blob: 30c216f209365b4580f4bb5344df6dee73485576 [file] [log] [blame]
/*
* Copyright (C) 2022 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 "ExtCamUtils"
// #define LOG_NDEBUG 0
#include "ExternalCameraUtils.h"
#include <aidlcommonsupport/NativeHandle.h>
#include <jpeglib.h>
#include <linux/videodev2.h>
#include <log/log.h>
#include <algorithm>
#include <cinttypes>
#include <cmath>
#define HAVE_JPEG // required for libyuv.h to export MJPEG decode APIs
#include <libyuv.h>
namespace android {
namespace hardware {
namespace camera {
namespace external {
namespace common {
namespace {
const int kDefaultCameraIdOffset = 100;
const int kDefaultJpegBufSize = 5 << 20; // 5MB
const int kDefaultNumVideoBuffer = 4;
const int kDefaultNumStillBuffer = 2;
const int kDefaultOrientation = 0; // suitable for natural landscape displays like tablet/TV
// For phone devices 270 is better
} // anonymous namespace
const char* ExternalCameraConfig::kDefaultCfgPath = "/vendor/etc/external_camera_config.xml";
ExternalCameraConfig ExternalCameraConfig::loadFromCfg(const char* cfgPath) {
using namespace tinyxml2;
ExternalCameraConfig ret;
XMLDocument configXml;
XMLError err = configXml.LoadFile(cfgPath);
if (err != XML_SUCCESS) {
ALOGE("%s: Unable to load external camera config file '%s'. Error: %s", __FUNCTION__,
cfgPath, XMLDocument::ErrorIDToName(err));
return ret;
} else {
ALOGI("%s: load external camera config succeeded!", __FUNCTION__);
}
XMLElement* extCam = configXml.FirstChildElement("ExternalCamera");
if (extCam == nullptr) {
ALOGI("%s: no external camera config specified", __FUNCTION__);
return ret;
}
XMLElement* providerCfg = extCam->FirstChildElement("Provider");
if (providerCfg == nullptr) {
ALOGI("%s: no external camera provider config specified", __FUNCTION__);
return ret;
}
XMLElement* cameraIdOffset = providerCfg->FirstChildElement("CameraIdOffset");
if (cameraIdOffset != nullptr) {
ret.cameraIdOffset = std::atoi(cameraIdOffset->GetText());
}
XMLElement* ignore = providerCfg->FirstChildElement("ignore");
if (ignore == nullptr) {
ALOGI("%s: no internal ignored device specified", __FUNCTION__);
return ret;
}
XMLElement* id = ignore->FirstChildElement("id");
while (id != nullptr) {
const char* text = id->GetText();
if (text != nullptr) {
ret.mInternalDevices.insert(text);
ALOGI("%s: device %s will be ignored by external camera provider", __FUNCTION__, text);
}
id = id->NextSiblingElement("id");
}
XMLElement* deviceCfg = extCam->FirstChildElement("Device");
if (deviceCfg == nullptr) {
ALOGI("%s: no external camera device config specified", __FUNCTION__);
return ret;
}
XMLElement* jpegBufSz = deviceCfg->FirstChildElement("MaxJpegBufferSize");
if (jpegBufSz == nullptr) {
ALOGI("%s: no max jpeg buffer size specified", __FUNCTION__);
} else {
ret.maxJpegBufSize = jpegBufSz->UnsignedAttribute("bytes", /*Default*/ kDefaultJpegBufSize);
}
XMLElement* numVideoBuf = deviceCfg->FirstChildElement("NumVideoBuffers");
if (numVideoBuf == nullptr) {
ALOGI("%s: no num video buffers specified", __FUNCTION__);
} else {
ret.numVideoBuffers =
numVideoBuf->UnsignedAttribute("count", /*Default*/ kDefaultNumVideoBuffer);
}
XMLElement* numStillBuf = deviceCfg->FirstChildElement("NumStillBuffers");
if (numStillBuf == nullptr) {
ALOGI("%s: no num still buffers specified", __FUNCTION__);
} else {
ret.numStillBuffers =
numStillBuf->UnsignedAttribute("count", /*Default*/ kDefaultNumStillBuffer);
}
XMLElement* fpsList = deviceCfg->FirstChildElement("FpsList");
if (fpsList == nullptr) {
ALOGI("%s: no fps list specified", __FUNCTION__);
} else {
if (!updateFpsList(fpsList, ret.fpsLimits)) {
return ret;
}
}
XMLElement* depth = deviceCfg->FirstChildElement("Depth16Supported");
