pipewire/src/modules/module-rtp-source.c

/* PipeWire */
/* SPDX-FileCopyrightText: Copyright © 2022 Wim Taymans <wim.taymans@gmail.com> */
/* SPDX-License-Identifier: MIT */

#include "config.h"

#include <limits.h>
#include <string.h>
#include <unistd.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/ioctl.h>
#include <arpa/inet.h>
#include <netinet/in.h>
#include <net/if.h>
#include <ctype.h>

#include <spa/utils/atomic.h>
#include <spa/utils/hook.h>
#include <spa/utils/result.h>
#include <spa/utils/ringbuffer.h>
#include <spa/utils/defs.h>
#include <spa/utils/dll.h>
#include <spa/utils/json.h>
#include <spa/utils/ratelimit.h>
#include <spa/param/audio/format-utils.h>
#include <spa/control/control.h>
#include <spa/debug/types.h>
#include <spa/debug/mem.h>

#include <pipewire/pipewire.h>
#include <pipewire/impl.h>

#include <module-rtp/stream.h>
#include "network-utils.h"

/** \page page_module_rtp_source RTP source
 *
 * The `rtp-source` module creates a PipeWire source that receives audio RTP packets.
 * These RTP packets may contain raw PCM data, Opus encoded audio, or midi audio.
 *
 * This module is usually loaded from the \ref page_module_rtp_sap so that the
 * source.ip and source.port and format parameters matches that of the sender that
 * is announced via SAP.
 *
 * ## Module Name
 *
 * `libpipewire-module-rtp-source`
 *
 * ## Module Options
 *
 * Options specific to the behavior of this module
 *
 * - `local.ifname = <str>`: interface name to use
 * - `source.ip = <str>`: the source ip address, default 224.0.0.56. Set this to the IP address
 *                you want to receive packets from or 0.0.0.0 to receive from any source address.
 * - `source.port = <int>`: the source port
 * - `node.always-process = <bool>`: true to receive even when not running
 * - `sess.latency.msec = <float>`: target network latency in milliseconds, default 100
 * - `sess.ignore-ssrc = <bool>`: ignore SSRC, default false
 * - `sess.media = <string>`: the media type audio|midi|opus, default audio
 * - `sess.ts-direct = <bool>`: use direct timestamp mode, default false
 *                (see the Buffer Modes section below)
 * - `stream.may-pause = <bool>`: pause the stream when no data is reveived, default false
 * - `stream.props = {}`: properties to be passed to the stream
 *
 * ## General options
 *
 * Options with well-known behavior:
 *
 * - \ref PW_KEY_REMOTE_NAME
 * - \ref PW_KEY_AUDIO_FORMAT
 * - \ref PW_KEY_AUDIO_RATE
 * - \ref PW_KEY_AUDIO_CHANNELS
 * - \ref SPA_KEY_AUDIO_POSITION
 * - \ref PW_KEY_MEDIA_NAME
 * - \ref PW_KEY_MEDIA_CLASS
 * - \ref PW_KEY_NODE_NAME
 * - \ref PW_KEY_NODE_DESCRIPTION
 * - \ref PW_KEY_NODE_GROUP
 * - \ref PW_KEY_NODE_LATENCY
 * - \ref PW_KEY_NODE_VIRTUAL
 *
 * ## Example configuration
 *\code{.unparsed}
 * # ~/.config/pipewire/pipewire.conf.d/my-rtp-source.conf
 *
 * context.modules = [
 * {   name = libpipewire-module-rtp-source
 *     args = {
 *         #local.ifname = eth0
 *         #source.ip = 224.0.0.56
 *         #source.port = 0
 *         sess.latency.msec = 100
 *         #sess.ignore-ssrc = false
 *         #node.always-process = false
 *         #sess.media = "audio"
 *         #audio.format = "S16BE"
 *         #audio.rate = 48000
 *         #audio.channels = 2
 *         #audio.position = [ FL FR ]
 *         stream.props = {
 *            #media.class = "Audio/Source"
 *            node.name = "rtp-source"
 *         }
 *     }
 * }
 * ]
 *\endcode
 *
 * ## Buffer modes
 *
 * RTP source nodes created by this module use an internal ring buffer. Received RTP audio
 * data is written into this ring buffer. When the node's process callback is run, it reads
 * from that ring buffer and provides audio data from it to the graph.
 *
 * The `sess.ts-direct` option controls the _buffer mode_, which defines how this ring buffer
 * is used. The RTP source nodes created by this module can operate in one of two of these
 * buffer modes. In both modes, the RTP source node uses the timestamps of incoming RTP
 * packets to write into the ring buffer (more specifically, at the position
 * `timestamp + latency from the sess.latency.msec option`). The modes are:
 *
 * -# *Constant latency mode*: This is the default mode. It is used when `sess.ts-direct`
 *    is set to false. `sess.latency.msec` then defines the ideal fill level of the ring
 *    buffer. If the fill level is above or below this, then a DLL is used to adjust the
 *    consumption of the buffer contents. If the fill level is below a critical value
 *    (that's the amount of data that is needed in a cycle), or if the fill level equals
 *    the total buffer size (meaning that no more data can be fed into the buffer), the
 *    buffer contents are resynchronized, meaning that the existing contents are thrown
 *    away, and the ring buffer is reset. This buffer mode is useful for when a constant
 *    latency is desired, and the actual moment playback starts is unimportant (meaning
 *    that playback is not necessarily in sync with other devices). This mode requires
 *    no special graph driver.
 * -# *Direct timestamp mode*: This is an alternate mode, used when `sess.ts-direct` is
 *    set to true. In this mode, ring buffer over- and underrun and fill level are not
 *    directly tracked; instead, they are handled implicitly. There is no constant latency
 *    maintained. The current time (more specifically, the \ref spa_io_clock::position field
 *    of \ref spa_io_position::clock) is directly used during playback to retrieve audio
 *    data. This assumes that a graph driver is used whose time is somehow synchronized
 *    to the sender's. Since the current time is directly used as an offset within the
 *    ring buffer, the correct data is always pulled from the ring buffer, that is, the
 *    data that shall be played now, in sync with the sender (and with other receivers).
 *    This buffer mode is useful for when receivers shall play in sync with each other,
 *    and shall use one common synchronized time, provided through the \ref spa_io_clock .
 *    `sess.latency.msec` functions as a configurable assumed maximum transport delay
 *    instead of a constant latency quantity in this mode. The DLL is not used in this
 *    mode, since the graph driver is assumed to be synchronized to the sender, as said,
 *    so any output sinks in the graph will already adjust their consumption pace to
 *    match the pace of the graph driver.
 *    AES67 sessions use this mode, for example.
 *
 * \since 0.3.60
 */

