Android 12红米设备MediaCodec原生崩溃问题求助

现象 #

运行Android 12的红米设备在应用内生成视频时触发运行时崩溃，三星、摩托罗拉、Pixel等品牌同版本Android设备无此问题。代码基于Google Grafika项目实现，核心编码逻辑如下：

val encoder = MediaCodec.createEncoderByType("video/avc")
val format = MediaFormat.createVideoFormat("video/avc", width, height)

format.setInteger(MediaFormat.KEY_COLOR_FORMAT, MediaCodecInfo.CodecCapabilities.COLOR_FormatSurface)
format.setInteger(MediaFormat.KEY_BIT_RATE, 100000)
format.setInteger(MediaFormat.KEY_FRAME_RATE, frameRate)
format.setInteger(MediaFormat.KEY_I_FRAME_INTERVAL, 2)
encoder.configure(format, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE)
val inputSurface: Surface
try {
    inputSurface = encoder.createInputSurface()
} catch (ise: IllegalStateException) {
    // catch exception
}

encoder.start()
val muxer = MediaMuxer(outputFile.toString(), MediaMuxer.OutputFormat.MUXER_OUTPUT_MPEG_4)
val bufferInfo = MediaCodec.BufferInfo()
trackIndex = drainEncoderToMuxer(encoder, muxer, bufferInfo, trackIndex, false, clock)

val canvas = inputSurface.lockHardwareCanvas()
try {
    drawFrame(canvas, frameIndex, frameCount) // 在画布上绘制帧的方法
} catch (e: Exception) {
    log.error(TAG, e) { "Caught exception while drawing frame" }
} finally {
    inputSurface.unlockCanvasAndPost(canvas) // 将输入帧提交给编码器
}

drainEncoderToMuxer(encoder, muxer, bufferInfo, trackIndex, true, clock)

encoder.stop()
encoder.release()
inputSurface.release()
muxer.stop()
muxer.release()

崩溃日志 #

2022-07-27 10:59:42.475 14534-14971/com.app.debug W/OMXUtils: do not know color format 0x7f000789 = 2130708361
2022-07-27 10:59:42.511 14534-14971/com.app.debug I/ACodec: Need to setDefaultCodecColorAspectsIfNeeded
2022-07-27 10:59:42.591 14534-14772/com.app.debug D/MPEG4Writer: PreAllocation disabled. fallocate : Operation not supported on transport endpoint, 95
2022-07-27 10:59:42.670 14534-14534/com.app.debug I/PlayButtonView: [PlayButton] called Play with indicator:false, looping:false
2022-07-27 10:59:42.853 14534-14637/com.app.debug A/HidlStatus: Status.cpp:143] Failed HIDL return status not checked. Usually this happens because of a transport error (error parceling, binder driver, or from unparceling). If you see this in code calling into "Bn" classes in for a HAL server process, then it is likely that the code there is returning transport errors there (as opposed to errors defined within its protocol). Error is: Status(EX_TRANSACTION_FAILED): 'DEAD_OBJECT: '
2022-07-27 10:59:42.853 14534-14971/ccom.app.debug E/ACodec: OMX/mediaserver died, signalling error!
2022-07-27 10:59:42.854 14534-14971/com.app.debug E/ACodec: signalError(omxError 0x8000100d, internalError -32)
2022-07-27 10:59:42.854 14534-14970/com.app.debug E/MediaCodec: Codec reported err 0xffffffe0, actionCode 0, while in state 6/STARTED
2022-07-27 10:59:43.059 14534-14637/com.app.debug: runtime.cc:669] Runtime aborting...
    runtime.cc:669] Dumping all threads without mutator lock held
    runtime.cc:669] All threads:
    runtime.cc:669] DALVIK THREADS (104):
    runtime.cc:669] "RenderThread" prio=10 tid=54 Runnable
    runtime.cc:669]   | group="" sCount=0 ucsCount=0 flags=0 obj=0x15fc90c8 self=0xb400007480344000
    runtime.cc:669]   | sysTid=14637 nice=-10 cgrp=system sched=0/0 handle=0x73ef42fcb0
    runtime.cc:669]   | state=R schedstat=( 2482457056 440300499 3164 ) utm=160 stm=87 core=6 HZ=100
    runtime.cc:669]   | stack=0x73ef338000-0x73ef33a000 stackSize=991KB
    runtime.cc:669]   | held mutexes= "abort lock" "mutator lock"(shared held)
    runtime.cc:669]   native: #00 pc 0000000000460620  /apex/com.android.art/lib64/libart.so (art::DumpNativeStack(std::__1::basic_ostream<char, std::__1::char_traits<char> >&, int, BacktraceMap*, char const*, art::ArtMethod*, void*, bool)+120)
    runtime.cc:669]   native: #01 pc 00000000006ffed0  /apex/com.android.art/lib64/libart.so (art::Thread::DumpStack(std::__1::basic_ostream<char, std::__1::char_traits<char> >&, bool, BacktraceMap*, bool) const+252)
    runtime.cc:669]   native: #02 pc 00000000007079a8  /apex/com.android.art/lib64/libart.so (art::DumpCheckpoint::Run(art::Thread*)+304)
    runtime.cc:669]   native: #03 pc 000000000031b7dc  /apex/com.android.art/lib64/libart.so (art::ThreadList::RunCheckpoint(art::Closure*, art::Closure*)+628)
    runtime.cc:669]   native: #04 pc 00000000006e6eb4  /apex/com.android.art/lib64/libart.so (art::ThreadList::Dump(std::__1::basic_ostream<char, std::__1::char_traits<char> >&, bool)+476)
    runtime.cc:669]   native: #05 pc 00000000006d8200  /apex/com.android.art/lib64/libart.so (art::AbortState::Dump(std::__1::basic_ostream<char, std::__1::char_traits<char> >&) const+204)
    runtime.cc:669]   native: #06 pc 00000000006d7fa8  /apex/com.android.art/lib64/libart.so (art::Runtime::Abort(char const*)+1672)