if (depth == nullptr) {
ret.depthEnabled = false;
ALOGI("%s: depth output is not enabled", __FUNCTION__);
} else {
ret.depthEnabled = depth->BoolAttribute("enabled", false);
}
if (ret.depthEnabled) {
XMLElement* depthFpsList = deviceCfg->FirstChildElement("DepthFpsList");
if (depthFpsList == nullptr) {
ALOGW("%s: no depth fps list specified", __FUNCTION__);
} else {
if (!updateFpsList(depthFpsList, ret.depthFpsLimits)) {
return ret;
}
}
}
XMLElement* minStreamSize = deviceCfg->FirstChildElement("MinimumStreamSize");
if (minStreamSize == nullptr) {
ALOGI("%s: no minimum stream size specified", __FUNCTION__);
} else {
ret.minStreamSize = {
static_cast<int32_t>(minStreamSize->UnsignedAttribute("width", /*Default*/ 0)),
static_cast<int32_t>(minStreamSize->UnsignedAttribute("height", /*Default*/ 0))};
}
XMLElement* orientation = deviceCfg->FirstChildElement("Orientation");
if (orientation == nullptr) {
ALOGI("%s: no sensor orientation specified", __FUNCTION__);
} else {
ret.orientation = orientation->IntAttribute("degree", /*Default*/ kDefaultOrientation);
}
ALOGI("%s: external camera cfg loaded: maxJpgBufSize %d,"
" num video buffers %d, num still buffers %d, orientation %d",
__FUNCTION__, ret.maxJpegBufSize, ret.numVideoBuffers, ret.numStillBuffers,
ret.orientation);
for (const auto& limit : ret.fpsLimits) {
ALOGI("%s: fpsLimitList: %dx%d@%f", __FUNCTION__, limit.size.width, limit.size.height,
limit.fpsUpperBound);
}
for (const auto& limit : ret.depthFpsLimits) {
ALOGI("%s: depthFpsLimitList: %dx%d@%f", __FUNCTION__, limit.size.width, limit.size.height,
limit.fpsUpperBound);
}
ALOGI("%s: minStreamSize: %dx%d", __FUNCTION__, ret.minStreamSize.width,
ret.minStreamSize.height);
return ret;
}
bool ExternalCameraConfig::updateFpsList(tinyxml2::XMLElement* fpsList,
std::vector<FpsLimitation>& fpsLimits) {
using namespace tinyxml2;
std::vector<FpsLimitation> limits;
XMLElement* row = fpsList->FirstChildElement("Limit");
while (row != nullptr) {
FpsLimitation prevLimit{{0, 0}, 1000.0};
FpsLimitation limit = {
{/* width */ static_cast<int32_t>(row->UnsignedAttribute("width", /*Default*/ 0)),
/* height */ static_cast<int32_t>(
row->UnsignedAttribute("height", /*Default*/ 0))},
/* fpsUpperBound */ row->DoubleAttribute("fpsBound", /*Default*/ 1000.0)};
if (limit.size.width <= prevLimit.size.width ||
limit.size.height <= prevLimit.size.height ||
limit.fpsUpperBound >= prevLimit.fpsUpperBound) {
ALOGE("%s: FPS limit list must have increasing size and decreasing fps!"
" Prev %dx%d@%f, Current %dx%d@%f",
__FUNCTION__, prevLimit.size.width, prevLimit.size.height,
prevLimit.fpsUpperBound, limit.size.width, limit.size.height,
limit.fpsUpperBound);
return false;
}
limits.push_back(limit);
row = row->NextSiblingElement("Limit");
}
fpsLimits = limits;
return true;
}
ExternalCameraConfig::ExternalCameraConfig()
: cameraIdOffset(kDefaultCameraIdOffset),
maxJpegBufSize(kDefaultJpegBufSize),
numVideoBuffers(kDefaultNumVideoBuffer),
numStillBuffers(kDefaultNumStillBuffer),
depthEnabled(false),
orientation(kDefaultOrientation) {
fpsLimits.push_back({/* size */ {/* width */ 640, /* height */ 480}, /* fpsUpperBound */ 30.0});
fpsLimits.push_back({/* size */ {/* width */ 1280, /* height */ 720}, /* fpsUpperBound */ 7.5});
fpsLimits.push_back(
{/* size */ {/* width */ 1920, /* height */ 1080}, /* fpsUpperBound */ 5.0});
minStreamSize = {0, 0};
}
} // namespace common
} // namespace external
namespace device {
namespace implementation {
double SupportedV4L2Format::FrameRate::getFramesPerSecond() const {
return static_cast<double>(durationDenominator) / durationNumerator;
}
Frame::Frame(uint32_t width, uint32_t height, uint32_t fourcc)
