Google Git
Sign in
coral / imx-gst-plugins-base / 3db0d15f07ffb1edaf7daac1d5337a5c787cfeab / . / gst / typefind / gsttypefindfunctions.c
blob: 25c8645e600cb20ca9a39a3343d7796fc5d5848c [file] [log] [blame]
/* GStreamer
* Copyright (C) 2003 Benjamin Otte <in7y118@public.uni-hamburg.de>
* Copyright (C) 2005-2009 Tim-Philipp Müller <tim centricular net>
* Copyright (C) 2009 Sebastian Dröge <sebastian.droege@collabora.co.uk>
*
* gsttypefindfunctions.c: collection of various typefind functions
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
* Boston, MA 02110-1301, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <glib.h>
#include <glib/gprintf.h>
/* don't want to add gio xdgmime typefinder if gio was disabled via configure */
#ifdef HAVE_GIO
#include <gio/gio.h>
#define USE_GIO
#endif
#include <gst/gst.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>
#include <gst/pbutils/pbutils.h>
#include <gst/base/gstbytereader.h>
GST_DEBUG_CATEGORY_STATIC (type_find_debug);
#define GST_CAT_DEFAULT type_find_debug
/* DataScanCtx: helper for typefind functions that scan through data
* step-by-step, to avoid doing a peek at each and every offset */
#define DATA_SCAN_CTX_CHUNK_SIZE 4096
typedef struct
{
guint64 offset;
const guint8 *data;
gint size;
} DataScanCtx;
static inline void
data_scan_ctx_advance (GstTypeFind * tf, DataScanCtx * c, guint bytes_to_skip)
{
c->offset += bytes_to_skip;
if (G_LIKELY (c->size > bytes_to_skip)) {
c->size -= bytes_to_skip;
c->data += bytes_to_skip;
} else {
c->data += c->size;
c->size = 0;
}
}
static inline gboolean
data_scan_ctx_ensure_data (GstTypeFind * tf, DataScanCtx * c, gint min_len)
{
const guint8 *data;
guint64 len;
guint chunk_len = MAX (DATA_SCAN_CTX_CHUNK_SIZE, min_len);
if (G_LIKELY (c->size >= min_len))
return TRUE;
data = gst_type_find_peek (tf, c->offset, chunk_len);
if (G_LIKELY (data != NULL)) {
c->data = data;
c->size = chunk_len;
return TRUE;
}
/* if there's less than our chunk size, try to get as much as we can, but
* always at least min_len bytes (we might be typefinding the first buffer
* of the stream and not have as much data available as we'd like) */
len = gst_type_find_get_length (tf);
if (len > 0) {
len = CLAMP (len - c->offset, min_len, chunk_len);
} else {
len = min_len;
}
data = gst_type_find_peek (tf, c->offset, len);
if (data != NULL) {
c->data = data;
c->size = len;
return TRUE;
}
return FALSE;
}
static inline gboolean
data_scan_ctx_memcmp (GstTypeFind * tf, DataScanCtx * c, guint offset,
const gchar * data, guint len)
{
if (!data_scan_ctx_ensure_data (tf, c, offset + len))
return FALSE;
return (memcmp (c->data + offset, data, len) == 0);
}
/*** text/plain ***/
static gboolean xml_check_first_element (GstTypeFind * tf,
const gchar * element, guint elen, gboolean strict);
static gboolean sdp_check_header (GstTypeFind * tf);
static GstStaticCaps utf8_caps = GST_STATIC_CAPS ("text/plain");
#define UTF8_CAPS gst_static_caps_get(&utf8_caps)
static gboolean
utf8_type_find_have_valid_utf8_at_offset (GstTypeFind * tf, guint64 offset,
GstTypeFindProbability * prob)
{
const guint8 *data;
/* randomly decided values */
guint min_size = 16; /* minimum size */
guint size = 32 * 1024; /* starting size */
guint probability = 95; /* starting probability */
guint step = 10; /* how much we reduce probability in each
* iteration */
while (probability > step && size > min_size) {
data = gst_type_find_peek (tf, offset, size);
if (data) {
gchar *end;
gchar *start = (gchar *) data;
if (g_utf8_validate (start, size, (const gchar **) &end) || (end - start + 4 > size)) { /* allow last char to be cut off */
*prob = probability;
return TRUE;
}
*prob = 0;
return FALSE;
}
size /= 2;
probability -= step;
}
*prob = 0;
return FALSE;
}
static void
utf8_type_find (GstTypeFind * tf, gpointer unused)
{
GstTypeFindProbability start_prob, mid_prob;
guint64 length;
/* leave xml to the xml typefinders */
if (xml_check_first_element (tf, "", 0, TRUE))
return;
/* leave sdp to the sdp typefinders */
if (sdp_check_header (tf))
return;
/* check beginning of stream */
if (!utf8_type_find_have_valid_utf8_at_offset (tf, 0, &start_prob))
return;
GST_LOG ("start is plain text with probability of %u", start_prob);
/* POSSIBLE is the highest probability we ever return if we can't
* probe into the middle of the file and don't know its length */
length = gst_type_find_get_length (tf);
if (length == 0 || length == (guint64) - 1) {
gst_type_find_suggest (tf, MIN (start_prob, GST_TYPE_FIND_POSSIBLE),
UTF8_CAPS);
return;
}
if (length < 64 * 1024) {
gst_type_find_suggest (tf, start_prob, UTF8_CAPS);
return;
}
/* check middle of stream */
if (!utf8_type_find_have_valid_utf8_at_offset (tf, length / 2, &mid_prob))
return;
GST_LOG ("middle is plain text with probability of %u", mid_prob);
gst_type_find_suggest (tf, (start_prob + mid_prob) / 2, UTF8_CAPS);
}
/*** text/utf-16 and text/utf-32} ***/
/* While UTF-8 is unicode too, using text/plain for UTF-16 and UTF-32
is going to break stuff. */
typedef struct
{
size_t bomlen;
const char *const bom;
gboolean (*checker) (const guint8 *, gint, gint);
int boost;
int endianness;
} GstUnicodeTester;
static gboolean
check_utf16 (const guint8 * data, gint len, gint endianness)
{
GstByteReader br;
guint16 high, low;
low = high = 0;
if (len & 1)
return FALSE;
gst_byte_reader_init (&br, data, len);
while (len >= 2) {
/* test first for a single 16 bit value in the BMP */
if (endianness == G_BIG_ENDIAN)
high = gst_byte_reader_get_uint16_be_unchecked (&br);
else
high = gst_byte_reader_get_uint16_le_unchecked (&br);
if (high >= 0xD800 && high <= 0xDBFF) {
/* start of a surrogate pair */
if (len < 4)
return FALSE;
len -= 2;
if (endianness == G_BIG_ENDIAN)
low = gst_byte_reader_get_uint16_be_unchecked (&br);
else
low = gst_byte_reader_get_uint16_le_unchecked (&br);
if (low >= 0xDC00 && low <= 0xDFFF) {
/* second half of the surrogate pair */
} else
return FALSE;
} else {
if (high >= 0xDC00 && high <= 0xDFFF)
return FALSE;
}
len -= 2;
}
return TRUE;
}
static gboolean
check_utf32 (const guint8 * data, gint len, gint endianness)
{
if (len & 3)
return FALSE;
while (len > 3) {
guint32 v;
if (endianness == G_BIG_ENDIAN)
v = (data[0] << 24) | (data[1] << 16) | (data[2] << 8) | data[3];
else
v = (data[3] << 24) | (data[2] << 16) | (data[1] << 8) | data[0];
if (v >= 0x10FFFF)
return FALSE;
data += 4;
len -= 4;
}
return TRUE;
}
static void
unicode_type_find (GstTypeFind * tf, const GstUnicodeTester * tester,
guint n_tester, const char *media_type, gboolean require_bom)
{
size_t n;
gint len = 4;
const guint8 *data = gst_type_find_peek (tf, 0, len);
int prob = -1;
const gint max_scan_size = 256 * 1024;
int endianness = 0;
if (!data) {
len = 2;
data = gst_type_find_peek (tf, 0, len);
if (!data)
return;
}
/* find a large enough size that works */
while (len < max_scan_size) {
size_t newlen = len << 1;
const guint8 *newdata = gst_type_find_peek (tf, 0, newlen);
if (!newdata)
break;
len = newlen;
data = newdata;
}
for (n = 0; n < n_tester; ++n) {
int bom_boost = 0, tmpprob;
if (len >= tester[n].bomlen) {
if (!memcmp (data, tester[n].bom, tester[n].bomlen))
bom_boost = tester[n].boost;
}
if (require_bom && bom_boost == 0)
continue;
if (!(*tester[n].checker) (data, len, tester[n].endianness))
continue;
tmpprob = GST_TYPE_FIND_POSSIBLE - 20 + bom_boost;
if (tmpprob > prob) {
prob = tmpprob;
endianness = tester[n].endianness;
}
}
if (prob > 0) {
GST_DEBUG ("This is valid %s %s", media_type,
endianness == G_BIG_ENDIAN ? "be" : "le");
gst_type_find_suggest_simple (tf, prob, media_type,
"endianness", G_TYPE_INT, endianness, NULL);
}
}
static GstStaticCaps utf16_caps = GST_STATIC_CAPS ("text/utf-16");
#define UTF16_CAPS gst_static_caps_get(&utf16_caps)
static void
utf16_type_find (GstTypeFind * tf, gpointer unused)
{
static const GstUnicodeTester utf16tester[2] = {
{2, "\xff\xfe", check_utf16, 10, G_LITTLE_ENDIAN},
{2, "\xfe\xff", check_utf16, 20, G_BIG_ENDIAN},
};
unicode_type_find (tf, utf16tester, G_N_ELEMENTS (utf16tester),
"text/utf-16", TRUE);
}
static GstStaticCaps utf32_caps = GST_STATIC_CAPS ("text/utf-32");
#define UTF32_CAPS gst_static_caps_get(&utf32_caps)
static void
utf32_type_find (GstTypeFind * tf, gpointer unused)
{
static const GstUnicodeTester utf32tester[2] = {
{4, "\xff\xfe\x00\x00", check_utf32, 10, G_LITTLE_ENDIAN},
{4, "\x00\x00\xfe\xff", check_utf32, 20, G_BIG_ENDIAN}
};
unicode_type_find (tf, utf32tester, G_N_ELEMENTS (utf32tester),
"text/utf-32", TRUE);
}
/*** text/uri-list ***/
static GstStaticCaps uri_caps = GST_STATIC_CAPS ("text/uri-list");
#define URI_CAPS (gst_static_caps_get(&uri_caps))
#define BUFFER_SIZE 16 /* If the string is < 16 bytes we're screwed */
#define INC_BUFFER { \
pos++; \
if (pos == BUFFER_SIZE) { \
pos = 0; \
offset += BUFFER_SIZE; \
data = gst_type_find_peek (tf, offset, BUFFER_SIZE); \
if (data == NULL) return; \
} else { \
data++; \
} \
}
static void
