encode h264 video using ffmpeg library memory issues - c

Question

I'm trying to do screen capture on OS X using ffmpeg's avfoundation library. I capture frames from the screen and encode it using H264 into an flv container.
Here's the command line output of the program:
Input #0, avfoundation, from 'Capture screen 0':
Duration: N/A, start: 9.253649, bitrate: N/A
Stream #0:0: Video: rawvideo (UYVY / 0x59565955), uyvy422, 1440x900, 14.58 tbr, 1000k tbn, 1000k tbc
raw video is inCodec
FLV (Flash Video)http://localhost:8090/test.flv
[libx264 # 0x102038e00] using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX
[libx264 # 0x102038e00] profile High, level 4.0
[libx264 # 0x102038e00] 264 - core 142 r2495 6a301b6 - H.264/MPEG-4 AVC codec - Copyleft 2003-2014 - http://www.videolan.org/x264.html - options: cabac=1 ref=3 deblock=1:0:0 analyse=0x3:0x113 me=hex subme=7 psy=1 psy_rd=1.00:0.00 mixed_ref=1 me_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 fast_pskip=1 chroma_qp_offset=-2 threads=6 lookahead_threads=1 sliced_threads=0 nr=0 decimate=1 interlaced=0 bluray_compat=0 constrained_intra=0 bframes=3 b_pyramid=2 b_adapt=1 b_bias=0 direct=1 weightb=1 open_gop=1 weightp=2 keyint=50 keyint_min=5 scenecut=40 intra_refresh=0 rc_lookahead=40 rc=abr mbtree=1 bitrate=400 ratetol=1.0 qcomp=0.60 qpmin=0 qpmax=69 qpstep=4 ip_ratio=1.40 aq=1:1.00
[tcp # 0x101a5fe70] Connection to tcp://localhost:8090 failed (Connection refused), trying next address
[tcp # 0x101a5fe70] Connection to tcp://localhost:8090 failed: Connection refused
url_fopen failed: Operation now in progress
[flv # 0x102038800] Using AVStream.codec.time_base as a timebase hint to the muxer is deprecated. Set AVStream.time_base instead.
encoded frame #0
encoded frame #1
......
encoded frame #49
encoded frame #50
testmee(8404,0x7fff7e05c300) malloc: *** error for object 0x102053e08: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
(lldb) bt
* thread #10: tid = 0x43873, 0x00007fff95639286 libsystem_kernel.dylib`__pthread_kill + 10, stop reason = signal SIGABRT
* frame #0: 0x00007fff95639286 libsystem_kernel.dylib`__pthread_kill + 10
frame #1: 0x00007fff9623742f libsystem_pthread.dylib`pthread_kill + 90
frame #2: 0x00007fff977ceb53 libsystem_c.dylib`abort + 129
frame #3: 0x00007fff9ab59e06 libsystem_malloc.dylib`szone_error + 625
frame #4: 0x00007fff9ab4f799 libsystem_malloc.dylib`small_malloc_from_free_list + 1105
frame #5: 0x00007fff9ab4d3bc libsystem_malloc.dylib`szone_malloc_should_clear + 1449
frame #6: 0x00007fff9ab4c877 libsystem_malloc.dylib`malloc_zone_malloc + 71
frame #7: 0x00007fff9ab4b395 libsystem_malloc.dylib`malloc + 42
frame #8: 0x00007fff94aa63d2 IOSurface`IOSurfaceClientLookupFromMachPort + 40
frame #9: 0x00007fff94aa6b38 IOSurface`IOSurfaceLookupFromMachPort + 12
frame #10: 0x00007fff92bfa6b2 CoreGraphics`_CGYDisplayStreamFrameAvailable + 342
frame #11: 0x00007fff92f6759c CoreGraphics`CGYDisplayStreamNotification_server + 336
frame #12: 0x00007fff92bfada6 CoreGraphics`display_stream_runloop_callout + 46
frame #13: 0x00007fff956eba07 CoreFoundation`__CFMachPortPerform + 247
frame #14: 0x00007fff956eb8f9 CoreFoundation`__CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE1_PERFORM_FUNCTION__ + 41
frame #15: 0x00007fff956eb86b CoreFoundation`__CFRunLoopDoSource1 + 475
frame #16: 0x00007fff956dd3e7 CoreFoundation`__CFRunLoopRun + 2375
frame #17: 0x00007fff956dc858 CoreFoundation`CFRunLoopRunSpecific + 296
frame #18: 0x00007fff95792ef1 CoreFoundation`CFRunLoopRun + 97
frame #19: 0x0000000105f79ff1 CMIOUnits`___lldb_unnamed_function2148$$CMIOUnits + 875
frame #20: 0x0000000105f6f2c2 CMIOUnits`___lldb_unnamed_function2127$$CMIOUnits + 14
frame #21: 0x00007fff97051765 CoreMedia`figThreadMain + 417
frame #22: 0x00007fff96235268 libsystem_pthread.dylib`_pthread_body + 131
frame #23: 0x00007fff962351e5 libsystem_pthread.dylib`_pthread_start + 176
frame #24: 0x00007fff9623341d libsystem_pthread.dylib`thread_start + 13
I've attached the code I used below.
