shithub: mcfs

ref: 74533c8ff1ab84ff0949d52c3726ccfcfc1bf76a
dir: mcfs/matroska.c

View raw version
#include <u.h>
#include <libc.h>
#include <bio.h>
#include "common.h"
#include "ebml.h"
#include "packet.h"

typedef struct Ebml Ebml;

struct Ebml {
	Packetctx;

	packet_f fpacket;

	vlong timestampscale;
	vlong perframe;
	int tracknum;
	int tracktype;
	vlong codecdelay;
	char lang[8];
	char name[256];
};

static Packet packets[256];

#define getnumber(expid, dest) \
	if(el.id == expid){ \
		vlong x; \
		if(ebmlrawuint(f, sz, &x) < 0) \
			return -1; \
		dest = x; \
		left -= sz; \
		continue; \
	}

#define getsigned(expid, dest) \
	if(el.id == expid){ \
		vlong x; \
		if(ebmlrawsint(f, sz, &x) < 0) \
			return -1; \
		dest = x; \
		left -= sz; \
		continue; \
	}

#define getfloat(expid, dest) \
	if(el.id == expid){ \
		double x; \
		if(ebmlfloat(f, sz, &x) < 0) \
			return -1; \
		dest = x; \
		left -= sz; \
		continue; \
	}

#define getstring(expid, dest) \
	if(el.id == expid){ \
		n = min(sizeof(dest)-1, sz); \
		if(Bread(f, dest, n) != n) \
			return -1; \
		dest[n] = 0; \
		if(n != sz) \
			Bseek(f, sz-n, 1); \
		left -= sz; \
		continue; \
	}

#define getbytes(expid, dest) \
	if(el.id == expid){ \
		dest.data = malloc(sz); \
		if(Bread(f, dest.data, sz) != sz) \
			return -1; \
		dest.sz = sz; \
		left -= sz; \
		continue; \
	}

static char *
format(Ebml *e)
{
	static char t[16];
	char *s;
	int n;

	if(e->tracktype == Etracksubtitles){
		if(strcmp(e->codec.name, "S_TEXT/UTF8") == 0)
			return "srt";
		else if(strcmp(e->codec.name, "S_TEXT/ASS") == 0)
			return "ass";
	}else if(e->tracktype == Etrackaudio){
		if(strcmp(e->codec.name, "A_MPEG/L3") == 0)
			return "mp3";
	}else if(e->tracktype == Etrackvideo){
		if(strcmp(e->codec.name, "V_MPEG4/ISO/AVC") == 0)
			return "h264";
	}

	n = snprint(t, sizeof(t), "%s", e->codec.name+2);
	for(n -= 1; n >= 0; n--)
		t[n] = tolower(t[n]);
	if((s = strchr(t, '/')) != nil)
		*s = 0;

	return t;
}

static int
asispacket(Biobuf *out, Packetctx *ctx, Packet *p, int np, uvlong, int)
{
	int i;

	if(ctx->frid == 0)
		Bwrite(out, ctx->codec.priv.data, ctx->codec.priv.sz);

	for(i = 0; i < np; i++, p++)
		Bwrite(out, p->data, p->sz);
	ctx->frid++;

	return 0;
}

static int
initctx(Ebml *e, double duration)
{
	char *c;