uri_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, BUFFER_SIZE);
guint pos = 0;
guint offset = 0;
if (data) {
/* Search for # comment lines */
while (*data == '#') {
/* Goto end of line */
while (*data != '\n') {
INC_BUFFER;
}
INC_BUFFER;
}
if (!g_ascii_isalpha (*data)) {
/* Had a non alpha char - can't be uri-list */
return;
}
INC_BUFFER;
while (g_ascii_isalnum (*data)) {
INC_BUFFER;
}
if (*data != ':') {
/* First non alpha char is not a : */
return;
}
/* Get the next 2 bytes as well */
data = gst_type_find_peek (tf, offset + pos, 3);
if (data == NULL)
return;
if (data[1] != '/' && data[2] != '/') {
return;
}
gst_type_find_suggest (tf, GST_TYPE_FIND_LIKELY, URI_CAPS);
}
}
/*** application/itc ***/
static GstStaticCaps itc_caps = GST_STATIC_CAPS ("application/itc");
#define ITC_CAPS (gst_static_caps_get(&itc_caps))
static void
itc_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
guint8 magic[8] = { 0x00, 0x00, 0x01, 0x1C, 0x69, 0x74, 0x63, 0x68 };
guint8 preamble[4] = { 0x00, 0x00, 0x00, 0x02 };
guint8 artwork_marker[8] = { 0x00, 0x00, 0x00, 0x00, 0x61, 0x72, 0x74, 0x77 };
guint8 item_marker[4] = { 0x69, 0x74, 0x65, 0x6D };
GstTypeFindProbability itc_prob = GST_TYPE_FIND_NONE;
int i;
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 8)))
return;
if (memcmp (c.data, magic, 8))
return;
/* At least we found the right magic */
itc_prob = GST_TYPE_FIND_MINIMUM;
data_scan_ctx_advance (tf, &c, 8);
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 12)))
goto done;
/* Check preamble 3 consecutive times */
for (i = 0; i < 3; i++) {
if (memcmp (c.data, preamble, 4))
goto done;
data_scan_ctx_advance (tf, &c, 4);
}
itc_prob = GST_TYPE_FIND_POSSIBLE;
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 8)))
goto done;
if (memcmp (c.data, artwork_marker, 8))
goto done;
itc_prob = GST_TYPE_FIND_LIKELY;
data_scan_ctx_advance (tf, &c, 8);
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 256)))
goto done;
/* ...and 256 0x00 padding bytes on what looks like the header's end */
for (i = 0; i < 256; i++) {
if (c.data[i])
goto done;
}
itc_prob = GST_TYPE_FIND_NEARLY_CERTAIN;
data_scan_ctx_advance (tf, &c, 256);
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 8)))
goto done;
if (memcmp (c.data + 4, item_marker, 4))
goto done;
itc_prob = GST_TYPE_FIND_MAXIMUM;
done:
gst_type_find_suggest (tf, itc_prob, ITC_CAPS);
}
/*** application/x-hls ***/
static GstStaticCaps hls_caps = GST_STATIC_CAPS ("application/x-hls");
#define HLS_CAPS (gst_static_caps_get(&hls_caps))
/* See http://tools.ietf.org/html/draft-pantos-http-live-streaming-05 */
static void
hls_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
/* Minimum useful size is #EXTM3U\n + 1 tag + ':' = 30 bytes */
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 30)))
return;
if (memcmp (c.data, "#EXTM3U", 7))
return;
data_scan_ctx_advance (tf, &c, 7);
/* Check only the first 4KB */
while (c.offset < 4096) {
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 21)))
return;
/* Search for # comment lines */
if (c.data[0] == '#' && (memcmp (c.data, "#EXT-X-TARGETDURATION", 21) == 0
|| memcmp (c.data, "#EXT-X-STREAM-INF", 17) == 0
|| memcmp (c.data, "#EXT-X-MEDIA", 12) == 0)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, HLS_CAPS);
return;
}
data_scan_ctx_advance (tf, &c, 1);
}
}
/*** application/xml **********************************************************/
#define XML_BUFFER_SIZE 16
#define XML_INC_BUFFER { \
pos++; \
if (pos == XML_BUFFER_SIZE) { \
pos = 0; \
offset += XML_BUFFER_SIZE; \
data = gst_type_find_peek (tf, offset, XML_BUFFER_SIZE); \
if (data == NULL) return FALSE; \
} else { \
data++; \
} \
}
#define XML_INC_BUFFER_DATA { \
pos++; \
if (pos >= length) { \
return FALSE; \
} else { \
data++; \
} \
}
static gboolean
xml_check_first_element_from_data (const guint8 * data, guint length,
const gchar * element, guint elen, gboolean strict)
{
gboolean got_xmldec;
guint pos = 0;
g_return_val_if_fail (data != NULL, FALSE);
if (length <= 5)
return FALSE;
/* look for the XMLDec
* see XML spec 2.8, Prolog and Document Type Declaration
* http://www.w3.org/TR/2004/REC-xml-20040204/#sec-prolog-dtd */
got_xmldec = (memcmp (data, "<?xml", 5) == 0);
if (strict && !got_xmldec)
return FALSE;
/* skip XMLDec in any case if we've got one */
if (got_xmldec) {
pos += 5;
data += 5;
}
/* look for the first element, it has to be the requested element. Bail
* out if it is not within the first 4kB. */
while (pos < MIN (4096, length)) {
while (*data != '<' && pos < MIN (4096, length)) {
XML_INC_BUFFER_DATA;
}
XML_INC_BUFFER_DATA;
if (!g_ascii_isalpha (*data)) {
/* if not alphabetic, it's a PI or an element / attribute declaration
* like <?xxx or <!xxx */
XML_INC_BUFFER_DATA;
continue;
}
/* the first normal element, check if it's the one asked for */
if (pos + elen + 1 >= length)
return FALSE;
return (element && strncmp ((const char *) data, element, elen) == 0);
}
return FALSE;
}
static gboolean
xml_check_first_element (GstTypeFind * tf, const gchar * element, guint elen,
gboolean strict)
{
gboolean got_xmldec;
const guint8 *data;
guint offset = 0;
guint pos = 0;
data = gst_type_find_peek (tf, 0, XML_BUFFER_SIZE);
if (!data)
return FALSE;
/* look for the XMLDec
* see XML spec 2.8, Prolog and Document Type Declaration
* http://www.w3.org/TR/2004/REC-xml-20040204/#sec-prolog-dtd */
got_xmldec = (memcmp (data, "<?xml", 5) == 0);
if (strict && !got_xmldec)
return FALSE;
/* skip XMLDec in any case if we've got one */
if (got_xmldec) {
pos += 5;
data += 5;
}
/* look for the first element, it has to be the requested element. Bail
* out if it is not within the first 4kB. */
while (data && (offset + pos) < 4096) {
while (*data != '<' && (offset + pos) < 4096) {
XML_INC_BUFFER;
}
XML_INC_BUFFER;
if (!g_ascii_isalpha (*data)) {
/* if not alphabetic, it's a PI or an element / attribute declaration
* like <?xxx or <!xxx */
XML_INC_BUFFER;
continue;
}
/* the first normal element, check if it's the one asked for */
data = gst_type_find_peek (tf, offset + pos, elen + 1);
return (data && element && strncmp ((char *) data, element, elen) == 0);
}
return FALSE;
}
static GstStaticCaps generic_xml_caps = GST_STATIC_CAPS ("application/xml");
#define GENERIC_XML_CAPS (gst_static_caps_get(&generic_xml_caps))
static void
xml_type_find (GstTypeFind * tf, gpointer unused)
{
if (xml_check_first_element (tf, "", 0, TRUE)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MINIMUM, GENERIC_XML_CAPS);
}
}
/*** application/dash+xml ****************************************************/
static GstStaticCaps dash_caps = GST_STATIC_CAPS ("application/dash+xml");
#define DASH_CAPS gst_static_caps_get (&dash_caps)
static void
dash_mpd_type_find (GstTypeFind * tf, gpointer unused)
{
if (xml_check_first_element (tf, "MPD", 3, FALSE) ||
xml_check_first_element (tf, "mpd", 3, FALSE)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, DASH_CAPS);
}
}
/*** application/sdp *********************************************************/
static GstStaticCaps sdp_caps = GST_STATIC_CAPS ("application/sdp");
#define SDP_CAPS (gst_static_caps_get(&sdp_caps))
static gboolean
sdp_check_header (GstTypeFind * tf)
{
const guint8 *data;
data = gst_type_find_peek (tf, 0, 5);
if (!data)
return FALSE;
/* sdp must start with v=0[\r]\n */
if (memcmp (data, "v=0", 3))
return FALSE;
if (data[3] == '\r' && data[4] == '\n')
return TRUE;
if (data[3] == '\n')
return TRUE;
return FALSE;
}
static void
sdp_type_find (GstTypeFind * tf, gpointer unused)
{
if (sdp_check_header (tf))
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, SDP_CAPS);
}
/*** application/smil *********************************************************/
static GstStaticCaps smil_caps = GST_STATIC_CAPS ("application/smil");
#define SMIL_CAPS (gst_static_caps_get(&smil_caps))
static void
smil_type_find (GstTypeFind * tf, gpointer unused)
{
if (xml_check_first_element (tf, "smil", 4, FALSE)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, SMIL_CAPS);
}
}
/*** application/ttml+xml *****************************************************/
static GstStaticCaps ttml_xml_caps = GST_STATIC_CAPS ("application/ttml+xml");
#define TTML_XML_CAPS (gst_static_caps_get(&ttml_xml_caps))
static void
ttml_xml_type_find (GstTypeFind * tf, gpointer unused)
{
if (xml_check_first_element (tf, "tt", 2, FALSE)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, TTML_XML_CAPS);
}
}
/*** text/html ***/
static GstStaticCaps html_caps = GST_STATIC_CAPS ("text/html");
#define HTML_CAPS gst_static_caps_get (&html_caps)
static void
html_type_find (GstTypeFind * tf, gpointer unused)
{
const gchar *d, *data;
data = (const gchar *) gst_type_find_peek (tf, 0, 16);
if (!data)
return;
if (!g_ascii_strncasecmp (data, "<!DOCTYPE HTML", 14)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, HTML_CAPS);
} else if (xml_check_first_element (tf, "html", 4, FALSE)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, HTML_CAPS);
