Files
UltraGrid/src/transmit.cpp
Martin Pulec 9b9105f6ae tx audio R-S warning: use the correct max_len
The data_len var used to compute packet_count may be lesser than its
initial value (tx->mtu - hdrs_len) because it might have been decreased
to number of remaining bytes of the last packet.

Returned back data_len to the do-loop and compute the value as
max_len.
2024-09-11 15:28:30 +02:00

1246 lines
50 KiB
C++

/*
* FILE: transmit.cpp
* AUTHOR: Colin Perkins <csp@csperkins.org>
* Ladan Gharai
* Martin Benes <martinbenesh@gmail.com>
* Lukas Hejtmanek <xhejtman@ics.muni.cz>
* Petr Holub <hopet@ics.muni.cz>
* Milos Liska <xliska@fi.muni.cz>
* Jiri Matela <matela@ics.muni.cz>
* Dalibor Matura <255899@mail.muni.cz>
* Ian Wesley-Smith <iwsmith@cct.lsu.edu>
* David Cassany <david.cassany@i2cat.net>
* Ignacio Contreras <ignacio.contreras@i2cat.net>
* Gerard Castillo <gerard.castillo@i2cat.net>
* Jordi "Txor" Casas Ríos <txorlings@gmail.com>
* Martin Pulec <pulec@cesnet.cz>
*
* Copyright (c) 2005-2010 Fundació i2CAT, Internet I Innovació Digital a Catalunya
* Copyright (c) 2001-2004 University of Southern California
* Copyright (c) 2005-2023 CESNET z.s.p.o.
*
* Redistribution and use in source and binary forms, with or without
* modification, is permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
*
* This product includes software developed by the University of Southern
* California Information Sciences Institute. This product also includes
* software developed by CESNET z.s.p.o.
*
* 4. Neither the name of the University, Institute, CESNET nor the names of
* its contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING,
* BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
* AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
* EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
*/
#include <algorithm>
#include <array>
#include <iostream>
#include <sstream>
#include <vector>
#include "audio/codec.h"
#include "audio/types.h"
#include "audio/utils.h"
#include "control_socket.h"
#include "crypto/openssl_encrypt.h"
#include "debug.h"
#include "host.h"
#include "lib_common.h"
#include "module.h"
#include "rtp/fec.h"
#include "rtp/rtp.h"
#include "rtp/rtp_callback.h"
#include "rtp/rtpenc_h264.h"
#include "transmit.h"
#include "tv.h"
#include "types.h"
#include "utils/jpeg_reader.h"
#include "utils/macros.h"
#include "utils/misc.h" // unit_evaluate
#include "utils/random.h"
#include "video.h"
#include "video_codec.h"
#define MOD_NAME "[transmit] "
#define TRANSMIT_MAGIC 0xe80ab15f
#define FEC_MAX_MULT 10
#define CONTROL_PORT_BANDWIDTH_REPORT_INTERVAL_NS NS_IN_SEC
#ifdef __APPLE__
#define GET_STARTTIME gettimeofday(&start, NULL)
#define GET_STOPTIME gettimeofday(&stop, NULL)
#define GET_DELTA delta = (stop.tv_sec - start.tv_sec) * 1000000000l + (stop.tv_usec - start.tv_usec) * 1000L
#elif defined __linux__
#define GET_STARTTIME clock_gettime(CLOCK_REALTIME, &start)
#define GET_STOPTIME clock_gettime(CLOCK_REALTIME, &stop)
#define GET_DELTA delta = (stop.tv_sec - start.tv_sec) * 1000000000l + stop.tv_nsec - start.tv_nsec
#else // Windows
#define GET_STARTTIME {QueryPerformanceFrequency(&freq); QueryPerformanceCounter(&start); }
#define GET_STOPTIME { QueryPerformanceCounter(&stop); }
#define GET_DELTA delta = (long)((double)(stop.QuadPart - start.QuadPart) * 1000 * 1000 * 1000 / freq.QuadPart);
#endif
using std::array;
using std::vector;
static void tx_update(struct tx *tx, struct video_frame *frame, int substream);
static void tx_done(struct module *tx);
static uint32_t format_interl_fps_hdr_row(enum interlacing_t interlacing, double input_fps);
static void
tx_send_base(struct tx *tx, struct video_frame *frame, struct rtp *rtp_session,
uint32_t ts, int send_m,
unsigned int substream,
int fragment_offset);
static bool set_fec(struct tx *tx, const char *fec);
static void fec_check_messages(struct tx *tx);
struct rate_limit_dyn {
unsigned long avg_frame_size; ///< moving average
long long last_excess; ///< nr of frames last excessive frame was emitted
static constexpr int EXCESS_GAP = 4; ///< minimal gap between excessive frames
};
struct tx {
struct module mod;
uint32_t magic;
enum tx_media_type media_type;
unsigned mtu;
double max_loss;
uint32_t last_ts;
int last_frame_fragment_id;
unsigned int buffer:22;
unsigned long int sent_frames;
int32_t avg_len;
int32_t avg_len_last;
enum fec_type fec_scheme;
bool fec_dup_1st_pkt;
int mult_count;
int last_fragment;
struct control_state *control = nullptr;
size_t sent_since_report = 0;
uint64_t last_stat_report = 0;
const struct openssl_encrypt_info *enc_funcs;
struct openssl_encrypt *encryption;
long long int bitrate;
struct rate_limit_dyn dyn_rate_limit_state;
char tmp_packet[RTP_MAX_MTU];
};
static void tx_update(struct tx *tx, struct video_frame *frame, int substream)
{
if(!frame) {
return;
}
uint64_t tmp_avg = tx->avg_len * tx->sent_frames + frame->tiles[substream].data_len *
(frame->fec_params.type != FEC_NONE ?