: mWidth(width), mHeight(height), mFourcc(fourcc) {}
Frame::~Frame() {}
V4L2Frame::V4L2Frame(uint32_t w, uint32_t h, uint32_t fourcc, int bufIdx, int fd, uint32_t dataSize,
uint64_t offset)
: Frame(w, h, fourcc), mBufferIndex(bufIdx), mFd(fd), mDataSize(dataSize), mOffset(offset) {}
V4L2Frame::~V4L2Frame() {
unmap();
}
int V4L2Frame::getData(uint8_t** outData, size_t* dataSize) {
return map(outData, dataSize);
}
int V4L2Frame::map(uint8_t** data, size_t* dataSize) {
if (data == nullptr || dataSize == nullptr) {
ALOGI("%s: V4L2 buffer map bad argument: data %p, dataSize %p", __FUNCTION__, data,
dataSize);
return -EINVAL;
}
std::lock_guard<std::mutex> lk(mLock);
if (!mMapped) {
void* addr = mmap(nullptr, mDataSize, PROT_READ, MAP_SHARED, mFd, mOffset);
if (addr == MAP_FAILED) {
ALOGE("%s: V4L2 buffer map failed: %s", __FUNCTION__, strerror(errno));
return -EINVAL;
}
mData = static_cast<uint8_t*>(addr);
mMapped = true;
}
*data = mData;
*dataSize = mDataSize;
ALOGV("%s: V4L map FD %d, data %p size %zu", __FUNCTION__, mFd, mData, mDataSize);
return 0;
}
int V4L2Frame::unmap() {
std::lock_guard<std::mutex> lk(mLock);
if (mMapped) {
ALOGV("%s: V4L unmap data %p size %zu", __FUNCTION__, mData, mDataSize);
if (munmap(mData, mDataSize) != 0) {
ALOGE("%s: V4L2 buffer unmap failed: %s", __FUNCTION__, strerror(errno));
return -EINVAL;
}
mMapped = false;
}
return 0;
}
AllocatedFrame::AllocatedFrame(uint32_t w, uint32_t h) : Frame(w, h, V4L2_PIX_FMT_YUV420) {}
AllocatedFrame::~AllocatedFrame() {}
int AllocatedFrame::getData(uint8_t** outData, size_t* dataSize) {
YCbCrLayout layout;
int ret = allocate(&layout);
if (ret != 0) {
return ret;
}
*outData = mData.data();
*dataSize = mBufferSize;
return 0;
}
int AllocatedFrame::allocate(YCbCrLayout* out) {
std::lock_guard<std::mutex> lk(mLock);
if ((mWidth % 2) || (mHeight % 2)) {
ALOGE("%s: bad dimension %dx%d (not multiple of 2)", __FUNCTION__, mWidth, mHeight);
return -EINVAL;
}
// This frame might be sent to jpeglib to be encoded. Since AllocatedFrame only contains YUV420,
// jpeglib expects height and width of Y component to be an integral multiple of 2*DCTSIZE,
// and heights and widths of Cb and Cr components to be an integral multiple of DCTSIZE. If the
// image size does not meet this requirement, libjpeg expects its input to be padded to meet the
// constraints. This padding is removed from the final encoded image so the content in the
// padding doesn't matter. What matters is that the memory is accessible to jpeglib at the time
// of encoding.
// For example, if the image size is 1500x844 and DCTSIZE is 8, jpeglib expects a YUV 420
// frame with components of following sizes:
// Y: 1504x848 because 1504 and 848 are the next smallest multiples of 2*8
// Cb/Cr: 752x424 which are the next smallest multiples of 8
// jpeglib takes an array of row pointers which makes vertical padding trivial when setting up
// the pointers. Padding horizontally is a bit more complicated. AllocatedFrame holds the data
// in a flattened buffer, which means memory accesses past a row will flow into the next logical
// row. For any row of a component, we can consider the first few bytes of the next row as
// padding for the current one. This is true for Y and Cb components and all but last row of the
// Cr component. Reading past the last row of Cr component will lead to undefined behavior as
// libjpeg attempts to read memory past the allocated buffer. To prevent undefined behavior,
// the buffer allocated here is padded such that libjpeg never accesses unallocated memory when
// reading the last row. Effectively, we only need to ensure that the last row of Cr component
// has width that is an integral multiple of DCTSIZE.