} else if ((d = memchr (data, '<', 16))) {
data = (const gchar *) gst_type_find_peek (tf, d - data, 6);
if (data && g_ascii_strncasecmp (data, "<html>", 6) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, HTML_CAPS);
}
}
}
/*** audio/midi ***/
static GstStaticCaps mid_caps = GST_STATIC_CAPS ("audio/midi");
#define MID_CAPS gst_static_caps_get(&mid_caps)
static void
mid_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, 4);
/* http://jedi.ks.uiuc.edu/~johns/links/music/midifile.html */
if (data && data[0] == 'M' && data[1] == 'T' && data[2] == 'h'
&& data[3] == 'd')
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, MID_CAPS);
}
/*** audio/mobile-xmf ***/
static GstStaticCaps mxmf_caps = GST_STATIC_CAPS ("audio/mobile-xmf");
#define MXMF_CAPS gst_static_caps_get(&mxmf_caps)
static void
mxmf_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = NULL;
/* Search FileId "XMF_" 4 bytes */
data = gst_type_find_peek (tf, 0, 4);
if (data && data[0] == 'X' && data[1] == 'M' && data[2] == 'F'
&& data[3] == '_') {
/* Search Format version "2.00" 4 bytes */
data = gst_type_find_peek (tf, 4, 4);
if (data && data[0] == '2' && data[1] == '.' && data[2] == '0'
&& data[3] == '0') {
/* Search TypeId 2 1 byte */
data = gst_type_find_peek (tf, 11, 1);
if (data && data[0] == 2) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, MXMF_CAPS);
}
}
}
}
/*** video/x-fli ***/
static GstStaticCaps flx_caps = GST_STATIC_CAPS ("video/x-fli");
#define FLX_CAPS gst_static_caps_get(&flx_caps)
static void
flx_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, 134);
if (data) {
/* check magic and the frame type of the first frame */
if ((data[4] == 0x11 || data[4] == 0x12 ||
data[4] == 0x30 || data[4] == 0x44) &&
data[5] == 0xaf &&
((data[132] == 0x00 || data[132] == 0xfa) && data[133] == 0xf1)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, FLX_CAPS);
}
return;
}
data = gst_type_find_peek (tf, 0, 6);
if (data) {
/* check magic only */
if ((data[4] == 0x11 || data[4] == 0x12 ||
data[4] == 0x30 || data[4] == 0x44) && data[5] == 0xaf) {
gst_type_find_suggest (tf, GST_TYPE_FIND_LIKELY, FLX_CAPS);
}
return;
}
}
/*** application/x-id3 ***/
static GstStaticCaps id3_caps = GST_STATIC_CAPS ("application/x-id3");
#define ID3_CAPS gst_static_caps_get(&id3_caps)
static void
id3v2_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, 10);
if (data && memcmp (data, "ID3", 3) == 0 &&
data[3] != 0xFF && data[4] != 0xFF &&
(data[6] & 0x80) == 0 && (data[7] & 0x80) == 0 &&
(data[8] & 0x80) == 0 && (data[9] & 0x80) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, ID3_CAPS);
}
}
static void
id3v1_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, -128, 3);
if (data && memcmp (data, "TAG", 3) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, ID3_CAPS);
}
}
/*** application/x-ape ***/
static GstStaticCaps apetag_caps = GST_STATIC_CAPS ("application/x-apetag");
#define APETAG_CAPS gst_static_caps_get(&apetag_caps)
static void
apetag_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data;
/* APEv1/2 at start of file */
data = gst_type_find_peek (tf, 0, 8);
if (data && !memcmp (data, "APETAGEX", 8)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, APETAG_CAPS);
return;
}
/* APEv1/2 at end of file */
data = gst_type_find_peek (tf, -32, 8);
if (data && !memcmp (data, "APETAGEX", 8)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, APETAG_CAPS);
return;
}
}
/*** audio/x-ttafile ***/
static GstStaticCaps tta_caps = GST_STATIC_CAPS ("audio/x-ttafile");
#define TTA_CAPS gst_static_caps_get(&tta_caps)
static void
tta_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, 3);
if (data) {
if (memcmp (data, "TTA", 3) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, TTA_CAPS);
return;
}
}
}
/*** audio/x-flac ***/
static GstStaticCaps flac_caps = GST_STATIC_CAPS ("audio/x-flac");
#define FLAC_CAPS (gst_static_caps_get(&flac_caps))
static void
flac_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 4)))
return;
/* standard flac (also old/broken flac-in-ogg with an initial 4-byte marker
* packet and without the usual packet framing) */
if (memcmp (c.data, "fLaC", 4) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, FLAC_CAPS);
return;
}
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 6)))
return;
/* flac-in-ogg, see http://flac.sourceforge.net/ogg_mapping.html */
if (memcmp (c.data, "\177FLAC\001", 6) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, FLAC_CAPS);
return;
}
/* disabled because it happily typefinds /dev/urandom as audio/x-flac, and
* because I yet have to see header-less flac in the wild */
#if 0
/* flac without headers (subset format) */
/* 64K should be enough */
while (c.offset < (64 * 1024)) {
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 4)))
break;
/* look for frame header,
* http://flac.sourceforge.net/format.html#frame_header
*/
if (c.data[0] == 0xff && (c.data[1] >> 2) == 0x3e) {
/* bit 15 in the header must be 0 */
if (((c.data[1] >> 1) & 0x01) == 0x01)
goto advance;
/* blocksize must be != 0x00 */
if ((c.data[2] >> 4) == 0x00)
goto advance;
/* samplerate must be != 0x0f */
if ((c.data[2] & 0x0f) == 0x0f)
goto advance;
/* also 0 is invalid, as it means get the info from the header and we
* don't have headers if we are here */
if ((c.data[2] & 0x0f) == 0x00)
goto advance;
/* channel assignment must be < 11 */
if ((c.data[3] >> 4) >= 11)
goto advance;
/* sample size must be != 0x07 and != 0x05 */
if (((c.data[3] >> 1) & 0x07) == 0x07)
goto advance;
if (((c.data[3] >> 1) & 0x07) == 0x05)
goto advance;
/* also 0 is invalid, as it means get the info from the header and we
* don't have headers if we are here */
if (((c.data[3] >> 1) & 0x07) == 0x00)
goto advance;
/* next bit must be 0 */
if ((c.data[3] & 0x01) == 0x01)
goto advance;
/* FIXME: shouldn't we include the crc check ? */
GST_DEBUG ("Found flac without headers at %d", (gint) c.offset);
gst_type_find_suggest (tf, GST_TYPE_FIND_POSSIBLE, FLAC_CAPS);
return;
}
advance:
data_scan_ctx_advance (tf, &c, 1);
}
#endif
}
/* TODO: we could probably make a generic function for this.. */
static gint
aac_type_find_scan_loas_frames_ep (GstTypeFind * tf, DataScanCtx * scan_ctx,
gint max_frames)
{
DataScanCtx c = *scan_ctx;
guint16 snc;
guint len;
gint count = 0;
do {
if (!data_scan_ctx_ensure_data (tf, &c, 5))
break;
/* EPAudioSyncStream */
len = ((c.data[2] & 0x0f) << 9) | (c.data[3] << 1) |
((c.data[4] & 0x80) >> 7);
if (len == 0 || !data_scan_ctx_ensure_data (tf, &c, len + 2)) {
GST_DEBUG ("Wrong sync or next frame not within reach, len=%u", len);
break;
}
/* check length of frame */
snc = GST_READ_UINT16_BE (c.data + len);
if (snc != 0x4de1) {
GST_DEBUG ("No sync found at 0x%" G_GINT64_MODIFIER "x", c.offset + len);
break;
}
++count;
GST_DEBUG ("Found LOAS syncword #%d at offset 0x%" G_GINT64_MODIFIER "x, "
"framelen %u", count, c.offset, len);
data_scan_ctx_advance (tf, &c, len);
} while (count < max_frames && (c.offset - scan_ctx->offset) < 64 * 1024);
GST_DEBUG ("found %d consecutive frames", count);
return count;
}
static gint
aac_type_find_scan_loas_frames (GstTypeFind * tf, DataScanCtx * scan_ctx,
gint max_frames)
{
DataScanCtx c = *scan_ctx;
guint16 snc;
guint len;
gint count = 0;
do {
if (!data_scan_ctx_ensure_data (tf, &c, 3))
break;
/* AudioSyncStream */
len = ((c.data[1] & 0x1f) << 8) | c.data[2];
/* add size of sync stream header */
len += 3;
if (len == 0 || !data_scan_ctx_ensure_data (tf, &c, len + 2)) {
GST_DEBUG ("Wrong sync or next frame not within reach, len=%u", len);
break;
}
/* check length of frame */
snc = GST_READ_UINT16_BE (c.data + len);
if ((snc & 0xffe0) != 0x56e0) {
GST_DEBUG ("No sync found at 0x%" G_GINT64_MODIFIER "x", c.offset + len);
break;
}
++count;
GST_DEBUG ("Found LOAS syncword #%d at offset 0x%" G_GINT64_MODIFIER "x, "
"framelen %u", count, c.offset, len);
data_scan_ctx_advance (tf, &c, len);
} while (count < max_frames && (c.offset - scan_ctx->offset) < 64 * 1024);
GST_DEBUG ("found %d consecutive frames", count);
return count;
}
/*** audio/mpeg version 2, 4 ***/
static GstStaticCaps aac_caps = GST_STATIC_CAPS ("audio/mpeg, "
"mpegversion = (int) { 2, 4 }, framed = (bool) false");
#define AAC_CAPS (gst_static_caps_get(&aac_caps))
#define AAC_AMOUNT (4096)
static void
aac_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
GstTypeFindProbability best_probability = GST_TYPE_FIND_NONE;
GstCaps *best_caps = NULL;
guint best_count = 0;
while (c.offset < AAC_AMOUNT) {
guint snc, len, offset, i;
/* detect adts header or adif header.
* The ADIF header is 4 bytes, that should be OK. The ADTS header, on
* the other hand, is 14 bits only, so we require one valid frame with
* again a valid syncpoint on the next one (28 bits) for certainty. We
* require 4 kB, which is quite a lot, since frames are generally 200-400
* bytes.
* LOAS has 2 possible syncwords, which are 11 bits and 16 bits long.
* The following stream syntax depends on which one is found.