(double) frame->fec_params.k / (frame->fec_params.k + frame->fec_params.m) :
1);
tx->sent_frames++;
tx->avg_len = tmp_avg / tx->sent_frames;
if(tx->sent_frames >= 100) {
if(tx->fec_scheme == FEC_LDGM && tx->max_loss > 0.0) {
if(abs(tx->avg_len_last - tx->avg_len) > tx->avg_len / 3) {
int data_len = tx->mtu - (40 + (sizeof(fec_payload_hdr_t)));
data_len = (data_len / 48) * 48;
//void *fec_state_old = tx->fec_state;
struct msg_sender *msg = (struct msg_sender *)
new_message(sizeof(struct msg_sender));
snprintf(msg->fec_cfg, sizeof(msg->fec_cfg), "LDGM percents %d %d %f",
data_len, tx->avg_len, tx->max_loss);
msg->type = SENDER_MSG_CHANGE_FEC;
struct response *resp = send_message_to_receiver(get_parent_module(&tx->mod),
(struct message *) msg);
free_response(resp);
tx->avg_len_last = tx->avg_len;
}
}
tx->avg_len = 0;
tx->sent_frames = 0;
}
}
struct tx *tx_init(struct module *parent, unsigned mtu, enum tx_media_type media_type,
const char *fec, const char *encryption, long long int bitrate)
{
if (mtu > RTP_MAX_MTU) {
log_msg(LOG_LEVEL_ERROR, "Requested MTU exceeds maximal value allowed by RTP library (%d B).\n", RTP_MAX_MTU);
return NULL;
}
if (bitrate < RATE_MIN) {
log_msg(LOG_LEVEL_ERROR, "Invalid bitrate value %lld passed (either positive bitrate or magic values from %d supported)!\n", bitrate, RATE_MIN);
return NULL;
}
struct tx *tx = (struct tx *) calloc(1, sizeof(struct tx));
if (tx == nullptr) {
return tx;
}
module_init_default(&tx->mod);
tx->mod.cls = MODULE_CLASS_TX;
tx->mod.priv_data = tx;
tx->mod.deleter = tx_done;
module_register(&tx->mod, parent);
tx->magic = TRANSMIT_MAGIC;
tx->media_type = media_type;
tx->mult_count = 1;
tx->max_loss = 0.0;
tx->mtu = mtu;
tx->buffer = ug_rand() & 0x3fffff;
tx->avg_len = tx->avg_len_last = tx->sent_frames = 0u;
tx->fec_scheme = FEC_NONE;
tx->last_frame_fragment_id = -1;
if (fec) {
if(!set_fec(tx, fec)) {
module_done(&tx->mod);
return NULL;
}
}
if (strlen(encryption) > 0) {
tx->enc_funcs = static_cast<const struct openssl_encrypt_info *>(load_library("openssl_encrypt",
LIBRARY_CLASS_UNDEFINED, OPENSSL_ENCRYPT_ABI_VERSION));
if (!tx->enc_funcs) {
fprintf(stderr, "UltraGrid was build without OpenSSL support!\n");
module_done(&tx->mod);
return NULL;
}
if (tx->enc_funcs->init(&tx->encryption,
encryption) != 0) {
log_msg(LOG_LEVEL_ERROR, MOD_NAME "Unable to initialize encryption\n");
module_done(&tx->mod);
return NULL;
}
}
tx->bitrate = bitrate;
if(parent)
tx->control = (struct control_state *) get_module(get_root_module(parent), "control");
return tx;
}
static bool set_fec(struct tx *tx, const char *fec_const)
{
char *fec = strdup(fec_const);
bool ret = true;
bool req_1st_pkt_dup = true;
// handle ...[:]nodup
char *end = fec + MAX(strlen(fec), 5) - 5;
if (strcmp(end, "nodup") == 0) {
req_1st_pkt_dup = true;
*end = '\0';
if (end > fec) { // ':'
end[-1] = '\0';
}
}
const char *fec_cfg = "";
if(strchr(fec, ':')) {
char *delim = strchr(fec, ':');
*delim = '\0';
fec_cfg = delim + 1;
}
struct msg_sender *msg = (struct msg_sender *)
new_message(sizeof(struct msg_sender));
msg->type = SENDER_MSG_CHANGE_FEC;
snprintf(msg->fec_cfg, sizeof(msg->fec_cfg), "flush");
tx->mult_count = 1; // default
if (strcasecmp(fec, "none") == 0) {
tx->fec_scheme = FEC_NONE;
} else if(strcasecmp(fec, "mult") == 0) {
tx->fec_scheme = FEC_MULT;
tx->mult_count = atoi(fec_cfg);
if (tx->mult_count <= 0 || tx->mult_count > FEC_MAX_MULT) {
MSG(ERROR,
"mult count must be between 1 and %d (%d given)!\n",
FEC_MAX_MULT, tx->mult_count);
ret = false;
}
} else if(strcasecmp(fec, "LDGM") == 0) {
tx->fec_dup_1st_pkt = req_1st_pkt_dup;
if(tx->media_type == TX_MEDIA_AUDIO) {
fprintf(stderr, "LDGM is not currently supported for audio!\n");
ret = false;
} else {
if (strlen(fec_cfg) == 0 ||
(strlen(fec_cfg) > 0 &&
strchr(fec_cfg, '%') == nullptr)) {
snprintf(msg->fec_cfg, sizeof(msg->fec_cfg),
"LDGM cfg %s", fec_cfg);
} else { // delay creation until we have avarage frame size
tx->max_loss = atof(fec_cfg);
}
tx->fec_scheme = FEC_LDGM;
}
} else if(strcasecmp(fec, "RS") == 0) {
tx->fec_dup_1st_pkt = req_1st_pkt_dup;
snprintf(msg->fec_cfg, sizeof(msg->fec_cfg), "RS cfg %s",
fec_cfg);
tx->fec_scheme = FEC_RS;
} else if(strcasecmp(fec, "help") == 0) {
color_printf("Usage:\n");
color_printf("\t" TBOLD("-f [A:|V:]{mult:count|ldgm[:params]|"
"rs[:params]}[:nodup]") "\n");
color_printf("\nIf neither A: or V: is speciefied, FEC is set "
"to the video (backward compat).\n\n");