size_t dataSize = mWidth * mHeight * 3 / 2; // YUV420
size_t cbWidth = mWidth / 2;
size_t requiredCbWidth = DCTSIZE * ((cbWidth + DCTSIZE - 1) / DCTSIZE);
size_t padding = requiredCbWidth - cbWidth;
size_t finalSize = dataSize + padding;
if (mData.size() != finalSize) {
mData.resize(finalSize);
mBufferSize = dataSize;
}
if (out != nullptr) {
out->y = mData.data();
out->yStride = mWidth;
uint8_t* cbStart = mData.data() + mWidth * mHeight;
uint8_t* crStart = cbStart + mWidth * mHeight / 4;
out->cb = cbStart;
out->cr = crStart;
out->cStride = mWidth / 2;
out->chromaStep = 1;
}
return 0;
}
int AllocatedFrame::getLayout(YCbCrLayout* out) {
IMapper::Rect noCrop = {0, 0, static_cast<int32_t>(mWidth), static_cast<int32_t>(mHeight)};
return getCroppedLayout(noCrop, out);
}
int AllocatedFrame::getCroppedLayout(const IMapper::Rect& rect, YCbCrLayout* out) {
if (out == nullptr) {
ALOGE("%s: null out", __FUNCTION__);
return -1;
}
std::lock_guard<std::mutex> lk(mLock);
if ((rect.left + rect.width) > static_cast<int>(mWidth) ||
(rect.top + rect.height) > static_cast<int>(mHeight) || (rect.left % 2) || (rect.top % 2) ||
(rect.width % 2) || (rect.height % 2)) {
ALOGE("%s: bad rect left %d top %d w %d h %d", __FUNCTION__, rect.left, rect.top,
rect.width, rect.height);
return -1;
}
out->y = mData.data() + mWidth * rect.top + rect.left;
out->yStride = mWidth;
uint8_t* cbStart = mData.data() + mWidth * mHeight;
uint8_t* crStart = cbStart + mWidth * mHeight / 4;
out->cb = cbStart + mWidth * rect.top / 4 + rect.left / 2;
out->cr = crStart + mWidth * rect.top / 4 + rect.left / 2;
out->cStride = mWidth / 2;
out->chromaStep = 1;
return 0;
}
bool isAspectRatioClose(float ar1, float ar2) {
constexpr float kAspectRatioMatchThres = 0.025f; // This threshold is good enough to
// distinguish 4:3/16:9/20:9 1.33/1.78/2
return std::abs(ar1 - ar2) < kAspectRatioMatchThres;
}
aidl::android::hardware::camera::common::Status importBufferImpl(
/*inout*/ std::map<int, CirculatingBuffers>& circulatingBuffers,
/*inout*/ HandleImporter& handleImporter, int32_t streamId, uint64_t bufId,
buffer_handle_t buf,
/*out*/ buffer_handle_t** outBufPtr) {
using ::aidl::android::hardware::camera::common::Status;
// AIDL does not have null NativeHandles. It sends empty handles instead.
// We check for when the buf is empty instead of when buf is null.
bool isBufEmpty = buf == nullptr || (buf->numFds == 0 && buf->numInts == 0);
if (isBufEmpty && bufId == BUFFER_ID_NO_BUFFER) {
ALOGE("%s: bufferId %" PRIu64 " has null buffer handle!", __FUNCTION__, bufId);
return Status::ILLEGAL_ARGUMENT;
}
CirculatingBuffers& cbs = circulatingBuffers[streamId];
if (cbs.count(bufId) == 0) {
if (buf == nullptr) {
ALOGE("%s: bufferId %" PRIu64 " has null buffer handle!", __FUNCTION__, bufId);
return Status::ILLEGAL_ARGUMENT;
}
// Register a newly seen buffer
buffer_handle_t importedBuf = buf;
handleImporter.importBuffer(importedBuf);
if (importedBuf == nullptr) {
ALOGE("%s: output buffer for stream %d is invalid!", __FUNCTION__, streamId);
return Status::INTERNAL_ERROR;
} else {
cbs[bufId] = importedBuf;
}
}
*outBufPtr = &cbs[bufId];
return Status::OK;
}
uint32_t getFourCcFromLayout(const YCbCrLayout& layout) {
intptr_t cb = reinterpret_cast<intptr_t>(layout.cb);
intptr_t cr = reinterpret_cast<intptr_t>(layout.cr);
if (std::abs(cb - cr) == 1 && layout.chromaStep == 2) {
// Interleaved format
if (layout.cb > layout.cr) {
return V4L2_PIX_FMT_NV21;
} else {
return V4L2_PIX_FMT_NV12;
}
} else if (layout.chromaStep == 1) {
// Planar format
if (layout.cb > layout.cr) {
return V4L2_PIX_FMT_YVU420; // YV12
} else {
return V4L2_PIX_FMT_YUV420; // YU12
}
} else {
return FLEX_YUV_GENERIC;
}
}
int getCropRect(CroppingType ct, const Size& inSize, const Size& outSize, IMapper::Rect* out) {
