTime is set to timeline[1] |
| // - At frames [833*2, 833*3-1] presentationTime is set to timeline[2] |
| // - At frame 833*3, no presentation time is found because we only |
| // search for timeline[0..2]. |
| // - Drop one buffer is better than further extend the presentation |
| // time. |
| // |
| // However, if frame 833*2 arrives 16.67ms early (right after frame |
| // 833*2-1), no presentation time can be found because |
| // getLatestVsyncEventData is called early. In that case, it's better to |
| // set presentation time by offseting last presentation time. |
| expectedPresentT += vsyncEventData.frameInterval; |
| } |
| |
| mLastCaptureTime = t; |
| mLastPresentTime = expectedPresentT; |
| |
| // Move the expected presentation time back by 1/3 of frame interval to |
| // mitigate the time drift. Due to time drift, if we directly use the |
| // expected presentation time, often times 2 expected presentation time |
| // falls into the same VSYNC interval. |
| return expectedPresentT - vsyncEventData.frameInterval/3; |
| } |
| |
| bool Camera3OutputStream::shouldLogError(status_t res) { |
| Mutex::Autolock l(mLock); |
| return shouldLogError(res, mState); |
| } |
| |
| }; // namespace camera3 |
| |
| }; // namespace android |