#define NAME "rtp-source"

PW_LOG_TOPIC(mod_topic, "mod." NAME);
#define PW_LOG_TOPIC_DEFAULT mod_topic

#define DEFAULT_IGMP_CHECK_INTERVAL_SEC	5
#define DEFAULT_IGMP_DEADLINE_SEC	30

#define DEFAULT_CLEANUP_SEC		60
#define DEFAULT_SOURCE_IP		"224.0.0.56"

#define DEFAULT_TS_OFFSET		-1

#define USAGE   "( local.ifname=<local interface name to use> ) "						\
		"( source.ip=<source IP address, default:"DEFAULT_SOURCE_IP"> ) "				\
 		"source.port=<int, source port> "								\
		"( sess.latency.msec=<target network latency, default "SPA_STRINGIFY(DEFAULT_SESS_LATENCY)"> ) "\
		"( sess.ignore-ssrc=<to ignore SSRC, default false> ) "\
 		"( sess.media=<string, the media type audio|midi|opus, default audio> ) "			\
		"( audio.format=<format, default:"DEFAULT_FORMAT"> ) "						\
		"( audio.rate=<sample rate, default:"SPA_STRINGIFY(DEFAULT_RATE)"> ) "				\
		"( audio.channels=<number of channels, default:"SPA_STRINGIFY(DEFAULT_CHANNELS)"> ) "		\
		"( audio.position=<channel map, default:"DEFAULT_POSITION"> ) "					\
		"( stream.props= { key=value ... } ) "

static const struct spa_dict_item module_info[] = {
	{ PW_KEY_MODULE_AUTHOR, "Wim Taymans <wim.taymans@gmail.com>" },
	{ PW_KEY_MODULE_DESCRIPTION, "RTP Source" },
	{ PW_KEY_MODULE_USAGE,	USAGE },
	{ PW_KEY_MODULE_VERSION, PACKAGE_VERSION },
};

struct igmp_recovery {
	struct pw_timer timer;
	int socket_fd;
	struct sockaddr_storage mcast_addr;
	socklen_t mcast_len;
	uint32_t if_index;
	bool is_ipv6;
	/* This is the interval the recovery timer runs at. The timer
	 * checks at each interval if recovery is required. This value
	 * is defined by the igmp.check.interval.sec property. */
	uint32_t check_interval;
	/* This is the deadline for packets to arrive. If the deadline
	 * is exceeded, an IGMP recovery is attempted. This value is
	 * defined by the igmp.deadline.sec property. */
	uint32_t deadline;
};

struct impl {
	struct pw_impl_module *module;
	struct spa_hook module_listener;
	struct pw_properties *props;
	struct pw_context *context;

	struct pw_loop *main_loop;
	struct pw_loop *data_loop;
	struct pw_timer_queue *timer_queue;

	struct pw_core *core;
	struct spa_hook core_listener;
	struct spa_hook core_proxy_listener;
	unsigned int do_disconnect:1;

	struct spa_ratelimit rate_limit;

	char *ifname;
	bool always_process;
	uint32_t cleanup_interval;

	/* IGMP recovery (triggers when no RTP packets are
	 * received after the recovery deadline is reached) */
	struct igmp_recovery igmp_recovery;

	/* Monotonic timestamp of the last time a packet was
	 * received. This is accessed with atomic accessors
	 * to avoid race conditions. */
	uint64_t last_packet_time;

	struct pw_timer standby_timer;
	/* This timer is used when the first stream_start() call fails because
	 * of an ENODEV error (see the stream_start() code for details) */
	struct pw_timer stream_start_retry_timer;

	struct pw_properties *stream_props;
	struct rtp_stream *stream;

	uint16_t src_port;
	struct sockaddr_storage src_addr;
	socklen_t src_len;
	struct spa_source *source;

	uint8_t *buffer;
	size_t buffer_size;

