bridge_softmix: Note why ast_stream_topology_set_stream cannot fail.
[asterisk/asterisk.git] / bridges / bridge_softmix.c
index 73ac296..8de88f2 100644 (file)
@@ -1,9 +1,10 @@
 /*
  * Asterisk -- An open source telephony toolkit.
  *
- * Copyright (C) 2007, Digium, Inc.
+ * Copyright (C) 2011, Digium, Inc.
  *
  * Joshua Colp <jcolp@digium.com>
+ * David Vossel <dvossel@digium.com>
  *
  * See http://www.asterisk.org for more information about
  * the Asterisk project. Please do not directly contact
  * \brief Multi-party software based channel mixing
  *
  * \author Joshua Colp <jcolp@digium.com>
+ * \author David Vossel <dvossel@digium.com>
  *
  * \ingroup bridges
- *
- * \todo This bridge operates in 8 kHz mode unless a define is uncommented.
- * This needs to be improved so the bridge moves between the dominant codec as needed depending
- * on channels present in the bridge and transcoding capabilities.
  */
 
+/*** MODULEINFO
+       <support_level>core</support_level>
+ ***/
+
 #include "asterisk.h"
 
-ASTERISK_FILE_VERSION(__FILE__, "$Revision$")
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <sys/time.h>
-#include <signal.h>
-#include <errno.h>
-#include <unistd.h>
-
-#include "asterisk/module.h"
-#include "asterisk/channel.h"
-#include "asterisk/bridging.h"
-#include "asterisk/bridging_technology.h"
-#include "asterisk/frame.h"
-#include "asterisk/options.h"
-#include "asterisk/logger.h"
-#include "asterisk/slinfactory.h"
-#include "asterisk/astobj2.h"
-#include "asterisk/timing.h"
+#include "asterisk/stream.h"
+#include "asterisk/test.h"
+#include "asterisk/vector.h"
+#include "bridge_softmix/include/bridge_softmix_internal.h"
+
+/*! The minimum sample rate of the bridge. */
+#define SOFTMIX_MIN_SAMPLE_RATE 8000   /* 8 kHz sample rate */
 
 /*! \brief Interval at which mixing will take place. Valid options are 10, 20, and 40. */
-#define SOFTMIX_INTERVAL 20
+#define DEFAULT_SOFTMIX_INTERVAL 20
 
 /*! \brief Size of the buffer used for sample manipulation */
-#define SOFTMIX_DATALEN (160 * (SOFTMIX_INTERVAL / 10))
+#define SOFTMIX_DATALEN(rate, interval) ((rate/50) * (interval / 10))
 
 /*! \brief Number of samples we are dealing with */
-#define SOFTMIX_SAMPLES (SOFTMIX_DATALEN / 2)
-
-/*! \brief Define used to turn on 16 kHz audio support */
-/* #define SOFTMIX_16_SUPPORT */
-
-/*! \brief Structure which contains per-channel mixing information */
-struct softmix_channel {
-       /*! Lock to protect this structure */
-       ast_mutex_t lock;
-       /*! Factory which contains audio read in from the channel */
-       struct ast_slinfactory factory;
-       /*! Frame that contains mixed audio to be written out to the channel */
-       struct ast_frame frame;
-       /*! Bit used to indicate that the channel provided audio for this mixing interval */
-       int have_audio:1;
-       /*! Bit used to indicate that a frame is available to be written out to the channel */
-       int have_frame:1;
-       /*! Buffer containing final mixed audio from all sources */
-       short final_buf[SOFTMIX_DATALEN];
-       /*! Buffer containing only the audio from the channel */
-       short our_buf[SOFTMIX_DATALEN];
+#define SOFTMIX_SAMPLES(rate, interval) (SOFTMIX_DATALEN(rate, interval) / 2)
+
+/*! \brief Number of mixing iterations to perform between gathering statistics. */
+#define SOFTMIX_STAT_INTERVAL 100
+
+/* This is the threshold in ms at which a channel's own audio will stop getting
+ * mixed out its own write audio stream because it is not talking. */
+#define DEFAULT_SOFTMIX_SILENCE_THRESHOLD 2500
+#define DEFAULT_SOFTMIX_TALKING_THRESHOLD 160
+
+#define SOFTBRIDGE_VIDEO_DEST_PREFIX "softbridge_dest"
+#define SOFTBRIDGE_VIDEO_DEST_LEN strlen(SOFTBRIDGE_VIDEO_DEST_PREFIX)
+#define SOFTBRIDGE_VIDEO_DEST_SEPARATOR '_'
+
+struct softmix_stats {
+       /*! Each index represents a sample rate used above the internal rate. */
+       unsigned int sample_rates[16];
+       /*! Each index represents the number of channels using the same index in the sample_rates array.  */
+       unsigned int num_channels[16];
+       /*! The number of channels above the internal sample rate */
+       unsigned int num_above_internal_rate;
+       /*! The number of channels at the internal sample rate */
+       unsigned int num_at_internal_rate;
+       /*! The absolute highest sample rate preferred by any channel in the bridge */
+       unsigned int highest_supported_rate;
+       /*! Is the sample rate locked by the bridge, if so what is that rate.*/
+       unsigned int locked_rate;
 };
 
-/*! \brief Function called when a bridge is created */
-static int softmix_bridge_create(struct ast_bridge *bridge)
+struct softmix_translate_helper_entry {
+       int num_times_requested; /*!< Once this entry is no longer requested, free the trans_pvt
+                                                                 and re-init if it was usable. */
+       struct ast_format *dst_format; /*!< The destination format for this helper */
+       struct ast_trans_pvt *trans_pvt; /*!< the translator for this slot. */
+       struct ast_frame *out_frame; /*!< The output frame from the last translation */
+       AST_LIST_ENTRY(softmix_translate_helper_entry) entry;
+};
+
+struct softmix_translate_helper {
+       struct ast_format *slin_src; /*!< the source format expected for all the translators */
+       AST_LIST_HEAD_NOLOCK(, softmix_translate_helper_entry) entries;
+};
+
+static struct softmix_translate_helper_entry *softmix_translate_helper_entry_alloc(struct ast_format *dst)
 {
-       int timingfd;
+       struct softmix_translate_helper_entry *entry;
+       if (!(entry = ast_calloc(1, sizeof(*entry)))) {
+               return NULL;
+       }
+       entry->dst_format = ao2_bump(dst);
+       /* initialize this to one so that the first time through the cleanup code after
+          allocation it won't be removed from the entry list */
+       entry->num_times_requested = 1;
+       return entry;
+}
 
-       if ((timingfd = ast_timer_open()) < 0) {
-               return -1;
+static void *softmix_translate_helper_free_entry(struct softmix_translate_helper_entry *entry)
+{
+       ao2_cleanup(entry->dst_format);
+
+       if (entry->trans_pvt) {
+               ast_translator_free_path(entry->trans_pvt);
+       }
+       if (entry->out_frame) {
+               ast_frfree(entry->out_frame);
        }
+       ast_free(entry);
+       return NULL;
+}
 
-       bridge->bridge_pvt = (void*)(unsigned long)timingfd;
+static void softmix_translate_helper_init(struct softmix_translate_helper *trans_helper, unsigned int sample_rate)
+{
+       memset(trans_helper, 0, sizeof(*trans_helper));
+       trans_helper->slin_src = ast_format_cache_get_slin_by_rate(sample_rate);
+}
 
-       return 0;
+static void softmix_translate_helper_destroy(struct softmix_translate_helper *trans_helper)
+{
+       struct softmix_translate_helper_entry *entry;
+
+       while ((entry = AST_LIST_REMOVE_HEAD(&trans_helper->entries, entry))) {
+               softmix_translate_helper_free_entry(entry);
+       }
 }
 
-/*! \brief Function called when a channel is joined into the bridge */
-static int softmix_bridge_join(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+static void softmix_translate_helper_change_rate(struct softmix_translate_helper *trans_helper, unsigned int sample_rate)
 {
-       struct softmix_channel *sc = NULL;
+       struct softmix_translate_helper_entry *entry;
 
-       /* Create a new softmix_channel structure and allocate various things on it */
-       if (!(sc = ast_calloc(1, sizeof(*sc)))) {
-               return -1;
+       trans_helper->slin_src = ast_format_cache_get_slin_by_rate(sample_rate);
+       AST_LIST_TRAVERSE_SAFE_BEGIN(&trans_helper->entries, entry, entry) {
+               if (entry->trans_pvt) {
+                       ast_translator_free_path(entry->trans_pvt);
+                       if (!(entry->trans_pvt = ast_translator_build_path(entry->dst_format, trans_helper->slin_src))) {
+                               AST_LIST_REMOVE_CURRENT(entry);
+                               entry = softmix_translate_helper_free_entry(entry);
+                       }
+               }
        }
+       AST_LIST_TRAVERSE_SAFE_END;
+}
 
-       /* Can't forget the lock */
-       ast_mutex_init(&sc->lock);
+/*!
+ * \internal
+ * \brief Get the next available audio on the softmix channel's read stream
+ * and determine if it should be mixed out or not on the write stream.
+ *
+ * \retval pointer to buffer containing the exact number of samples requested on success.
+ * \retval NULL if no samples are present
+ */
+static int16_t *softmix_process_read_audio(struct softmix_channel *sc, unsigned int num_samples)
+{
+       if ((ast_slinfactory_available(&sc->factory) >= num_samples) &&
+               ast_slinfactory_read(&sc->factory, sc->our_buf, num_samples)) {
+               sc->have_audio = 1;
+               return sc->our_buf;
+       }
+       sc->have_audio = 0;
+       return NULL;
+}
 
-       /* Setup smoother */
-       ast_slinfactory_init(&sc->factory);
+/*!
+ * \internal
+ * \brief Process a softmix channel's write audio
+ *
+ * \details This function will remove the channel's talking from its own audio if present and
+ * possibly even do the channel's write translation for it depending on how many other
+ * channels use the same write format.
+ */
+static void softmix_process_write_audio(struct softmix_translate_helper *trans_helper,
+       struct ast_format *raw_write_fmt,
+       struct softmix_channel *sc, unsigned int default_sample_size)
+{
+       struct softmix_translate_helper_entry *entry = NULL;
+       int i;
 
-       /* Setup frame parameters */
-       sc->frame.frametype = AST_FRAME_VOICE;
-#ifdef SOFTMIX_16_SUPPORT
-       sc->frame.subclass = AST_FORMAT_SLINEAR16;
-#else
-       sc->frame.subclass = AST_FORMAT_SLINEAR;
-#endif
-       sc->frame.data.ptr = sc->final_buf;
-       sc->frame.datalen = SOFTMIX_DATALEN;
-       sc->frame.samples = SOFTMIX_SAMPLES;
+       /* If we provided audio that was not determined to be silence,
+        * then take it out while in slinear format. */
+       if (sc->have_audio && sc->talking && !sc->binaural) {
+               for (i = 0; i < sc->write_frame.samples; i++) {
+                       ast_slinear_saturated_subtract(&sc->final_buf[i], &sc->our_buf[i]);
+               }
+               /* check to see if any entries exist for the format. if not we'll want
+                  to remove it during cleanup */
+               AST_LIST_TRAVERSE(&trans_helper->entries, entry, entry) {
+                       if (ast_format_cmp(entry->dst_format, raw_write_fmt) == AST_FORMAT_CMP_EQUAL) {
+                               ++entry->num_times_requested;
+                               break;
+                       }
+               }
+               /* do not do any special write translate optimization if we had to make
+                * a special mix for them to remove their own audio. */
+               return;
+       } else if (sc->have_audio && sc->talking && sc->binaural > 0) {
+               /*
+                * Binaural audio requires special saturated substract since we have two
+                * audio signals per channel now.
+                */
+               softmix_process_write_binaural_audio(sc, default_sample_size);
+               return;
+       }
 