*/
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 6)))
break;
snc = GST_READ_UINT16_BE (c.data);
if (G_UNLIKELY ((snc & 0xfff6) == 0xfff0)) {
/* ADTS header - find frame length */
GST_DEBUG ("Found one ADTS syncpoint at offset 0x%" G_GINT64_MODIFIER
"x, tracing next...", c.offset);
len = ((c.data[3] & 0x03) << 11) |
(c.data[4] << 3) | ((c.data[5] & 0xe0) >> 5);
if (len == 0 || !data_scan_ctx_ensure_data (tf, &c, len + 6)) {
GST_DEBUG ("Wrong sync or next frame not within reach, len=%u", len);
goto next;
}
offset = len;
/* check if there's a second ADTS frame */
snc = GST_READ_UINT16_BE (c.data + offset);
if ((snc & 0xfff6) == 0xfff0) {
GstCaps *caps;
guint mpegversion, sample_freq_idx, channel_config, profile_idx, rate;
guint8 audio_config[2];
mpegversion = (c.data[1] & 0x08) ? 2 : 4;
profile_idx = c.data[2] >> 6;
sample_freq_idx = ((c.data[2] & 0x3c) >> 2);
channel_config = ((c.data[2] & 0x01) << 2) + (c.data[3] >> 6);
GST_DEBUG ("Found second ADTS-%d syncpoint at offset 0x%"
G_GINT64_MODIFIER "x, framelen %u", mpegversion, c.offset, len);
/* 0xd and 0xe are reserved. 0xf means the sample frequency is directly
* specified in the header, but that's not allowed for ADTS */
if (sample_freq_idx > 0xc) {
GST_DEBUG ("Unexpected sample frequency index %d or wrong sync",
sample_freq_idx);
goto next;
}
rate = gst_codec_utils_aac_get_sample_rate_from_index (sample_freq_idx);
GST_LOG ("ADTS: profile=%u, rate=%u", profile_idx, rate);
/* The ADTS frame header is slightly different from the
* AudioSpecificConfig defined for the MPEG-4 container, so we just
* construct enough of it for getting the level here. */
/* ADTS counts profiles from 0 instead of 1 to save bits */
audio_config[0] = (profile_idx + 1) << 3;
audio_config[0] |= (sample_freq_idx >> 1) & 0x7;
audio_config[1] = (sample_freq_idx & 0x1) << 7;
audio_config[1] |= (channel_config & 0xf) << 3;
caps = gst_caps_new_simple ("audio/mpeg",
"framed", G_TYPE_BOOLEAN, FALSE,
"mpegversion", G_TYPE_INT, mpegversion,
"stream-format", G_TYPE_STRING, "adts", NULL);
gst_codec_utils_aac_caps_set_level_and_profile (caps, audio_config, 2);
/* add rate and number of channels if we can */
if (channel_config != 0 && channel_config <= 7) {
const guint channels_map[] = { 0, 1, 2, 3, 4, 5, 6, 8 };
gst_caps_set_simple (caps, "channels", G_TYPE_INT,
channels_map[channel_config], "rate", G_TYPE_INT, rate, NULL);
}
/* length of the second ADTS frame */
len = ((c.data[offset + 3] & 0x03) << 11) |
(c.data[offset + 4] << 3) | ((c.data[offset + 5] & 0xe0) >> 5);
if (len == 0 || !data_scan_ctx_ensure_data (tf, &c, offset + len + 6)) {
GST_DEBUG ("Wrong sync or next frame not within reach, len=%u", len);
gst_type_find_suggest (tf, GST_TYPE_FIND_LIKELY, caps);
} else {
offset += len;
/* find more aac sync to select correctly */
/* check if there's a third/fourth/fifth/sixth ADTS frame, if there is a sixth frame, set probability to maximum:100% */
for (i = 3; i <= 6; i++) {
len = ((c.data[offset + 3] & 0x03) << 11) |
(c.data[offset + 4] << 3) | ((c.data[offset + 5] & 0xe0) >> 5);
if (len == 0
|| !data_scan_ctx_ensure_data (tf, &c, offset + len + 6)) {
GST_DEBUG ("Wrong sync or next frame not within reach, len=%u",
len);
break;
}
snc = GST_READ_UINT16_BE (c.data + offset);
if ((snc & 0xfff6) == 0xfff0) {
GST_DEBUG ("Find %und Sync..probability is %u ", i,
GST_TYPE_FIND_LIKELY + 5 * (i - 2));
offset += len;
} else {
break;
}
}
gst_type_find_suggest (tf, GST_TYPE_FIND_LIKELY + 5 * (i - 3), caps);
}
gst_caps_unref (caps);
break;
}
GST_DEBUG ("No next frame found... (should have been at 0x%x)", len);
} else if (G_UNLIKELY ((snc & 0xffe0) == 0x56e0 || snc == 0x4de1)) {
gint count;
/* LOAS frame */
GST_INFO ("Possible LOAS syncword at offset 0x%" G_GINT64_MODIFIER
"x, scanning for more frames...", c.offset);
if (snc == 0x4de1)
count = aac_type_find_scan_loas_frames_ep (tf, &c, 20);
else
count = aac_type_find_scan_loas_frames (tf, &c, 20);
if (count >= 3 && count > best_count) {
gst_caps_replace (&best_caps, NULL);
best_caps = gst_caps_new_simple ("audio/mpeg",
"framed", G_TYPE_BOOLEAN, FALSE,
"mpegversion", G_TYPE_INT, 4,
"stream-format", G_TYPE_STRING, "loas", NULL);
best_count = count;
best_probability = GST_TYPE_FIND_POSSIBLE - 10 + count * 3;
if (best_probability >= GST_TYPE_FIND_LIKELY)
break;
}
} else if (!memcmp (c.data, "ADIF", 4)) {
/* ADIF header */
gst_type_find_suggest_simple (tf, GST_TYPE_FIND_LIKELY, "audio/mpeg",
"framed", G_TYPE_BOOLEAN, FALSE, "mpegversion", G_TYPE_INT, 4,
"stream-format", G_TYPE_STRING, "adif", NULL);
break;
}
next:
data_scan_ctx_advance (tf, &c, 1);
}
if (best_probability > GST_TYPE_FIND_NONE) {
gst_type_find_suggest (tf, best_probability, best_caps);
gst_caps_unref (best_caps);
}
}
/*** audio/mpeg version 1 ***/
/*
* The chance that random data is identified as a valid mp3 header is 63 / 2^18
* (0.024%) per try. This makes the function for calculating false positives
* 1 - (1 - ((63 / 2 ^18) ^ GST_MP3_TYPEFIND_MIN_HEADERS)) ^ buffersize)
* This has the following probabilities of false positives:
* datasize MIN_HEADERS
* (bytes) 1 2 3 4
* 4096 62.6% 0.02% 0% 0%
* 16384 98% 0.09% 0% 0%
* 1 MiB 100% 5.88% 0% 0%
* 1 GiB 100% 100% 1.44% 0%
* 1 TiB 100% 100% 100% 0.35%
* This means that the current choice (3 headers by most of the time 4096 byte
* buffers is pretty safe for now.
*
* The max. size of each frame is 1440 bytes, which means that for N frames to
* be detected, we need 1440 * GST_MP3_TYPEFIND_MIN_HEADERS + 3 bytes of data.
* Assuming we step into the stream right after the frame header, this
* means we need 1440 * (GST_MP3_TYPEFIND_MIN_HEADERS + 1) - 1 + 3 bytes
* of data (5762) to always detect any mp3.
*/
static const guint mp3types_bitrates[2][3][16] =
{ {{0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
{0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
{0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}},
{{0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}},
};
static const guint mp3types_freqs[3][3] = { {11025, 12000, 8000},
{22050, 24000, 16000},
{44100, 48000, 32000}
};
static inline guint
mp3_type_frame_length_from_header (guint32 header, guint * put_layer,
guint * put_channels, guint * put_bitrate, guint * put_samplerate,
gboolean * may_be_free_format, gint possible_free_framelen)
{
guint bitrate, layer, length, mode, samplerate, version, channels;
if ((header & 0xffe00000) != 0xffe00000)
return 0;
/* we don't need extension, copyright, original or
* emphasis for the frame length */
header >>= 6;
/* mode */
mode = header & 0x3;
header >>= 3;
/* padding */
length = header & 0x1;
header >>= 1;
/* sampling frequency */
samplerate = header & 0x3;
if (samplerate == 3)
return 0;
header >>= 2;
/* bitrate index */
bitrate = header & 0xF;
if (bitrate == 0 && possible_free_framelen == -1) {
GST_LOG ("Possibly a free format mp3 - signaling");
*may_be_free_format = TRUE;
}
if (bitrate == 15 || (bitrate == 0 && possible_free_framelen == -1))
return 0;
/* ignore error correction, too */
header >>= 5;
/* layer */
layer = 4 - (header & 0x3);
if (layer == 4)
return 0;
header >>= 2;
/* version 0=MPEG2.5; 2=MPEG2; 3=MPEG1 */
version = header & 0x3;
if (version == 1)
return 0;
/* lookup */
channels = (mode == 3) ? 1 : 2;
samplerate = mp3types_freqs[version > 0 ? version - 1 : 0][samplerate];
if (bitrate == 0) {
/* possible freeform mp3 */
if (layer == 1) {
length *= 4;
length += possible_free_framelen;
bitrate = length * samplerate / 48000;
} else {
length += possible_free_framelen;
bitrate = length * samplerate /
((layer == 3 && version != 3) ? 72000 : 144000);
}
/* freeform mp3 should have a higher-than-usually-allowed bitrate */
GST_LOG ("calculated bitrate: %u, max usually: %u", bitrate,
mp3types_bitrates[version == 3 ? 0 : 1][layer - 1][14]);
if (bitrate < mp3types_bitrates[version == 3 ? 0 : 1][layer - 1][14])
return 0;
} else {
/* calculating */
bitrate = mp3types_bitrates[version == 3 ? 0 : 1][layer - 1][bitrate];
if (layer == 1) {
length = ((12000 * bitrate / samplerate) + length) * 4;
} else {
length += ((layer == 3
&& version != 3) ? 72000 : 144000) * bitrate / samplerate;
}
}
GST_LOG ("mp3typefind: calculated mp3 frame length of %u bytes", length);
GST_LOG
("mp3typefind: samplerate = %u - bitrate = %u - layer = %u - version = %u"
" - channels = %u", samplerate, bitrate, layer, version, channels);
if (put_layer)
*put_layer = layer;
if (put_channels)
*put_channels = channels;
if (put_bitrate)
*put_bitrate = bitrate;
if (put_samplerate)
*put_samplerate = samplerate;
return length;
}
static GstStaticCaps mp3_caps = GST_STATIC_CAPS ("audio/mpeg, "
"mpegversion = (int) 1, layer = (int) [ 1, 3 ]");
#define MP3_CAPS (gst_static_caps_get(&mp3_caps))
/*
* random values for typefinding
* if no more data is available, we will return a probability of
* (found_headers/TRY_HEADERS) * (MAXIMUM * (TRY_SYNC - bytes_skipped)
* / TRY_SYNC)
* if found_headers >= MIN_HEADERS
*/
#define GST_MP3_TYPEFIND_MIN_HEADERS (2)
#define GST_MP3_TYPEFIND_TRY_HEADERS (5)
#define GST_MP3_TYPEFIND_TRY_SYNC (GST_TYPE_FIND_MAXIMUM * 100) /* 10kB */
#define GST_MP3_TYPEFIND_SYNC_SIZE (2048)
#define GST_MP3_WRONG_HEADER (10)
static void
mp3_type_find_at_offset (GstTypeFind * tf, guint64 start_off,
guint * found_layer, GstTypeFindProbability * found_prob)
{
const guint8 *data = NULL;
const guint8 *data_end = NULL;
guint size;
guint64 skipped;
gint last_free_offset = -1;
gint last_free_framelen = -1;
gboolean headerstart = TRUE;
*found_layer = 0;
*found_prob = 0;
size = 0;
skipped = 0;
while (skipped < GST_MP3_TYPEFIND_TRY_SYNC) {
if (size <= 0) {