	e->duration = duration;
	e->duration *= e->timestampscale;
	c = format(e);
	if(e->tracktype == Etrackvideo){
		e->fpacket = ivfpacket;
		if(strcmp(c, "vp9") == 0)
			e->fmt = FmtVp09;
		else if(strcmp(c, "vp8") == 0)
			e->fmt = FmtVp08;
		else if(strcmp(c, "h264") == 0)
			e->fmt = FmtAvc1;
		else if(strcmp(c, "av1") == 0)
			e->fmt = FmtAv01;
		else if(strcmp(c, "theora") == 0){
			e->fmt = FmtTheora;
			e->fpacket = oggpacket;
		}else
			goto err;
		return 0;
	}else if(e->tracktype == Etrackaudio){
		if(strcmp(c, "vorbis") == 0){
			e->fmt = FmtVorbis;
			e->fpacket = oggpacket;
		}else if(strcmp(c, "opus") == 0){
			e->fmt = FmtOpus;
			e->fpacket = oggpacket;
		}else if(strcmp(c, "mp3") == 0){
			e->fmt = FmtMp3;
			e->fpacket = asispacket;
		}else if(strcmp(c, "aac") == 0){
			e->fmt = FmtMp4a;
			e->fpacket = aacpacket;
		}else if(strcmp(c, "flac") == 0){
			e->fmt = FmtFlac;
			e->fpacket = asispacket;
		}else
			goto err;
		return 0;
	}else if(e->tracktype == Etracksubtitles){
		if(strcmp(c, "srt") == 0){
			e->fmt = FmtSrt;
			e->fpacket = srtpacket;
		}else if(strcmp(c, "ass") == 0){
			e->fmt = FmtAss;
			e->fpacket = asspacket;
		}else
			goto err;
		return 0;
	}

err:
	werrstr("don't know how to remux %s (track type %d)", e->codec.name, e->tracktype);

	return -1;
}

static void
trackinfo(Biobuf *o, Ebml *e)
{
	char *t;
	int rate;

	rate = e->audio.outsamplerate;
	if(rate == 0)
		rate = e->audio.samplerate;
	t = ebmltracktype(e->tracktype);
	Bprint(o, "%d\t%s\t%s\t", e->tracknum, t, format(e));
	if(e->tracktype == Etrackvideo)
		Bprint(o, "%d\t%d", e->video.width, e->video.height);
	else if(e->tracktype == Etrackaudio)
		Bprint(o, "%d\t%d\t%s\t%s", e->audio.channels, rate, e->lang, e->name);
	else if(e->tracktype == Etracksubtitles)
		Bprint(o, "%s\t%s", e->lang, e->name);
	else
		Bprint(o, "???");
	Bprint(o, "\n");
}

int
matroskarun(Biobuf *f)
{
	vlong left, n, sz, nsz, bufsz, track, off, packetsz, x, endtracks, bgend;
	int isebml, npackets, i, skipdata, lacing, refblock, key;
	uvlong ts, timestamp, timestampscale;
	double duration;
	s16int timecode;
	uchar *buf, *b;
	Ebml e, te;
	Elspec el;

	buf = nil;
	bufsz = 0;
	track = -1;
	timestamp = 0;
	left = (1ULL<<63)-1;
	endtracks = -1;
	te.tracknum = -1;
	timestampscale = 1000000;
	e.tracknum = -1;
	e.comp.algo = -1;
	duration = 0;
	skipdata = trackdump == Nodump;
	el.id = 0;
	refblock = 0;
	bgend = 0;
	npackets = 0;
	ts = 0;
	key = 0;
	for(isebml = 0; left != 0;){
		if(el.id == EBlockDuration)
			te.blockdur *= timestampscale;
		if(endtracks > 0 && left < endtracks && skipdata){
			/* early exit */
			left = 0;
			break;
		}

		off = Boffset(f);

		if(off >= bgend && !skipdata){
			if(npackets > 0){
				if(te.fpacket(&out, &te, packets, npackets, ts, key || !refblock) != 0)
					goto err;
				npackets = 0;
			}
		}

		n = ebmlel(f, left, &el, &sz);
		if(n < 0){
			werrstr("invalid ebml: %r at %#llx (size %lld)", off, sz);
			goto err;
		}
		if(el.type < 0)
			continue;
		if(n == 0){ /* eof */
			left = 0;
			break;
		}
		left -= n;