#define STATE_IDLE	0
#define STATE_PROBE	1
#define STATE_RECEIVING	2
#define STATE_STOPPING	3
	int state;
	bool may_pause;
	bool standby;
	bool waiting;
};

static int do_start(struct spa_loop *loop, bool async, uint32_t seq, const void *data,
		size_t size, void *user_data)
{
	struct impl *impl = user_data;
	if (impl->waiting) {
		struct spa_dict_item item[1];

		impl->waiting = false;
		impl->standby = false;

		pw_log_info("resume RTP source");

		item[0] = SPA_DICT_ITEM_INIT("rtp.receiving", "true");
		rtp_stream_update_properties(impl->stream, &SPA_DICT_INIT(item, 1));

		if (impl->may_pause)
			rtp_stream_set_active(impl->stream, true);
	}
	return 0;
}

static void
on_rtp_io(void *data, int fd, uint32_t mask)
{
	struct impl *impl = data;
	ssize_t len;
	int suppressed;
	uint64_t current_time;

	current_time = rtp_stream_get_nsec(impl->stream);

	if (mask & SPA_IO_IN) {
		if ((len = recv(fd, impl->buffer, impl->buffer_size, 0)) < 0)
			goto receive_error;

		if (len < 12)
			goto short_packet;

		if (SPA_LIKELY(impl->stream)) {
			if (rtp_stream_receive_packet(impl->stream, impl->buffer, len,
							current_time) < 0)
				goto receive_error;
		}

		/* Update last packet timestamp for IGMP recovery.
		 * The recovery timer will check this to see if recovery
		 * is necessary. Do this _before_ invoking do_start()
		 * in case the stream is waking up from standby. */
		SPA_ATOMIC_STORE(impl->last_packet_time, current_time);

		if (SPA_ATOMIC_LOAD(impl->state) != STATE_RECEIVING) {
			if (!SPA_ATOMIC_CAS(impl->state, STATE_PROBE, STATE_RECEIVING)) {
				if (SPA_ATOMIC_CAS(impl->state, STATE_IDLE, STATE_RECEIVING))
					pw_loop_invoke(impl->main_loop, do_start, 1, NULL, 0, false, impl);
			}
		}
	}
	return;

receive_error:
	if ((suppressed = spa_ratelimit_test(&impl->rate_limit, current_time)) >= 0)
		pw_log_warn("(%d suppressed) recv() error: %m", suppressed);
	return;
short_packet:
	if ((suppressed = spa_ratelimit_test(&impl->rate_limit, current_time)) >= 0)
		pw_log_warn("(%d suppressed) short packet of len %zd received",
				suppressed, len);
	return;
}

static int rejoin_igmp_group(struct spa_loop *loop, bool async, uint32_t seq,
				const void *data, size_t size, void *user_data)
{
	/* IMPORTANT: This must be run from within the data loop. */

	int res;
	struct impl *impl = user_data;
	uint64_t current_time;

	/* Force IGMP membership refresh by leaving the group first, then rejoin */
	if (impl->igmp_recovery.is_ipv6) {
		struct ipv6_mreq mr6;
		memset(&mr6, 0, sizeof(mr6));
		mr6.ipv6mr_multiaddr = ((struct sockaddr_in6*)&impl->igmp_recovery.mcast_addr)->sin6_addr;
		mr6.ipv6mr_interface = impl->igmp_recovery.if_index;

		/* Leave the group first */
		res = setsockopt(impl->igmp_recovery.socket_fd, IPPROTO_IPV6, IPV6_LEAVE_GROUP,
					&mr6, sizeof(mr6));
		if (SPA_LIKELY(res == 0)) {
			pw_log_info("left IPv6 multicast group");
		} else {
			if (errno == EADDRNOTAVAIL) {
				pw_log_info("attempted to leave IPv6 multicast group, but "
						"membership was already silently dropped");
			} else {
				pw_log_warn("failed to leave IPv6 multicast group: %m");
			}
		}

		res = setsockopt(impl->igmp_recovery.socket_fd, IPPROTO_IPV6, IPV6_JOIN_GROUP,
				&mr6, sizeof(mr6));
		if (res < 0) {
			pw_log_warn("failed to re-join IPv6 multicast group: %m");
		} else {
			pw_log_info("re-joined IPv6 multicast group successfully");
		}
	} else {
		struct ip_mreqn mr4;
		memset(&mr4, 0, sizeof(mr4));
		mr4.imr_multiaddr = ((struct sockaddr_in*)&impl->igmp_recovery.mcast_addr)->sin_addr;
		mr4.imr_ifindex = impl->igmp_recovery.if_index;