size = GST_MP3_TYPEFIND_SYNC_SIZE * 2;
do {
size /= 2;
data = gst_type_find_peek (tf, skipped + start_off, size);
} while (size > 10 && !data);
if (!data)
break;
data_end = data + size;
}
if (*data == 0xFF) {
const guint8 *head_data = NULL;
guint layer = 0, bitrate, samplerate, channels;
guint found = 0; /* number of valid headers found */
guint64 offset = skipped;
gboolean changed = FALSE;
while (found < GST_MP3_TYPEFIND_TRY_HEADERS) {
guint32 head;
guint length;
guint prev_layer = 0;
guint prev_channels = 0, prev_samplerate = 0;
gboolean free = FALSE;
if ((gint64) (offset - skipped + 4) >= 0 &&
data + offset - skipped + 4 < data_end) {
head_data = data + offset - skipped;
} else {
head_data = gst_type_find_peek (tf, offset + start_off, 4);
}
if (!head_data)
break;
head = GST_READ_UINT32_BE (head_data);
if (!(length = mp3_type_frame_length_from_header (head, &layer,
&channels, &bitrate, &samplerate, &free,
last_free_framelen))) {
if (free) {
if (last_free_offset == -1)
last_free_offset = offset;
else {
last_free_framelen = offset - last_free_offset;
offset = last_free_offset;
continue;
}
} else {
last_free_framelen = -1;
}
/* Mark the fact that we didn't find a valid header at the beginning */
if (found == 0)
headerstart = FALSE;
GST_LOG ("%d. header at offset %" G_GUINT64_FORMAT
" (0x%" G_GINT64_MODIFIER "x) was not an mp3 header "
"(possibly-free: %s)", found + 1, start_off + offset,
start_off + offset, free ? "yes" : "no");
break;
}
if ((prev_layer && prev_layer != layer) ||
/* (prev_bitrate && prev_bitrate != bitrate) || <-- VBR */
(prev_samplerate && prev_samplerate != samplerate) ||
(prev_channels && prev_channels != channels)) {
/* this means an invalid property, or a change, which might mean
* that this is not a mp3 but just a random bytestream. It could
* be a freaking funky encoded mp3 though. We'll just not count
* this header*/
if (prev_layer)
changed = TRUE;
prev_layer = layer;
prev_channels = channels;
prev_samplerate = samplerate;
} else {
found++;
GST_LOG ("found %d. header at offset %" G_GUINT64_FORMAT " (0x%"
G_GINT64_MODIFIER "X)", found, start_off + offset,
start_off + offset);
}
offset += length;
}
g_assert (found <= GST_MP3_TYPEFIND_TRY_HEADERS);
if (found != 0 && head_data == NULL &&
gst_type_find_peek (tf, offset + start_off - 1, 1) == NULL)
/* Incomplete last frame - don't count it. */
found--;
if (found == GST_MP3_TYPEFIND_TRY_HEADERS ||
(found >= GST_MP3_TYPEFIND_MIN_HEADERS && head_data == NULL)) {
/* we can make a valid guess */
guint probability = found * GST_TYPE_FIND_MAXIMUM *
(GST_MP3_TYPEFIND_TRY_SYNC - skipped) /
GST_MP3_TYPEFIND_TRY_HEADERS / GST_MP3_TYPEFIND_TRY_SYNC;
if (!headerstart
&& probability > (GST_TYPE_FIND_MINIMUM + GST_MP3_WRONG_HEADER))
probability -= GST_MP3_WRONG_HEADER;
if (probability < GST_TYPE_FIND_MINIMUM)
probability = GST_TYPE_FIND_MINIMUM;
if (start_off > 0)
probability /= 2;
if (!changed)
probability = (probability + GST_TYPE_FIND_MAXIMUM) / 2;
GST_INFO
("audio/mpeg calculated %u = %u * %u / %u * (%u - %"
G_GUINT64_FORMAT ") / %u", probability, GST_TYPE_FIND_MAXIMUM,
found, GST_MP3_TYPEFIND_TRY_HEADERS, GST_MP3_TYPEFIND_TRY_SYNC,
(guint64) skipped, GST_MP3_TYPEFIND_TRY_SYNC);
/* make sure we're not id3 tagged */
head_data = gst_type_find_peek (tf, -128, 3);
if (head_data && (memcmp (head_data, "TAG", 3) == 0)) {
probability = 0;
}
g_assert (probability <= GST_TYPE_FIND_MAXIMUM);
*found_prob = probability;
if (probability > 0)
*found_layer = layer;
return;
}
}
data++;
skipped++;
size--;
}
}
static void
mp3_type_find (GstTypeFind * tf, gpointer unused)
{
GstTypeFindProbability prob, mid_prob;
const guint8 *data;
guint layer, mid_layer;
guint64 length;
mp3_type_find_at_offset (tf, 0, &layer, &prob);
length = gst_type_find_get_length (tf);
if (length == 0 || length == (guint64) - 1) {
if (prob != 0)
goto suggest;
return;
}
/* if we're pretty certain already, skip the additional check */
if (prob >= GST_TYPE_FIND_LIKELY)
goto suggest;
mp3_type_find_at_offset (tf, length / 2, &mid_layer, &mid_prob);
if (mid_prob > 0) {
if (prob == 0) {
GST_LOG ("detected audio/mpeg only in the middle (p=%u)", mid_prob);
layer = mid_layer;
prob = mid_prob;
goto suggest;
}
if (layer != mid_layer) {
GST_WARNING ("audio/mpeg layer discrepancy: %u vs. %u", layer, mid_layer);
return; /* FIXME: or should we just go with the one in the middle? */
}
/* detected mpeg audio both in middle of the file and at the start */
prob = (prob + mid_prob) / 2;
goto suggest;
}
/* a valid header right at the start makes it more likely
* that this is actually plain mpeg-1 audio */
if (prob > 0) {
data = gst_type_find_peek (tf, 0, 4); /* use min. frame size? */
if (data && mp3_type_frame_length_from_header (GST_READ_UINT32_BE (data),
&layer, NULL, NULL, NULL, NULL, 0) != 0) {
prob = MIN (prob + 10, GST_TYPE_FIND_MAXIMUM);
}
}
if (prob > 0)
goto suggest;
return;
suggest:
{
g_return_if_fail (layer >= 1 && layer <= 3);
gst_type_find_suggest_simple (tf, prob, "audio/mpeg",
"mpegversion", G_TYPE_INT, 1, "layer", G_TYPE_INT, layer,
"parsed", G_TYPE_BOOLEAN, FALSE, NULL);
}
}
/*** audio/x-musepack ***/
static GstStaticCaps musepack_caps =
GST_STATIC_CAPS ("audio/x-musepack, streamversion= (int) { 7, 8 }");
#define MUSEPACK_CAPS (gst_static_caps_get(&musepack_caps))
static void
musepack_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, 4);
GstTypeFindProbability prop = GST_TYPE_FIND_MINIMUM;
gint streamversion = -1;
if (data && memcmp (data, "MP+", 3) == 0) {
streamversion = 7;
if ((data[3] & 0x7f) == 7) {
prop = GST_TYPE_FIND_MAXIMUM;
} else {
prop = GST_TYPE_FIND_LIKELY + 10;
}
} else if (data && memcmp (data, "MPCK", 4) == 0) {
streamversion = 8;
prop = GST_TYPE_FIND_MAXIMUM;
}
if (streamversion != -1) {
gst_type_find_suggest_simple (tf, prop, "audio/x-musepack",
"streamversion", G_TYPE_INT, streamversion, NULL);
}
}
/*** audio/x-ac3 ***/
/* FIXME 0.11: should be audio/ac3, but isn't for backwards compatibility */
static GstStaticCaps ac3_caps = GST_STATIC_CAPS ("audio/x-ac3");
#define AC3_CAPS (gst_static_caps_get(&ac3_caps))
static GstStaticCaps eac3_caps = GST_STATIC_CAPS ("audio/x-eac3");
#define EAC3_CAPS (gst_static_caps_get(&eac3_caps))
struct ac3_frmsize
{
unsigned short bit_rate;
unsigned short frm_size[3];
};
static const struct ac3_frmsize ac3_frmsizecod_tbl[] = {
{32, {64, 69, 96}},
{32, {64, 70, 96}},
{40, {80, 87, 120}},
{40, {80, 88, 120}},
{48, {96, 104, 144}},
{48, {96, 105, 144}},
{56, {112, 121, 168}},
{56, {112, 122, 168}},
{64, {128, 139, 192}},
{64, {128, 140, 192}},
{80, {160, 174, 240}},
{80, {160, 175, 240}},
{96, {192, 208, 288}},
{96, {192, 209, 288}},
{112, {224, 243, 336}},
{112, {224, 244, 336}},
{128, {256, 278, 384}},
{128, {256, 279, 384}},
{160, {320, 348, 480}},
{160, {320, 349, 480}},
{192, {384, 417, 576}},
{192, {384, 418, 576}},
{224, {448, 487, 672}},
{224, {448, 488, 672}},
{256, {512, 557, 768}},
{256, {512, 558, 768}},
{320, {640, 696, 960}},
{320, {640, 697, 960}},
{384, {768, 835, 1152}},
{384, {768, 836, 1152}},
{448, {896, 975, 1344}},
{448, {896, 976, 1344}},
{512, {1024, 1114, 1536}},
{512, {1024, 1115, 1536}},
{576, {1152, 1253, 1728}},
{576, {1152, 1254, 1728}},
{640, {1280, 1393, 1920}},
{640, {1280, 1394, 1920}}
};
static void
ac3_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
/* Search for an ac3 frame; not necessarily right at the start, but give it
* a lower probability if not found right at the start. Check that the
* frame is followed by a second frame at the expected offset.
* We could also check the two ac3 CRCs, but we don't do that right now */
while (c.offset < 1024) {
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 5)))
break;
if (c.data[0] == 0x0b && c.data[1] == 0x77) {
guint bsid = c.data[5] >> 3;
if (bsid <= 8) {
/* ac3 */
guint fscod = c.data[4] >> 6;
guint frmsizecod = c.data[4] & 0x3f;
if (fscod < 3 && frmsizecod < 38) {
DataScanCtx c_next = c;
guint frame_size;
frame_size = ac3_frmsizecod_tbl[frmsizecod].frm_size[fscod];
GST_LOG ("possible AC3 frame sync at offset %"
G_GUINT64_FORMAT ", size=%u", c.offset, frame_size);
if (data_scan_ctx_ensure_data (tf, &c_next, (frame_size * 2) + 5)) {
data_scan_ctx_advance (tf, &c_next, frame_size * 2);
if (c_next.data[0] == 0x0b && c_next.data[1] == 0x77) {
fscod = c_next.data[4] >> 6;
frmsizecod = c_next.data[4] & 0x3f;
if (fscod < 3 && frmsizecod < 38) {
GstTypeFindProbability prob;
GST_LOG ("found second AC3 frame (size=%u), looks good",
ac3_frmsizecod_tbl[frmsizecod].frm_size[fscod]);
if (c.offset == 0)
prob = GST_TYPE_FIND_MAXIMUM;
else
prob = GST_TYPE_FIND_NEARLY_CERTAIN;
gst_type_find_suggest (tf, prob, AC3_CAPS);
return;
}
} else {
GST_LOG ("no second AC3 frame found, false sync");
}
}
}
} else if (bsid <= 16 && bsid > 10) {
/* eac3 */
DataScanCtx c_next = c;
guint frame_size;
frame_size = (((c.data[2] & 0x07) << 8) + c.data[3]) + 1;
GST_LOG ("possible E-AC3 frame sync at offset %"
G_GUINT64_FORMAT ", size=%u", c.offset, frame_size);
if (data_scan_ctx_ensure_data (tf, &c_next, (frame_size * 2) + 5)) {
data_scan_ctx_advance (tf, &c_next, frame_size * 2);
if (c_next.data[0] == 0x0b && c_next.data[1] == 0x77) {
GstTypeFindProbability prob;
GST_LOG ("found second E-AC3 frame, looks good");
if (c.offset == 0)
prob = GST_TYPE_FIND_MAXIMUM;
else
prob = GST_TYPE_FIND_NEARLY_CERTAIN;
gst_type_find_suggest (tf, prob, EAC3_CAPS);
return;
} else {
GST_LOG ("no second E-AC3 frame found, false sync");
}
}
} else {
GST_LOG ("invalid AC3 BSID: %u", bsid);
}
}
data_scan_ctx_advance (tf, &c, 1);
}
}
/*** audio/x-dts ***/
static GstStaticCaps dts_caps = GST_STATIC_CAPS ("audio/x-dts");