		if(el.id == EEBML){ /* EBML comes first */
			if(isebml != 0){
				werrstr("double EBML?");
				goto err;
			}
			isebml++;
		}else if(el.id == ESegment){
			if(sz > 0)
				left = sz;
			if(isebml != 1){
				werrstr("invalid ebml");
				goto err;
			}
			isebml++; /* make sure we don't see more segments */
			continue; /* go in */
		}else if(el.id == EInfo){ /* segment info */
			continue;
		}else if(el.id == ETracks){
			endtracks = left - sz; /* to skip early in case track dump not required */
			continue;
		}else if(el.id == ECluster){
			if(!skipdata) /* skip it entirely if no dump required */
				continue;
		}else if(el.id == ETrackEntry){ /* track entry */
			if(e.tracknum > 0){
				if(trackdump == Nodump)
					trackinfo(&out, &e);
				else if(trackdump == Dumpvideo && e.tracktype == Etrackvideo)
					trackdump = e.tracknum;
				else if(trackdump == Dumpaudio && e.tracktype == Etrackaudio)
					trackdump = e.tracknum;
			}
			if(e.tracknum == trackdump)
				memmove(&te, &e, sizeof(e));
			memset(&e, 0, sizeof(e));
			e.comp.algo = -1;
			e.timestampscale = timestampscale;
			continue;
		}else if(el.id == EVideo){
			continue;
		}else if(el.id == EAudio){
			e.audio.samplerate = 8000;
			e.audio.channels = 1;
			continue;
		}else if(el.id == EContentEncodings || el.id == EContentEncoding || el.id == EContentCompression){
			continue;
		}else if(el.id == EBlockGroup && !skipdata){
			refblock = 0;
			bgend = off+n+sz;
			npackets = 0;
			continue;
		}else if((el.id == ESimpleBlock || el.id == EBlock) && !skipdata){
			if(te.tracknum == -1)
				memmove(&te, &e, sizeof(e));
			n = ebmluint(f, sz, &track);
			if(n < 0){
				werrstr("block: %r");
				goto err;
			}
			left -= n;
			sz -= n;
			if(trackdump == Dumpvideo && e.tracktype == Etrackvideo)
				trackdump = te.tracknum;
			else if(trackdump == Dumpaudio && e.tracktype == Etrackaudio)
				trackdump = te.tracknum;
			if(track == trackdump && track == te.tracknum){
				if(te.fpacket == nil && initctx(&te, duration) != 0){
					werrstr("packet: %r");
					goto err;
				}

				if(bufsz < sz){
					buf = realloc(buf, sz);
					bufsz = sz;
				}
				if(Bread(f, buf, sz) != sz){
					werrstr("short read");
					goto err;
				}
				left -= sz;
				sz -= 3;
				timecode = buf[0]<<8 | buf[1];
				lacing = (buf[2] >> 1) & 3;
				key = buf[2] & 0x80;
				npackets = buf[3]+1;

				if(te.comp.algo == 3){ /* header stripping, need to put bytes back */
					if(te.comp.sz == 0)
						te.comp.algo = -1;
					else{
						nsz = sz + npackets*te.comp.sz;
						if(bufsz < nsz){
							buf = realloc(buf, nsz);
							bufsz = nsz;
						}
					}
				}else if(te.comp.algo >= 0){
					werrstr("unsupported: comp algo %d, please report a bug", te.comp.algo);
					goto err;
				}