三星Android 12设备正常日志 #

:59.672 30929-31816/com.app.debug W/OMXUtils: do not know color format 0x7f000789 = 2130708361
2022-07-27 16:28:59.697 30929-31816/com.app.debug I/ACodec: [OMX.qcom.video.encoder.avc] Now Loaded->Idle
2022-07-27 16:28:59.725 30929-31815/com.app.debug I/MediaCodec: setCodecState state(0), called in 6
2022-07-27 16:28:59.725 30929-31816/com.app.debug I/ACodec: [OMX.qcom.video.encoder.avc] Now Idle->Executing
2022-07-27 16:28:59.730 30929-31816/com.app.debug I/ACodec: [OMX.qcom.video.encoder.avc] Now Executing
2022-07-27 16:28:59.748 30929-31452/com.app.debug D/MPEG4Writer: PreAllocation disabled. fallocate : Operation not supported on transport endpoint, 95
2022-07-27 16:28:59.770 30929-31313/com.app.debug D/OpenGLRenderer: eglCreateWindowSurface
2022-07-27 16:28:59.784 30929-30929/com.app.debug I/PlayButtonView: [PlayButton] called Play with indicator:false, looping:false
2022-07-27 16:28:59.806 30929-31816/com.app.debug D/ACodec: dataspace changed to 0x10c10000 (R:2(Limited), P:3(BT601_6_625), M:3(BT601_6), T:3(SMPTE170M)) (R:2(Limited), S:2(BT601_625), T:3(SMPTE_170M))
2022-07-27 16:28:59.807 30929-31313/com.app.debug E/FrameEvents: updateAcquireFence: Did not find frame.
2022-07-27 16:28:59.833 30929-31313/com.app.debug E/FrameEvents: updateAcquireFence: Did not find frame.
2022-07-27 16:28:59.835 30929-31452/com.app.debug D/MPEG4Writer: fpathconf _PC_FILESIZEBITS:64
2022-07-27 16:28:59.835 30929-31452/com.app.debug D/MPEG4Writer: File size limit set to 4503599627370495 bytes implicitly
2022-07-27 16:28:59.835 30929-31452/com.app.debug D/MPEG4Writer: MP4WtrCtrlHlpLooper Started
2022-07-27 16:28:59.837 30929-31815/com.app.debug I/MediaCodec: setCodecState state(1), called in 6
2022-07-27 16:28:59.838 30929-31829/com.app.debug I/MPEG4Writer: setStartTimestampUs: 0 from Video track
2022-07-27 16:28:59.838 30929-31829/com.app.debug I/MPEG4Writer: Earliest track starting time: 0

已尝试的方案 #

• 将KEY_COLOR_FORMAT改为COLOR_FormatYUV420Flexible和COLOR_FormatYUV422Flexible，可避免崩溃，但无法生成有效视频文件。

解决方案 #

1. 分离编码线程与RenderThread #

崩溃发生在RenderThread，推测是线程资源冲突导致OMX服务崩溃。所有MediaCodec、MediaMuxer相关操作必须放在独立的后台线程执行，禁止在UI线程或RenderThread中处理编码逻辑。

2. 替换Canvas渲染为EGL OpenGL流程 #

部分设备的硬件编码器对lockHardwareCanvas()兼容性较差，改用Grafika标准的EGL渲染方案：

• 创建EGLContext关联编码器的InputSurface
• 使用OpenGL ES API绘制帧内容
• 绘制完成后调用eglSwapBuffers()提交帧到编码器

3. 补充颜色空间配置参数 #

红米设备编码器可能需要明确的颜色空间参数，在配置MediaFormat时添加：

format.setInteger(MediaFormat.KEY_COLOR_STANDARD, MediaFormat.COLOR_STANDARD_BT601)
format.setInteger(MediaFormat.KEY_COLOR_TRANSFER, MediaFormat.COLOR_TRANSFER_SMPTE_170M)
format.setInteger(MediaFormat.KEY_COLOR_RANGE, MediaFormat.COLOR_RANGE_LIMITED)

4. 优化编码器数据导出逻辑 #

确保drainEncoderToMuxer方法处理逻辑正确：

• 等待编码器输出有效数据后，再将轨道添加到MediaMuxer
• 处理EOS信号时，确保所有编码器缓存数据都写入Muxer
• 避免在编码器未输出任何数据时调用muxer.stop()

5. 动态验证编码器格式支持 #

提前校验编码器对Surface格式的支持，避免硬编码格式导致兼容性问题：

val codecInfo = MediaCodecList(MediaCodecList.REGULAR_CODECS).findEncoderForFormat(format)
val capabilities = codecInfo.getCapabilitiesForType("video/avc")
if (!capabilities.colorFormats.contains(MediaCodecInfo.CodecCapabilities.COLOR_FormatSurface)) {
    // 降级到编码器支持的其他格式
}

内容的提问来源于stack exchange，提问作者Madhan