-       /* Can't forget to record our pvt structure within the bridged channel structure */
-       bridge_channel->bridge_pvt = sc;
+       /* Attempt to optimize channels using the same translation path/codec. Build a list of entries
+          of translation paths and track the number of references for each type. Each one of the same
+          type should be able to use the same out_frame. Since the optimization is only necessary for
+          multiple channels (>=2) using the same codec make sure resources are allocated only when
+          needed and released when not (see also softmix_translate_helper_cleanup */
+       AST_LIST_TRAVERSE(&trans_helper->entries, entry, entry) {
+               if (sc->binaural != 0) {
+                       continue;
+               }
+               if (ast_format_cmp(entry->dst_format, raw_write_fmt) == AST_FORMAT_CMP_EQUAL) {
+                       entry->num_times_requested++;
+               } else {
+                       continue;
+               }
+               if (!entry->trans_pvt && (entry->num_times_requested > 1)) {
+                       entry->trans_pvt = ast_translator_build_path(entry->dst_format, trans_helper->slin_src);
+               }
+               if (entry->trans_pvt && !entry->out_frame) {
+                       entry->out_frame = ast_translate(entry->trans_pvt, &sc->write_frame, 0);
+               }
+               if (entry->out_frame && entry->out_frame->frametype == AST_FRAME_VOICE
+                               && entry->out_frame->datalen < MAX_DATALEN) {
+                       ao2_replace(sc->write_frame.subclass.format, entry->out_frame->subclass.format);
+                       memcpy(sc->final_buf, entry->out_frame->data.ptr, entry->out_frame->datalen);
+                       sc->write_frame.datalen = entry->out_frame->datalen;
+                       sc->write_frame.samples = entry->out_frame->samples;
+               }
+               break;
+       }
 
-       return 0;
+       /* add new entry into list if this format destination was not matched. */
+       if (!entry && (entry = softmix_translate_helper_entry_alloc(raw_write_fmt))) {
+               AST_LIST_INSERT_HEAD(&trans_helper->entries, entry, entry);
+       }
 }
 
-/*! \brief Function called when a channel leaves the bridge */
-static int softmix_bridge_leave(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+static void softmix_translate_helper_cleanup(struct softmix_translate_helper *trans_helper)
 {
-       struct softmix_channel *sc = bridge_channel->bridge_pvt;
+       struct softmix_translate_helper_entry *entry;
 
-       /* Drop mutex lock */
-       ast_mutex_destroy(&sc->lock);
+       AST_LIST_TRAVERSE_SAFE_BEGIN(&trans_helper->entries, entry, entry) {
+               /* if it hasn't been requested then remove it */
+               if (!entry->num_times_requested) {
+                       AST_LIST_REMOVE_CURRENT(entry);
+                       softmix_translate_helper_free_entry(entry);
+                       continue;
+               }
 
-       /* Drop the factory */
-       ast_slinfactory_destroy(&sc->factory);
+               if (entry->out_frame) {
+                       ast_frfree(entry->out_frame);
+                       entry->out_frame = NULL;
+               }
 
-       /* Eep! drop ourselves */
-       ast_free(sc);
+               /* nothing is optimized for a single path reference, so there is
+                  no reason to continue to hold onto the codec */
+               if (entry->num_times_requested == 1 && entry->trans_pvt) {
+                       ast_translator_free_path(entry->trans_pvt);
+                       entry->trans_pvt = NULL;
+               }
 
-       return 0;
+               /* for each iteration (a mixing run) in the bridge softmix thread the number
+                  of references to a given entry is recalculated, so reset the number of
+                  times requested */
+               entry->num_times_requested = 0;
+       }
+       AST_LIST_TRAVERSE_SAFE_END;
 }
 
-/*! \brief Function called when a channel writes a frame into the bridge */
-static enum ast_bridge_write_result softmix_bridge_write(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel, struct ast_frame *frame)
+static void set_softmix_bridge_data(int rate, int interval, struct ast_bridge_channel *bridge_channel, int reset, int set_binaural, int binaural_pos_id, int is_announcement)
 {
-       struct softmix_channel *sc = bridge_channel->bridge_pvt;
+       struct softmix_channel *sc = bridge_channel->tech_pvt;
+       struct ast_format *slin_format;
+       int setup_fail;
 
-       /* Only accept audio frames, all others are unsupported */
-       if (frame->frametype != AST_FRAME_VOICE) {
-               return AST_BRIDGE_WRITE_UNSUPPORTED;
+#ifdef BINAURAL_RENDERING
+       if (interval != BINAURAL_MIXING_INTERVAL) {
+               interval = BINAURAL_MIXING_INTERVAL;
        }
+#endif
+
+       /* The callers have already ensured that sc is never NULL. */
+       ast_assert(sc != NULL);
+
+       slin_format = ast_format_cache_get_slin_by_rate(rate);
 
        ast_mutex_lock(&sc->lock);
+       if (reset) {
+               ast_slinfactory_destroy(&sc->factory);
+               ast_dsp_free(sc->dsp);
+       }
 
-       /* If a frame was provided add it to the smoother */
-#ifdef SOFTMIX_16_SUPPORT
-       if (frame->frametype == AST_FRAME_VOICE && frame->subclass == AST_FORMAT_SLINEAR16) {
-#else
-       if (frame->frametype == AST_FRAME_VOICE && frame->subclass == AST_FORMAT_SLINEAR) {
-#endif
-               ast_slinfactory_feed(&sc->factory, frame);
+       /* Setup write frame parameters */
+       sc->write_frame.frametype = AST_FRAME_VOICE;
+       /*
+        * NOTE: The write_frame format holds a reference because translation
+        * could be needed and the format changed to the translated format
+        * for the channel.  The translated format may not be a
+        * static cached format.
+        */
+       ao2_replace(sc->write_frame.subclass.format, slin_format);
+       sc->write_frame.data.ptr = sc->final_buf;
+       sc->write_frame.datalen = SOFTMIX_DATALEN(rate, interval);
+       sc->write_frame.samples = SOFTMIX_SAMPLES(rate, interval);
+
+       /* We will store the rate here cause we need to set the data again when a channel is unsuspended */
+       sc->rate = rate;
+
+       /* If the channel will contain binaural data we will set a identifier in the channel
+        * if set_binaural == -1 this is just a sample rate update, will ignore it. */
+       if (set_binaural == 1) {
+               sc->binaural = 1;
+       } else if (set_binaural == 0) {
+               sc->binaural = 0;
        }
 
-       /* If a frame is ready to be written out, do so */
-       if (sc->have_frame) {
-               ast_write(bridge_channel->chan, &sc->frame);
-               sc->have_frame = 0;
+       /* Setting the binaural position. This doesn't require a change of the overlaying channel infos
+        * and doesn't have to be done if we just updating sample rates. */
+       if (binaural_pos_id != -1) {
+               sc->binaural_pos = binaural_pos_id;
+       }
+       if (is_announcement != -1) {
+               sc->is_announcement = is_announcement;
+       }
+
+       /*
+        * NOTE: The read_slin_format does not hold a reference because it
+        * will always be a signed linear format.
+        */
+       sc->read_slin_format = slin_format;
+
+       /* Setup smoother */
+       setup_fail = ast_slinfactory_init_with_format(&sc->factory, slin_format);
+
+       /* set new read and write formats on channel. */
+       ast_channel_lock(bridge_channel->chan);
+       setup_fail |= ast_set_read_format_path(bridge_channel->chan,
+               ast_channel_rawreadformat(bridge_channel->chan), slin_format);
+       ast_channel_unlock(bridge_channel->chan);
+
+       /* If channel contains binaural data we will set it here for the trans_pvt. */
+       if (set_binaural == 1 || (set_binaural == -1 && sc->binaural == 1)) {
+               setup_fail |= ast_set_write_format_interleaved_stereo(bridge_channel->chan, slin_format);
+       } else if (set_binaural == 0) {
+               setup_fail |= ast_set_write_format(bridge_channel->chan, slin_format);
+       }
+
+       /* set up new DSP.  This is on the read side only right before the read frame enters the smoother.  */
+       sc->dsp = ast_dsp_new_with_rate(rate);
+       if (setup_fail || !sc->dsp) {
+               /* Bad news.  Could not setup the channel for softmix. */
+               ast_mutex_unlock(&sc->lock);
+               ast_bridge_channel_leave_bridge(bridge_channel, BRIDGE_CHANNEL_STATE_END, 0);
+               return;
+       }
+
+       /* we want to aggressively detect silence to avoid feedback */
+       if (bridge_channel->tech_args.talking_threshold) {
+               ast_dsp_set_threshold(sc->dsp, bridge_channel->tech_args.talking_threshold);
+       } else {
+               ast_dsp_set_threshold(sc->dsp, DEFAULT_SOFTMIX_TALKING_THRESHOLD);
        }
 
-       /* Alllll done */
        ast_mutex_unlock(&sc->lock);
+}
+
+/*!
+ * \internal
+ * \brief Poke the mixing thread in case it is waiting for an active channel.
+ * \since 12.0.0
+ *
+ * \param softmix_data Bridge mixing data.
+ *
+ * \return Nothing
+ */
+static void softmix_poke_thread(struct softmix_bridge_data *softmix_data)
+{
+       ast_mutex_lock(&softmix_data->lock);
+       ast_cond_signal(&softmix_data->cond);
+       ast_mutex_unlock(&softmix_data->lock);
+}
+
+/*! \brief Function called when a channel is unsuspended from the bridge */
+static void softmix_bridge_unsuspend(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+{
+#ifdef BINAURAL_RENDERING
+       struct softmix_channel *sc = bridge_channel->tech_pvt;
+       if (sc->binaural) {
+               /* Restore some usefull data if it was a binaural channel */
+               struct ast_format *slin_format;
 
-       return AST_BRIDGE_WRITE_SUCCESS;
+               slin_format = ast_format_cache_get_slin_by_rate(sc->rate);
+               ast_set_write_format_interleaved_stereo(bridge_channel->chan, slin_format);
+       }
+#endif
+       if (bridge->tech_pvt) {
+               softmix_poke_thread(bridge->tech_pvt);
+       }
 }
 
-/*! \brief Function called when the channel's thread is poked */
-static int softmix_bridge_poke(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+/*!
+ * \brief Determine if a stream is a video source stream.
+ *
+ * \param stream The stream to test
+ * \retval 1 The stream is a video source
+ * \retval 0 The stream is not a video source
+ */
+static int is_video_source(const struct ast_stream *stream)
 {
-       struct softmix_channel *sc = bridge_channel->bridge_pvt;
+       if (ast_stream_get_type(stream) == AST_MEDIA_TYPE_VIDEO &&
+               strncmp(ast_stream_get_name(stream), SOFTBRIDGE_VIDEO_DEST_PREFIX,
+                       SOFTBRIDGE_VIDEO_DEST_LEN)) {
+               return 1;
+       }
 
-       ast_mutex_lock(&sc->lock);
+       return 0;
+}
 
-       if (sc->have_frame) {
-               ast_write(bridge_channel->chan, &sc->frame);
-               sc->have_frame = 0;
+/*!
+ * \brief Determine if a stream is a video destination stream.
+ *
+ * A source channel name can be provided to narrow this to a destination stream
+ * for a particular source channel. Further, a source stream name can be provided
+ * to narrow this to a particular source stream's destination. However, empty strings
+ * can be provided to match any destination video stream, regardless of source channel
+ * or source stream.
+ *
+ * \param stream The stream to test
+ * \param source_channel_name The name of a source video channel to match
+ * \param source_stream_name The name of the source video stream to match
+ * \retval 1 The stream is a video destination stream
+ * \retval 0 The stream is not a video destination stream
+ */
+static int is_video_dest(const struct ast_stream *stream, const char *source_channel_name,
+       const char *source_stream_name)
+{
+       char *dest_video_name;
+       size_t dest_video_name_len;
+
+       if (ast_stream_get_type(stream) != AST_MEDIA_TYPE_VIDEO) {
+               return 0;
        }
 