ret = false;
} else {
fprintf(stderr, "Unknown FEC: %s\n", fec);
ret = false;
}
if (tx->fec_dup_1st_pkt) {
MSG(VERBOSE, "Duplicating 1st packet of every frame for better "
"error resiliency.\n");
}
if (ret) {
struct response *resp = send_message_to_receiver(get_parent_module(&tx->mod),
(struct message *) msg);
free_response(resp);
} else {
free_message((struct message *) msg, NULL);
}
free(fec);
return ret;
}
static void fec_check_messages(struct tx *tx)
{
struct message *msg;
while ((msg = check_message(&tx->mod))) {
auto *data = reinterpret_cast<struct msg_universal *>(msg);
const char *text = data->text;
if (strstr(text, MSG_UNIVERSAL_TAG_TX) != text) {
LOG(LOG_LEVEL_ERROR) << "[Transmit] Unexpected TX message: " << text << "\n";
free_message(msg, new_response(RESPONSE_BAD_REQUEST, "Unexpected message"));
continue;
}
text += strlen(MSG_UNIVERSAL_TAG_TX);
struct response *r = nullptr;
if (strstr(text, "fec ") == text) {
text += strlen("fec ");
if (set_fec(tx, text)) {
r = new_response(RESPONSE_OK, nullptr);
LOG(LOG_LEVEL_NOTICE) << "[Transmit] FEC set to new setting: " << text << "\n";
} else {
r = new_response(RESPONSE_INT_SERV_ERR, "cannot set FEC");
LOG(LOG_LEVEL_ERROR) << "[Transmit] Unable to reconfiure FEC to: " << text << "\n";
}
} else if (strstr(text, "rate ") == text) {
text += strlen("rate ");
auto new_rate = unit_evaluate(text, nullptr);
if (new_rate >= RATE_MIN) {
tx->bitrate = new_rate;
r = new_response(RESPONSE_OK, nullptr);
LOG(LOG_LEVEL_NOTICE) << "[Transmit] Bitrate set to: " << text << (new_rate > 0 ? "B" : "") << "\n";
} else {
r = new_response(RESPONSE_BAD_REQUEST, "Wrong value for bitrate");
LOG(LOG_LEVEL_ERROR) << "[Transmit] Wrong bitrate: " << text << "\n";
}
} else {
r = new_response(RESPONSE_BAD_REQUEST, "Unknown TX message");
LOG(LOG_LEVEL_ERROR) << "[Transmit] Unknown TX message: " << text << "\n";
}
free_message(msg, r);
}
}
static void tx_done(struct module *mod)
{
struct tx *tx = (struct tx *) mod->priv_data;
assert(tx->magic == TRANSMIT_MAGIC);
free(tx);
}
/*
* sends one or more frames (tiles) with same TS in one RTP stream. Only one m-bit is set.
*/
void
tx_send(struct tx *tx, struct video_frame *frame, struct rtp *rtp_session)
{
unsigned int i;
assert(!frame->fragment || tx->fec_scheme == FEC_NONE); // currently no support for FEC with fragments
assert(!frame->fragment || frame->tile_count); // multiple tile are not currently supported for fragmented send
fec_check_messages(tx);
uint32_t ts =
(frame->flags & TIMESTAMP_VALID) == 0
? get_local_mediatime()
: get_local_mediatime_offset() + frame->timestamp;
if(frame->fragment &&
tx->last_frame_fragment_id == frame->frame_fragment_id) {
ts = tx->last_ts;
} else {
tx->last_frame_fragment_id = frame->frame_fragment_id;
tx->last_ts = ts;
}
for(i = 0; i < frame->tile_count; ++i)
{
int last = FALSE;
int fragment_offset = 0;
if (i == frame->tile_count - 1) {
if(!frame->fragment || frame->last_fragment)
last = TRUE;
}
if(frame->fragment)
fragment_offset = vf_get_tile(frame, i)->offset;
tx_send_base(tx, frame, rtp_session, ts, last,
i, fragment_offset);
}
tx->buffer++;
}
void format_video_header(struct video_frame *frame, int tile_idx, int buffer_idx, uint32_t *video_hdr)
{
uint32_t tmp;
video_hdr[3] = htonl(frame->tiles[tile_idx].width << 16 | frame->tiles[tile_idx].height);
video_hdr[4] = get_fourcc(frame->color_spec);
video_hdr[2] = htonl(frame->tiles[tile_idx].data_len);
tmp = tile_idx << 22;
tmp |= 0x3fffff & buffer_idx;
video_hdr[0] = htonl(tmp);
/* word 6 */
video_hdr[5] = format_interl_fps_hdr_row(frame->interlacing, frame->fps);
}
void format_audio_header(const audio_frame2 *frame, int channel, int buffer_idx, uint32_t *audio_hdr)
{
uint32_t tmp = 0;
tmp = channel << 22U; /* bits 0-9 */
tmp |= buffer_idx; /* bits 10-31 */
audio_hdr[0] = htonl(tmp);
audio_hdr[2] = htonl(frame->get_data_len(channel));
/* fourth word */
tmp = (frame->get_bps() * 8) << 26U;
tmp |= frame->get_sample_rate();
audio_hdr[3] = htonl(tmp);
/* fifth word */
audio_hdr[4] = htonl(get_audio_tag(frame->get_codec()));
}
static uint32_t format_interl_fps_hdr_row(enum interlacing_t interlacing, double input_fps)
{
unsigned int fpsd, fd, fps, fi;
uint32_t tmp;
tmp = interlacing << 29;
fps = round(input_fps);
fpsd = 1; /// @todo make use of this value (for now it is always one)
fd = 0;
fi = 0;
if (input_fps > 1.0 && fabs(input_fps - round(input_fps) / 1.001) < 0.005) { // 29.97 etc.