		/* Leave the group first */
		res = setsockopt(impl->igmp_recovery.socket_fd, IPPROTO_IP, IP_DROP_MEMBERSHIP,
					&mr4, sizeof(mr4));
		if (SPA_LIKELY(res == 0)) {
			pw_log_info("left IPv4 multicast group");
		} else {
			if (errno == EADDRNOTAVAIL) {
				pw_log_info("attempted to leave IPv4 multicast group, but "
						"membership was already silently dropped");
			} else {
				pw_log_warn("failed to leave IPv4 multicast group: %m");
			}
		}

		res = setsockopt(impl->igmp_recovery.socket_fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
				&mr4, sizeof(mr4));
		if (res < 0) {
			pw_log_warn("failed to re-join IPv4 multicast group: %m");
		} else {
			pw_log_info("re-joined IPv4 multicast group successfully");
		}
	}

	current_time = rtp_stream_get_nsec(impl->stream);
	SPA_ATOMIC_STORE(impl->last_packet_time, current_time);

	return res;
}

static void on_igmp_recovery_timer_event(void *data)
{
	int res;
	struct impl *impl = data;
	char addr[128];
	uint64_t current_time, elapsed_seconds, last_packet_time;

	/* Only attempt recovery if we have a valid socket and multicast address */
	if (SPA_UNLIKELY(impl->igmp_recovery.socket_fd < 0)) {
		pw_log_trace("no socket, skipping IGMP recovery");
		goto finish;
	}

	/* This check if performed even if standby = false or
	 * receiving != STATE_RECEIVING , because the very reason
	 * for these states may be that the receiver socket was
	 * silently kicked out of the IGMP group (which causes data
	 * to no longer arrive, thus leading to these states). */

	current_time = rtp_stream_get_nsec(impl->stream);
	last_packet_time = SPA_ATOMIC_LOAD(impl->last_packet_time);
	elapsed_seconds = (current_time - last_packet_time) / SPA_NSEC_PER_SEC;

	/* Only trigger recovery if enough time has elapsed since last packet */
	if (elapsed_seconds < impl->igmp_recovery.deadline) {
		pw_log_trace("IGMP recovery check: %" PRIu64 " seconds elapsed, "
				"need %" PRIu32 " seconds", elapsed_seconds,
				impl->igmp_recovery.deadline);
		goto finish;
	}

	pw_net_get_ip(&impl->igmp_recovery.mcast_addr, addr, sizeof(addr), NULL, NULL);
	pw_log_info("starting IGMP recovery for %s", addr);

	/* Run the actual recovery in the data loop, since recovery involves
	 * rejoining the socket to the IGMP group. By running this in the
	 * data loop, race conditions due to stray packets causing an on_rtp_io()
	 * invocation at the same time when the IGMP group rejoining takes place
	 * is avoided, since on_rtp_io() too runs in the data loop.
	 * This is a blocking call to make sure the rejoin attempt was fully
	 * done by the time this callback ends. (rejoin_igmp_group() does not
	 * do work that takes a long time to finish. )*/
	res = pw_loop_locked(impl->data_loop, rejoin_igmp_group, 1, NULL, 0, impl);

	if (SPA_LIKELY(res == 0)) {
		pw_log_info("IGMP recovery for %s finished", addr);
	} else {
		pw_log_error("error while finishing IGMP recovery for %s: %s",
				addr, spa_strerror(res));
	}

finish:
	pw_timer_queue_add(impl->timer_queue, &impl->igmp_recovery.timer,
			&impl->igmp_recovery.timer.timeout,
			impl->igmp_recovery.check_interval * SPA_NSEC_PER_SEC,
			on_igmp_recovery_timer_event, impl);
}

static int make_socket(const struct sockaddr* sa, socklen_t salen, char *ifname,
			struct igmp_recovery *igmp_recovery)
{
	int af, fd, val, res;
	struct ifreq req;
	struct sockaddr_storage ba = *(struct sockaddr_storage *)sa;
	bool do_connect = false;
	char addr[128];

	af = sa->sa_family;
	if ((fd = socket(af, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK, 0)) < 0) {
		res = -errno;
		pw_log_error("socket failed: %m");
		return res;
	}
#ifdef SO_TIMESTAMP
	val = 1;
	if (setsockopt(fd, SOL_SOCKET, SO_TIMESTAMP, &val, sizeof(val)) < 0) {
		res = -errno;
		pw_log_error("setsockopt failed: %m");
		goto error;
	}
#endif
	val = 1;
	if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val)) < 0) {
		res = -errno;
		pw_log_error("setsockopt failed: %m");
		goto error;
	}