-       ast_mutex_unlock(&sc->lock);
+       dest_video_name_len = SOFTBRIDGE_VIDEO_DEST_LEN + 1;
+
+       if (!ast_strlen_zero(source_channel_name)) {
+               dest_video_name_len += strlen(source_channel_name) + 1;
+               if (!ast_strlen_zero(source_stream_name)) {
+                       dest_video_name_len += strlen(source_stream_name) + 1;
+               }
+       }
+       dest_video_name = ast_alloca(dest_video_name_len);
+
+       if (!ast_strlen_zero(source_channel_name)) {
+               if (!ast_strlen_zero(source_stream_name)) {
+                       snprintf(dest_video_name, dest_video_name_len, "%s%c%s%c%s",
+                               SOFTBRIDGE_VIDEO_DEST_PREFIX, SOFTBRIDGE_VIDEO_DEST_SEPARATOR,
+                               source_channel_name, SOFTBRIDGE_VIDEO_DEST_SEPARATOR,
+                               source_stream_name);
+                       return !strcmp(ast_stream_get_name(stream), dest_video_name);
+               } else {
+                       snprintf(dest_video_name, dest_video_name_len, "%s%c%s",
+                               SOFTBRIDGE_VIDEO_DEST_PREFIX, SOFTBRIDGE_VIDEO_DEST_SEPARATOR,
+                               source_channel_name);
+                       return !strncmp(ast_stream_get_name(stream), dest_video_name, dest_video_name_len - 1);
+               }
+       } else {
+               snprintf(dest_video_name, dest_video_name_len, "%s",
+                       SOFTBRIDGE_VIDEO_DEST_PREFIX);
+               return !strncmp(ast_stream_get_name(stream), dest_video_name, dest_video_name_len - 1);
+       }
 
        return 0;
 }
 
-/*! \brief Function which acts as the mixing thread */
-static int softmix_bridge_thread(struct ast_bridge *bridge)
+static int append_source_streams(struct ast_stream_topology *dest,
+       const char *channel_name,
+       const struct ast_stream_topology *source)
 {
-       int timingfd = (unsigned short)(unsigned long)bridge->bridge_pvt;
+       int i;
 
-       ast_timer_set_rate(timingfd, (1000 / SOFTMIX_INTERVAL));
+       for (i = 0; i < ast_stream_topology_get_count(source); ++i) {
+               struct ast_stream *stream;
+               struct ast_stream *stream_clone;
+               char *stream_clone_name = NULL;
 
-       while (!bridge->stop && !bridge->refresh && bridge->array_num) {
-               struct ast_bridge_channel *bridge_channel = NULL;
-               short buf[SOFTMIX_DATALEN] = {0, };
-               int timeout = -1;
+               stream = ast_stream_topology_get_stream(source, i);
+               if (!is_video_source(stream)) {
+                       continue;
+               }
 
-               /* Go through pulling audio from each factory that has it available */
-               AST_LIST_TRAVERSE(&bridge->channels, bridge_channel, entry) {
-                       struct softmix_channel *sc = bridge_channel->bridge_pvt;
+               if (ast_asprintf(&stream_clone_name, "%s_%s_%s", SOFTBRIDGE_VIDEO_DEST_PREFIX,
+                       channel_name, ast_stream_get_name(stream)) < 0) {
+                       return -1;
+               }
 
-                       ast_mutex_lock(&sc->lock);
+               stream_clone = ast_stream_clone(stream, stream_clone_name);
+               ast_free(stream_clone_name);
+               if (!stream_clone) {
+                       return -1;
+               }
+               if (ast_stream_topology_append_stream(dest, stream_clone) < 0) {
+                       ast_stream_free(stream_clone);
+                       return -1;
+               }
+       }
 
-                       /* Try to get audio from the factory if available */
-                       if (ast_slinfactory_available(&sc->factory) >= SOFTMIX_SAMPLES && ast_slinfactory_read(&sc->factory, sc->our_buf, SOFTMIX_SAMPLES)) {
-                               short *data1, *data2;
-                               int i;
-
-                               /* Put into the local final buffer */
-                               for (i = 0, data1 = buf, data2 = sc->our_buf; i < SOFTMIX_DATALEN; i++, data1++, data2++)
-                                       ast_slinear_saturated_add(data1, data2);
-                               /* Yay we have our own audio */
-                               sc->have_audio = 1;
-                       } else {
-                               /* Awww we don't have audio ;( */
-                               sc->have_audio = 0;
-                       }
-                       ast_mutex_unlock(&sc->lock);
+       return 0;
+}
+
+static int append_all_streams(struct ast_stream_topology *dest,
+       const struct ast_stream_topology *source)
+{
+       int i;
+       int dest_index = 0;
+
+       for (i = 0; i < ast_stream_topology_get_count(source); ++i) {
+               struct ast_stream *clone;
+               int added = 0;
+
+               clone = ast_stream_clone(ast_stream_topology_get_stream(source, i), NULL);
+               if (!clone) {
+                       return -1;
                }
 
-               /* Next step go through removing the channel's own audio and creating a good frame... */
-               AST_LIST_TRAVERSE(&bridge->channels, bridge_channel, entry) {
-                       struct softmix_channel *sc = bridge_channel->bridge_pvt;
-                       int i = 0;
+               /* If we can reuse an existing removed stream then do so */
+               while (dest_index < ast_stream_topology_get_count(dest)) {
+                       struct ast_stream *stream = ast_stream_topology_get_stream(dest, dest_index);
 
-                       /* Copy from local final buffer to our final buffer */
-                       memcpy(sc->final_buf, buf, sizeof(sc->final_buf));
+                       dest_index++;
 
-                       /* If we provided audio then take it out */
-                       if (sc->have_audio) {
-                               for (i = 0; i < SOFTMIX_DATALEN; i++) {
-                                       ast_slinear_saturated_subtract(&sc->final_buf[i], &sc->our_buf[i]);
-                               }
+                       if (ast_stream_get_state(stream) == AST_STREAM_STATE_REMOVED) {
+                               /* This cannot fail because dest_index - 1 is less than the
+                                * current count in dest. */
+                               ast_stream_topology_set_stream(dest, dest_index - 1, clone);
+                               added = 1;
+                               break;
                        }
+               }
+
+               /* If no removed stream exists that we took the place of append the stream */
+               if (!added && ast_stream_topology_append_stream(dest, clone) < 0) {
+                       ast_stream_free(clone);
+                       return -1;
+               }
+       }
+
+       return 0;
+}
+
+/*!
+ * \brief Issue channel stream topology change requests.
+ *
+ * When in SFU mode, each participant needs to be able to
+ * send video directly to other participants in the bridge.
+ * This means that all participants need to have their topologies
+ * updated. The joiner needs to have destination streams for
+ * all current participants, and the current participants need
+ * to have destinations streams added for the joiner's sources.
+ *
+ * \param joiner The channel that is joining the softmix bridge
+ * \param participants The current participants in the softmix bridge
+ */
+static void sfu_topologies_on_join(struct ast_bridge_channel *joiner, struct ast_bridge_channels_list *participants)
+{
+       struct ast_stream_topology *joiner_video = NULL;
+       struct ast_bridge_channel *participant;
+       int res;
+       struct softmix_channel *sc;
+
+       joiner_video = ast_stream_topology_alloc();
+       if (!joiner_video) {
+               return;
+       }
 
-                       /* The frame is now ready for use... */
-                       sc->have_frame = 1;
+       sc = joiner->tech_pvt;
 
-                       /* Poke bridged channel thread just in case */
-                       pthread_kill(bridge_channel->thread, SIGURG);
+       ast_channel_lock(joiner->chan);
+       res = append_source_streams(joiner_video, ast_channel_name(joiner->chan), ast_channel_get_stream_topology(joiner->chan));
+       sc->topology = ast_stream_topology_clone(ast_channel_get_stream_topology(joiner->chan));
+       ast_channel_unlock(joiner->chan);
+
+       if (res || !sc->topology) {
+               goto cleanup;
+       }
+
+       AST_LIST_TRAVERSE(participants, participant, entry) {
+               if (participant == joiner) {
+                       continue;
+               }
+               ast_channel_lock(participant->chan);
+               res = append_source_streams(sc->topology, ast_channel_name(participant->chan),
+                               ast_channel_get_stream_topology(participant->chan));
+               ast_channel_unlock(participant->chan);
+               if (res) {
+                       goto cleanup;
                }
+       }
 
-               ao2_unlock(bridge);
+       ast_channel_request_stream_topology_change(joiner->chan, sc->topology, NULL);
 
-               /* Wait for the timing source to tell us to wake up and get things done */
-               ast_waitfor_n_fd(&timingfd, 1, &timeout, NULL);
+       AST_LIST_TRAVERSE(participants, participant, entry) {
+               if (participant == joiner) {
+                       continue;
+               }
+
+               sc = participant->tech_pvt;
+               if (append_all_streams(sc->topology, joiner_video)) {
+                       goto cleanup;
+               }
+               ast_channel_request_stream_topology_change(participant->chan, sc->topology, NULL);
+       }
+
+cleanup:
+       ast_stream_topology_free(joiner_video);
+}
+
+/*! \brief Function called when a channel is joined into the bridge */
+static int softmix_bridge_join(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+{
+       struct softmix_channel *sc;
+       struct softmix_bridge_data *softmix_data;
+       int set_binaural = 0;
+       /*
+        * If false, the channel will be convolved, but since it is a non stereo channel, output
+        * will be mono.
+        */
+       int skip_binaural_output = 1;
+       int pos_id;
+       int is_announcement = 0;
+       int samplerate_change;
+
+       softmix_data = bridge->tech_pvt;
+       if (!softmix_data) {
+               return -1;
+       }
+
+       /* Create a new softmix_channel structure and allocate various things on it */
+       if (!(sc = ast_calloc(1, sizeof(*sc)))) {
+               return -1;
+       }
+
+       samplerate_change = softmix_data->internal_rate;
+       pos_id = -1;
+       if (bridge->softmix.binaural_active) {
+               if (strncmp(ast_channel_name(bridge_channel->chan), "CBAnn", 5) != 0) {
+                       set_binaural = ast_format_get_channel_count(bridge_channel->write_format) > 1 ? 1 : 0;
+                       if (set_binaural) {
+                               softmix_data->internal_rate = samplerate_change;
+                       }
+                       skip_binaural_output = 0;
+               } else {
+                       is_announcement = 1;
+               }
+               if (set_binaural) {
+                       softmix_data->convolve.binaural_active = 1;
+               }
+               if (!skip_binaural_output)      {
+                       pos_id = set_binaural_data_join(&softmix_data->convolve, softmix_data->default_sample_size);
+                       if (pos_id == -1) {
+                               ast_log(LOG_ERROR, "Bridge %s: Failed to join channel %s. "
+                                               "Could not allocate enough memory.\n", bridge->uniqueid,
+                                               ast_channel_name(bridge_channel->chan));
+                               return -1;
+                       }
+               }
+       }
 
-               ast_timer_ack(timingfd, 1);
+       /* Can't forget the lock */
+       ast_mutex_init(&sc->lock);
+
+       /* Can't forget to record our pvt structure within the bridged channel structure */
+       bridge_channel->tech_pvt = sc;
+
+       set_softmix_bridge_data(softmix_data->internal_rate,
+               softmix_data->internal_mixing_interval
+                       ? softmix_data->internal_mixing_interval
+                       : DEFAULT_SOFTMIX_INTERVAL,
+               bridge_channel, 0, set_binaural, pos_id, is_announcement);
+
+       if (bridge->softmix.video_mode.mode == AST_BRIDGE_VIDEO_MODE_SFU) {
+               sfu_topologies_on_join(bridge_channel, &bridge->channels);
+       }
+
+       softmix_poke_thread(softmix_data);
+       return 0;
+}
+
+static void remove_destination_streams(struct ast_stream_topology *topology,
+       const char *channel_name)
+{
+       int i;
+
+       for (i = 0; i < ast_stream_topology_get_count(topology); ++i) {
+               struct ast_stream *stream;
+
+               stream = ast_stream_topology_get_stream(topology, i);
+
+               if (is_video_dest(stream, channel_name, NULL)) {
+                       ast_stream_set_state(stream, AST_STREAM_STATE_REMOVED);
+               }
+       }
+}
+
+static int sfu_topologies_on_leave(struct ast_bridge_channel *leaver, struct ast_bridge_channels_list *participants)
+{
+       struct ast_bridge_channel *participant;
+       struct softmix_channel *sc;
 