fd = 1;
} else if (fps < 1.0) {
fps = round(1.0 / input_fps);
fi = 1;
}
tmp |= fps << 19;
tmp |= fpsd << 15;
tmp |= fd << 14;
tmp |= fi << 13;
return htonl(tmp);
}
static inline void check_symbol_size(int fec_symbol_size, int payload_len)
{
thread_local static bool status_printed = false;
if (status_printed && log_level < LOG_LEVEL_DEBUG2) {
return;
}
if (fec_symbol_size > payload_len) {
LOG(LOG_LEVEL_WARNING) << MOD_NAME
"Warning: FEC symbol size exceeds payload size! "
"FEC symbol size: " << fec_symbol_size
<< "\n";
} else {
const int ll =
status_printed ? LOG_LEVEL_DEBUG2 : LOG_LEVEL_INFO;
LOG(ll) << MOD_NAME "FEC symbol size: " << fec_symbol_size
<< ", symbols per packet: "
<< payload_len / fec_symbol_size << ", payload size: "
<< payload_len / fec_symbol_size * fec_symbol_size
<< "\n";
}
status_printed = true;
}
/**
* Splits symbol (FEC symbol or uncompressed line) to 1 or more MTUs. Symbol starts
* always on beginning of packet.
*
* If symbol_size is longer than MTU (more symbols fit one packet), the aligned
* packet size is always the same.
*
* @param symbol_size FEC symbol size or linesize for uncompressed
*/
static inline int get_video_pkt_len(int mtu,
int symbol_size, int *symbol_offset)
{
if (symbol_size > mtu) {
if (symbol_size - *symbol_offset <= mtu) {
mtu = symbol_size - *symbol_offset;
*symbol_offset = 0;
} else {
*symbol_offset += mtu;
}
return mtu;
}
return mtu / symbol_size * symbol_size;
}
/// @param mtu is tx->mtu - hdrs_len
static vector<int> get_packet_sizes(struct video_frame *frame, int substream, int mtu) {
if (frame->fec_params.type != FEC_NONE) {
check_symbol_size(frame->fec_params.symbol_size, mtu);
}
unsigned int symbol_size = 1;
int symbol_offset = 0;
if (frame->fec_params.type == FEC_NONE && !is_codec_opaque(frame->color_spec)) {
symbol_size = vc_get_linesize(frame->tiles[substream].width, frame->color_spec);
int pf_block_size = PIX_BLOCK_LCM / get_pf_block_pixels(frame->color_spec) * get_pf_block_bytes(frame->color_spec);
assert(pf_block_size <= mtu);
mtu = mtu / pf_block_size * pf_block_size;
} else if (frame->fec_params.type != FEC_NONE) {
symbol_size = frame->fec_params.symbol_size;
}
vector<int> ret;
unsigned pos = 0;
do {
int len = symbol_size == 1
? mtu
: get_video_pkt_len(mtu, symbol_size, &symbol_offset);
pos += len;
ret.push_back(len);
} while (pos < frame->tiles[substream].data_len);
if (pos > frame->tiles[substream].data_len) {
ret[ret.size() - 1] -=
(int) (pos - frame->tiles[substream].data_len);
}
return ret;
}
static void
report_stats(struct tx *tx, struct rtp *rtp_session, long data_sent)
{
if (!tx->control || !control_stats_enabled(tx->control)) {
return;
}
tx->sent_since_report += data_sent;
const time_ns_t current_time_ns = get_time_in_ns();
if (current_time_ns - tx->last_stat_report <
CONTROL_PORT_BANDWIDTH_REPORT_INTERVAL_NS) {
return;
}
const char *media =
tx->media_type == TX_MEDIA_VIDEO ? "video" : "audio";
std::ostringstream oss;
oss << "tx_send " << std::hex << rtp_my_ssrc(rtp_session) << std::dec
<< " " << media << " " << tx->sent_since_report;
control_report_stats(tx->control, oss.str());
tx->last_stat_report = current_time_ns;
tx->sent_since_report = 0;
}
/**
* Returns inter-packet interval in nanoseconds.
*/
static long
get_packet_rate(struct tx *tx, struct video_frame *frame, int substream, long packet_count)
{
if (tx->bitrate == RATE_UNLIMITED) {
return 0;
}
double time_for_frame = 1.0 / frame->fps / frame->tile_count;
double interval_between_pkts = time_for_frame / tx->mult_count / packet_count;
// use only 75% of the time - we less likely overshot the frame time and
// can minimize risk of swapping packets between 2 frames (out-of-order ones)
interval_between_pkts = interval_between_pkts * 0.75;
// prevent bitrate to be "too low", here 1 Mbps at minimum
interval_between_pkts = std::min<double>(interval_between_pkts, tx->mtu / 1000'000.0);
long packet_rate_auto = interval_between_pkts * 1000'000'000L;
if (tx->bitrate == RATE_AUTO) { // adaptive (spread packets to 75% frame time)
return packet_rate_auto;
}
if (tx->bitrate == RATE_DYNAMIC) {
if (frame->tiles[substream].data_len > 2 * tx->dyn_rate_limit_state.avg_frame_size
&& tx->dyn_rate_limit_state.last_excess > rate_limit_dyn::EXCESS_GAP) {
packet_rate_auto /= 2; // double packet rate for this frame
tx->dyn_rate_limit_state.last_excess = 0;
} else {
tx->dyn_rate_limit_state.last_excess += 1;
}
tx->dyn_rate_limit_state.avg_frame_size = (9 * tx->dyn_rate_limit_state.avg_frame_size + frame->tiles[substream].data_len) / 10;
return packet_rate_auto;
}
long long int bitrate = tx->bitrate & ~RATE_FLAG_FIXED_RATE;