if (out == nullptr) {
ALOGE("%s: out is null", __FUNCTION__);
return -1;
}
uint32_t inW = inSize.width;
uint32_t inH = inSize.height;
uint32_t outW = outSize.width;
uint32_t outH = outSize.height;
// Handle special case where aspect ratio is close to input but scaled
// dimension is slightly larger than input
float arIn = ASPECT_RATIO(inSize);
float arOut = ASPECT_RATIO(outSize);
if (isAspectRatioClose(arIn, arOut)) {
out->left = 0;
out->top = 0;
out->width = static_cast<int32_t>(inW);
out->height = static_cast<int32_t>(inH);
return 0;
}
if (ct == VERTICAL) {
uint64_t scaledOutH = static_cast<uint64_t>(outH) * inW / outW;
if (scaledOutH > inH) {
ALOGE("%s: Output size %dx%d cannot be vertically cropped from input size %dx%d",
__FUNCTION__, outW, outH, inW, inH);
return -1;
}
scaledOutH = scaledOutH & ~0x1; // make it multiple of 2
out->left = 0;
out->top = static_cast<int32_t>((inH - scaledOutH) / 2) & ~0x1;
out->width = static_cast<int32_t>(inW);
out->height = static_cast<int32_t>(scaledOutH);
ALOGV("%s: crop %dx%d to %dx%d: top %d, scaledH %d", __FUNCTION__, inW, inH, outW, outH,
out->top, static_cast<int32_t>(scaledOutH));
} else {
uint64_t scaledOutW = static_cast<uint64_t>(outW) * inH / outH;
if (scaledOutW > inW) {
ALOGE("%s: Output size %dx%d cannot be horizontally cropped from input size %dx%d",
__FUNCTION__, outW, outH, inW, inH);
return -1;
}
scaledOutW = scaledOutW & ~0x1; // make it multiple of 2
out->left = static_cast<int32_t>((inW - scaledOutW) / 2) & ~0x1;
out->top = 0;
out->width = static_cast<int32_t>(scaledOutW);
out->height = static_cast<int32_t>(inH);
ALOGV("%s: crop %dx%d to %dx%d: top %d, scaledW %d", __FUNCTION__, inW, inH, outW, outH,
out->top, static_cast<int32_t>(scaledOutW));
}
return 0;
}
int formatConvert(const YCbCrLayout& in, const YCbCrLayout& out, Size sz, uint32_t format) {
int ret = 0;
switch (format) {
case V4L2_PIX_FMT_NV21:
ret = libyuv::I420ToNV21(
static_cast<uint8_t*>(in.y), static_cast<int32_t>(in.yStride),
static_cast<uint8_t*>(in.cb), static_cast<int32_t>(in.cStride),
static_cast<uint8_t*>(in.cr), static_cast<int32_t>(in.cStride),
static_cast<uint8_t*>(out.y), static_cast<int32_t>(out.yStride),
static_cast<uint8_t*>(out.cr), static_cast<int32_t>(out.cStride),
static_cast<int32_t>(sz.width), static_cast<int32_t>(sz.height));
if (ret != 0) {
ALOGE("%s: convert to NV21 buffer failed! ret %d", __FUNCTION__, ret);
return ret;
}
break;
case V4L2_PIX_FMT_NV12:
ret = libyuv::I420ToNV12(
static_cast<uint8_t*>(in.y), static_cast<int32_t>(in.yStride),
static_cast<uint8_t*>(in.cb), static_cast<int32_t>(in.cStride),
static_cast<uint8_t*>(in.cr), static_cast<int32_t>(in.cStride),
static_cast<uint8_t*>(out.y), static_cast<int32_t>(out.yStride),
static_cast<uint8_t*>(out.cb), static_cast<int32_t>(out.cStride),
static_cast<int32_t>(sz.width), static_cast<int32_t>(sz.height));
if (ret != 0) {
ALOGE("%s: convert to NV12 buffer failed! ret %d", __FUNCTION__, ret);
return ret;
}
break;
case V4L2_PIX_FMT_YVU420: // YV12
case V4L2_PIX_FMT_YUV420: // YU12
// TODO: maybe we can speed up here by somehow save this copy?
ret = libyuv::I420Copy(static_cast<uint8_t*>(in.y), static_cast<int32_t>(in.yStride),
static_cast<uint8_t*>(in.cb), static_cast<int32_t>(in.cStride),
static_cast<uint8_t*>(in.cr), static_cast<int32_t>(in.cStride),
static_cast<uint8_t*>(out.y), static_cast<int32_t>(out.yStride),
static_cast<uint8_t*>(out.cb), static_cast<int32_t>(out.cStride),
static_cast<uint8_t*>(out.cr), static_cast<int32_t>(out.cStride),
static_cast<int32_t>(sz.width), static_cast<int32_t>(sz.height));
if (ret != 0) {
ALOGE("%s: copy to YV12 or YU12 buffer failed! ret %d", __FUNCTION__, ret);
return ret;
}
break;
case FLEX_YUV_GENERIC:
// TODO: b/72261744 write to arbitrary flexible YUV layout. Slow.