	spa_zero(req);
	if (ifname) {
		snprintf(req.ifr_name, sizeof(req.ifr_name), "%s", ifname);
		res = ioctl(fd, SIOCGIFINDEX, &req);
	        if (res < 0)
	                pw_log_warn("SIOCGIFINDEX %s failed: %m", ifname);
	}
	res = 0;
	if (af == AF_INET) {
		static const uint32_t ipv4_mcast_mask = 0xe0000000;
		struct sockaddr_in *sa4 = (struct sockaddr_in*)sa;
		if ((ntohl(sa4->sin_addr.s_addr) & ipv4_mcast_mask) == ipv4_mcast_mask) {
			struct ip_mreqn mr4;
			memset(&mr4, 0, sizeof(mr4));
			mr4.imr_multiaddr = sa4->sin_addr;
			mr4.imr_ifindex = req.ifr_ifindex;
			pw_net_get_ip((struct sockaddr_storage*)sa, addr, sizeof(addr), NULL, NULL);
			pw_log_info("join IPv4 group: %s", addr);
			res = setsockopt(fd, IPPROTO_IP, IP_ADD_MEMBERSHIP, &mr4, sizeof(mr4));
		} else {
			struct sockaddr_in *ba4 = (struct sockaddr_in*)&ba;
			if (ba4->sin_addr.s_addr != INADDR_ANY) {
				ba4->sin_addr.s_addr = INADDR_ANY;
				do_connect = true;
			}
		}
	} else if (af == AF_INET6) {
		struct sockaddr_in6 *sa6 = (struct sockaddr_in6*)sa;
		if (sa6->sin6_addr.s6_addr[0] == 0xff) {
			struct ipv6_mreq mr6;
			memset(&mr6, 0, sizeof(mr6));
			mr6.ipv6mr_multiaddr = sa6->sin6_addr;
			mr6.ipv6mr_interface = req.ifr_ifindex;
			pw_net_get_ip((struct sockaddr_storage*)sa, addr, sizeof(addr), NULL, NULL);
			pw_log_info("join IPv6 group: %s", addr);
			res = setsockopt(fd, IPPROTO_IPV6, IPV6_JOIN_GROUP, &mr6, sizeof(mr6));
		} else {
			struct sockaddr_in6 *ba6 = (struct sockaddr_in6*)&ba;
			ba6->sin6_addr = in6addr_any;
		}
	} else {
		res = -EINVAL;
		goto error;
	}

	if (res < 0) {
		res = -errno;
		pw_log_error("join mcast failed: %m");
		goto error;
	}

	/* Store multicast info for recovery */
	igmp_recovery->socket_fd = fd;
	igmp_recovery->mcast_addr = ba;
	igmp_recovery->mcast_len = salen;
	igmp_recovery->if_index = req.ifr_ifindex;
	igmp_recovery->is_ipv6 = (af == AF_INET6);
	pw_log_debug("stored %s multicast info: socket_fd=%d, "
			"if_index=%d", igmp_recovery->is_ipv6 ?
			"IPv6" : "IPv4", fd, req.ifr_ifindex);

	if (bind(fd, (struct sockaddr*)&ba, salen) < 0) {
		res = -errno;
		pw_log_error("bind() failed: %m");
		goto error;
	}
	if (do_connect) {
		if (connect(fd, sa, salen) < 0) {
			res = -errno;
			pw_log_error("connect() failed: %m");
			goto error;
		}
	}
	return fd;
error:
	close(fd);
	return res;
}

static void stream_report_error(void *data, const char *error)
{
	struct impl *impl = data;
	if (error) {
		pw_log_error("stream error: %s", error);
		pw_impl_module_schedule_destroy(impl->module);
	}
}

static void stream_open_connection(void *data, int *result);

static void on_open_connection_retry_timer_event(void *data)
{
	struct impl *impl = data;
	pw_log_debug("trying again to open connection after previous attempt failed with ENODEV");
	stream_open_connection(impl, NULL);
}

static void stream_open_connection(void *data, int *result)
{
	int res = 0;
	int fd;
	struct impl *impl = data;

	if (impl->source != NULL)
		goto finish;

	pw_log_info("starting RTP listener");

	if ((fd = make_socket((const struct sockaddr *)&impl->src_addr,
					impl->src_len, impl->ifname,
					&(impl->igmp_recovery))) < 0) {
		/* If make_socket() tries to create a socket and join to a multicast
		 * group while the network interfaces are not ready yet to do so
		 * (usually because a network manager component is still setting up
		 * those network interfaces), ENODEV will be returned. This is essentially
		 * a race condition. There is no discernible way to be notified when the
		 * network interfaces are ready for that operation, so the next best
		 * approach is to essentially do a form of polling by retrying the
		 * stream_start() call after some time. The stream_start_retry_timer exists
		 * precisely for that purpose. This means that ENODEV is not treated as
		 * an error, but instead, it triggers the creation of that timer. */
		if (fd == -ENODEV) {
			pw_log_warn("failed to create socket because network device is not ready "
				"and present yet; will try again");

			pw_timer_queue_cancel(&impl->stream_start_retry_timer);
			/* Use a 1-second retry interval. The network interfaces
			 * are likely to be up and running then. */
			pw_timer_queue_add(impl->timer_queue, &impl->stream_start_retry_timer,
					NULL, 1 * SPA_NSEC_PER_SEC,
					on_open_connection_retry_timer_event, impl);

			/* It is important to return 0 in this case. Otherwise, the nonzero return
			 * value will later be propagated through the core as an error. */
			res = 0;
			goto finish;
		} else {
			pw_log_error("failed to create socket: %s", spa_strerror(fd));
			/* If ENODEV was returned earlier, and the stream_start_retry_timer
			 * was consequently created, but then a non-ENODEV error occurred,
			 * the timer must be stopped and removed. */
			pw_timer_queue_cancel(&impl->stream_start_retry_timer);
			res = fd;
			goto finish;
		}
	}