-               ao2_lock(bridge);
+       AST_LIST_TRAVERSE(participants, participant, entry) {
+               sc = participant->tech_pvt;
+               remove_destination_streams(sc->topology, ast_channel_name(leaver->chan));
+               ast_channel_request_stream_topology_change(participant->chan, sc->topology, NULL);
        }
 
-       ast_timer_close(timingfd);
+       sc = leaver->tech_pvt;
+       remove_destination_streams(sc->topology, "");
+       ast_channel_request_stream_topology_change(leaver->chan, sc->topology, NULL);
 
        return 0;
 }
 
-static struct ast_bridge_technology softmix_bridge = {
-       .name = "softmix",
-       .capabilities = AST_BRIDGE_CAPABILITY_MULTIMIX | AST_BRIDGE_CAPABILITY_THREAD | AST_BRIDGE_CAPABILITY_MULTITHREADED,
-       .preference = AST_BRIDGE_PREFERENCE_LOW,
-#ifdef SOFTMIX_16_SUPPORT
-       .formats = AST_FORMAT_SLINEAR16,
-#else
-       .formats = AST_FORMAT_SLINEAR,
-#endif
-       .create = softmix_bridge_create,
-       .join = softmix_bridge_join,
-       .leave = softmix_bridge_leave,
-       .write = softmix_bridge_write,
-       .thread = softmix_bridge_thread,
-       .poke = softmix_bridge_poke,
-};
+/*! \brief Function called when a channel leaves the bridge */
+static void softmix_bridge_leave(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+{
+       struct softmix_channel *sc;
+       struct softmix_bridge_data *softmix_data;
 
-static int unload_module(void)
+       softmix_data = bridge->tech_pvt;
+       sc = bridge_channel->tech_pvt;
+       if (!sc) {
+               return;
+       }
+
+       if (bridge->softmix.video_mode.mode == AST_BRIDGE_VIDEO_MODE_SFU) {
+               sfu_topologies_on_leave(bridge_channel, &bridge->channels);
+       }
+
+       if (bridge->softmix.binaural_active) {
+               if (sc->binaural) {
+                       set_binaural_data_leave(&softmix_data->convolve, sc->binaural_pos,
+                                       softmix_data->default_sample_size);
+               }
+       }
+
+       bridge_channel->tech_pvt = NULL;
+
+       ast_stream_topology_free(sc->topology);
+
+       /* Drop mutex lock */
+       ast_mutex_destroy(&sc->lock);
+
+       /* Drop the factory */
+       ast_slinfactory_destroy(&sc->factory);
+
+       /* Drop any formats on the frames */
+       ao2_cleanup(sc->write_frame.subclass.format);
+
+       /* Drop the DSP */
+       ast_dsp_free(sc->dsp);
+
+       /* Eep! drop ourselves */
+       ast_free(sc);
+}
+
+static void softmix_pass_video_top_priority(struct ast_bridge *bridge, struct ast_frame *frame)
 {
-       return ast_bridge_technology_unregister(&softmix_bridge);
+       struct ast_bridge_channel *cur;
+
+       AST_LIST_TRAVERSE(&bridge->channels, cur, entry) {
+               if (cur->suspended) {
+                       continue;
+               }
+               if (ast_bridge_is_video_src(bridge, cur->chan) == 1) {
+                       ast_bridge_channel_queue_frame(cur, frame);
+                       break;
+               }
+       }
 }
 
