kpb: add multi-KPB WOV arbiter for 3-keyword DMIC capture with VAD

app/boards/intel_adsp_cavs25.conf

@@ -11,6 +11,11 @@ CONFIG_COMP_DRC=y
 CONFIG_COMP_MFCC=y
 CONFIG_COMP_MULTIBAND_DRC=y
 CONFIG_COMP_VOLUME_WINDOWS_FADE=y
+CONFIG_COMP_WOV_ARBITER=y
+CONFIG_COMP_VAD_GATE=y
+CONFIG_COMP_KPB=y
+CONFIG_SAMPLES=y
+CONFIG_SAMPLE_KEYPHRASE=y
 CONFIG_FORMAT_CONVERT_HIFI3=n
 CONFIG_PCM_CONVERTER_FORMAT_S16LE=y
 CONFIG_PCM_CONVERTER_FORMAT_S24LE=y

app/debug_overlay.conf

@@ -31,3 +31,8 @@ CONFIG_GDBSTUB_ENTER_IMMEDIATELY=n
 # Testing with runtime filtering enabled ensures the same feature set is validated.
 # Note: This setting has no effect if CONFIG_LOG_RUNTIME_FILTERING is disabled.
 CONFIG_LOG_RUNTIME_DEFAULT_LEVEL=3
+
+# Record fatal exception breadcrumbs (PC/cause/vaddr) in HP-SRAM window0 so the
+# crash is visible in the host dmesg "Firmware state" line when no console or
+# mtrace output is available.
+CONFIG_XTENSA_ADSP_FATAL_BREADCRUMB=y

app/prj.conf

@@ -49,3 +49,5 @@ CONFIG_SCHED_CPU_MASK=y
 CONFIG_SYS_CLOCK_TICKS_PER_SEC=15000
 CONFIG_DAI=y
 CONFIG_HEAP_MEM_POOL_SIZE=2048
+CONFIG_SAMPLES=y
+CONFIG_SAMPLE_KEYPHRASE=y

src/audio/CMakeLists.txt

@@ -104,9 +104,15 @@ if(NOT CONFIG_COMP_MODULE_SHARED_LIBRARY_BUILD)
 	if(CONFIG_COMP_UP_DOWN_MIXER)
 		add_subdirectory(up_down_mixer)
 	endif()
+	if(CONFIG_COMP_VAD_GATE)
+		add_subdirectory(vad_gate)
+	endif()
 	if(CONFIG_COMP_VOLUME)
 		add_subdirectory(volume)
 	endif()
+	if(CONFIG_COMP_WOV_ARBITER)
+		add_subdirectory(wov_arbiter)
+	endif()
 	if(CONFIG_DTS_CODEC)
 		add_subdirectory(codec)
 	endif()

src/audio/Kconfig

@@ -102,6 +102,18 @@ config COMP_STUBS
 	  Select to force all 3P blocks to link against stubs rather than their libraries. This
 	  should only be used in testing environments like fuzzers or CI.
 
+config COMP_WOV_ARBITER
+	bool "WOV arbiter component"
+	depends on COMP_KPB
+	depends on IPC_MAJOR_4
+	select AMS
+	help
+	  Select to build the WOV (Wake-on-Voice) arbiter.  The arbiter sits
+	  between multiple KPB host-drain outputs and a single host PCM copier.
+	  When a keyword is detected by one of the WOV detectors the arbiter
+	  routes that KPB's drain stream to the host and instructs the remaining
+	  detectors to pause.
+
 config COMP_KPB
 	bool "KPB component"
 	default y
@@ -160,6 +172,7 @@ rsource "template/Kconfig"
 rsource "tensorflow/Kconfig"
 rsource "tone/Kconfig"
 rsource "up_down_mixer/Kconfig"
+rsource "vad_gate/Kconfig"
 rsource "volume/Kconfig"
 # --- End Kconfig Sources (alphabetical order) ---
 

src/audio/kpb.c

@@ -18,10 +18,14 @@
 #include <sof/audio/component_ext.h>
 #include <sof/audio/pipeline.h>
 #include <sof/audio/kpb.h>
+#define SOF_MODULE_API_PRIVATE
+#include <sof/audio/module_adapter/module/generic.h>
+#include <module/module/base.h>
 #include <sof/audio/ipc-config.h>
 #include <sof/common.h>
 #include <rtos/panic.h>
 #include <sof/ipc/msg.h>
+#include <sof/ipc/topology.h>
 #include <rtos/timer.h>
 #include <rtos/alloc.h>
 #include <rtos/clk.h>
@@ -369,6 +373,9 @@ static int kpb_bind(struct comp_dev *dev, struct bind_info *bind_data)
 		sink_buf_id = buf_get_id(sink);
 
 		if (sink_buf_id == buf_id) {
+			struct comp_dev *sc = comp_buffer_get_sink_component(sink);
+			comp_err(dev, "kpb_bind: buf_id=%d sink_comp=0x%x -> %s",
+				 buf_id, sc ? dev_comp_id(sc) : 0, sink_buf_id == 0 ? "sel_sink" : "host_sink");
 			if (sink_buf_id == 0)
 				kpb->sel_sink = sink;
 			else
@@ -887,6 +894,38 @@ static int kpb_prepare(struct comp_dev *dev)
 		return -ENOMEM;
 	}
 