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
#include <libswscale/swscale.h>
#include <libavdevice/avdevice.h>
#include <libavutil/opt.h>
#include <stdio.h>
#include <stdint.h>
#include <unistd.h>
/* compile using
gcc -g -o stream test.c -lavformat -lavutil -lavcodec -lavdevice -lswscale
*/
// void show_av_device() {
// inFmt->get_device_list(inFmtCtx, device_list);
// printf("Device Info=============\n");
// //avformat_open_input(&inFmtCtx,"video=Capture screen 0",inFmt,&inOptions);
// printf("===============================\n");
// }
void AVFAIL (int code, const char *what) {
char msg[500];
av_strerror(code, msg, sizeof(msg));
fprintf(stderr, "failed: %s\nerror: %s\n", what, msg);
exit(2);
}
#define AVCHECK(f) do { int e = (f); if (e < 0) AVFAIL(e, #f); } while (0)
#define AVCHECKPTR(p,f) do { p = (f); if (!p) AVFAIL(AVERROR_UNKNOWN, #f); } while (0)
void registerLibs() {
av_register_all();
avdevice_register_all();
avformat_network_init();
avcodec_register_all();
}
int main(int argc, char *argv[]) {
//conversion variables
struct SwsContext *swsCtx = NULL;
//input stream variables
AVFormatContext *inFmtCtx = NULL;
AVCodecContext *inCodecCtx = NULL;
AVCodec *inCodec = NULL;
AVInputFormat *inFmt = NULL;
AVFrame *inFrame = NULL;
AVDictionary *inOptions = NULL;
const char *streamURL = "http://localhost:8090/test.flv";
const char *name = "avfoundation";
// AVFrame *inFrameYUV = NULL;
AVPacket inPacket;
//output stream variables
AVCodecContext *outCodecCtx = NULL;
AVCodec *outCodec;
AVFormatContext *outFmtCtx = NULL;
AVOutputFormat *outFmt = NULL;
AVFrame *outFrameYUV = NULL;
AVStream *stream = NULL;
int i, videostream, ret;
int numBytes, frameFinished;
registerLibs();
inFmtCtx = avformat_alloc_context(); //alloc input context
av_dict_set(&inOptions, "pixel_format", "uyvy422", 0);
av_dict_set(&inOptions, "probesize", "7000000", 0);
inFmt = av_find_input_format(name);
ret = avformat_open_input(&inFmtCtx, "Capture screen 0:", inFmt, &inOptions);
if (ret < 0) {
printf("Could not load the context for the input device\n");
return -1;
}
if (avformat_find_stream_info(inFmtCtx, NULL) < 0) {
printf("Could not find stream info for screen\n");
return -1;
}
av_dump_format(inFmtCtx, 0, "Capture screen 0", 0);
// inFmtCtx->streams is an array of pointers of size inFmtCtx->nb_stream
videostream = av_find_best_stream(inFmtCtx, AVMEDIA_TYPE_VIDEO, -1, -1, &inCodec, 0);
if (videostream == -1) {
printf("no video stream found\n");
return -1;
} else {
printf("%s is inCodec\n", inCodec->long_name);
}
inCodecCtx = inFmtCtx->streams[videostream]->codec;
// open codec
if (avcodec_open2(inCodecCtx, inCodec, NULL) > 0) {
printf("Couldn't open codec");
return -1; // couldn't open codec
}
//setup output params
outFmt = av_guess_format(NULL, streamURL, NULL);
if(outFmt == NULL) {
printf("output format was not guessed properly");
return -1;
}
if((outFmtCtx = avformat_alloc_context()) < 0) {
printf("output context not allocated. ERROR");
return -1;
}
printf("%s", outFmt->long_name);
outFmtCtx->oformat = outFmt;
snprintf(outFmtCtx->filename, sizeof(outFmtCtx->filename), streamURL);
printf("%s\n", outFmtCtx->filename);
outCodec = avcodec_find_encoder(AV_CODEC_ID_H264);