ALOGE("%s: unsupported flexible yuv layout"
" y %p cb %p cr %p y_str %d c_str %d c_step %d",
__FUNCTION__, out.y, out.cb, out.cr, out.yStride, out.cStride, out.chromaStep);
return -1;
default:
ALOGE("%s: unknown YUV format 0x%x!", __FUNCTION__, format);
return -1;
}
return 0;
}
int encodeJpegYU12(const Size& inSz, const YCbCrLayout& inLayout, int jpegQuality,
const void* app1Buffer, size_t app1Size, void* out, size_t maxOutSize,
size_t& actualCodeSize) {
/* libjpeg is a C library so we use C-style "inheritance" by
* putting libjpeg's jpeg_destination_mgr first in our custom
* struct. This allows us to cast jpeg_destination_mgr* to
* CustomJpegDestMgr* when we get it passed to us in a callback */
struct CustomJpegDestMgr {
struct jpeg_destination_mgr mgr;
JOCTET* mBuffer;
size_t mBufferSize;
size_t mEncodedSize;
bool mSuccess;
} dmgr;
jpeg_compress_struct cinfo = {};
jpeg_error_mgr jerr;
/* Initialize error handling with standard callbacks, but
* then override output_message (to print to ALOG) and
* error_exit to set a flag and print a message instead
* of killing the whole process */
cinfo.err = jpeg_std_error(&jerr);
cinfo.err->output_message = [](j_common_ptr cinfo) {
char buffer[JMSG_LENGTH_MAX];
/* Create the message */
(*cinfo->err->format_message)(cinfo, buffer);
ALOGE("libjpeg error: %s", buffer);
};
cinfo.err->error_exit = [](j_common_ptr cinfo) {
(*cinfo->err->output_message)(cinfo);
if (cinfo->client_data) {
auto& dmgr = *reinterpret_cast<CustomJpegDestMgr*>(cinfo->client_data);
dmgr.mSuccess = false;
}
};
/* Now that we initialized some callbacks, let's create our compressor */
jpeg_create_compress(&cinfo);
/* Initialize our destination manager */
dmgr.mBuffer = static_cast<JOCTET*>(out);
dmgr.mBufferSize = maxOutSize;
dmgr.mEncodedSize = 0;
dmgr.mSuccess = true;
cinfo.client_data = static_cast<void*>(&dmgr);
/* These lambdas become C-style function pointers and as per C++11 spec
* may not capture anything */
dmgr.mgr.init_destination = [](j_compress_ptr cinfo) {
auto& dmgr = reinterpret_cast<CustomJpegDestMgr&>(*cinfo->dest);
dmgr.mgr.next_output_byte = dmgr.mBuffer;
dmgr.mgr.free_in_buffer = dmgr.mBufferSize;
ALOGV("%s:%d jpeg start: %p [%zu]", __FUNCTION__, __LINE__, dmgr.mBuffer, dmgr.mBufferSize);
};
dmgr.mgr.empty_output_buffer = [](j_compress_ptr cinfo __unused) {
ALOGV("%s:%d Out of buffer", __FUNCTION__, __LINE__);
return 0;
};
dmgr.mgr.term_destination = [](j_compress_ptr cinfo) {
auto& dmgr = reinterpret_cast<CustomJpegDestMgr&>(*cinfo->dest);
dmgr.mEncodedSize = dmgr.mBufferSize - dmgr.mgr.free_in_buffer;
ALOGV("%s:%d Done with jpeg: %zu", __FUNCTION__, __LINE__, dmgr.mEncodedSize);
};
cinfo.dest = reinterpret_cast<struct jpeg_destination_mgr*>(&dmgr);
/* We are going to be using JPEG in raw data mode, so we are passing
* straight subsampled planar YCbCr and it will not touch our pixel
* data or do any scaling or anything */
cinfo.image_width = inSz.width;
cinfo.image_height = inSz.height;
cinfo.input_components = 3;
cinfo.in_color_space = JCS_YCbCr;
/* Initialize defaults and then override what we want */
jpeg_set_defaults(&cinfo);
jpeg_set_quality(&cinfo, jpegQuality, 1);
jpeg_set_colorspace(&cinfo, JCS_YCbCr);
cinfo.raw_data_in = 1;
cinfo.dct_method = JDCT_IFAST;
/* Configure sampling factors. The sampling factor is JPEG subsampling 420
* because the source format is YUV420. Note that libjpeg sampling factors
* are... a little weird. Sampling of Y=2,U=1,V=1 means there is 1 U and
* 1 V value for each 2 Y values */
cinfo.comp_info[0].h_samp_factor = 2;
cinfo.comp_info[0].v_samp_factor = 2;
cinfo.comp_info[1].h_samp_factor = 1;
cinfo.comp_info[1].v_samp_factor = 1;
cinfo.comp_info[2].h_samp_factor = 1;
cinfo.comp_info[2].v_samp_factor = 1;
/* Start the compressor */
jpeg_start_compress(&cinfo, TRUE);
/* Let's not hardcode YUV420 in 6 places... 5 was enough */
int maxVSampFactor = cinfo.max_v_samp_factor;