	/* Cleanup the timer in case ENODEV occurred earlier, and this time,
	 * the socket creation succeeded. */
	pw_timer_queue_cancel(&impl->stream_start_retry_timer);

	impl->source = pw_loop_add_io(impl->data_loop, fd,
				SPA_IO_IN, true, on_rtp_io, impl);
	if (impl->source == NULL) {
		pw_log_error("can't create io source: %m");
		close(fd);
		res = -errno;
		goto finish;
	}

	if ((res = pw_timer_queue_add(impl->timer_queue, &impl->igmp_recovery.timer,
			NULL, impl->igmp_recovery.check_interval * SPA_NSEC_PER_SEC,
			on_igmp_recovery_timer_event, impl)) < 0) {
		pw_log_error("can't add timer: %s", spa_strerror(res));
		goto finish;
	}

finish:
	if (res != 0) {
		pw_log_error("failed to start RTP stream: %s", spa_strerror(res));
		rtp_stream_set_error(impl->stream, res, "Can't start RTP stream");
	}

	if (result)
		*result = res;
}

static void stream_close_connection(void *data, int *result)
{
	struct impl *impl = data;

	if (result)
		*result = 0;

	if (!impl->source)
		return;

	pw_log_info("stopping RTP listener");

	pw_timer_queue_cancel(&impl->stream_start_retry_timer);
	pw_timer_queue_cancel(&impl->igmp_recovery.timer);

	pw_loop_destroy_source(impl->data_loop, impl->source);
	impl->source = NULL;
}

static void stream_destroy(void *d)
{
	struct impl *impl = d;
	impl->stream = NULL;
}

static void stream_props_changed(struct impl *impl, uint32_t id, const struct spa_pod *param)
{
	struct spa_pod_object *obj = (struct spa_pod_object *)param;
	struct spa_pod_prop *prop;

	if (param == NULL)
		return;

	SPA_POD_OBJECT_FOREACH(obj, prop) {
		if (prop->key == SPA_PROP_params) {
			struct spa_pod *params = NULL;
			struct spa_pod_parser prs;
			struct spa_pod_frame f;
			const char *key;
			struct spa_pod *pod;
			const char *value;

			if (spa_pod_parse_object(param, SPA_TYPE_OBJECT_Props, NULL, SPA_PROP_params,
					SPA_POD_OPT_Pod(&params)) < 0)
				return;
			spa_pod_parser_pod(&prs, params);
			if (spa_pod_parser_push_struct(&prs, &f) < 0)
				return;

			while (true) {
				if (spa_pod_parser_get_string(&prs, &key) < 0)
					break;
				if (spa_pod_parser_get_pod(&prs, &pod) < 0)
					break;
				if (spa_pod_get_string(pod, &value) < 0)
					continue;
				pw_log_info("key '%s', value '%s'", key, value);
				if (!spa_streq(key, "source.ip"))
					continue;
				if (pw_net_parse_address(value, impl->src_port, &impl->src_addr,
						&impl->src_len) < 0) {
					pw_log_error("invalid source.ip: '%s'", value);
					break;
				}
				pw_properties_set(impl->stream_props, "rtp.source.ip", value);
				struct spa_dict_item item[1];
				item[0] = SPA_DICT_ITEM_INIT("rtp.source.ip", value);
				rtp_stream_update_properties(impl->stream, &SPA_DICT_INIT(item, 1));
				break;
			}
		}
	}
}

static void stream_param_changed(void *data, uint32_t id, const struct spa_pod *param)
{
	struct impl *impl = data;

	switch (id) {
	case SPA_PARAM_Props:
		if (param != NULL)
			stream_props_changed(impl, id, param);
		break;
	}
}

static const struct rtp_stream_events stream_events = {
	RTP_VERSION_STREAM_EVENTS,
	.destroy = stream_destroy,
	.report_error = stream_report_error,
	.open_connection = stream_open_connection,
	.close_connection = stream_close_connection,
	.param_changed = stream_param_changed,
};

static void on_standby_timer_event(void *data)
{
	struct impl *impl = data;

	pw_log_debug("standby timer event; state: %d standby: %d waiting: %d",
			impl->state, impl->standby, impl->waiting);

	if (SPA_ATOMIC_CAS(impl->state, STATE_PROBE, STATE_STOPPING)) {
		if (!impl->standby) {
			struct spa_dict_item item[1];

			pw_log_info("timeout, standby RTP source");
			impl->standby = true;
			impl->waiting = true;

			item[0] = SPA_DICT_ITEM_INIT("rtp.receiving", "false");
			rtp_stream_update_properties(impl->stream, &SPA_DICT_INIT(item, 1));

			if (impl->may_pause)
				rtp_stream_set_active(impl->stream, false);
		}
		//pw_impl_module_schedule_destroy(impl->module);
		SPA_ATOMIC_STORE(impl->state, STATE_IDLE);
	} else {
		pw_log_debug("timeout, keeping active RTP source");
		SPA_ATOMIC_CAS(impl->state, STATE_RECEIVING, STATE_PROBE);
	}

	pw_timer_queue_add(impl->timer_queue, &impl->standby_timer,
			&impl->standby_timer.timeout, impl->cleanup_interval * SPA_NSEC_PER_SEC,
			on_standby_timer_event, impl);
}

static void core_destroy(void *d)
{
	struct impl *impl = d;
	spa_hook_remove(&impl->core_listener);
	impl->core = NULL;
	pw_impl_module_schedule_destroy(impl->module);
}