-static int load_module(void)
+/*!
+ * \internal
+ * \brief Determine what to do with a video frame.
+ * \since 12.0.0
+ *
+ * \param bridge Which bridge is getting the frame
+ * \param bridge_channel Which channel is writing the frame.
+ * \param frame What is being written.
+ *
+ * \return Nothing
+ */
+static void softmix_bridge_write_video(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel, struct ast_frame *frame)
 {
-       return ast_bridge_technology_register(&softmix_bridge);
+       struct softmix_channel *sc;
+       int video_src_priority;
+
+       /* Determine if the video frame should be distributed or not */
+       switch (bridge->softmix.video_mode.mode) {
+       case AST_BRIDGE_VIDEO_MODE_NONE:
+               break;
+       case AST_BRIDGE_VIDEO_MODE_SINGLE_SRC:
+               video_src_priority = ast_bridge_is_video_src(bridge, bridge_channel->chan);
+               if (video_src_priority == 1) {
+                       /* Pass to me and everyone else. */
+                       ast_bridge_queue_everyone_else(bridge, NULL, frame);
+               }
+               break;
+       case AST_BRIDGE_VIDEO_MODE_TALKER_SRC:
+               sc = bridge_channel->tech_pvt;
+               ast_mutex_lock(&sc->lock);
+               ast_bridge_update_talker_src_video_mode(bridge, bridge_channel->chan,
+                       sc->video_talker.energy_average,
+                       frame->subclass.frame_ending);
+               ast_mutex_unlock(&sc->lock);
+               video_src_priority = ast_bridge_is_video_src(bridge, bridge_channel->chan);
+               if (video_src_priority == 1) {
+                       int num_src = ast_bridge_number_video_src(bridge);
+                       int echo = num_src > 1 ? 0 : 1;
+
+                       ast_bridge_queue_everyone_else(bridge, echo ? NULL : bridge_channel, frame);
+               } else if (video_src_priority == 2) {
+                       softmix_pass_video_top_priority(bridge, frame);
+               }
+               break;
+       case AST_BRIDGE_VIDEO_MODE_SFU:
+               /* Nothing special to do here, the bridge channel stream map will ensure the
+                * video goes everywhere it needs to
+                */
+               ast_bridge_queue_everyone_else(bridge, bridge_channel, frame);
+               break;
+       }
+}
+
+/*!
+ * \internal
+ * \brief Determine what to do with a voice frame.
+ * \since 12.0.0
+ *
+ * \param bridge Which bridge is getting the frame
+ * \param bridge_channel Which channel is writing the frame.
+ * \param frame What is being written.
+ *
+ * \return Nothing
+ */
+static void softmix_bridge_write_voice(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel, struct ast_frame *frame)
+{
+       struct softmix_channel *sc = bridge_channel->tech_pvt;
+       struct softmix_bridge_data *softmix_data = bridge->tech_pvt;
+       int totalsilence = 0;
+       int cur_energy = 0;
+       int silence_threshold = bridge_channel->tech_args.silence_threshold ?
+               bridge_channel->tech_args.silence_threshold :
+               DEFAULT_SOFTMIX_SILENCE_THRESHOLD;
+       char update_talking = -1;  /* if this is set to 0 or 1, tell the bridge that the channel has started or stopped talking. */
+
+       /* Write the frame into the conference */
+       ast_mutex_lock(&sc->lock);
+
+       if (ast_format_cmp(frame->subclass.format, sc->read_slin_format) != AST_FORMAT_CMP_EQUAL) {
+               /*
+                * The incoming frame is not the expected format.  Update
+                * the channel's translation path to get us slinear from
+                * the new format for the next frame.
+                *
+                * There is the possibility that this frame is an old slinear
+                * rate frame that was in flight when the softmix bridge
+                * changed rates.  If so it will self correct on subsequent
+                * frames.
+                */
+               ast_channel_lock(bridge_channel->chan);
+               ast_debug(1, "Channel %s wrote unexpected format into bridge.  Got %s, expected %s.\n",
+                       ast_channel_name(bridge_channel->chan),
+                       ast_format_get_name(frame->subclass.format),
+                       ast_format_get_name(sc->read_slin_format));
+               ast_set_read_format_path(bridge_channel->chan, frame->subclass.format,
+                       sc->read_slin_format);
+               ast_channel_unlock(bridge_channel->chan);
+       }
+
+       /* The channel will be leaving soon if there is no dsp. */
+       if (sc->dsp) {
+               ast_dsp_silence_with_energy(sc->dsp, frame, &totalsilence, &cur_energy);
+       }
+
+       if (bridge->softmix.video_mode.mode == AST_BRIDGE_VIDEO_MODE_TALKER_SRC) {
+               int cur_slot = sc->video_talker.energy_history_cur_slot;
+
+               sc->video_talker.energy_accum -= sc->video_talker.energy_history[cur_slot];
+               sc->video_talker.energy_accum += cur_energy;
+               sc->video_talker.energy_history[cur_slot] = cur_energy;
+               sc->video_talker.energy_average = sc->video_talker.energy_accum / DEFAULT_ENERGY_HISTORY_LEN;
+               sc->video_talker.energy_history_cur_slot++;
+               if (sc->video_talker.energy_history_cur_slot == DEFAULT_ENERGY_HISTORY_LEN) {
+                       sc->video_talker.energy_history_cur_slot = 0; /* wrap around */
+               }
+       }
+
+       if (totalsilence < silence_threshold) {
+               if (!sc->talking) {
+                       update_talking = 1;
+               }
+               sc->talking = 1; /* tell the write process we have audio to be mixed out */
+       } else {
+               if (sc->talking) {
+                       update_talking = 0;
+               }
+               sc->talking = 0;
+       }
+
+       /* Before adding audio in, make sure we haven't fallen behind. If audio has fallen
+        * behind 4 times the amount of samples mixed on every iteration of the mixer, Re-sync
+        * the audio by flushing the buffer before adding new audio in. */
+       if (ast_slinfactory_available(&sc->factory) > (4 * SOFTMIX_SAMPLES(softmix_data->internal_rate, softmix_data->internal_mixing_interval))) {
+               ast_slinfactory_flush(&sc->factory);
+       }
+
+       /* If a frame was provided add it to the smoother, unless drop silence is enabled and this frame
+        * is not determined to be talking. */
+       if (!(bridge_channel->tech_args.drop_silence && !sc->talking)) {
+               ast_slinfactory_feed(&sc->factory, frame);
+       }
+
+       /* Alllll done */
+       ast_mutex_unlock(&sc->lock);
+
+       if (update_talking != -1) {
+               ast_bridge_channel_notify_talking(bridge_channel, update_talking);
+       }
+}
+
+static int remove_all_original_streams(struct ast_stream_topology *dest,
+       const struct ast_stream_topology *source,
+       const struct ast_stream_topology *original)
+{
+       int i;
+
+       for (i = 0; i < ast_stream_topology_get_count(source); ++i) {
+               struct ast_stream *stream;
+               int original_index;
+
+               stream = ast_stream_topology_get_stream(source, i);
+
+               /* Mark the existing stream as removed so we get a new one, this will get
+                * reused on a subsequent renegotiation.
+                */
+               for (original_index = 0; original_index < ast_stream_topology_get_count(original); ++original_index) {
+                       struct ast_stream *original_stream = ast_stream_topology_get_stream(original, original_index);
+
+                       if (!strcmp(ast_stream_get_name(stream), ast_stream_get_name(original_stream))) {
+                               struct ast_stream *removed;
+
+                               /* Since the participant is still going to be in the bridge we
+                                * change the name so that routing does not attempt to route video
+                                * to this stream.
+                                */
+                               removed = ast_stream_clone(stream, "removed");
+                               if (!removed) {
+                                       return -1;
+                               }
+
+                               ast_stream_set_state(removed, AST_STREAM_STATE_REMOVED);
+
+                               /* The destination topology can only ever contain the same, or more,
+                                * streams than the original so this is safe.
+                                */
+                               if (ast_stream_topology_set_stream(dest, original_index, removed)) {
+                                       ast_stream_free(removed);
+                                       return -1;
+                               }
+
+                               break;
+                       }
+               }
+       }
+
+       return 0;
+}
+
+static void sfu_topologies_on_source_change(struct ast_bridge_channel *source, struct ast_bridge_channels_list *participants)
+{
+       struct ast_stream_topology *source_video = NULL;
+       struct ast_bridge_channel *participant;
+       int res;
+
+       source_video = ast_stream_topology_alloc();
+       if (!source_video) {
+               return;
+       }
+
+       ast_channel_lock(source->chan);
+       res = append_source_streams(source_video, ast_channel_name(source->chan), ast_channel_get_stream_topology(source->chan));
+       ast_channel_unlock(source->chan);
+       if (res) {
+               goto cleanup;
+       }
+
+       AST_LIST_TRAVERSE(participants, participant, entry) {
+               struct ast_stream_topology *original_topology;
+               struct softmix_channel *sc;
+
+               if (participant == source) {
+                       continue;
+               }
+
+               sc = participant->tech_pvt;
+
+               original_topology = ast_stream_topology_clone(sc->topology);
+               if (!original_topology) {
+                       goto cleanup;
+               }
+
+               /* We add all the source streams back in, if any removed streams are already present they will
+                * get used first followed by appending new ones.
+                */
+               if (append_all_streams(sc->topology, source_video)) {
+                       ast_stream_topology_free(original_topology);
+                       goto cleanup;
+               }
+
+               /* And the original existing streams get marked as removed. This causes the remote side to see
+                * a new stream for the source streams.
+                */
+               if (remove_all_original_streams(sc->topology, source_video, original_topology)) {
+                       ast_stream_topology_free(original_topology);
+                       goto cleanup;
+               }
+
+               ast_channel_request_stream_topology_change(participant->chan, sc->topology, NULL);
+               ast_stream_topology_free(original_topology);
+       }
+
+cleanup:
+       ast_stream_topology_free(source_video);
+}
+
+/*!
+ * \internal
+ * \brief Determine what to do with a control frame.
+ * \since 12.0.0
+ *
+ * \param bridge Which bridge is getting the frame
+ * \param bridge_channel Which channel is writing the frame.
+ * \param frame What is being written.
+ *
+ * \retval 0 Frame accepted into the bridge.
+ * \retval -1 Frame needs to be deferred.
+ */
+static int softmix_bridge_write_control(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel, struct ast_frame *frame)
+{
+       struct softmix_bridge_data *softmix_data = bridge->tech_pvt;
+
+       /*
+        * XXX Softmix needs to use channel roles to determine what to
+        * do with control frames.
+        */
+
+       switch (frame->subclass.integer) {
+       case AST_CONTROL_VIDUPDATE:
+               if (!bridge->softmix.video_mode.video_update_discard ||
+                       ast_tvdiff_ms(ast_tvnow(), softmix_data->last_video_update) > bridge->softmix.video_mode.video_update_discard) {
+                       ast_bridge_queue_everyone_else(bridge, NULL, frame);
+                       softmix_data->last_video_update = ast_tvnow();
+               }
+               break;
+       case AST_CONTROL_STREAM_TOPOLOGY_SOURCE_CHANGED:
+               if (bridge->softmix.video_mode.mode == AST_BRIDGE_VIDEO_MODE_SFU) {
+                       sfu_topologies_on_source_change(bridge_channel, &bridge->channels);
+               }
+               break;
+       default:
+               break;
+       }
+
+       return 0;
+}
+
+/*!
+ * \internal
+ * \brief Determine what to do with a frame written into the bridge.
+ * \since 12.0.0
+ *
+ * \param bridge Which bridge is getting the frame
+ * \param bridge_channel Which channel is writing the frame.
+ * \param frame What is being written.
+ *
+ * \retval 0 Frame accepted into the bridge.
+ * \retval -1 Frame needs to be deferred.
+ *
+ * \note On entry, bridge is already locked.
+ */
+static int softmix_bridge_write(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel, struct ast_frame *frame)
+{
+       int res = 0;
+
+       if (!bridge->tech_pvt || !bridge_channel || !bridge_channel->tech_pvt) {
+               /* "Accept" the frame and discard it. */
+               return 0;
+       }
+
+       /*
+        * XXX Softmix needs to use channel roles to determine who gets
+        * what frame.  Possible roles: announcer, recorder, agent,
+        * supervisor.
+        */
+       switch (frame->frametype) {
+       case AST_FRAME_NULL:
+               /* "Accept" the frame and discard it. */
+               break;
+       case AST_FRAME_DTMF_BEGIN:
+       case AST_FRAME_DTMF_END:
+               res = ast_bridge_queue_everyone_else(bridge, bridge_channel, frame);
+               break;
+       case AST_FRAME_VOICE:
+               softmix_bridge_write_voice(bridge, bridge_channel, frame);
+               break;
+       case AST_FRAME_VIDEO:
+               softmix_bridge_write_video(bridge, bridge_channel, frame);
+               break;
+       case AST_FRAME_CONTROL:
+               res = softmix_bridge_write_control(bridge, bridge_channel, frame);
+               break;
+       case AST_FRAME_BRIDGE_ACTION:
+               res = ast_bridge_queue_everyone_else(bridge, bridge_channel, frame);
+               break;
+       case AST_FRAME_BRIDGE_ACTION_SYNC:
+               ast_log(LOG_ERROR, "Synchronous bridge action written to a softmix bridge.\n");
+               ast_assert(0);
+       default:
+               ast_debug(3, "Frame type %u unsupported\n", frame->frametype);
+               /* "Accept" the frame and discard it. */
+               break;
+       }
+
+       return res;
+}
+
+static void gather_softmix_stats(struct softmix_stats *stats,
+       const struct softmix_bridge_data *softmix_data,
+       struct ast_bridge_channel *bridge_channel)
+{
+       int channel_native_rate;
+
+       /* Gather stats about channel sample rates. */