int avg_packet_size = frame->tiles[substream].data_len / packet_count;
long packet_rate = 1000'000'000L * avg_packet_size * 8 / bitrate; // fixed rate
if ((tx->bitrate & RATE_FLAG_FIXED_RATE) == 0) { // adaptive capped rate
packet_rate = std::max(packet_rate, packet_rate_auto);
}
return packet_rate;
}
static int
get_tx_hdr_len(bool is_ipv6)
{
return (is_ipv6 ? IPV6_HDR_LEN : IPV4_HDR_LEN) + UDP_HDR_LEN +
RTP_HDR_LEN;
}
static void
tx_send_base(struct tx *tx, struct video_frame *frame, struct rtp *rtp_session,
uint32_t ts, int send_m,
unsigned int substream,
int fragment_offset)
{
assert(fragment_offset == 0); // no longer supported
if (!rtp_has_receiver(rtp_session)) {
return;
}
struct tile *tile = &frame->tiles[substream];
// see definition in rtp_callback.h
uint32_t rtp_hdr[100];
int rtp_hdr_len;
int pt = fec_pt_from_fec_type(TX_MEDIA_VIDEO, frame->fec_params.type, tx->encryption); /* A value specified in our packet format */
#ifdef __linux__
struct timespec start, stop;
#elif defined __APPLE__
struct timeval start, stop;
#else // Windows
LARGE_INTEGER start, stop, freq;
#endif
long delta, overslept = 0;
int hdrs_len = get_tx_hdr_len(rtp_is_ipv6(rtp_session));
assert(tx->magic == TRANSMIT_MAGIC);
tx_update(tx, frame, substream);
if (frame->fec_params.type == FEC_NONE) {
hdrs_len += (sizeof(video_payload_hdr_t));
rtp_hdr_len = sizeof(video_payload_hdr_t);
format_video_header(frame, substream, tx->buffer, rtp_hdr);
} else {
hdrs_len += (sizeof(fec_payload_hdr_t));
rtp_hdr_len = sizeof(fec_payload_hdr_t);
uint32_t tmp = substream << 22;
tmp |= 0x3fffff & tx->buffer;
// see definition in rtp_callback.h
rtp_hdr[0] = htonl(tmp);
rtp_hdr[2] = htonl(tile->data_len);
rtp_hdr[3] = htonl(
frame->fec_params.k << 19 |
frame->fec_params.m << 6 |
frame->fec_params.c);
rtp_hdr[4] = htonl(frame->fec_params.seed);
}
if (tx->encryption) {
hdrs_len += sizeof(crypto_payload_hdr_t) + tx->enc_funcs->get_overhead(tx->encryption);
rtp_hdr[rtp_hdr_len / sizeof(uint32_t)] =
htonl(tx->enc_funcs->get_cipher(tx->encryption) << 24);
rtp_hdr_len += sizeof(crypto_payload_hdr_t);
}
vector<int> packet_sizes = get_packet_sizes(frame, substream, tx->mtu - hdrs_len);
long mult_pkt_cnt = (long) packet_sizes.size() * tx->mult_count;
const long packet_rate =
get_packet_rate(tx, frame, (int) substream, mult_pkt_cnt);
// initialize header array with values (except offset which is different among
// different packts)
void *rtp_headers = malloc((mult_pkt_cnt + 1) * rtp_hdr_len);
uint32_t *rtp_hdr_packet = (uint32_t *) rtp_headers;
for (int m = 0; m < tx->mult_count; ++m) {
unsigned pos = 0;
for (unsigned i = 0; i < packet_sizes.size(); ++i) {
memcpy(rtp_hdr_packet, rtp_hdr, rtp_hdr_len);
rtp_hdr_packet[1] = htonl(pos);
rtp_hdr_packet += rtp_hdr_len / sizeof(uint32_t);
pos += packet_sizes.at(i);
}
}
if (tx->fec_dup_1st_pkt) { // dup 1st pkt with RS/LDGM
mult_pkt_cnt += 1;
memcpy(rtp_hdr_packet, rtp_hdr, rtp_hdr_len);
rtp_hdr_packet[1] = htonl(0);
}
if (!tx->encryption) {
rtp_async_start(rtp_session, mult_pkt_cnt);
}
rtp_hdr_packet = (uint32_t *) rtp_headers;
for (long i = 0; i < mult_pkt_cnt; ++i) {
GET_STARTTIME;
const int m = i == mult_pkt_cnt - 1 ? send_m : 0;
char *data = tile->data + ntohl(rtp_hdr_packet[1]);
int data_len = packet_sizes.at(i % packet_sizes.size());
char encrypted_data[RTP_MAX_PACKET_LEN + MAX_CRYPTO_EXCEED];
if (tx->encryption != nullptr) {
data_len = tx->enc_funcs->encrypt(
tx->encryption, data, data_len,
(char *) rtp_hdr_packet,
frame->fec_params.type != FEC_NONE
? sizeof(fec_payload_hdr_t)
: sizeof(video_payload_hdr_t),
encrypted_data);
if (data_len <= 0) {
return;
}
data = encrypted_data;
}
rtp_send_data_hdr(rtp_session, ts, pt, m, 0, nullptr,
(char *) rtp_hdr_packet, rtp_hdr_len, data,
data_len, nullptr, 0, 0);
rtp_hdr_packet += rtp_hdr_len / sizeof(uint32_t);
// TRAFFIC SHAPER
if (m != 1) { // wait for all but last packet
do {
GET_STOPTIME;
GET_DELTA;
} while (packet_rate - delta - overslept > 0);
overslept = -(packet_rate - delta - overslept);
//fprintf(stdout, "%ld ", overslept);
}
}
const long data_sent = tile->data_len + rtp_hdr_len * mult_pkt_cnt;
report_stats(tx, rtp_session, data_sent);
if (!tx->encryption) {
rtp_async_wait(rtp_session);
}
free(rtp_headers);
}
static void audio_tx_send_chan(struct tx *tx, struct rtp *rtp_session,
uint32_t timestamp, const audio_frame2 *buffer,
int channel, bool send_m);
/*
* This multiplication scheme relies upon the fact, that our RTP/pbuf implementation is
* not sensitive to packet duplication. Otherwise, we can get into serious problems.