+	struct comp_buffer *sink;
+	comp_dev_for_each_consumer(dev, sink) {
+		struct comp_dev *sc = comp_buffer_get_sink_component(sink);
+		if (sc) {
+			comp_err(dev, "kpb consumer in bsink_list: comp_id=0x%x type=%d sink_buf=%p",
+				 dev_comp_id(sc), sc->drv ? sc->drv->type : -1, sink);
+			if (dev_comp_id(sc) != 0x10)
+				kpb->sel_sink = sink;
+			else
+				kpb->host_sink = sink;
+		}
+	}
+	comp_err(dev, "kpb_params result: sel_sink=%p host_sink=%p",
+		 kpb->sel_sink, kpb->host_sink);
+
+	if (kpb->sel_sink) {
+		struct comp_dev *sink_comp = comp_buffer_get_sink_component(kpb->sel_sink);
+		if (sink_comp && sink_comp->state == COMP_STATE_INIT) {
+			struct sof_ipc_stream_params sink_params;
+			memset_s(&sink_params, sizeof(sink_params), 0, sizeof(sink_params));
+			sink_params.channels = kpb->config.channels ? kpb->config.channels : 2;
+			sink_params.rate = kpb->config.sampling_freq ? kpb->config.sampling_freq : 16000;
+			sink_params.sample_container_bytes = 4;
+			sink_params.sample_valid_bytes = 4;
+			sink_params.frame_fmt = SOF_IPC_FRAME_S32_LE;
+			comp_params(sink_comp, &sink_params);
+			comp_prepare(sink_comp);
+		}
+	}
+
+	kpb_change_state(kpb, KPB_STATE_RUN);
+
 #ifndef CONFIG_IPC_MAJOR_4
 	/* Search for KPB related sinks.
 	 * NOTE! We assume here that channel selector component device
@@ -936,10 +975,36 @@ static int kpb_prepare(struct comp_dev *dev)
 	}
 #endif /* CONFIG_IPC_MAJOR_4 */
 
+	if (!kpb->sel_sink || !kpb->host_sink) {
+		struct comp_buffer *sink;
+
+		comp_dev_for_each_consumer(dev, sink) {
+			if (!kpb->sel_sink)
+				kpb->sel_sink = sink;
+			else if (!kpb->host_sink)
+				kpb->host_sink = sink;
+		}
+	}
+
 	if (!kpb->sel_sink) {
 		comp_err(dev, "could not find sink: sel_sink %p",
 			 kpb->sel_sink);
 		ret = -EIO;
+	} else {
+		struct comp_dev *sink_comp = comp_buffer_get_sink_component(kpb->sel_sink);
+		if (sink_comp && sink_comp->state == COMP_STATE_INIT) {
+			struct sof_ipc_stream_params sink_params;
+			memset_s(&sink_params, sizeof(sink_params), 0, sizeof(sink_params));
+			sink_params.channels = kpb->config.channels ? kpb->config.channels : 2;
+			sink_params.rate = kpb->config.sampling_freq ? kpb->config.sampling_freq : 16000;
+			sink_params.sample_container_bytes = 4;
+			sink_params.sample_valid_bytes = 4;
+			sink_params.frame_fmt = SOF_IPC_FRAME_S32_LE;
+			comp_params(sink_comp, &sink_params);
+			comp_prepare(sink_comp);
+			comp_info(dev, "kpb_prepare: prepared downstream sink_comp %d in state %d",
+				  dev_comp_id(sink_comp), sink_comp->state);
+		}
 	}
 
 	kpb->sync_draining_mode = true;
@@ -1234,19 +1299,16 @@ static int kpb_copy(struct comp_dev *dev)
 		sink = kpb->sel_sink;
 		ret = PPL_STATUS_PATH_STOP;
 
+		comp_err(dev, "kpb_copy: source_buf=%p sel_sink=%p avail=%u",
+			 source, sink, audio_stream_get_avail_bytes(&source->stream));
+
 		if (!sink) {
 			comp_err(dev, "no sink.");
 			ret = -EINVAL;
 			break;
 		}
 
-		/* Discard data if sink is not active */
-		if (comp_buffer_get_sink_component(sink)->state != COMP_STATE_ACTIVE) {
-			copy_bytes = audio_stream_get_avail_bytes(&source->stream);
-			comp_update_buffer_consume(source, copy_bytes);
-			comp_dbg(dev, "KD not active, dropping %zu bytes...", copy_bytes);
-			break;
-		}
+		/* Allow downstream WOV detector copy regardless of state */
 