int cVSubSampling = cinfo.comp_info[0].v_samp_factor / cinfo.comp_info[1].v_samp_factor;
/* Compute our macroblock height, so we can pad our input to be vertically
* macroblock aligned. No need to for horizontal alignment since AllocatedFrame already
* pads horizontally */
size_t mcuV = DCTSIZE * maxVSampFactor;
size_t paddedHeight = mcuV * ((inSz.height + mcuV - 1) / mcuV);
/* libjpeg uses arrays of row pointers, which makes it really easy to pad
* data vertically (unfortunately doesn't help horizontally) */
std::vector<JSAMPROW> yLines(paddedHeight);
std::vector<JSAMPROW> cbLines(paddedHeight / cVSubSampling);
std::vector<JSAMPROW> crLines(paddedHeight / cVSubSampling);
uint8_t* py = static_cast<uint8_t*>(inLayout.y);
uint8_t* pcb = static_cast<uint8_t*>(inLayout.cb);
uint8_t* pcr = static_cast<uint8_t*>(inLayout.cr);
for (int32_t i = 0; i < paddedHeight; i++) {
/* Once we are in the padding territory we still point to the last line
* effectively replicating it several times ~ CLAMP_TO_EDGE */
int li = std::min(i, inSz.height - 1);
yLines[i] = static_cast<JSAMPROW>(py + li * inLayout.yStride);
if (i < paddedHeight / cVSubSampling) {
li = std::min(i, (inSz.height - 1) / cVSubSampling);
cbLines[i] = static_cast<JSAMPROW>(pcb + li * inLayout.cStride);
crLines[i] = static_cast<JSAMPROW>(pcr + li * inLayout.cStride);
}
}
/* If APP1 data was passed in, use it */
if (app1Buffer && app1Size) {
jpeg_write_marker(&cinfo, JPEG_APP0 + 1, static_cast<const JOCTET*>(app1Buffer), app1Size);
}
/* While we still have padded height left to go, keep giving it one
* macroblock at a time. */
while (cinfo.next_scanline < cinfo.image_height) {
const uint32_t batchSize = DCTSIZE * maxVSampFactor;
const uint32_t nl = cinfo.next_scanline;
JSAMPARRAY planes[3]{&yLines[nl], &cbLines[nl / cVSubSampling],
&crLines[nl / cVSubSampling]};
uint32_t done = jpeg_write_raw_data(&cinfo, planes, batchSize);
if (done != batchSize) {
ALOGE("%s: compressed %u lines, expected %u (total %u/%u)", __FUNCTION__, done,
batchSize, cinfo.next_scanline, cinfo.image_height);
return -1;
}
}
/* This will flush everything */
jpeg_finish_compress(&cinfo);
/* Grab the actual code size and set it */
actualCodeSize = dmgr.mEncodedSize;
return 0;
}
Size getMaxThumbnailResolution(const common::V1_0::helper::CameraMetadata& chars) {
Size thumbSize{0, 0};
camera_metadata_ro_entry entry = chars.find(ANDROID_JPEG_AVAILABLE_THUMBNAIL_SIZES);
for (uint32_t i = 0; i < entry.count; i += 2) {
Size sz{.width = entry.data.i32[i], .height = entry.data.i32[i + 1]};
if (sz.width * sz.height > thumbSize.width * thumbSize.height) {
thumbSize = sz;
}
}
if (thumbSize.width * thumbSize.height == 0) {
ALOGW("%s: non-zero thumbnail size not available", __FUNCTION__);
}
return thumbSize;
}
void freeReleaseFences(std::vector<CaptureResult>& results) {
for (auto& result : results) {
native_handle_t* inputReleaseFence =
::android::makeFromAidl(result.inputBuffer.releaseFence);
if (inputReleaseFence != nullptr) {
native_handle_close(inputReleaseFence);
native_handle_delete(inputReleaseFence);
}
for (auto& buf : result.outputBuffers) {
native_handle_t* outReleaseFence = ::android::makeFromAidl(buf.releaseFence);
if (outReleaseFence != nullptr) {
native_handle_close(outReleaseFence);
native_handle_delete(outReleaseFence);
}
}
}
}
#define ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
#define UPDATE(md, tag, data, size) \
do { \
if ((md).update((tag), (data), (size))) { \
ALOGE("Update " #tag " failed!"); \
return BAD_VALUE; \
} \
} while (0)
status_t fillCaptureResultCommon(CameraMetadata& md, nsecs_t timestamp,
camera_metadata_ro_entry& activeArraySize) {
if (activeArraySize.count < 4) {
ALOGE("%s: cannot find active array size!", __FUNCTION__);
return -EINVAL;
}
// android.control
// For USB camera, we don't know the AE state. Set the state to converged to
// indicate the frame should be good to use. Then apps don't have to wait the
// AE state.