*/
void
audio_tx_send(struct tx *tx, struct rtp *rtp_session,
const audio_frame2 *buffer)
{
if (!rtp_has_receiver(rtp_session)) {
return;
}
fec_check_messages(tx);
const uint32_t timestamp =
buffer->get_timestamp() == -1
? get_local_mediatime()
: get_local_mediatime_offset() + buffer->get_timestamp();
for (int iter = 0; iter < tx->mult_count; ++iter) {
for (int chan = 0; chan < buffer->get_channel_count(); ++chan) {
bool send_m = iter == tx->mult_count - 1 &&
chan == buffer->get_channel_count() - 1;
audio_tx_send_chan(tx, rtp_session, timestamp, buffer,
chan, send_m);
}
}
tx->buffer++;
}
static void
audio_tx_send_chan(struct tx *tx, struct rtp *rtp_session, uint32_t timestamp,
const audio_frame2 *buffer, int channel, bool send_m)
{
int pt = fec_pt_from_fec_type(
TX_MEDIA_AUDIO, buffer->get_fec_params(0).type,
tx->encryption); /* PT set for audio in our packet format */
unsigned m = 0U;
// see definition in rtp_callback.h
uint32_t rtp_hdr[100];
int rtp_hdr_len = 0;
int hdrs_len = get_tx_hdr_len(rtp_is_ipv6(rtp_session));
unsigned int fec_symbol_size =
buffer->get_fec_params(channel).symbol_size;
const char *chan_data = buffer->get_data(channel);
unsigned pos = 0U;
if (buffer->get_fec_params(0).type == FEC_NONE) {
hdrs_len += (sizeof(audio_payload_hdr_t));
rtp_hdr_len = sizeof(audio_payload_hdr_t);
format_audio_header(buffer, channel, tx->buffer, rtp_hdr);
} else {
hdrs_len += (sizeof(fec_payload_hdr_t));
rtp_hdr_len = sizeof(fec_payload_hdr_t);
uint32_t tmp = channel << 22;
tmp |= 0x3fffff & tx->buffer;
// see definition in rtp_callback.h
rtp_hdr[0] = htonl(tmp);
rtp_hdr[2] = htonl(buffer->get_data_len(channel));
rtp_hdr[3] = htonl(buffer->get_fec_params(channel).k << 19 |
buffer->get_fec_params(channel).m << 6 |
buffer->get_fec_params(channel).c);
rtp_hdr[4] = htonl(buffer->get_fec_params(channel).seed);
}
if (tx->encryption) {
hdrs_len += sizeof(crypto_payload_hdr_t) +
tx->enc_funcs->get_overhead(tx->encryption);
rtp_hdr[rtp_hdr_len / sizeof(uint32_t)] =
htonl(tx->enc_funcs->get_cipher(tx->encryption) << 24);
rtp_hdr_len += sizeof(crypto_payload_hdr_t);
}
if (buffer->get_fec_params(0).type != FEC_NONE) {
check_symbol_size(fec_symbol_size, tx->mtu - hdrs_len);
}
long data_sent = 0;
const int max_len = tx->mtu - hdrs_len;
do {
const char *data = chan_data + pos;
int data_len = max_len;
if (pos + data_len >=
(unsigned int) buffer->get_data_len(channel)) {
data_len = buffer->get_data_len(channel) - pos;
if (send_m) {
m = 1;
}
}
rtp_hdr[1] = htonl(pos);
pos += data_len;
char encrypted_data[RTP_MAX_PACKET_LEN + MAX_CRYPTO_EXCEED];
if (tx->encryption) {
data_len = tx->enc_funcs->encrypt(
tx->encryption, const_cast<char *>(data), data_len,
(char *) rtp_hdr,
rtp_hdr_len - sizeof(crypto_payload_hdr_t),
encrypted_data);
if (data_len <= 0) {
return;
}
data = encrypted_data;
}
data_sent += data_len + rtp_hdr_len;
rtp_send_data_hdr(rtp_session, timestamp, pt, m,
0, /* contributing sources */
0, /* contributing sources length */
(char *) rtp_hdr, rtp_hdr_len,
const_cast<char *>(data), data_len, 0, 0, 0);
} while (pos < buffer->get_data_len(channel));
report_stats(tx, rtp_session, data_sent);
if (buffer->get_fec_params(0).type == FEC_NONE) {
return;
}
// issue a warning if R-S is inadequate
const int packet_count =
(buffer->get_data_len(channel) + max_len - 1) / max_len;
if (packet_count > 3) {
return;
}
const char *pl_suffix = packet_count == 1 ? "" : "s";
log_msg_once(LOG_LEVEL_WARNING, to_fourcc('t', 'x', 'a', 'F'),
MOD_NAME
"[audio] %d packet%s per audio channel may be too low "
"for Reed Solomon, consider mult instead!%s\n",
packet_count, pl_suffix,
packet_count == 3 ? " (Or increase the redundancy.)" : "");
}
static bool
validate_std_audio(const audio_frame2 * buffer, int payload_size)
{
if ((buffer->get_codec() == AC_MP3 ||
buffer->get_codec() == AC_OPUS) &&
buffer->get_channel_count() > 1) { // we cannot interleave Opus here
const uint32_t msg_id = to_fourcc('t', 'x', 'v', 'a');
log_msg_once(LOG_LEVEL_ERROR, msg_id,
MOD_NAME
"%s can currently have only 1 channel in "
"RFC-compliant mode! Discarding channels but the "
"first one...\n",
get_name_to_audio_codec(buffer->get_codec()));
}
if (buffer->get_codec() == AC_OPUS &&
payload_size < (int) buffer->get_data_len(0)) {
MSG(ERROR, "Opus frame larger than packet! Discarding...\n");
return false;
}
return true;
}
/**
* audio_tx_send_standard - Send interleaved channels from the audio_frame2,
* as the mulaw and A-law standards (dynamic or std PT).