				b = buf+3;
				switch(lacing){ /* lacing */
				case 0: /* none */
					packets[0].data = b;
					packets[0].sz = sz;
					npackets = 1;
					break;
				case 1: /* xiph */
					sz--;
					off = 4;
					for(i = 0; i < npackets-1; i++){
						packets[i].sz = 0;
						do{
							packets[i].sz += buf[off];
							sz--;
						}while(buf[off++] == 0xff);
					}
					for(i = 0; i < npackets-1; i++){
						packets[i].data = buf+off;
						off += packets[i].sz;
						sz -= packets[i].sz;
					}
					packets[i].data = buf+off;
					packets[i].sz = sz;
					break;
				case 2: /* fixed-size */
					sz--;
					if((sz % npackets) != 0){
						werrstr("invalid number of frames with fixed-size lacing");
						goto err;
					}
					packets[0].sz = sz / npackets;
					for(i = 0; i < npackets; i++){
						packets[i].data = buf+4 + i*packets[0].sz;
						packets[i].sz = packets[0].sz;
						sz -= packets[0].sz;
					}
					break;
				case 3: /* ebml */
					sz--;
					packetsz = 0;
					off = 4;
					for(i = 0; i < npackets-1; i++){
						n = (i == 0 ? ebmluintb : ebmlsintb)(buf+off, sz, &x);
						if(n < 0)
							goto err;
						packetsz += x;
						if(packetsz < 0){
							werrstr("invalid frame size %lld", packetsz);
							goto err;
						}
						packets[i].sz = packetsz;
						off += n;
						sz -= n;
					}
					for(i = 0; i < npackets-1; i++){
						if(packets[i].sz > sz){
							werrstr("frame %d/%d out of bounds: %d > %lld", i, npackets-1, packets[i].sz, sz);
							goto err;
						}
						packets[i].data = buf+off;
						off += packets[i].sz;
						sz -= packets[i].sz;
					}
					packets[i].data = buf+off;
					packets[i].sz = sz;
					break;
				}

				if(te.comp.algo == 3){ /* prepend stripped data to each packet */
					b = buf+bufsz;
					for(i = npackets-1; i >= 0; i--){
						b -= packets[i].sz;
						memmove(b, packets[i].data, packets[i].sz);
						b -= te.comp.sz;
						memmove(b, te.comp.data, te.comp.sz);
						packets[i].data = b;
						packets[i].sz += te.comp.sz;
					}
				}

				/* ns timestamp */
				ts = (timestamp + timecode) * timestampscale - te.codec.delay;
				if(el.id != EBlock){
					if(te.fpacket(&out, &te, packets, npackets, ts, key) != 0)
						goto err;
					npackets = 0;
				}
				continue;
			}
		}else
			getnumber(EReferenceBlock, refblock)
		else
			getnumber(ETimestampScale, timestampscale)
		else
			getfloat(ESamplingFrequency, e.audio.samplerate)
		else
			getfloat(EOutputSamplingFrequency, e.audio.outsamplerate)
		else
			getnumber(EChannels, e.audio.channels)
		else
			getnumber(EBitDepth, e.audio.bps)
		else
			getnumber(ETrackNumber, e.tracknum)
		else
			getnumber(ETrackType, e.tracktype)
		else
			getstring(ECodecID, e.codec.name)
		else
			getbytes(ECodecPrivate, e.codec.priv)
		else
			getnumber(ECodecDelay, e.codec.delay)
		else
			getnumber(EContentCompAlgo, e.comp.algo)
		else
			getbytes(EContentCompSettings, e.comp)
		else
			getnumber(EPixelWidth, e.video.width)
		else
			getnumber(EPixelHeight, e.video.height)
		else
			getnumber(ETimestamp, timestamp)
		else
			getnumber(EDefaultDuration, e.perframe)
		else
			getnumber(ESeekPreRoll, e.seekpreroll)
		else
			getfloat(EDuration, duration)
		else
			getnumber(ETrackUID, e.trackuid)
		else
			getsigned(EDiscardPadding, te.discardpad)
		else
			getnumber(EBlockDuration, te.blockdur)
		else
			getstring(ELanguage, e.lang)
		else
			getstring(EName, e.name)

		if(sz > 0){
			if(Bseek(f, sz, 1) < 0)
				return -1;
			left -= sz;
		}
	}

	if(isebml == 2 && left == 0){
		if(e.tracknum > 0){
			if(trackdump == Nodump)
				trackinfo(&out, &e);
		}
		return 0;
	}

err:
	return -1;
}