if(!outCodec) {
printf("could not find encoder for H264 \n" );
return -1;
}
stream = avformat_new_stream(outFmtCtx, outCodec);
outCodecCtx = stream->codec;
avcodec_get_context_defaults3(outCodecCtx, outCodec);
outCodecCtx->codec_id = AV_CODEC_ID_H264;
outCodecCtx->codec_type = AVMEDIA_TYPE_VIDEO;
outCodecCtx->flags = CODEC_FLAG_GLOBAL_HEADER;
outCodecCtx->width = inCodecCtx->width;
outCodecCtx->height = inCodecCtx->height;
outCodecCtx->time_base.den = 25;
outCodecCtx->time_base.num = 1;
outCodecCtx->pix_fmt = AV_PIX_FMT_YUV420P;
outCodecCtx->gop_size = 50;
outCodecCtx->bit_rate = 400000;
//setup output encoders etc
if(stream) {
ret = avcodec_open2(outCodecCtx, outCodec, NULL);
if (ret < 0) {
printf("Could not open output encoder");
return -1;
}
}
if (avio_open(&outFmtCtx->pb, outFmtCtx->filename, AVIO_FLAG_WRITE ) < 0) {
perror("url_fopen failed");
}
avio_open_dyn_buf(&outFmtCtx->pb);
ret = avformat_write_header(outFmtCtx, NULL);
if (ret != 0) {
printf("was not able to write header to output format");
return -1;
}
unsigned char *pb_buffer;
int len = avio_close_dyn_buf(outFmtCtx->pb, (unsigned char **)(&pb_buffer));
avio_write(outFmtCtx->pb, (unsigned char *)pb_buffer, len);
numBytes = avpicture_get_size(PIX_FMT_UYVY422, inCodecCtx->width, inCodecCtx->height);
// Allocate video frame
inFrame = av_frame_alloc();
swsCtx = sws_getContext(inCodecCtx->width, inCodecCtx->height, inCodecCtx->pix_fmt, inCodecCtx->width,
inCodecCtx->height, PIX_FMT_YUV420P, SWS_BILINEAR, NULL, NULL, NULL);
int frame_count = 0;
while(av_read_frame(inFmtCtx, &inPacket) >= 0) {
if(inPacket.stream_index == videostream) {
avcodec_decode_video2(inCodecCtx, inFrame, &frameFinished, &inPacket);
// 1 Frame might need more than 1 packet to be filled
if(frameFinished) {
outFrameYUV = av_frame_alloc();
uint8_t *buffer = (uint8_t *)av_malloc(numBytes);
int ret = avpicture_fill((AVPicture *)outFrameYUV, buffer, PIX_FMT_YUV420P,
inCodecCtx->width, inCodecCtx->height);
if(ret < 0){
printf("%d is return val for fill\n", ret);
return -1;
}
//convert image to YUV
sws_scale(swsCtx, (uint8_t const * const* )inFrame->data,
inFrame->linesize, 0, inCodecCtx->height,
outFrameYUV->data, outFrameYUV->linesize);
//outFrameYUV now holds the YUV scaled frame/picture
outFrameYUV->format = outCodecCtx->pix_fmt;
outFrameYUV->width = outCodecCtx->width;
outFrameYUV->height = outCodecCtx->height;
AVPacket pkt;
int got_output;
av_init_packet(&pkt);
pkt.data = NULL;
pkt.size = 0;
outFrameYUV->pts = frame_count;
ret = avcodec_encode_video2(outCodecCtx, &pkt, outFrameYUV, &got_output);
if (ret < 0) {
fprintf(stderr, "Error encoding video frame: %s\n", av_err2str(ret));
return -1;
}
if(got_output) {
if(stream->codec->coded_frame->key_frame) {
pkt.flags |= AV_PKT_FLAG_KEY;
}
pkt.stream_index = stream->index;
if(pkt.pts != AV_NOPTS_VALUE)
pkt.pts = av_rescale_q(pkt.pts, stream->codec->time_base, stream->time_base);
if(pkt.dts != AV_NOPTS_VALUE)
pkt.dts = av_rescale_q(pkt.dts, stream->codec->time_base, stream->time_base);
if(avio_open_dyn_buf(&outFmtCtx->pb)!= 0) {
printf("ERROR: Unable to open dynamic buffer\n");
}
ret = av_interleaved_write_frame(outFmtCtx, &pkt);
unsigned char *pb_buffer;