#define DTS_CAPS (gst_static_caps_get (&dts_caps))
#define DTS_MIN_FRAMESIZE 96
#define DTS_MAX_FRAMESIZE 18725 /* 16384*16/14 */
static gboolean
dts_parse_frame_header (DataScanCtx * c, guint * frame_size,
guint * sample_rate, guint * channels, guint * depth, guint * endianness)
{
static const int sample_rates[16] = { 0, 8000, 16000, 32000, 0, 0, 11025,
22050, 44100, 0, 0, 12000, 24000, 48000, 96000, 192000
};
static const guint8 channels_table[16] = { 1, 2, 2, 2, 2, 3, 3, 4, 4, 5,
6, 6, 6, 7, 8, 8
};
guint16 hdr[8];
guint32 marker;
guint num_blocks, chans, lfe, i;
marker = GST_READ_UINT32_BE (c->data);
/* raw big endian or 14-bit big endian */
if (marker == 0x7FFE8001 || marker == 0x1FFFE800) {
*endianness = G_BIG_ENDIAN;
for (i = 0; i < G_N_ELEMENTS (hdr); ++i)
hdr[i] = GST_READ_UINT16_BE (c->data + (i * sizeof (guint16)));
} else
/* raw little endian or 14-bit little endian */
if (marker == 0xFE7F0180 || marker == 0xFF1F00E8) {
*endianness = G_LITTLE_ENDIAN;
for (i = 0; i < G_N_ELEMENTS (hdr); ++i)
hdr[i] = GST_READ_UINT16_LE (c->data + (i * sizeof (guint16)));
} else {
return FALSE;
}
GST_LOG ("dts sync marker 0x%08x at offset %u", marker, (guint) c->offset);
/* 14-bit mode */
if (marker == 0x1FFFE800 || marker == 0xFF1F00E8) {
if ((hdr[2] & 0xFFF0) != 0x07F0)
return FALSE;
/* discard top 2 bits (2 void), shift in 2 */
hdr[0] = (hdr[0] << 2) | ((hdr[1] >> 12) & 0x0003);
/* discard top 4 bits (2 void, 2 shifted into hdr[0]), shift in 4 etc. */
hdr[1] = (hdr[1] << 4) | ((hdr[2] >> 10) & 0x000F);
hdr[2] = (hdr[2] << 6) | ((hdr[3] >> 8) & 0x003F);
hdr[3] = (hdr[3] << 8) | ((hdr[4] >> 6) & 0x00FF);
hdr[4] = (hdr[4] << 10) | ((hdr[5] >> 4) & 0x03FF);
hdr[5] = (hdr[5] << 12) | ((hdr[6] >> 2) & 0x0FFF);
hdr[6] = (hdr[6] << 14) | ((hdr[7] >> 0) & 0x3FFF);
g_assert (hdr[0] == 0x7FFE && hdr[1] == 0x8001);
*depth = 14;
} else {
*depth = 16;
}
GST_LOG ("frame header: %04x%04x%04x%04x", hdr[2], hdr[3], hdr[4], hdr[5]);
num_blocks = (hdr[2] >> 2) & 0x7F;
*frame_size = (((hdr[2] & 0x03) << 12) | (hdr[3] >> 4)) + 1;
chans = ((hdr[3] & 0x0F) << 2) | (hdr[4] >> 14);
*sample_rate = sample_rates[(hdr[4] >> 10) & 0x0F];
lfe = (hdr[5] >> 9) & 0x03;
if (num_blocks < 5 || *frame_size < 96 || *sample_rate == 0)
return FALSE;
if (marker == 0x1FFFE800 || marker == 0xFF1F00E8)
*frame_size = (*frame_size * 16) / 14; /* FIXME: round up? */
if (chans < G_N_ELEMENTS (channels_table))
*channels = channels_table[chans] + ((lfe) ? 1 : 0);
else
*channels = 0;
return TRUE;
}
static void
dts_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
/* Search for an dts frame; not necessarily right at the start, but give it
* a lower probability if not found right at the start. Check that the
* frame is followed by a second frame at the expected offset. */
while (c.offset <= DTS_MAX_FRAMESIZE) {
guint frame_size = 0, rate = 0, chans = 0, depth = 0, endianness = 0;
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, DTS_MIN_FRAMESIZE)))
return;
if (G_UNLIKELY (dts_parse_frame_header (&c, &frame_size, &rate, &chans,
&depth, &endianness))) {
GstTypeFindProbability prob;
DataScanCtx next_c;
prob = (c.offset == 0) ? GST_TYPE_FIND_LIKELY : GST_TYPE_FIND_POSSIBLE;
/* check for second frame sync */
next_c = c;
data_scan_ctx_advance (tf, &next_c, frame_size);
if (data_scan_ctx_ensure_data (tf, &next_c, 4)) {
GST_LOG ("frame size: %u 0x%04x", frame_size, frame_size);
GST_MEMDUMP ("second frame sync", next_c.data, 4);
if (GST_READ_UINT32_BE (c.data) == GST_READ_UINT32_BE (next_c.data))
prob = GST_TYPE_FIND_MAXIMUM;
}
if (chans > 0) {
gst_type_find_suggest_simple (tf, prob, "audio/x-dts",
"rate", G_TYPE_INT, rate, "channels", G_TYPE_INT, chans,
"depth", G_TYPE_INT, depth, "endianness", G_TYPE_INT, endianness,
"framed", G_TYPE_BOOLEAN, FALSE, NULL);
} else {
gst_type_find_suggest_simple (tf, prob, "audio/x-dts",
"rate", G_TYPE_INT, rate, "depth", G_TYPE_INT, depth,
"endianness", G_TYPE_INT, endianness,
"framed", G_TYPE_BOOLEAN, FALSE, NULL);
}
return;
}
data_scan_ctx_advance (tf, &c, 1);
}
}
/*** gsm ***/
/* can only be detected by using the extension, in which case we use the default
* GSM properties */
static GstStaticCaps gsm_caps =
GST_STATIC_CAPS ("audio/x-gsm, rate=8000, channels=1");
#define GSM_CAPS (gst_static_caps_get(&gsm_caps))
/*** wavpack ***/
static GstStaticCaps wavpack_caps =
GST_STATIC_CAPS ("audio/x-wavpack, framed = (boolean) false");
#define WAVPACK_CAPS (gst_static_caps_get(&wavpack_caps))
static GstStaticCaps wavpack_correction_caps =
GST_STATIC_CAPS ("audio/x-wavpack-correction, framed = (boolean) false");
#define WAVPACK_CORRECTION_CAPS (gst_static_caps_get(&wavpack_correction_caps))
static void
wavpack_type_find (GstTypeFind * tf, gpointer unused)
{
GstTypeFindProbability base_prob = GST_TYPE_FIND_POSSIBLE;
guint64 offset;
guint32 blocksize;
const guint8 *data;
guint count_wv, count_wvc;
data = gst_type_find_peek (tf, 0, 32);
if (!data)
return;
if (data[0] != 'w' || data[1] != 'v' || data[2] != 'p' || data[3] != 'k')
return;
/* Note: wavpack blocks can be fairly large (easily 60-110k), possibly
* larger than the max. limits imposed by certain typefinding elements
* like id3demux or apedemux, so typefinding is most likely only going to
* work in pull-mode */
blocksize = GST_READ_UINT32_LE (data + 4);
GST_LOG ("wavpack header, blocksize=0x%04x", blocksize);
count_wv = 0;
count_wvc = 0;
offset = 32;
while (offset < 8 + blocksize) {
guint32 sublen;
/* get chunk header */
GST_LOG ("peeking at chunk at offset 0x%04x", (guint) offset);
data = gst_type_find_peek (tf, offset, 4);
if (data == NULL)
break;
sublen = ((guint32) data[1]) << 1;
if (data[0] & 0x80) {
sublen |= (((guint32) data[2]) << 9) | (((guint32) data[3]) << 17);
sublen += 1 + 3; /* id + length */
} else {
sublen += 1 + 1; /* id + length */
}
if (offset + sublen > 8 + blocksize) {
GST_LOG ("chunk length too big (%u > %" G_GUINT64_FORMAT ")", sublen,
blocksize - offset);
break;
}
if ((data[0] & 0x20) == 0) {
switch (data[0] & 0x0f) {
case 0xa: /* ID_WV_BITSTREAM */
case 0xc: /* ID_WVX_BITSTREAM */
++count_wv;
break;
case 0xb: /* ID_WVC_BITSTREAM */
++count_wvc;
break;
default:
break;
}
if (count_wv >= 5 || count_wvc >= 5)
break;
}
offset += sublen;
}
/* check for second block header */
data = gst_type_find_peek (tf, 8 + blocksize, 4);
if (data != NULL && memcmp (data, "wvpk", 4) == 0) {
GST_DEBUG ("found second block sync");
base_prob = GST_TYPE_FIND_LIKELY;
}
GST_DEBUG ("wvc=%d, wv=%d", count_wvc, count_wv);
if (count_wvc > 0 && count_wvc > count_wv) {
gst_type_find_suggest (tf,
MIN (base_prob + 5 * count_wvc, GST_TYPE_FIND_NEARLY_CERTAIN),
WAVPACK_CORRECTION_CAPS);
} else if (count_wv > 0) {
gst_type_find_suggest (tf,
MIN (base_prob + 5 * count_wv, GST_TYPE_FIND_NEARLY_CERTAIN),
WAVPACK_CAPS);
}
}
/*** application/postscrip ***/
static GstStaticCaps postscript_caps =
GST_STATIC_CAPS ("application/postscript");
#define POSTSCRIPT_CAPS (gst_static_caps_get(&postscript_caps))
static void
postscript_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data = gst_type_find_peek (tf, 0, 3);
if (!data)
return;
if (data[0] == 0x04)
data++;
if (data[0] == '%' && data[1] == '!')
gst_type_find_suggest (tf, GST_TYPE_FIND_POSSIBLE, POSTSCRIPT_CAPS);
}
/*** image/svg+xml ***/
static GstStaticCaps svg_caps = GST_STATIC_CAPS ("image/svg+xml");
#define SVG_CAPS (gst_static_caps_get(&svg_caps))
static void
svg_type_find (GstTypeFind * tf, gpointer unused)
{
static const gchar svg_doctype[] = "!DOCTYPE svg";
static const gchar svg_tag[] = "<svg";
DataScanCtx c = { 0, NULL, 0 };
while (c.offset <= 1024) {
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 12)))
break;
if (memcmp (svg_doctype, c.data, 12) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, SVG_CAPS);
return;
} else if (memcmp (svg_tag, c.data, 4) == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_LIKELY, SVG_CAPS);
return;
}
data_scan_ctx_advance (tf, &c, 1);
}
}
/*** multipart/x-mixed-replace mimestream ***/
static GstStaticCaps multipart_caps =
GST_STATIC_CAPS ("multipart/x-mixed-replace");
#define MULTIPART_CAPS gst_static_caps_get(&multipart_caps)
/* multipart/x-mixed replace is:
* <maybe some whitespace>--<some ascii chars>[\r]\n
* <more ascii chars>[\r]\nContent-type:<more ascii>[\r]\n */
static void
multipart_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data;
const guint8 *x;
#define MULTIPART_MAX_BOUNDARY_OFFSET 16
data = gst_type_find_peek (tf, 0, MULTIPART_MAX_BOUNDARY_OFFSET);
if (!data)
return;
for (x = data;
x - data < MULTIPART_MAX_BOUNDARY_OFFSET - 2 && g_ascii_isspace (*x);
x++);
if (x[0] != '-' || x[1] != '-')
return;
/* Could be okay, peek what should be enough for a complete header */
#define MULTIPART_MAX_HEADER_SIZE 256
data = gst_type_find_peek (tf, 0, MULTIPART_MAX_HEADER_SIZE);
if (!data)
return;
for (x = data; x - data < MULTIPART_MAX_HEADER_SIZE - 14; x++) {
if (!isascii (*x)) {
return;
}
if (*x == '\n' &&
!g_ascii_strncasecmp ("content-type:", (gchar *) x + 1, 13)) {
gst_type_find_suggest (tf, GST_TYPE_FIND_MAXIMUM, MULTIPART_CAPS);
return;
}
}
}
/*** video/mpeg systemstream ***/
static GstStaticCaps mpeg_sys_caps = GST_STATIC_CAPS ("video/mpeg, "
"systemstream = (boolean) true, mpegversion = (int) [ 1, 2 ]");
#define MPEG_SYS_CAPS gst_static_caps_get(&mpeg_sys_caps)
#define IS_MPEG_HEADER(data) (G_UNLIKELY((((guint8 *)(data))[0] == 0x00) && \
(((guint8 *)(data))[1] == 0x00) && \
(((guint8 *)(data))[2] == 0x01)))
#define IS_MPEG_PACK_CODE(b) ((b) == 0xBA)
#define IS_MPEG_SYS_CODE(b) ((b) == 0xBB)
#define IS_MPEG_PACK_HEADER(data) (IS_MPEG_HEADER (data) && \