+       ast_channel_lock(bridge_channel->chan);
+       channel_native_rate = MAX(SOFTMIX_MIN_SAMPLE_RATE,
+               ast_format_get_sample_rate(ast_channel_rawreadformat(bridge_channel->chan)));
+       ast_channel_unlock(bridge_channel->chan);
+
+       if (stats->highest_supported_rate < channel_native_rate) {
+               stats->highest_supported_rate = channel_native_rate;
+       }
+       if (softmix_data->internal_rate < channel_native_rate) {
+               int i;
+
+               for (i = 0; i < ARRAY_LEN(stats->sample_rates); i++) {
+                       if (stats->sample_rates[i] == channel_native_rate) {
+                               stats->num_channels[i]++;
+                               break;
+                       } else if (!stats->sample_rates[i]) {
+                               stats->sample_rates[i] = channel_native_rate;
+                               stats->num_channels[i]++;
+                               break;
+                       }
+               }
+               stats->num_above_internal_rate++;
+       } else if (softmix_data->internal_rate == channel_native_rate) {
+               stats->num_at_internal_rate++;
+       }
+}
+
+/*!
+ * \internal
+ * \brief Analyse mixing statistics and change bridges internal rate
+ * if necessary.
+ *
+ * \retval 0, no changes to internal rate
+ * \retval 1, internal rate was changed, update all the channels on the next mixing iteration.
+ */
+static unsigned int analyse_softmix_stats(struct softmix_stats *stats,
+               struct softmix_bridge_data *softmix_data, int binaural_active)
+{
+       int i;
+
+       if (binaural_active) {
+               stats->locked_rate = SOFTMIX_BINAURAL_SAMPLE_RATE;
+       }
+
+       /*
+        * Re-adjust the internal bridge sample rate if
+        * 1. The bridge's internal sample rate is locked in at a sample
+        *    rate other than the current sample rate being used.
+        * 2. two or more channels support a higher sample rate
+        * 3. no channels support the current sample rate or a higher rate
+        */
+       if (stats->locked_rate) {
+               /* if the rate is locked by the bridge, only update it if it differs
+                * from the current rate we are using. */
+               if (softmix_data->internal_rate != stats->locked_rate) {
+                       ast_debug(1, "Locking at new rate.  Bridge changed from %u to %u.\n",
+                               softmix_data->internal_rate, stats->locked_rate);
+                       softmix_data->internal_rate = stats->locked_rate;
+                       return 1;
+               }
+       } else if (stats->num_above_internal_rate >= 2) {
+               /* the highest rate is just used as a starting point */
+               unsigned int best_rate = stats->highest_supported_rate;
+               int best_index = -1;
+
+               for (i = 0; i < ARRAY_LEN(stats->num_channels); i++) {
+                       if (stats->num_channels[i]) {
+                               break;
+                       }
+                       if (2 <= stats->num_channels[i]) {
+                               /* Two or more channels support this rate. */
+                               if (best_index == -1
+                                       || stats->sample_rates[best_index] < stats->sample_rates[i]) {
+                                       /*
+                                        * best_rate starts out being the first sample rate
+                                        * greater than the internal sample rate that two or
+                                        * more channels support.
+                                        *
+                                        * or
+                                        *
+                                        * There are multiple rates above the internal rate
+                                        * and this rate is higher than the previous rate two
+                                        * or more channels support.
+                                        */
+                                       best_rate = stats->sample_rates[i];
+                                       best_index = i;
+                               }
+                       } else if (best_index == -1) {
+                               /*
+                                * It is possible that multiple channels exist with native sample
+                                * rates above the internal sample rate, but none of those channels
+                                * have the same rate in common.  In this case, the lowest sample
+                                * rate among those channels is picked. Over time as additional
+                                * statistic runs are made the internal sample rate number will
+                                * adjust to the most optimal sample rate, but it may take multiple
+                                * iterations.
+                                */
+                               best_rate = MIN(best_rate, stats->sample_rates[i]);
+                       }
+               }
+
+               ast_debug(1, "Multiple above internal rate.  Bridge changed from %u to %u.\n",
+                       softmix_data->internal_rate, best_rate);
+               softmix_data->internal_rate = best_rate;
+               return 1;
+       } else if (!stats->num_at_internal_rate && !stats->num_above_internal_rate) {
+               /* In this case, the highest supported rate is actually lower than the internal rate */
+               ast_debug(1, "All below internal rate.  Bridge changed from %u to %u.\n",
+                       softmix_data->internal_rate, stats->highest_supported_rate);
+               softmix_data->internal_rate = stats->highest_supported_rate;
+               return 1;
+       }
+       return 0;
+}
+
+static int softmix_mixing_array_init(struct softmix_mixing_array *mixing_array,
+               unsigned int starting_num_entries, unsigned int binaural_active)
+{
+       memset(mixing_array, 0, sizeof(*mixing_array));
+       mixing_array->max_num_entries = starting_num_entries;
+       if (!(mixing_array->buffers = ast_calloc(mixing_array->max_num_entries, sizeof(int16_t *)))) {
+               ast_log(LOG_NOTICE, "Failed to allocate softmix mixing structure.\n");
+               return -1;
+       }
+       if (binaural_active) {
+               if (!(mixing_array->chan_pairs = ast_calloc(mixing_array->max_num_entries,
+                               sizeof(struct convolve_channel_pair *)))) {
+                       ast_log(LOG_NOTICE, "Failed to allocate softmix mixing structure.\n");
+                       return -1;
+               }
+       }
+       return 0;
+}
+
+static void softmix_mixing_array_destroy(struct softmix_mixing_array *mixing_array,
+               unsigned int binaural_active)
+{
+       ast_free(mixing_array->buffers);
+       if (binaural_active) {
+               ast_free(mixing_array->chan_pairs);
+       }
+}
+
+static int softmix_mixing_array_grow(struct softmix_mixing_array *mixing_array,
+               unsigned int num_entries, unsigned int binaural_active)
+{
+       int16_t **tmp;
+
+       /* give it some room to grow since memory is cheap but allocations can be expensive */
+       mixing_array->max_num_entries = num_entries;
+       if (!(tmp = ast_realloc(mixing_array->buffers, (mixing_array->max_num_entries * sizeof(int16_t *))))) {
+               ast_log(LOG_NOTICE, "Failed to re-allocate softmix mixing structure.\n");
+               return -1;
+       }
+       if (binaural_active) {
+               struct convolve_channel_pair **tmp2;
+               if (!(tmp2 = ast_realloc(mixing_array->chan_pairs,
+                               (mixing_array->max_num_entries * sizeof(struct convolve_channel_pair *))))) {
+                       ast_log(LOG_NOTICE, "Failed to re-allocate softmix mixing structure.\n");
+                       return -1;
+               }
+               mixing_array->chan_pairs = tmp2;
+       }
+       mixing_array->buffers = tmp;
+       return 0;
+}
+
+/*!
+ * \brief Mixing loop.
+ *
+ * \retval 0 on success
+ * \retval -1 on failure
+ */
+static int softmix_mixing_loop(struct ast_bridge *bridge)
+{
+       struct softmix_stats stats = { { 0 }, };
+       struct softmix_mixing_array mixing_array;
+       struct softmix_bridge_data *softmix_data = bridge->tech_pvt;
+       struct ast_timer *timer;
+       struct softmix_translate_helper trans_helper;
+       int16_t buf[MAX_DATALEN];
+#ifdef BINAURAL_RENDERING
+       int16_t bin_buf[MAX_DATALEN];
+       int16_t ann_buf[MAX_DATALEN];
+#endif
+       unsigned int stat_iteration_counter = 0; /* counts down, gather stats at zero and reset. */
+       int timingfd;
+       int update_all_rates = 0; /* set this when the internal sample rate has changed */
+       unsigned int idx;
+       unsigned int x;
+       int res = -1;
+
+       timer = softmix_data->timer;
+       timingfd = ast_timer_fd(timer);
+       softmix_translate_helper_init(&trans_helper, softmix_data->internal_rate);
+       ast_timer_set_rate(timer, (1000 / softmix_data->internal_mixing_interval));
+
+       /* Give the mixing array room to grow, memory is cheap but allocations are expensive. */
+       if (softmix_mixing_array_init(&mixing_array, bridge->num_channels + 10,
+                       bridge->softmix.binaural_active)) {
+               goto softmix_cleanup;
+       }
+
+       /*
+        * XXX Softmix needs to use channel roles to determine who gets
+        * what audio mixed.
+        */
+       while (!softmix_data->stop && bridge->num_active) {
+               struct ast_bridge_channel *bridge_channel;
+               int timeout = -1;
+               struct ast_format *cur_slin = ast_format_cache_get_slin_by_rate(softmix_data->internal_rate);
+               unsigned int softmix_samples = SOFTMIX_SAMPLES(softmix_data->internal_rate, softmix_data->internal_mixing_interval);
+               unsigned int softmix_datalen = SOFTMIX_DATALEN(softmix_data->internal_rate, softmix_data->internal_mixing_interval);
+
+               if (softmix_datalen > MAX_DATALEN) {
+                       /* This should NEVER happen, but if it does we need to know about it. Almost
+                        * all the memcpys used during this process depend on this assumption.  Rather
+                        * than checking this over and over again through out the code, this single
+                        * verification is done on each iteration. */
+                       ast_log(LOG_WARNING,
+                               "Bridge %s: Conference mixing error, requested mixing length greater than mixing buffer.\n",
+                               bridge->uniqueid);
+                       goto softmix_cleanup;
+               }
+
+               /* Grow the mixing array buffer as participants are added. */
+               if (mixing_array.max_num_entries < bridge->num_channels
+                       && softmix_mixing_array_grow(&mixing_array, bridge->num_channels + 5,
+                                       bridge->softmix.binaural_active)) {
+                       goto softmix_cleanup;
+               }
+
+               /* init the number of buffers stored in the mixing array to 0.
+                * As buffers are added for mixing, this number is incremented. */
+               mixing_array.used_entries = 0;
+
+               /* These variables help determine if a rate change is required */
+               if (!stat_iteration_counter) {
+                       memset(&stats, 0, sizeof(stats));
+                       stats.locked_rate = bridge->softmix.internal_sample_rate;
+               }
+
+               /* If the sample rate has changed, update the translator helper */
+               if (update_all_rates) {
+                       softmix_translate_helper_change_rate(&trans_helper, softmix_data->internal_rate);
+               }
+
+#ifdef BINAURAL_RENDERING
+               check_binaural_position_change(bridge, softmix_data, bridge_channel);
+#endif
+
+               /* Go through pulling audio from each factory that has it available */
+               AST_LIST_TRAVERSE(&bridge->channels, bridge_channel, entry) {
+                       struct softmix_channel *sc = bridge_channel->tech_pvt;
+
+                       if (!sc) {
+                               /* This channel failed to join successfully. */
+                               continue;
+                       }
+
+                       /* Update the sample rate to match the bridge's native sample rate if necessary. */
+                       if (update_all_rates) {
+                               set_softmix_bridge_data(softmix_data->internal_rate,
+                                               softmix_data->internal_mixing_interval, bridge_channel, 1, -1, -1, -1);
+                       }
+
+                       /* If stat_iteration_counter is 0, then collect statistics during this mixing interation */
+                       if (!stat_iteration_counter) {
+                               gather_softmix_stats(&stats, softmix_data, bridge_channel);
+                       }
+
+                       /* if the channel is suspended, don't check for audio, but still gather stats */
+                       if (bridge_channel->suspended) {
+                               continue;
+                       }
+
+                       /* Try to get audio from the factory if available */
+                       ast_mutex_lock(&sc->lock);
+                       if ((mixing_array.buffers[mixing_array.used_entries] = softmix_process_read_audio(sc, softmix_samples))) {
+#ifdef BINAURAL_RENDERING
+                               add_binaural_mixing(bridge, softmix_data, softmix_samples, &mixing_array, sc,
+                                               ast_channel_name(bridge_channel->chan));
+#endif
+                               mixing_array.used_entries++;
+                       }
+                       ast_mutex_unlock(&sc->lock);
+               }
+
+               /* mix it like crazy (non binaural channels)*/
+               memset(buf, 0, softmix_datalen);
+               for (idx = 0; idx < mixing_array.used_entries; ++idx) {
+                       for (x = 0; x < softmix_samples; ++x) {
+                               ast_slinear_saturated_add(buf + x, mixing_array.buffers[idx] + x);
+                       }
+               }
+
+#ifdef BINAURAL_RENDERING
+               binaural_mixing(bridge, softmix_data, &mixing_array, bin_buf, ann_buf);
+#endif
+
+               /* Next step go through removing the channel's own audio and creating a good frame... */
+               AST_LIST_TRAVERSE(&bridge->channels, bridge_channel, entry) {
+                       struct softmix_channel *sc = bridge_channel->tech_pvt;
+
+                       if (!sc || bridge_channel->suspended) {
+                               /* This channel failed to join successfully or is suspended. */