static const struct pw_proxy_events core_proxy_events = {
	.destroy = core_destroy,
};

static void impl_destroy(struct impl *impl)
{
	if (impl->stream)
		rtp_stream_destroy(impl->stream);
	if (impl->source)
		pw_loop_destroy_source(impl->data_loop, impl->source);

	if (impl->core && impl->do_disconnect)
		pw_core_disconnect(impl->core);

	pw_timer_queue_cancel(&impl->standby_timer);
	pw_timer_queue_cancel(&impl->stream_start_retry_timer);
	pw_timer_queue_cancel(&impl->igmp_recovery.timer);

	if (impl->data_loop)
		pw_context_release_loop(impl->context, impl->data_loop);

	pw_properties_free(impl->stream_props);
	pw_properties_free(impl->props);

	free(impl->buffer);
	free(impl->ifname);
	free(impl);
}

static void module_destroy(void *d)
{
	struct impl *impl = d;
	spa_hook_remove(&impl->module_listener);
	impl_destroy(impl);
}

static const struct pw_impl_module_events module_events = {
	PW_VERSION_IMPL_MODULE_EVENTS,
	.destroy = module_destroy,
};

static void on_core_error(void *d, uint32_t id, int seq, int res, const char *message)
{
	struct impl *impl = d;

	pw_log_error("error id:%u seq:%d res:%d (%s): %s",
			id, seq, res, spa_strerror(res), message);

	if (id == PW_ID_CORE && res == -EPIPE)
		pw_impl_module_schedule_destroy(impl->module);
}

static const struct pw_core_events core_events = {
	PW_VERSION_CORE_EVENTS,
	.error = on_core_error,
};

static void copy_props(struct impl *impl, struct pw_properties *props, const char *key)
{
	const char *str;
	if ((str = pw_properties_get(props, key)) != NULL) {
		if (pw_properties_get(impl->stream_props, key) == NULL)
			pw_properties_set(impl->stream_props, key, str);
	}
}

SPA_EXPORT
int pipewire__module_init(struct pw_impl_module *module, const char *args)
{
	struct pw_context *context = pw_impl_module_get_context(module);
	struct impl *impl;
	const char *str, *sess_name;
	struct pw_properties *props, *stream_props;
	int64_t ts_offset;
	char addr[128];
	int res = 0;
	uint32_t header_size;

	PW_LOG_TOPIC_INIT(mod_topic);

	impl = calloc(1, sizeof(struct impl));
	if (impl == NULL)
		return -errno;

	if (args == NULL)
		args = "";

	props = impl->props = pw_properties_new_string(args);
	stream_props = impl->stream_props = pw_properties_new(NULL, NULL);
	if (props == NULL || stream_props == NULL) {
		res = -errno;
		pw_log_error( "can't create properties: %m");
		goto out;
	}

	impl->module = module;
	impl->context = context;
	impl->main_loop = pw_context_get_main_loop(context);
	impl->timer_queue = pw_context_get_timer_queue(context);
	impl->data_loop = pw_context_acquire_loop(context, &props->dict);

	impl->rate_limit.interval = 2 * SPA_NSEC_PER_SEC;
	impl->rate_limit.burst = 1;

	if ((sess_name = pw_properties_get(props, "sess.name")) == NULL)
		sess_name = pw_get_host_name();

	pw_properties_set(props, PW_KEY_NODE_LOOP_NAME, impl->data_loop->name);
	if (pw_properties_get(props, PW_KEY_NODE_NAME) == NULL)
		pw_properties_setf(props, PW_KEY_NODE_NAME, "rtp_session.%s", sess_name);
	if (pw_properties_get(props, PW_KEY_NODE_DESCRIPTION) == NULL)
		pw_properties_setf(props, PW_KEY_NODE_DESCRIPTION, "%s", sess_name);
	if (pw_properties_get(props, PW_KEY_MEDIA_NAME) == NULL)
		pw_properties_setf(props, PW_KEY_MEDIA_NAME, "RTP Session with %s",
				sess_name);

	if ((str = pw_properties_get(props, "stream.props")) != NULL)
		pw_properties_update_string(stream_props, str, strlen(str));

	copy_props(impl, props, PW_KEY_NODE_LOOP_NAME);
	copy_props(impl, props, PW_KEY_AUDIO_FORMAT);
	copy_props(impl, props, PW_KEY_AUDIO_RATE);
	copy_props(impl, props, PW_KEY_AUDIO_CHANNELS);
	copy_props(impl, props, SPA_KEY_AUDIO_POSITION);
	copy_props(impl, props, PW_KEY_NODE_NAME);
	copy_props(impl, props, PW_KEY_NODE_DESCRIPTION);
	copy_props(impl, props, PW_KEY_NODE_GROUP);
	copy_props(impl, props, PW_KEY_NODE_LATENCY);
	copy_props(impl, props, PW_KEY_NODE_VIRTUAL);
	copy_props(impl, props, PW_KEY_NODE_CHANNELNAMES);
	copy_props(impl, props, PW_KEY_MEDIA_NAME);
	copy_props(impl, props, PW_KEY_MEDIA_CLASS);
	copy_props(impl, props, "net.mtu");
	copy_props(impl, props, "sess.media");
	copy_props(impl, props, "sess.name");
	copy_props(impl, props, "sess.min-ptime");
	copy_props(impl, props, "sess.max-ptime");
	copy_props(impl, props, "sess.latency.msec");
	copy_props(impl, props, "sess.ts-direct");
	copy_props(impl, props, "sess.ignore-ssrc");
	copy_props(impl, props, "stream.may-pause");