const uint8_t aeState = ANDROID_CONTROL_AE_STATE_CONVERGED;
UPDATE(md, ANDROID_CONTROL_AE_STATE, &aeState, 1);
const uint8_t ae_lock = ANDROID_CONTROL_AE_LOCK_OFF;
UPDATE(md, ANDROID_CONTROL_AE_LOCK, &ae_lock, 1);
// Set AWB state to converged to indicate the frame should be good to use.
const uint8_t awbState = ANDROID_CONTROL_AWB_STATE_CONVERGED;
UPDATE(md, ANDROID_CONTROL_AWB_STATE, &awbState, 1);
const uint8_t awbLock = ANDROID_CONTROL_AWB_LOCK_OFF;
UPDATE(md, ANDROID_CONTROL_AWB_LOCK, &awbLock, 1);
const uint8_t flashState = ANDROID_FLASH_STATE_UNAVAILABLE;
UPDATE(md, ANDROID_FLASH_STATE, &flashState, 1);
// This means pipeline latency of X frame intervals. The maximum number is 4.
const uint8_t requestPipelineMaxDepth = 4;
UPDATE(md, ANDROID_REQUEST_PIPELINE_DEPTH, &requestPipelineMaxDepth, 1);
// android.scaler
const int32_t crop_region[] = {
activeArraySize.data.i32[0],
activeArraySize.data.i32[1],
activeArraySize.data.i32[2],
activeArraySize.data.i32[3],
};
UPDATE(md, ANDROID_SCALER_CROP_REGION, crop_region, ARRAY_SIZE(crop_region));
// android.sensor
UPDATE(md, ANDROID_SENSOR_TIMESTAMP, &timestamp, 1);
// android.statistics
const uint8_t lensShadingMapMode = ANDROID_STATISTICS_LENS_SHADING_MAP_MODE_OFF;
UPDATE(md, ANDROID_STATISTICS_LENS_SHADING_MAP_MODE, &lensShadingMapMode, 1);
const uint8_t sceneFlicker = ANDROID_STATISTICS_SCENE_FLICKER_NONE;
UPDATE(md, ANDROID_STATISTICS_SCENE_FLICKER, &sceneFlicker, 1);
return OK;
}
#undef ARRAY_SIZE
#undef UPDATE
AllocatedV4L2Frame::AllocatedV4L2Frame(std::shared_ptr<V4L2Frame> frameIn)
: Frame(frameIn->mWidth, frameIn->mHeight, frameIn->mFourcc) {
uint8_t* dataIn;
size_t dataSize;
if (frameIn->getData(&dataIn, &dataSize) != 0) {
ALOGE("%s: map input V4L2 frame failed!", __FUNCTION__);
return;
}
mData.resize(dataSize);
std::memcpy(mData.data(), dataIn, dataSize);
}
AllocatedV4L2Frame::~AllocatedV4L2Frame() {}
int AllocatedV4L2Frame::getData(uint8_t** outData, size_t* dataSize) {
if (outData == nullptr || dataSize == nullptr) {
ALOGE("%s: outData(%p)/dataSize(%p) must not be null", __FUNCTION__, outData, dataSize);
return -1;
}
*outData = mData.data();
*dataSize = mData.size();
return 0;
}
} // namespace implementation
} // namespace device
} // namespace camera
} // namespace hardware
} // namespace android