+                               continue;
+                       }
+
+                       ast_mutex_lock(&sc->lock);
+
+                       /* Make SLINEAR write frame from local buffer */
+                       ao2_t_replace(sc->write_frame.subclass.format, cur_slin,
+                               "Replace softmix channel slin format");
+#ifdef BINAURAL_RENDERING
+                       if (bridge->softmix.binaural_active && softmix_data->convolve.binaural_active
+                                       && sc->binaural) {
+                               create_binaural_frame(bridge_channel, sc, bin_buf, ann_buf, softmix_datalen,
+                                               softmix_samples, buf);
+                       } else
+#endif
+                       {
+                               sc->write_frame.datalen = softmix_datalen;
+                               sc->write_frame.samples = softmix_samples;
+                               memcpy(sc->final_buf, buf, softmix_datalen);
+                       }
+                       /* process the softmix channel's new write audio */
+                       softmix_process_write_audio(&trans_helper,
+                                       ast_channel_rawwriteformat(bridge_channel->chan), sc,
+                                       softmix_data->default_sample_size);
+
+                       ast_mutex_unlock(&sc->lock);
+
+                       /* A frame is now ready for the channel. */
+                       ast_bridge_channel_queue_frame(bridge_channel, &sc->write_frame);
+               }
+
+               update_all_rates = 0;
+               if (!stat_iteration_counter) {
+                       update_all_rates = analyse_softmix_stats(&stats, softmix_data,
+                                       bridge->softmix.binaural_active);
+                       stat_iteration_counter = SOFTMIX_STAT_INTERVAL;
+               }
+               stat_iteration_counter--;
+
+               ast_bridge_unlock(bridge);
+               /* cleanup any translation frame data from the previous mixing iteration. */
+               softmix_translate_helper_cleanup(&trans_helper);
+               /* Wait for the timing source to tell us to wake up and get things done */
+               ast_waitfor_n_fd(&timingfd, 1, &timeout, NULL);
+               if (ast_timer_ack(timer, 1) < 0) {
+                       ast_log(LOG_ERROR, "Bridge %s: Failed to acknowledge timer in softmix.\n",
+                               bridge->uniqueid);
+                       ast_bridge_lock(bridge);
+                       goto softmix_cleanup;
+               }
+               ast_bridge_lock(bridge);
+
+               /* make sure to detect mixing interval changes if they occur. */
+               if (bridge->softmix.internal_mixing_interval
+                       && (bridge->softmix.internal_mixing_interval != softmix_data->internal_mixing_interval)) {
+                       softmix_data->internal_mixing_interval = bridge->softmix.internal_mixing_interval;
+                       ast_timer_set_rate(timer, (1000 / softmix_data->internal_mixing_interval));
+                       update_all_rates = 1; /* if the interval changes, the rates must be adjusted as well just to be notified new interval.*/
+               }
+       }
+
+       res = 0;
+
+softmix_cleanup:
+       softmix_translate_helper_destroy(&trans_helper);
+       softmix_mixing_array_destroy(&mixing_array, bridge->softmix.binaural_active);
+       return res;
+}
+
+/*!
+ * \internal
+ * \brief Mixing thread.
+ * \since 12.0.0
+ *
+ * \note The thread does not have its own reference to the
+ * bridge.  The lifetime of the thread is tied to the lifetime
+ * of the mixing technology association with the bridge.
+ */
+static void *softmix_mixing_thread(void *data)
+{
+       struct softmix_bridge_data *softmix_data = data;
+       struct ast_bridge *bridge = softmix_data->bridge;
+
+       ast_bridge_lock(bridge);
+       if (bridge->callid) {
+               ast_callid_threadassoc_add(bridge->callid);
+       }
+
+       ast_debug(1, "Bridge %s: starting mixing thread\n", bridge->uniqueid);
+
+       while (!softmix_data->stop) {
+               if (!bridge->num_active) {
+                       /* Wait for something to happen to the bridge. */
+                       ast_bridge_unlock(bridge);
+                       ast_mutex_lock(&softmix_data->lock);
+                       if (!softmix_data->stop) {
+                               ast_cond_wait(&softmix_data->cond, &softmix_data->lock);
+                       }
+                       ast_mutex_unlock(&softmix_data->lock);
+                       ast_bridge_lock(bridge);
+                       continue;
+               }
+
+               if (bridge->softmix.binaural_active && !softmix_data->binaural_init) {
+#ifndef BINAURAL_RENDERING
+                       ast_bridge_lock(bridge);
+                       bridge->softmix.binaural_active = 0;
+                       ast_bridge_unlock(bridge);
+                       ast_log(LOG_WARNING, "Bridge: %s: Binaural rendering active by config but not "
+                                       "compiled.\n", bridge->uniqueid);
+#else
+                       /* Set and init binaural data if binaural is activated in the configuration. */
+                       softmix_data->internal_rate = SOFTMIX_BINAURAL_SAMPLE_RATE;
+                       softmix_data->default_sample_size = SOFTMIX_SAMPLES(softmix_data->internal_rate,
+                                       softmix_data->internal_mixing_interval);
+                       /* If init for binaural processing fails we will fall back to mono audio processing. */
+                       if (init_convolve_data(&softmix_data->convolve, softmix_data->default_sample_size)
+                                       == -1) {
+                               ast_bridge_lock(bridge);
+                               bridge->softmix.binaural_active = 0;
+                               ast_bridge_unlock(bridge);
+                               ast_log(LOG_ERROR, "Bridge: %s: Unable to allocate memory for "
+                                               "binaural processing,  Will only process mono audio.\n",
+                                               bridge->uniqueid);
+                       }
+                       softmix_data->binaural_init = 1;
+#endif
+               }
+
+               if (softmix_mixing_loop(bridge)) {
+                       /*
+                        * A mixing error occurred.  Sleep and try again later so we
+                        * won't flood the logs.
+                        */
+                       ast_bridge_unlock(bridge);
+                       sleep(1);
+                       ast_bridge_lock(bridge);
+               }
+       }
+
+       ast_bridge_unlock(bridge);
+
+       ast_debug(1, "Bridge %s: stopping mixing thread\n", bridge->uniqueid);
+
+       return NULL;
+}
+
+static void softmix_bridge_data_destroy(struct softmix_bridge_data *softmix_data)
+{
+       if (softmix_data->timer) {
+               ast_timer_close(softmix_data->timer);
+               softmix_data->timer = NULL;
+       }
+       ast_mutex_destroy(&softmix_data->lock);
+       ast_cond_destroy(&softmix_data->cond);
+       ast_free(softmix_data);
+}
+
+/*! \brief Function called when a bridge is created */
+static int softmix_bridge_create(struct ast_bridge *bridge)
+{
+       struct softmix_bridge_data *softmix_data;
+
+       softmix_data = ast_calloc(1, sizeof(*softmix_data));
+       if (!softmix_data) {
+               return -1;
+       }
+       softmix_data->bridge = bridge;
+       ast_mutex_init(&softmix_data->lock);
+       ast_cond_init(&softmix_data->cond, NULL);
+       softmix_data->timer = ast_timer_open();
+       if (!softmix_data->timer) {
+               ast_log(AST_LOG_WARNING, "Failed to open timer for softmix bridge\n");
+               softmix_bridge_data_destroy(softmix_data);
+               return -1;
+       }
+       /* start at minimum rate, let it grow from there */
+       softmix_data->internal_rate = SOFTMIX_MIN_SAMPLE_RATE;
+       softmix_data->internal_mixing_interval = DEFAULT_SOFTMIX_INTERVAL;
+
+#ifdef BINAURAL_RENDERING
+       softmix_data->default_sample_size = SOFTMIX_SAMPLES(softmix_data->internal_rate,
+                       softmix_data->internal_mixing_interval);
+#endif
+
+       bridge->tech_pvt = softmix_data;
+
+       /* Start the mixing thread. */
+       if (ast_pthread_create(&softmix_data->thread, NULL, softmix_mixing_thread,
+               softmix_data)) {
+               softmix_data->thread = AST_PTHREADT_NULL;
+               softmix_bridge_data_destroy(softmix_data);
+               bridge->tech_pvt = NULL;
+               return -1;
+       }
+
+       return 0;
+}
+
+/*!
+ * \internal
+ * \brief Request the softmix mixing thread stop.
+ * \since 12.0.0
+ *
+ * \param bridge Which bridge is being stopped.
+ *
+ * \return Nothing
+ */
+static void softmix_bridge_stop(struct ast_bridge *bridge)
+{
+       struct softmix_bridge_data *softmix_data;
+
+       softmix_data = bridge->tech_pvt;
+       if (!softmix_data) {
+               return;
+       }
+
+       ast_mutex_lock(&softmix_data->lock);
+       softmix_data->stop = 1;
+       ast_mutex_unlock(&softmix_data->lock);
+}
+
+/*! \brief Function called when a bridge is destroyed */
+static void softmix_bridge_destroy(struct ast_bridge *bridge)
+{
+       struct softmix_bridge_data *softmix_data;
+       pthread_t thread;
+
+       softmix_data = bridge->tech_pvt;
+       if (!softmix_data) {
+               return;
+       }
+
+       /* Stop the mixing thread. */
+       ast_mutex_lock(&softmix_data->lock);
+       softmix_data->stop = 1;
+       ast_cond_signal(&softmix_data->cond);
+       thread = softmix_data->thread;
+       softmix_data->thread = AST_PTHREADT_NULL;
+       ast_mutex_unlock(&softmix_data->lock);
+       if (thread != AST_PTHREADT_NULL) {
+               ast_debug(1, "Bridge %s: Waiting for mixing thread to die.\n", bridge->uniqueid);
+               pthread_join(thread, NULL);
+       }
+#ifdef BINAURAL_RENDERING
+       free_convolve_data(&softmix_data->convolve);
+#endif
+       softmix_bridge_data_destroy(softmix_data);
+       bridge->tech_pvt = NULL;
+}
+
+/*!
+ * \brief Map a source stream to all of its destination streams.
+ *
+ * \param source_stream_name Name of the source stream
+ * \param source_channel_name Name of channel where the source stream originates
+ * \param bridge_stream_position The slot in the bridge where source video will come from
+ * \param participants The bridge_channels in the bridge
+ */
+static void map_source_to_destinations(const char *source_stream_name, const char *source_channel_name,
+       size_t bridge_stream_position, struct ast_bridge_channels_list *participants)
+{
+       struct ast_bridge_channel *participant;
+
+       AST_LIST_TRAVERSE(participants, participant, entry) {
+               int i;
+               struct ast_stream_topology *topology;
+
+               if (!strcmp(source_channel_name, ast_channel_name(participant->chan))) {
+                       continue;
+               }
+
+               ast_bridge_channel_lock(participant);
+               ast_channel_lock(participant->chan);
+               topology = ast_channel_get_stream_topology(participant->chan);
+
+               for (i = 0; i < ast_stream_topology_get_count(topology); ++i) {
+                       struct ast_stream *stream;
+
+                       stream = ast_stream_topology_get_stream(topology, i);
+                       if (is_video_dest(stream, source_channel_name, source_stream_name)) {
+                               AST_VECTOR_REPLACE(&participant->stream_map.to_channel, bridge_stream_position, i);
+                               break;
+                       }
+               }
+               ast_channel_unlock(participant->chan);
+               ast_bridge_channel_unlock(participant);
+       }
+}
+
+/*!
+ * \brief stream_topology_changed callback
+ *
+ * For most video modes, nothing beyond the ordinary is required.
+ * For the SFU case, though, we need to completely remap the streams
+ * in order to ensure video gets directed where it is expected to go.
+ *
+ * \param bridge The bridge
+ * \param bridge_channel Channel whose topology has changed
+ */
+static void softmix_bridge_stream_topology_changed(struct ast_bridge *bridge, struct ast_bridge_channel *bridge_channel)
+{
+       struct ast_bridge_channel *participant;
+       struct ast_vector_int media_types;
+       int nths[AST_MEDIA_TYPE_END] = {0};
+
+       switch (bridge->softmix.video_mode.mode) {
+       case AST_BRIDGE_VIDEO_MODE_NONE:
+       case AST_BRIDGE_VIDEO_MODE_SINGLE_SRC:
+       case AST_BRIDGE_VIDEO_MODE_TALKER_SRC:
+       default:
+               ast_bridge_channel_stream_map(bridge_channel);
+               return;
+       case AST_BRIDGE_VIDEO_MODE_SFU:
+               break;
+       }
+
+       AST_VECTOR_INIT(&media_types, AST_MEDIA_TYPE_END);
+
+       /* First traversal: re-initialize all of the participants' stream maps */
+       AST_LIST_TRAVERSE(&bridge->channels, participant, entry) {
+               ast_bridge_channel_lock(participant);
+               AST_VECTOR_RESET(&participant->stream_map.to_channel, AST_VECTOR_ELEM_CLEANUP_NOOP);
+               AST_VECTOR_RESET(&participant->stream_map.to_bridge, AST_VECTOR_ELEM_CLEANUP_NOOP);
+               ast_bridge_channel_unlock(participant);
+       }
+
+       /* Second traversal: Map specific video channels from their source to their destinations.
+        *
+        * This is similar to what is done in ast_stream_topology_map(),
+        * except that video channels are handled differently.  Each video
+        * source has it's own unique index on the bridge.  This way, a
+        * particular channel's source video can be distributed to the
+        * appropriate destination streams on the other channels.
+        */
+       AST_LIST_TRAVERSE(&bridge->channels, participant, entry) {
+               int i;
+               struct ast_stream_topology *topology;
+
+               ast_bridge_channel_lock(participant);
+               ast_channel_lock(participant->chan);
+
+               topology = ast_channel_get_stream_topology(participant->chan);
+               if (topology) {
+                       /*
+                        * Sigh.  We have to clone to avoid deadlock in
+                        * map_source_to_destinations() because topology
+                        * is not an ao2 object.
+                        */
+                       topology = ast_stream_topology_clone(topology);
+               }
+               if (!topology) {
+                       /* Oh, my, we are in trouble. */
+                       ast_channel_unlock(participant->chan);
+                       ast_bridge_channel_unlock(participant);
+                       continue;
+               }
+
+               for (i = 0; i < ast_stream_topology_get_count(topology); ++i) {