	str = pw_properties_get(props, "local.ifname");
	impl->ifname = str ? strdup(str) : NULL;

	impl->src_port = pw_properties_get_uint32(props, "source.port", 0);
	if (impl->src_port == 0) {
		res = -EINVAL;
		pw_log_error("invalid source.port");
		goto out;
	}
	if ((str = pw_properties_get(props, "source.ip")) == NULL)
		str = DEFAULT_SOURCE_IP;
	if ((res = pw_net_parse_address(str, impl->src_port, &impl->src_addr, &impl->src_len)) < 0) {
		pw_log_error("invalid source.ip %s: %s", str, spa_strerror(res));
		goto out;
	}
	pw_net_get_ip(&impl->src_addr, addr, sizeof(addr), NULL, NULL);
	pw_properties_set(stream_props, "rtp.source.ip", addr);
	pw_properties_setf(stream_props, "rtp.source.port", "%u", impl->src_port);

	header_size = impl->src_addr.ss_family == AF_INET ?
                        IP4_HEADER_SIZE : IP6_HEADER_SIZE;
	header_size += UDP_HEADER_SIZE;
	pw_properties_setf(stream_props, "net.header", "%u", header_size);

	ts_offset = pw_properties_get_int64(props, "sess.ts-offset", DEFAULT_TS_OFFSET);
	if (ts_offset == -1)
		ts_offset = pw_rand32();
	pw_properties_setf(stream_props, "rtp.receiver-ts-offset", "%u", (uint32_t)ts_offset);

	impl->always_process = pw_properties_get_bool(stream_props,
			PW_KEY_NODE_ALWAYS_PROCESS, true);
	impl->may_pause = pw_properties_get_bool(stream_props,
			"stream.may-pause", false);
	impl->standby = false;
	impl->waiting = true;
	/* Because we don't know the stream receiving state at the start, we try to fake it
	 * till we make it (or get timed out) */
	pw_properties_set(stream_props, "rtp.receiving", "true");

	impl->cleanup_interval = pw_properties_get_uint32(stream_props,
			"cleanup.sec", DEFAULT_CLEANUP_SEC);

	impl->igmp_recovery.check_interval = SPA_MAX(pw_properties_get_uint32(stream_props,
					"igmp.check.interval.sec",
					DEFAULT_IGMP_CHECK_INTERVAL_SEC), 1u);
	pw_log_info("using IGMP check interval of %" PRIu32 " second(s)",
			impl->igmp_recovery.check_interval);

	impl->igmp_recovery.deadline = SPA_MAX(pw_properties_get_uint32(stream_props,
					"igmp.deadline.sec", DEFAULT_IGMP_DEADLINE_SEC), 5u);
	pw_log_info("using IGMP deadline of %" PRIu32 " second(s)",
			impl->igmp_recovery.deadline);

	impl->core = pw_context_get_object(impl->context, PW_TYPE_INTERFACE_Core);
	if (impl->core == NULL) {
		str = pw_properties_get(props, PW_KEY_REMOTE_NAME);
		impl->core = pw_context_connect(impl->context,
				pw_properties_new(
					PW_KEY_REMOTE_NAME, str,
					NULL),
				0);
		impl->do_disconnect = true;
	}
	if (impl->core == NULL) {
		res = -errno;
		pw_log_error("can't connect: %m");
		goto out;
	}

	pw_proxy_add_listener((struct pw_proxy*)impl->core,
			&impl->core_proxy_listener,
			&core_proxy_events, impl);
	pw_core_add_listener(impl->core,
			&impl->core_listener,
			&core_events, impl);

	if ((res = pw_timer_queue_add(impl->timer_queue, &impl->standby_timer,
			NULL, impl->cleanup_interval * SPA_NSEC_PER_SEC,
			on_standby_timer_event, impl)) < 0) {
		pw_log_error("can't add timer: %s", spa_strerror(res));
		goto out;
	}

	impl->stream = rtp_stream_new(impl->core,
			PW_DIRECTION_OUTPUT, pw_properties_copy(stream_props),
			&stream_events, impl);
	if (impl->stream == NULL) {
		res = -errno;
		pw_log_error("can't create stream: %m");
		goto out;
	}

	impl->buffer_size = rtp_stream_get_mtu(impl->stream);
	impl->buffer = calloc(1, impl->buffer_size);
	if (impl->buffer == NULL) {
		res = -errno;
		pw_log_error("can't create packet buffer of size %zd: %m", impl->buffer_size);
		goto out;
	}

	pw_impl_module_add_listener(module, &impl->module_listener, &module_events, impl);

	pw_impl_module_update_properties(module, &SPA_DICT_INIT_ARRAY(module_info));

	pw_log_info("Successfully loaded module-rtp-source");

	return 0;
out:
	impl_destroy(impl);
	return res;
}