int len = avio_close_dyn_buf(outFmtCtx->pb, (unsigned char **)&pb_buffer);
avio_write(outFmtCtx->pb, (unsigned char *)pb_buffer, len);
} else {
ret = 0;
}
if(ret != 0) {
fprintf(stderr, "Error while writing video frame: %s\n", av_err2str(ret));
exit(1);
}
fprintf(stderr, "encoded frame #%d\n", frame_count);
frame_count++;
av_free_packet(&pkt);
av_frame_free(&outFrameYUV);
av_free(buffer);
}
}
av_free_packet(&inPacket);
}
av_write_trailer(outFmtCtx);
//close video stream
if(stream) {
avcodec_close(outCodecCtx);
}
for (i = 0; i < outFmtCtx->nb_streams; i++) {
av_freep(&outFmtCtx->streams[i]->codec);
av_freep(&outFmtCtx->streams[i]);
}
if (!(outFmt->flags & AVFMT_NOFILE))
/* Close the output file. */
avio_close(outFmtCtx->pb);
/* free the output format context */
avformat_free_context(outFmtCtx);
// Free the YUV frame populated by the decoder
av_free(inFrame);
// Close the video codec (decoder)
avcodec_close(inCodecCtx);
// Close the input video file
avformat_close_input(&inFmtCtx);
return 1;
}
I'm not sure what I've done wrong here. But, what I've observed is that for each frame thats been encoded, my memory usage goes up by about 6MB. Backtracking afterward usually leads one of the following two culprits:
avf_read_frame function in avfoundation.m
av_dup_packet function in avpacket.h
Can I also get advice on the way I'm using avio_open_dyn_buff function to be able to stream over http? I've also attached my ffmpeg library versions below:
ffmpeg version N-70876-g294bb6c Copyright (c) 2000-2015 the FFmpeg developers
built with Apple LLVM version 6.0 (clang-600.0.56) (based on LLVM 3.5svn)
configuration: --prefix=/usr/local --enable-gpl --enable-postproc --enable-pthreads --enable-libmp3lame --enable-libtheora --enable-libx264 --enable-libvorbis --disable-mmx --disable-ssse3 --disable-armv5te --disable-armv6 --disable-neon --enable-shared --disable-static --disable-stripping
libavutil 54. 20.100 / 54. 20.100
libavcodec 56. 29.100 / 56. 29.100
libavformat 56. 26.101 / 56. 26.101
libavdevice 56. 4.100 / 56. 4.100
libavfilter 5. 13.101 / 5. 13.101
libswscale 3. 1.101 / 3. 1.101
libswresample 1. 1.100 / 1. 1.100
libpostproc 53. 3.100 / 53. 3.100
Hyper fast Audio and Video encoder
Valgrind analysis attached here because I exceeded stack overflow's character limit. http://pastebin.com/MPeRhjhN

Answer 1

Your code contains the following allocation:
uint8_t *buffer = (uint8_t *)av_malloc(numBytes*sizeof(uint8_t));
But has no matching av_free(). That's probably where you're losing 6MB/frame. av_free(outFrameYUV) won't free the memory inside the struct, just the struct itself. More correct would be to use av_frame_unref() instead of these individual av_free()s.
I also see you're calling this in your loop:
av_free(inFrame);
But the allocation for that frame is done outside your loop, so after the first run, it would presumably be dead. You want to allocate and free the frame both inside the loop. And for unreferencing, use av_frame_unref() instead of av_free().
I'd recommend running your program using asan or valgrind to detect more such issues, it'll keep track of what memory is accessed when and if incorrect, where it was free'ed.