IS_MPEG_PACK_CODE (((guint8 *)(data))[3]))
#define IS_MPEG_PES_CODE(b) (((b) & 0xF0) == 0xE0 || ((b) & 0xF0) == 0xC0 || \
(b) >= 0xBD)
#define IS_MPEG_PES_HEADER(data) (IS_MPEG_HEADER (data) && \
IS_MPEG_PES_CODE (((guint8 *)(data))[3]))
#define MPEG2_MAX_PROBE_LENGTH (128 * 1024) /* 128kB should be 64 packs of the
* most common 2kB pack size. */
#define MPEG2_MIN_SYS_HEADERS 2
#define MPEG2_MAX_SYS_HEADERS 5
static gboolean
mpeg_sys_is_valid_pack (GstTypeFind * tf, const guint8 * data, guint len,
guint * pack_size)
{
/* Check the pack header @ offset for validity, assuming that the 4 byte header
* itself has already been checked. */
guint8 stuff_len;
if (len < 12)
return FALSE;
/* Check marker bits */
if ((data[4] & 0xC4) == 0x44) {
/* MPEG-2 PACK */
if (len < 14)
return FALSE;
if ((data[6] & 0x04) != 0x04 ||
(data[8] & 0x04) != 0x04 ||
(data[9] & 0x01) != 0x01 || (data[12] & 0x03) != 0x03)
return FALSE;
stuff_len = data[13] & 0x07;
/* Check the following header bytes, if we can */
if ((14 + stuff_len + 4) <= len) {
if (!IS_MPEG_HEADER (data + 14 + stuff_len))
return FALSE;
}
if (pack_size)
*pack_size = 14 + stuff_len;
return TRUE;
} else if ((data[4] & 0xF1) == 0x21) {
/* MPEG-1 PACK */
if ((data[6] & 0x01) != 0x01 ||
(data[8] & 0x01) != 0x01 ||
(data[9] & 0x80) != 0x80 || (data[11] & 0x01) != 0x01)
return FALSE;
/* Check the following header bytes, if we can */
if ((12 + 4) <= len) {
if (!IS_MPEG_HEADER (data + 12))
return FALSE;
}
if (pack_size)
*pack_size = 12;
return TRUE;
}
return FALSE;
}
static gboolean
mpeg_sys_is_valid_pes (GstTypeFind * tf, const guint8 * data, guint len,
guint * pack_size)
{
guint pes_packet_len;
/* Check the PES header at the given position, assuming the header code itself
* was already checked */
if (len < 6)
return FALSE;
/* For MPEG Program streams, unbounded PES is not allowed, so we must have a
* valid length present */
pes_packet_len = GST_READ_UINT16_BE (data + 4);
if (pes_packet_len == 0)
return FALSE;
/* Check the following header, if we can */
if (6 + pes_packet_len + 4 <= len) {
if (!IS_MPEG_HEADER (data + 6 + pes_packet_len))
return FALSE;
}
if (pack_size)
*pack_size = 6 + pes_packet_len;
return TRUE;
}
static gboolean
mpeg_sys_is_valid_sys (GstTypeFind * tf, const guint8 * data, guint len,
guint * pack_size)
{
guint sys_hdr_len;
/* Check the System header at the given position, assuming the header code itself
* was already checked */
if (len < 6)
return FALSE;
sys_hdr_len = GST_READ_UINT16_BE (data + 4);
if (sys_hdr_len < 6)
return FALSE;
/* Check the following header, if we can */
if (6 + sys_hdr_len + 4 <= len) {
if (!IS_MPEG_HEADER (data + 6 + sys_hdr_len))
return FALSE;
}
if (pack_size)
*pack_size = 6 + sys_hdr_len;
return TRUE;
}
/* calculation of possibility to identify random data as mpeg systemstream:
* bits that must match in header detection: 32 (or more)
* chance that random data is identifed: 1/2^32
* chance that MPEG2_MIN_PACK_HEADERS headers are identified:
* 1/2^(32*MPEG2_MIN_PACK_HEADERS)
* chance that this happens in MPEG2_MAX_PROBE_LENGTH bytes:
* 1-(1+1/2^(32*MPEG2_MIN_PACK_HEADERS)^MPEG2_MAX_PROBE_LENGTH)
* for current values:
* 1-(1+1/2^(32*4)^101024)
* = <some_number>
* Since we also check marker bits and pes packet lengths, this probability is a
* very coarse upper bound.
*/
static void
mpeg_sys_type_find (GstTypeFind * tf, gpointer unused)
{
const guint8 *data, *data0, *first_sync, *end;
gint mpegversion = 0;
guint pack_headers = 0;
guint pes_headers = 0;
guint pack_size;
guint since_last_sync = 0;
guint32 sync_word = 0xffffffff;
guint potential_headers = 0;
G_STMT_START {
gint len;
len = MPEG2_MAX_PROBE_LENGTH;
while (len >= 16) {
data = gst_type_find_peek (tf, 0, 5 + len);
if (data != NULL)
break;
len = len / 2;
}
if (!data)
return;
end = data + len;
}
G_STMT_END;
data0 = data;
first_sync = NULL;
while (data < end) {
sync_word <<= 8;
if (sync_word == 0x00000100) {
/* Found potential sync word */
if (first_sync == NULL)
first_sync = data - 3;
if (since_last_sync > 4) {
/* If more than 4 bytes since the last sync word, reset our counters,
* as we're only interested in counting contiguous packets */
pes_headers = pack_headers = 0;
}
pack_size = 0;
potential_headers++;
if (IS_MPEG_PACK_CODE (data[0])) {
if ((data[1] & 0xC0) == 0x40) {
/* MPEG-2 */
mpegversion = 2;
} else if ((data[1] & 0xF0) == 0x20) {
mpegversion = 1;
}
if (mpegversion != 0 &&
mpeg_sys_is_valid_pack (tf, data - 3, end - data + 3, &pack_size)) {
pack_headers++;
}
} else if (IS_MPEG_PES_CODE (data[0])) {
/* PES stream */
if (mpeg_sys_is_valid_pes (tf, data - 3, end - data + 3, &pack_size)) {
pes_headers++;
if (mpegversion == 0)
mpegversion = 2;
}
} else if (IS_MPEG_SYS_CODE (data[0])) {
if (mpeg_sys_is_valid_sys (tf, data - 3, end - data + 3, &pack_size)) {
pack_headers++;
}
}
/* If we found a packet with a known size, skip the bytes in it and loop
* around to check the next packet. */
if (pack_size != 0) {
data += pack_size - 3;
sync_word = 0xffffffff;
since_last_sync = 0;
continue;
}
}
sync_word |= data[0];
since_last_sync++;
data++;
/* If we have found MAX headers, and *some* were pes headers (pack headers
* are optional in an mpeg system stream) then return our high-probability
* result */
if (pes_headers > 0 && (pack_headers + pes_headers) > MPEG2_MAX_SYS_HEADERS)
goto suggest;
}
/* If we at least saw MIN headers, and *some* were pes headers (pack headers
* are optional in an mpeg system stream) then return a lower-probability
* result */
if (pes_headers > 0 && (pack_headers + pes_headers) > MPEG2_MIN_SYS_HEADERS)
goto suggest;
return;
suggest:
{
guint prob;
prob = GST_TYPE_FIND_POSSIBLE + (10 * (pack_headers + pes_headers));
prob = MIN (prob, GST_TYPE_FIND_MAXIMUM);
/* With the above test, we get into problems when we try to typefind
a MPEG stream from a small amount of data, which can happen when
we get data pushed from a HTTP source. We thus make a second test
to give higher probability if all the potential headers were either
pack or pes headers (ie, no potential header was unrecognized). */
if (potential_headers == pack_headers + pes_headers) {
GST_LOG ("Only %u headers, but all were recognized", potential_headers);
prob += 10;
prob = MIN (prob, GST_TYPE_FIND_MAXIMUM);
}
/* lower probability if the first packet wasn't right at the start */
if (data0 != first_sync && prob >= 10)
prob -= 10;
GST_LOG ("Suggesting MPEG %d system stream, %d packs, %d pes, prob %u%%",
mpegversion, pack_headers, pes_headers, prob);
gst_type_find_suggest_simple (tf, prob, "video/mpeg",
"systemstream", G_TYPE_BOOLEAN, TRUE,
"mpegversion", G_TYPE_INT, mpegversion, NULL);
}
};
/*** video/mpegts Transport Stream ***/
static GstStaticCaps mpegts_caps = GST_STATIC_CAPS ("video/mpegts, "
"systemstream = (boolean) true, packetsize = (int) [ 188, 208 ]");
#define MPEGTS_CAPS gst_static_caps_get(&mpegts_caps)
#define GST_MPEGTS_TYPEFIND_MIN_HEADERS 4
#define GST_MPEGTS_TYPEFIND_MAX_HEADERS 10
#define GST_MPEGTS_MAX_PACKET_SIZE 208
#define GST_MPEGTS_TYPEFIND_SYNC_SIZE \
(GST_MPEGTS_TYPEFIND_MIN_HEADERS * GST_MPEGTS_MAX_PACKET_SIZE)
#define GST_MPEGTS_TYPEFIND_MAX_SYNC \
(GST_MPEGTS_TYPEFIND_MAX_HEADERS * GST_MPEGTS_MAX_PACKET_SIZE)
#define GST_MPEGTS_TYPEFIND_SCAN_LENGTH \
(GST_MPEGTS_TYPEFIND_MAX_SYNC * 4)
#define MPEGTS_HDR_SIZE 4
/* Check for sync byte, error_indicator == 0 and packet has payload.
* Adaptation control field (data[3] & 0x30) may be zero for TS packets with
* null PIDs. Still, these streams are valid TS streams (for null packets,
* AFC is supposed to be 0x1, but the spec also says decoders should just
* discard any packets with AFC = 0x00) */
#define IS_MPEGTS_HEADER(data) (data[0] == 0x47 && \
(data[1] & 0x80) == 0x00 && \
((data[3] & 0x30) != 0x00 || \
((data[3] & 0x30) == 0x00 && (data[1] & 0x1f) == 0x1f && (data[2] & 0xff) == 0xff)))
/* Helper function to search ahead at intervals of packet_size for mpegts
* headers */
static gint
mpeg_ts_probe_headers (GstTypeFind * tf, guint64 offset, gint packet_size)
{
/* We always enter this function having found at least one header already */
gint found = 1;
const guint8 *data = NULL;
GST_LOG ("looking for mpeg-ts packets of size %u", packet_size);
while (found < GST_MPEGTS_TYPEFIND_MAX_HEADERS) {
offset += packet_size;
data = gst_type_find_peek (tf, offset, MPEGTS_HDR_SIZE);
if (data == NULL || !IS_MPEGTS_HEADER (data))
return found;
found++;
GST_LOG ("mpeg-ts sync #%2d at offset %" G_GUINT64_FORMAT, found, offset);
}
return found;
}
/* Try and detect at least 4 packets in at most 10 packets worth of
* data. Need to try several possible packet sizes */
static void
mpeg_ts_type_find (GstTypeFind * tf, gpointer unused)
{
/* TS packet sizes to test: normal, DVHS packet size and
* FEC with 16 or 20 byte codes packet size. */
const gint pack_sizes[] = { 188, 192, 204, 208 };
const guint8 *data = NULL;
guint size = 0;
guint64 skipped = 0;
while (skipped < GST_MPEGTS_TYPEFIND_SCAN_LENGTH) {
if (size < MPEGTS_HDR_SIZE) {
data = gst_type_find_peek (tf, skipped, GST_MPEGTS_TYPEFIND_SYNC_SIZE);
if (!data)
break;
size = GST_MPEGTS_TYPEFIND_SYNC_SIZE;
}
/* Have at least MPEGTS_HDR_SIZE bytes at this point */
if (IS_MPEGTS_HEADER (data)) {
gint p;
GST_LOG ("possible mpeg-ts sync at offset %" G_GUINT64_FORMAT, skipped);
for (p = 0; p < G_N_ELEMENTS (pack_sizes); p++) {
gint found;
/* Probe ahead at size pack_sizes[p] */
found = mpeg_ts_probe_headers (tf, skipped, pack_sizes[p]);
if (found >= GST_MPEGTS_TYPEFIND_MIN_HEADERS) {
gint probability;
/* found at least 4 headers. 10 headers = MAXIMUM probability.