*/
void audio_tx_send_standard(struct tx* tx, struct rtp *rtp_session,
const audio_frame2 * buffer) {
//TODO to be more abstract in order to accept A-law too and other supported standards with such implementation
assert(buffer->get_codec() == AC_MULAW ||
buffer->get_codec() == AC_ALAW ||
buffer->get_codec() == AC_MP3 ||
buffer->get_codec() == AC_OPUS);
uint32_t ts;
static uint32_t ts_prev = 0;
enum {
PCMA_U_BPS = 1, // data BPS - get_bps() is 2 to expand to 16-bit
// format on the receiver. The semantic of 2 is
// the same as for other codecs like Opus to
// asses the decmopressed sample size.
};
const bool is_pcma_u =
buffer->get_codec() == AC_MULAW || buffer->get_codec() == AC_ALAW;
// Configure the right Payload type,
// 8000 Hz, 1 channel PCMU/A is the ITU-T G.711 standard
// Other channels or Hz goes to DynRTP-Type97
int pt = PT_DynRTP_Type97;
if (is_pcma_u && buffer->get_channel_count() == 1 &&
buffer->get_sample_rate() == kHz8) {
pt = buffer->get_codec() == AC_MULAW ? PT_ITU_T_G711_PCMU
: PT_ITU_T_G711_PCMA;
}
if (buffer->get_codec() == AC_MP3) {
pt = PT_MPA;
}
int data_len = buffer->get_data_len(0); /* Number of samples to send */
int payload_size = tx->mtu - 40 - 8 - 12; /* Max size of an RTP payload field (minus IPv6, UDP and RTP header lengths) */
if (is_pcma_u) { // we may split the data into more packets, compute
// chunk size
const int frame_size = buffer->get_channel_count() * PCMA_U_BPS;
payload_size = payload_size / frame_size * frame_size; // align to frame size
// The sizes for the different channels must be the same.
for (int i = 1; i < buffer->get_channel_count(); i++) {
assert(buffer->get_data_len(0) ==
buffer->get_data_len(i));
}
data_len *= buffer->get_channel_count();
} else if (pt == PT_MPA) {
payload_size -= sizeof(mpa_hdr_t);
}
if (!validate_std_audio(buffer, payload_size)) {
return;
}
int pos = 0;
do {
int pkt_len = std::min(payload_size, data_len - pos);
if (buffer->get_codec() == AC_OPUS) {
memcpy(tx->tmp_packet, buffer->get_data(0), pkt_len);
} else if (buffer->get_codec() == AC_MP3) {
memset(tx->tmp_packet, 0, 2);
const uint16_t offset = htons(pos);
memcpy(tx->tmp_packet + 2, &offset, sizeof offset);
pkt_len += sizeof(mpa_hdr_t);
memcpy(tx->tmp_packet + 4, buffer->get_data(0), pkt_len);
} else { // interleave
for (int ch = 0; ch < buffer->get_channel_count(); ch++) {
remux_channel(
tx->tmp_packet,
buffer->get_data(ch) +
pos / buffer->get_channel_count(),
PCMA_U_BPS,
pkt_len / buffer->get_channel_count(), 1,
buffer->get_channel_count(), 0, ch);
}
}
// Update first sample timestamp
if (buffer->get_codec() == AC_OPUS) {
/* OPUS packet will be the whole contained in one packet
* according to RFC 7587. For PCMA/PCMU there may be more
* packets so we cannot use the whole frame duration. */
ts = get_std_audio_local_mediatime(buffer->get_duration(), 48000);
} else {
ts = get_std_audio_local_mediatime((double) pkt_len / (double) buffer->get_channel_count() / (double) buffer->get_sample_rate(), buffer->get_sample_rate());
}
rtp_send_ctrl(rtp_session, ts_prev, 0, get_time_in_ns()); //send RTCP SR
ts_prev = ts;
// Send the packet
rtp_send_data(rtp_session, ts, pt, 0, 0, /* contributing sources */
0, /* contributing sources length */
tx->tmp_packet, pkt_len, 0, 0, 0);
pos += pkt_len;
} while (pos < data_len);
}
/**
* H.264 standard transmission
*/
void tx_send_h264(struct tx *tx, struct video_frame *frame,
struct rtp *rtp_session) {
assert(frame->tile_count == 1); // std transmit doesn't handle more than one tile
assert(!frame->fragment || tx->fec_scheme == FEC_NONE); // currently no support for FEC with fragments
assert(!frame->fragment || frame->tile_count); // multiple tiles are not currently supported for fragmented send
uint32_t ts = get_std_video_local_mediatime();
struct tile *tile = &frame->tiles[0];
char pt = PT_DynRTP_Type96;
unsigned char hdr[2];
int cc = 0;
uint32_t csrc = 0;
int m = 0;
char *extn = 0;
uint16_t extn_len = 0;
uint16_t extn_type = 0;
const uint8_t *start = (uint8_t *) tile->data;
int data_len = tile->data_len;
unsigned maxPacketSize = tx->mtu - 40;
const unsigned char *endptr = 0;
const unsigned char *nal = start;
while ((nal = rtpenc_get_next_nal(nal, data_len - (nal - start), &endptr))) {
unsigned int nalsize = endptr - nal;
bool eof = endptr == start + data_len;
bool lastNALUnitFragment = false; // by default
unsigned curNALOffset = 0;
char *nalc = const_cast<char *>(reinterpret_cast<const char *>(nal));
while(!lastNALUnitFragment){
// We have NAL unit data in the buffer. There are three cases to consider:
// 1. There is a new NAL unit in the buffer, and it's small enough to deliver
// to the RTP sink (as is).
// 2. There is a new NAL unit in the buffer, but it's too large to deliver to
// the RTP sink in its entirety. Deliver the first fragment of this data,
// as a FU packet, with one extra preceding header byte (for the "FU header").