+                       struct ast_stream *stream = ast_stream_topology_get_stream(topology, i);
+
+                       if (is_video_source(stream)) {
+                               AST_VECTOR_APPEND(&media_types, AST_MEDIA_TYPE_VIDEO);
+                               AST_VECTOR_REPLACE(&participant->stream_map.to_bridge, i, AST_VECTOR_SIZE(&media_types) - 1);
+                               AST_VECTOR_REPLACE(&participant->stream_map.to_channel, AST_VECTOR_SIZE(&media_types) - 1, -1);
+                               /*
+                                * Unlock the channel and participant to prevent
+                                * potential deadlock in map_source_to_destinations().
+                                */
+                               ast_channel_unlock(participant->chan);
+                               ast_bridge_channel_unlock(participant);
+                               map_source_to_destinations(ast_stream_get_name(stream), ast_channel_name(participant->chan),
+                                       AST_VECTOR_SIZE(&media_types) - 1, &bridge->channels);
+                               ast_bridge_channel_lock(participant);
+                               ast_channel_lock(participant->chan);
+                       } else if (is_video_dest(stream, NULL, NULL)) {
+                               /* We expect to never read media from video destination channels, but just
+                                * in case, we should set their to_bridge value to -1.
+                                */
+                               AST_VECTOR_REPLACE(&participant->stream_map.to_bridge, i, -1);
+                       } else {
+                               /* XXX This is copied from ast_stream_topology_map(). This likely could
+                                * be factored out in some way
+                                */
+                               enum ast_media_type type = ast_stream_get_type(stream);
+                               int index = AST_VECTOR_GET_INDEX_NTH(&media_types, ++nths[type],
+                                       type, AST_VECTOR_ELEM_DEFAULT_CMP);
+
+                               if (index == -1) {
+                                       AST_VECTOR_APPEND(&media_types, type);
+                                       index = AST_VECTOR_SIZE(&media_types) - 1;
+                               }
+
+                               AST_VECTOR_REPLACE(&participant->stream_map.to_bridge, i, index);
+                               AST_VECTOR_REPLACE(&participant->stream_map.to_channel, index, i);
+                       }
+               }
+
+               ast_stream_topology_free(topology);
+
+               ast_channel_unlock(participant->chan);
+               ast_bridge_channel_unlock(participant);
+       }
+}
+
+static struct ast_bridge_technology softmix_bridge = {
+       .name = "softmix",
+       .capabilities = AST_BRIDGE_CAPABILITY_MULTIMIX,
+       .preference = AST_BRIDGE_PREFERENCE_BASE_MULTIMIX,
+       .create = softmix_bridge_create,
+       .stop = softmix_bridge_stop,
+       .destroy = softmix_bridge_destroy,
+       .join = softmix_bridge_join,
+       .leave = softmix_bridge_leave,
+       .unsuspend = softmix_bridge_unsuspend,
+       .write = softmix_bridge_write,
+       .stream_topology_changed = softmix_bridge_stream_topology_changed,
+};
+
+#ifdef TEST_FRAMEWORK
+struct stream_parameters {
+       const char *name;
+       const char *formats;
+       enum ast_media_type type;
+};
+
+static struct ast_stream_topology *build_topology(const struct stream_parameters *params, size_t num_streams)
+{
+       struct ast_stream_topology *topology;
+       size_t i;
+
+       topology = ast_stream_topology_alloc();
+       if (!topology) {
+               return NULL;
+       }
+
+       for (i = 0; i < num_streams; ++i) {
+               RAII_VAR(struct ast_format_cap *, caps, NULL, ao2_cleanup);
+               struct ast_stream *stream;
+
+               caps = ast_format_cap_alloc(AST_FORMAT_CAP_FLAG_DEFAULT);
+               if (!caps) {
+                       goto fail;
+               }
+               if (ast_format_cap_update_by_allow_disallow(caps, params[i].formats, 1) < 0) {
+                       goto fail;
+               }
+               stream = ast_stream_alloc(params[i].name, params[i].type);
+               if (!stream) {
+                       goto fail;
+               }
+               ast_stream_set_formats(stream, caps);
+               if (ast_stream_topology_append_stream(topology, stream) < 0) {
+                       ast_stream_free(stream);
+                       goto fail;
+               }
+       }
+
+       return topology;
+
+fail:
+       ast_stream_topology_free(topology);
+       return NULL;
+}
+
+static int validate_stream(struct ast_test *test, struct ast_stream *stream,
+       const struct stream_parameters *params)
+{
+       struct ast_format_cap *stream_caps;
+       struct ast_format_cap *params_caps;
+
+       if (ast_stream_get_type(stream) != params->type) {
+               ast_test_status_update(test, "Expected stream type '%s' but got type '%s'\n",
+                       ast_codec_media_type2str(params->type),
+                       ast_codec_media_type2str(ast_stream_get_type(stream)));
+               return -1;
+       }
+       if (strcmp(ast_stream_get_name(stream), params->name)) {
+               ast_test_status_update(test, "Expected stream name '%s' but got type '%s'\n",
+                       params->name, ast_stream_get_name(stream));
+               return -1;
+       }
+
+       stream_caps = ast_stream_get_formats(stream);
+       params_caps = ast_format_cap_alloc(AST_FORMAT_CAP_FLAG_DEFAULT);
+       if (!params_caps) {
+               ast_test_status_update(test, "Allocation error on capabilities\n");
+               return -1;
+       }
+       ast_format_cap_update_by_allow_disallow(params_caps, params->formats, 1);
+
+       if (ast_format_cap_identical(stream_caps, params_caps)) {
+               ast_test_status_update(test, "Formats are not as expected on stream '%s'\n",
+                       ast_stream_get_name(stream));
+               ao2_cleanup(params_caps);
+               return -1;
+       }
+
+       ao2_cleanup(params_caps);
+       return 0;
+}
+
+static int validate_original_streams(struct ast_test *test, struct ast_stream_topology *topology,
+       const struct stream_parameters *params, size_t num_streams)
+{
+       int i;
+
+       if (ast_stream_topology_get_count(topology) < num_streams) {
+               ast_test_status_update(test, "Topology only has %d streams. Needs to have at least %zu\n",
+                       ast_stream_topology_get_count(topology), num_streams);
+               return -1;
+       }
+
+       for (i = 0; i < ARRAY_LEN(params); ++i) {
+               if (validate_stream(test, ast_stream_topology_get_stream(topology, i), &params[i])) {
+                       return -1;
+               }
+       }
+
+       return 0;
+}
+
+AST_TEST_DEFINE(sfu_append_source_streams)
+{
+       enum ast_test_result_state res = AST_TEST_FAIL;
+       static const struct stream_parameters bob_streams[] = {
+               { "bob_audio", "ulaw,alaw,g722,opus", AST_MEDIA_TYPE_AUDIO, },
+               { "bob_video", "h264,vp8", AST_MEDIA_TYPE_VIDEO, },
+       };
+       static const struct stream_parameters alice_streams[] = {
+               { "alice_audio", "ulaw,opus", AST_MEDIA_TYPE_AUDIO, },
+               { "alice_video", "vp8", AST_MEDIA_TYPE_VIDEO, },
+       };
+       static const struct stream_parameters alice_dest_stream = {
+               "softbridge_dest_PJSIP/Bob-00000001_bob_video", "vp8", AST_MEDIA_TYPE_VIDEO,
+       };
+       static const struct stream_parameters bob_dest_stream = {
+               "softbridge_dest_PJSIP/Alice-00000000_alice_video", "h264,vp8", AST_MEDIA_TYPE_VIDEO,
+       };
+       struct ast_stream_topology *topology_alice = NULL;
+       struct ast_stream_topology *topology_bob = NULL;
+
+       switch (cmd) {
+       case TEST_INIT:
+               info->name = "sfu_append_source_streams";
+               info->category = "/bridges/bridge_softmix/";
+               info->summary = "Test appending of video streams";
+               info->description =
+                       "This tests does stuff.";
+               return AST_TEST_NOT_RUN;
+       case TEST_EXECUTE:
+               break;
+       }
+
+       topology_alice = build_topology(alice_streams, ARRAY_LEN(alice_streams));
+       if (!topology_alice) {
+               goto end;
+       }
+
+       topology_bob = build_topology(bob_streams, ARRAY_LEN(bob_streams));
+       if (!topology_bob) {
+               goto end;
+       }
+
+       if (append_source_streams(topology_alice, "PJSIP/Bob-00000001", topology_bob)) {
+               ast_test_status_update(test, "Failed to append Bob's streams to Alice\n");
+               goto end;
+       }
+
+       if (ast_stream_topology_get_count(topology_alice) != 3) {
+               ast_test_status_update(test, "Alice's topology isn't large enough! It's %d but needs to be %d\n",
+                       ast_stream_topology_get_count(topology_alice), 3);
+               goto end;
+       }
+
+       if (validate_original_streams(test, topology_alice, alice_streams, ARRAY_LEN(alice_streams))) {
+               goto end;
+       }
+
+       if (validate_stream(test, ast_stream_topology_get_stream(topology_alice, 2), &alice_dest_stream)) {
+               goto end;
+       }
+
+       if (append_source_streams(topology_bob, "PJSIP/Alice-00000000", topology_alice)) {
+               ast_test_status_update(test, "Failed to append Alice's streams to Bob\n");
+               goto end;
+       }
+
+       if (ast_stream_topology_get_count(topology_bob) != 3) {
+               ast_test_status_update(test, "Bob's topology isn't large enough! It's %d but needs to be %d\n",
+                       ast_stream_topology_get_count(topology_bob), 3);
+               goto end;
+       }
+
+       if (validate_original_streams(test, topology_bob, bob_streams, ARRAY_LEN(bob_streams))) {
+               goto end;
+       }
+
+       if (validate_stream(test, ast_stream_topology_get_stream(topology_bob, 2), &bob_dest_stream)) {
+               goto end;
+       }
+
+       res = AST_TEST_PASS;
+
+end:
+       ast_stream_topology_free(topology_alice);
+       ast_stream_topology_free(topology_bob);
+       return res;
+}
+
+AST_TEST_DEFINE(sfu_remove_destination_streams)
+{
+       enum ast_test_result_state res = AST_TEST_FAIL;
+       static const struct stream_parameters params[] = {
+               { "alice_audio", "ulaw,alaw,g722,opus", AST_MEDIA_TYPE_AUDIO, },
+               { "alice_video", "h264,vp8", AST_MEDIA_TYPE_VIDEO, },
+               { "softbridge_dest_PJSIP/Bob-00000001_video", "vp8", AST_MEDIA_TYPE_VIDEO, },
+               { "softbridge_dest_PJSIP/Carol-00000002_video", "h264", AST_MEDIA_TYPE_VIDEO, },
+       };
+       static const struct {
+               const char *channel_name;
+               int num_streams;
+               int params_index[4];
+       } removal_results[] = {
+               { "PJSIP/Bob-00000001", 4, { 0, 1, 2, 3 }, },
+               { "PJSIP/Edward-00000004", 4, { 0, 1, 2, 3 }, },
+               { "", 4, { 0, 1, 2, 3 }, },
+       };
+       struct ast_stream_topology *orig = NULL;
+       int i;
+
+       switch (cmd) {
+       case TEST_INIT:
+               info->name = "sfu_remove_destination_streams";
+               info->category = "/bridges/bridge_softmix/";
+               info->summary = "Test removal of destination video streams";
+               info->description =
+                       "This tests does stuff.";
+               return AST_TEST_NOT_RUN;
+       case TEST_EXECUTE:
+               break;
+       }
+
+       orig = build_topology(params, ARRAY_LEN(params));
+       if (!orig) {
+               ast_test_status_update(test, "Unable to build initial stream topology\n");
+               goto end;
+       }
+
+       for (i = 0; i < ARRAY_LEN(removal_results); ++i) {
+               int j;
+
+               remove_destination_streams(orig, removal_results[i].channel_name);
+
+               if (ast_stream_topology_get_count(orig) != removal_results[i].num_streams) {
+                       ast_test_status_update(test, "Resulting topology has %d streams, when %d are expected\n",
+                               ast_stream_topology_get_count(orig), removal_results[i].num_streams);
+                       goto end;
+               }
+
+               for (j = 0; j < removal_results[i].num_streams; ++j) {
+                       struct ast_stream *actual;
+                       struct ast_stream *expected;
+                       int orig_index;
+
+                       actual = ast_stream_topology_get_stream(orig, j);
+
+                       orig_index = removal_results[i].params_index[j];
+                       expected = ast_stream_topology_get_stream(orig, orig_index);
+
+                       if (!ast_format_cap_identical(ast_stream_get_formats(actual),
+                               ast_stream_get_formats(expected))) {
+                               struct ast_str *expected_str;
+                               struct ast_str *actual_str;
+
+                               expected_str = ast_str_alloca(64);
+                               actual_str = ast_str_alloca(64);
+
+                               ast_test_status_update(test, "Mismatch between expected (%s) and actual (%s) stream formats\n",
+                                       ast_format_cap_get_names(ast_stream_get_formats(expected), &expected_str),
+                                       ast_format_cap_get_names(ast_stream_get_formats(actual), &actual_str));
+                               goto end;
+                       }
+
+                       if (is_video_dest(actual, removal_results[i].channel_name, NULL) &&
+                               ast_stream_get_state(actual) != AST_STREAM_STATE_REMOVED) {
+                               ast_test_status_update(test, "Removed stream %s does not have a state of removed\n", ast_stream_get_name(actual));
+                               goto end;
+                       }
+               }
+       }
+
+       res = AST_TEST_PASS;
+
+end:
+       ast_stream_topology_free(orig);
+       return res;
+}
+
+#endif
+
+static int unload_module(void)
+{
+       ast_bridge_technology_unregister(&softmix_bridge);
+       AST_TEST_UNREGISTER(sfu_append_source_streams);
+       AST_TEST_UNREGISTER(sfu_remove_destination_streams);
+       return 0;
+}
+
+static int load_module(void)
+{
+       if (ast_bridge_technology_register(&softmix_bridge)) {
+               unload_module();
+               return AST_MODULE_LOAD_DECLINE;
+       }
+       AST_TEST_REGISTER(sfu_append_source_streams);
+       AST_TEST_REGISTER(sfu_remove_destination_streams);
+       return AST_MODULE_LOAD_SUCCESS;
 }
 
 AST_MODULE_INFO_STANDARD(ASTERISK_GPL_KEY, "Multi-party software based channel mixing");