* Arbitrarily, I assigned 10% probability for each header we
* found, 40% -> 100% */
probability = MIN (10 * found, GST_TYPE_FIND_MAXIMUM);
gst_type_find_suggest_simple (tf, probability, "video/mpegts",
"systemstream", G_TYPE_BOOLEAN, TRUE,
"packetsize", G_TYPE_INT, pack_sizes[p], NULL);
return;
}
}
}
data++;
skipped++;
size--;
}
}
#define GST_MPEGVID_TYPEFIND_TRY_PICTURES 6
#define GST_MPEGVID_TYPEFIND_TRY_SYNC (100 * 1024) /* 100 kB */
/* Scan ahead a maximum of max_extra_offset bytes until the next IS_MPEG_HEADER
* offset. After the call, offset will be after the 0x000001, i.e. at the 4th
* byte of the MPEG header. Returns TRUE if a header was found, FALSE if not.
*/
static gboolean
mpeg_find_next_header (GstTypeFind * tf, DataScanCtx * c,
guint64 max_extra_offset)
{
guint64 extra_offset;
for (extra_offset = 0; extra_offset <= max_extra_offset; ++extra_offset) {
if (!data_scan_ctx_ensure_data (tf, c, 4))
return FALSE;
if (IS_MPEG_HEADER (c->data)) {
data_scan_ctx_advance (tf, c, 3);
return TRUE;
}
data_scan_ctx_advance (tf, c, 1);
}
return FALSE;
}
/*** video/mpeg MPEG-4 elementary video stream ***/
static GstStaticCaps mpeg4_video_caps = GST_STATIC_CAPS ("video/mpeg, "
"systemstream=(boolean)false, mpegversion=4, parsed=(boolean)false");
#define MPEG4_VIDEO_CAPS gst_static_caps_get(&mpeg4_video_caps)
/*
* This typefind is based on the elementary video header defined in
* http://xhelmboyx.tripod.com/formats/mpeg-layout.txt
* In addition, it allows the visual object sequence header to be
* absent, and even the VOS header to be absent. In the latter case,
* a number of VOPs have to be present.
*/
static void
mpeg4_video_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
gboolean seen_vios_at_0 = FALSE;
gboolean seen_vios = FALSE;
gboolean seen_vos = FALSE;
gboolean seen_vol = FALSE;
guint num_vop_headers = 0;
guint8 sc;
while (c.offset < GST_MPEGVID_TYPEFIND_TRY_SYNC) {
if (num_vop_headers >= GST_MPEGVID_TYPEFIND_TRY_PICTURES)
break;
if (!mpeg_find_next_header (tf, &c,
GST_MPEGVID_TYPEFIND_TRY_SYNC - c.offset))
break;
sc = c.data[0];
/* visual_object_sequence_start_code */
if (sc == 0xB0) {
if (seen_vios)
break; /* Terminate at second vios */
if (c.offset == 0)
seen_vios_at_0 = TRUE;
seen_vios = TRUE;
data_scan_ctx_advance (tf, &c, 2);
if (!mpeg_find_next_header (tf, &c, 0))
break;
sc = c.data[0];
/* Optional metadata */
if (sc == 0xB2)
if (!mpeg_find_next_header (tf, &c, 24))
break;
}
/* visual_object_start_code (consider it optional) */
if (sc == 0xB5) {
data_scan_ctx_advance (tf, &c, 2);
/* may contain ID marker and YUV clamping */
if (!mpeg_find_next_header (tf, &c, 7))
break;
sc = c.data[0];
}
/* video_object_start_code */
if (sc <= 0x1F) {
if (seen_vos)
break; /* Terminate at second vos */
seen_vos = TRUE;
data_scan_ctx_advance (tf, &c, 2);
continue;
}
/* video_object_layer_start_code */
if (sc >= 0x20 && sc <= 0x2F) {
seen_vol = TRUE;
data_scan_ctx_advance (tf, &c, 5);
continue;
}
/* video_object_plane_start_code */
if (sc == 0xB6) {
num_vop_headers++;
data_scan_ctx_advance (tf, &c, 2);
continue;
}
/* Unknown start code. */
}
if (num_vop_headers > 0 || seen_vol) {
GstTypeFindProbability probability = 0;
GST_LOG ("Found %d pictures, vios: %d, vos:%d, vol:%d", num_vop_headers,
seen_vios, seen_vos, seen_vol);
if (num_vop_headers >= GST_MPEGVID_TYPEFIND_TRY_PICTURES && seen_vios_at_0
&& seen_vos && seen_vol)
probability = GST_TYPE_FIND_MAXIMUM - 1;
else if (num_vop_headers >= GST_MPEGVID_TYPEFIND_TRY_PICTURES && seen_vios
&& seen_vos && seen_vol)
probability = GST_TYPE_FIND_NEARLY_CERTAIN - 1;
else if (seen_vios_at_0 && seen_vos && seen_vol)
probability = GST_TYPE_FIND_NEARLY_CERTAIN - 6;
else if (num_vop_headers >= GST_MPEGVID_TYPEFIND_TRY_PICTURES && seen_vos
&& seen_vol)
probability = GST_TYPE_FIND_NEARLY_CERTAIN - 6;
else if (num_vop_headers >= GST_MPEGVID_TYPEFIND_TRY_PICTURES && seen_vol)
probability = GST_TYPE_FIND_NEARLY_CERTAIN - 9;
else if (num_vop_headers >= GST_MPEGVID_TYPEFIND_TRY_PICTURES)
probability = GST_TYPE_FIND_LIKELY - 1;
else if (num_vop_headers > 2 && seen_vios && seen_vos && seen_vol)
probability = GST_TYPE_FIND_LIKELY - 9;
else if (seen_vios && seen_vos && seen_vol)
probability = GST_TYPE_FIND_LIKELY - 20;
else if (num_vop_headers > 0 && seen_vos && seen_vol)
probability = GST_TYPE_FIND_POSSIBLE;
else if (num_vop_headers > 0)
probability = GST_TYPE_FIND_POSSIBLE - 10;
else if (seen_vos && seen_vol)
probability = GST_TYPE_FIND_POSSIBLE - 20;
gst_type_find_suggest (tf, probability, MPEG4_VIDEO_CAPS);
}
}
/*** video/x-h263 H263 video stream ***/
static GstStaticCaps h263_video_caps =
GST_STATIC_CAPS ("video/x-h263, variant=(string)itu");
#define H263_VIDEO_CAPS gst_static_caps_get(&h263_video_caps)
#define H263_MAX_PROBE_LENGTH (128 * 1024)
static void
h263_video_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
guint64 data = 0xffff; /* prevents false positive for first 2 bytes */
guint64 psc = 0;
guint8 ptype = 0;
guint format;
guint good = 0;
guint bad = 0;
guint pc_type, pb_mode;
while (c.offset < H263_MAX_PROBE_LENGTH) {
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 4)))
break;
/* Find the picture start code */
data = (data << 8) + c.data[0];
psc = data & G_GUINT64_CONSTANT (0xfffffc0000);
if (psc == 0x800000) {
/* Found PSC */
/* PTYPE */
ptype = (data & 0x3fc) >> 2;
/* Source Format */
format = ptype & 0x07;
/* Now that we have a Valid PSC, check if we also have a valid PTYPE and
the Source Format, which should range between 1 and 5 */
if (((ptype >> 6) == 0x2) && (format > 0 && format < 6)) {
pc_type = data & 0x02;
pb_mode = c.data[1] & 0x20 >> 4;
if (!pc_type && pb_mode)
bad++;
else
good++;
} else
bad++;
/* FIXME: maybe bail out early if we get mostly bad syncs ? */
}
data_scan_ctx_advance (tf, &c, 1);
}
GST_LOG ("good: %d, bad: %d", good, bad);
if (good > 2 * bad)
gst_type_find_suggest (tf, GST_TYPE_FIND_POSSIBLE, H263_VIDEO_CAPS);
return;
}
/*** video/x-h264 H264 elementary video stream ***/
static GstStaticCaps h264_video_caps =
GST_STATIC_CAPS ("video/x-h264,stream-format=byte-stream");
#define H264_VIDEO_CAPS gst_static_caps_get(&h264_video_caps)
#define H264_MAX_PROBE_LENGTH (128 * 1024) /* 128kB for HD should be enough. */
static void
h264_video_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
/* Stream consists of: a series of sync codes (00 00 00 01) followed
* by NALs
*/
gboolean seen_idr = FALSE;
gboolean seen_sps = FALSE;
gboolean seen_pps = FALSE;
gboolean seen_ssps = FALSE;
int nut, ref;
int good = 0;
int bad = 0;
while (c.offset < H264_MAX_PROBE_LENGTH) {
if (G_UNLIKELY (!data_scan_ctx_ensure_data (tf, &c, 4)))
break;
if (IS_MPEG_HEADER (c.data)) {
nut = c.data[3] & 0x9f; /* forbiden_zero_bit | nal_unit_type */
ref = c.data[3] & 0x60; /* nal_ref_idc */
/* if forbidden bit is different to 0 won't be h264 */
if (nut > 0x1f) {
bad++;
break;
}
/* collect statistics about the NAL types */
if ((nut >= 1 && nut <= 13) || nut == 19) {
if ((nut == 5 && ref == 0) ||
((nut == 6 || (nut >= 9 && nut <= 12)) && ref != 0)) {
bad++;
} else {
if (nut == 7)
seen_sps = TRUE;
else if (nut == 8)
seen_pps = TRUE;
else if (nut == 5)
seen_idr = TRUE;
good++;
}
} else if (nut >= 14 && nut <= 33) {
if (nut == 15) {
seen_ssps = TRUE;
good++;
} else if (nut == 14 || nut == 20) {
/* Sometimes we see NAL 14 or 20 without SSPS
* if dropped into the middle of a stream -
* just ignore those (don't add to bad count) */
if (seen_ssps)
good++;
} else {
/* reserved */
/* Theoretically these are good, since if they exist in the
stream it merely means that a newer backwards-compatible
h.264 stream. But we should be identifying that separately. */
bad++;
}
} else {
/* unspecified, application specific */
/* don't consider these bad */
}
GST_LOG ("good:%d, bad:%d, pps:%d, sps:%d, idr:%d ssps:%d", good, bad,
seen_pps, seen_sps, seen_idr, seen_ssps);
if (seen_sps && seen_pps && seen_idr && good >= 10 && bad < 4) {
gst_type_find_suggest (tf, GST_TYPE_FIND_LIKELY, H264_VIDEO_CAPS);
return;
}
data_scan_ctx_advance (tf, &c, 4);
}
data_scan_ctx_advance (tf, &c, 1);
}
GST_LOG ("good:%d, bad:%d, pps:%d, sps:%d, idr:%d ssps=%d", good, bad,
seen_pps, seen_sps, seen_idr, seen_ssps);
if (good >= 2 && bad == 0) {
gst_type_find_suggest (tf, GST_TYPE_FIND_POSSIBLE, H264_VIDEO_CAPS);
}
}
/*** video/x-h265 H265 elementary video stream ***/
static GstStaticCaps h265_video_caps =
GST_STATIC_CAPS ("video/x-h265,stream-format=byte-stream");
#define H265_VIDEO_CAPS gst_static_caps_get(&h265_video_caps)
#define H265_MAX_PROBE_LENGTH (128 * 1024) /* 128kB for HD should be enough. */
static void
h265_video_type_find (GstTypeFind * tf, gpointer unused)
{
DataScanCtx c = { 0, NULL, 0 };
/* Stream consists of: a series of sync codes (00 00 00 01) followed
* by NALs
*/
gboolean seen_irap = FALSE;
gboolean seen_vps = FALSE;