// 3. There is a NAL unit in the buffer, and we've already delivered some
// fragment(s) of this. Deliver the next fragment of this data,
// as a FU packet, with two (H.264) extra preceding header bytes
// (for the "NAL header" and the "FU header").
if (curNALOffset == 0) { // case 1 or 2
if (nalsize <= maxPacketSize) { // case 1
if (eof) m = 1;
if (rtp_send_data(rtp_session, ts, pt, m, cc, &csrc,
nalc, nalsize,
extn, extn_len, extn_type) < 0) {
error_msg("There was a problem sending the RTP packet\n");
}
lastNALUnitFragment = true;
} else { // case 2
// We need to send the NAL unit data as FU packets. Deliver the first
// packet now. Note that we add "NAL header" and "FU header" bytes to the front
// of the packet (overwriting the existing "NAL header").
hdr[0] = (nal[0] & 0xE0) | 28; //FU indicator
hdr[1] = 0x80 | (nal[0] & 0x1F); // FU header (with S bit)
if (rtp_send_data_hdr(rtp_session, ts, pt, m, cc, &csrc,
(char *) hdr, 2,
nalc + 1, maxPacketSize - 2,
extn, extn_len, extn_type) < 0) {
error_msg("There was a problem sending the RTP packet\n");
}
curNALOffset += maxPacketSize - 1;
lastNALUnitFragment = false;
nalsize -= maxPacketSize - 1;
}
} else { // case 3
// We are sending this NAL unit data as FU packets. We've already sent the
// first packet (fragment). Now, send the next fragment. Note that we add
// "NAL header" and "FU header" bytes to the front. (We reuse these bytes that
// we already sent for the first fragment, but clear the S bit, and add the E
// bit if this is the last fragment.)
hdr[1] = hdr[1] & ~0x80;// FU header (no S bit)
if (nalsize + 1 > maxPacketSize) {
// We can't send all of the remaining data this time:
if (rtp_send_data_hdr(rtp_session, ts, pt, m, cc, &csrc,
(char *) hdr, 2,
nalc + curNALOffset,
maxPacketSize - 2, extn, extn_len,
extn_type) < 0) {
error_msg("There was a problem sending the RTP packet\n");
}
curNALOffset += maxPacketSize - 2;
lastNALUnitFragment = false;
nalsize -= maxPacketSize - 2;
} else {
// This is the last fragment:
if (eof) m = 1;
hdr[1] |= 0x40;// set the E bit in the FU header
if (rtp_send_data_hdr(rtp_session, ts, pt, m, cc, &csrc,
(char *) hdr, 2,
nalc + curNALOffset,
nalsize, extn, extn_len, extn_type) < 0) {
error_msg("There was a problem sending the RTP packet\n");
}
lastNALUnitFragment = true;
}
}
}
}
if (endptr != start + data_len) {
error_msg("No NAL found!\n");
}
}
void tx_send_jpeg(struct tx *tx, struct video_frame *frame,
struct rtp *rtp_session) {
uint32_t ts = 0;
assert(frame->tile_count == 1); // std transmit doesn't handle more than one tile
assert(!frame->fragment || tx->fec_scheme == FEC_NONE); // currently no support for FEC with fragments
assert(!frame->fragment || frame->tile_count); // multiple tiles are not currently supported for fragmented send
ts = get_std_video_local_mediatime();
struct tile *tile = &frame->tiles[0];
char pt = PT_JPEG;
struct jpeg_rtp_data d;
if (!jpeg_get_rtp_hdr_data((uint8_t *) frame->tiles[0].data, frame->tiles[0].data_len, &d)) {
exit_uv(1);
return;
}
uint32_t jpeg_hdr[2 /* JPEG hdr */ + 1 /* RM hdr */ + 129 /* QT hdr */];
int hdr_off = 0;
unsigned int type_spec = 0u;
jpeg_hdr[hdr_off++] = htonl(type_spec << 24u);
jpeg_hdr[hdr_off++] = htonl(d.type << 24u | d.q << 16u | d.width / 8u << 8u | d.height / 8u);
if (d.restart_interval != 0) {
// we do not align restart interval on packet boundaries yet
jpeg_hdr[hdr_off++] = htonl(d.restart_interval << 16u | 1u << 15u | 1u << 14u | 0x3fffu);
}
// quantization headers
if (d.q == 255u) { // we must include the tables
unsigned int mbz = 0u; // must be zero
unsigned int precision = 0u;
unsigned int qt_len = 2 * 64u;
jpeg_hdr[hdr_off++] = htonl(mbz << 24u | precision << 16u | qt_len);
memcpy(&jpeg_hdr[hdr_off], d.quantization_tables[0], 64);
hdr_off += 64 / sizeof(uint32_t);
memcpy(&jpeg_hdr[hdr_off], d.quantization_tables[1], 64);
hdr_off += 64 / sizeof(uint32_t);
}
char *data = (char *) d.data;
int bytes_left = tile->data_len - ((char *) d.data - tile->data);
int max_mtu = tx->mtu - ((rtp_is_ipv6(rtp_session) ? 40 : 20) + 8 + 12); // IP hdr size + UDP hdr size + RTP hdr size
int fragment_offset = 0;
do {
int hdr_len;
if (fragment_offset == 0) { // include quantization header only in 1st pkt
hdr_len = hdr_off * sizeof(uint32_t);
} else {
hdr_len = 8 + (d.restart_interval > 0 ? 4 : 0);
}
int data_len = max_mtu - hdr_len;
int m = 0;
if (bytes_left <= data_len) {
data_len = bytes_left;
m = 1;
}
jpeg_hdr[0] = htonl(type_spec << 24u | fragment_offset);
int ret = rtp_send_data_hdr(rtp_session, ts, pt, m, 0, 0,
(char *) &jpeg_hdr, hdr_len,
data, data_len, 0, 0, 0);
if (ret < 0) {
log_msg(LOG_LEVEL_ERROR, "Error sending RTP/JPEG packet!\n");
}
data += data_len;
bytes_left -= data_len;
fragment_offset += data_len;
} while (bytes_left > 0);
}
int tx_get_buffer_id(struct tx *tx)
{
return tx->buffer;
}