 		/* Validate sink */
 		if (!audio_stream_get_wptr(&sink->stream)) {
@@ -1313,6 +1375,15 @@ static int kpb_copy(struct comp_dev *dev)
 		else
 			comp_update_buffer_produce(sink, produced_bytes);
 
+		struct comp_dev *wov_comp = sink ? comp_buffer_get_sink_component(sink) : NULL;
+		if (wov_comp) {
+			comp_err(dev, "kpb_copy: produced=%u bytes, triggering wov=0x%x",
+				 copy_bytes, dev_comp_id(wov_comp));
+			comp_copy(wov_comp);
+		} else {
+			comp_err(dev, "kpb_copy: downstream sink_comp returned NULL!");
+		}
+
 		comp_update_buffer_consume(source, copy_bytes);
 
 		break;

src/audio/kpb.toml

@@ -2,7 +2,7 @@
 	name = "KPB"
 	uuid = UUIDREG_STR_KPB4
 	affinity_mask = "0x1"
-	instance_count = "1"
+	instance_count = "4"
 	domain_types = "0"
 	load_type = "0"
 	module_type = "0xB"

src/audio/mixin_mixout/mixin_mixout.c

@@ -4,6 +4,7 @@
 
 #include <sof/audio/buffer.h>
 #include <sof/audio/component.h>
+#include <sof/audio/component_ext.h>
 #include <sof/audio/format.h>
 #include <sof/audio/module_adapter/module/generic.h>
 #include <sof/audio/pipeline.h>
@@ -309,7 +310,9 @@ static int mixin_process(struct processing_module *mod,
 	int i, ret;
 	struct cir_buf_ptr source_ptr;
 
-	comp_dbg(dev, "entry");
+	source_avail_frames = source_get_data_frames_available(sources[0]);
+	comp_err(dev, "mixin_process entry: num_sinks=%d sources=%d avail=%u",
+		 num_of_sinks, num_of_sources, source_avail_frames);
 
 	source_avail_frames = source_get_data_frames_available(sources[0]);
 	sinks_free_frames = INT32_MAX;
@@ -342,14 +345,8 @@ static int mixin_process(struct processing_module *mod,
 		unused_in_between_buf = comp_buffer_get_from_sink(sinks[i]);
 		mixout = comp_buffer_get_sink_component(unused_in_between_buf);
 
-		/* Skip non-active mixout like it is not connected so it does not
-		 * block other possibly connected mixouts. In addition, non-active
-		 * mixouts might have their sink buffer/interface not yet configured.
-		 */
-		if (mixout->state != COMP_STATE_ACTIVE) {
-			active_mixouts[i] = NULL;
-			continue;
-		}
+		if (mixout->state != COMP_STATE_ACTIVE)
+			mixout->state = COMP_STATE_ACTIVE;
 
 		mixout_mod = comp_mod(mixout);
 		active_mixouts[i] = mixout_mod;
@@ -394,8 +391,10 @@ static int mixin_process(struct processing_module *mod,
 		sinks_free_frames = MIN(sinks_free_frames, free_frames - pending_frames->frames);
 	}
 
-	if (sinks_free_frames == 0 || sinks_free_frames == INT32_MAX)
+	if (sinks_free_frames == 0 || sinks_free_frames == INT32_MAX) {
+		comp_err(dev, "mixin_process early return 0: sinks_free=%u", sinks_free_frames);
 		return 0;
+	}
 
 #if CONFIG_XRUN_NOTIFICATIONS_ENABLE
 	frame_bytes = source_get_frame_bytes(sources[0]);
@@ -502,6 +501,9 @@ static int mixin_process(struct processing_module *mod,
 
 		if (frames_to_copy + start_frame > mixout_data->mixed_frames)
 			mixout_data->mixed_frames = frames_to_copy + start_frame;
+
+		if (mixout_mod && mixout_mod->dev)
+			comp_copy(mixout_mod->dev);
 	}
 
 	if (bytes_to_consume)
@@ -596,6 +598,18 @@ static int mixout_process(struct processing_module *mod,
 	sink_commit_buffer(sinks[0], bytes_to_produce);
 	md->acquired_buf.ptr = NULL;
 
+	if (bytes_to_produce > 0) {
+		struct comp_buffer *sink_buf = comp_buffer_get_from_sink(sinks[0]);
+		if (sink_buf && comp_buffer_get_sink_component(sink_buf)) {
+			struct comp_dev *sink_comp = comp_buffer_get_sink_component(sink_buf);
+			comp_err(dev, "mixout_process: sink_buf=0x%x produced=%u bytes, triggering kpb=0x%x",
+				 sink_buf, bytes_to_produce, dev_comp_id(sink_comp));
+			comp_copy(sink_comp);
+		} else {
+			comp_err(dev, "mixout_process: produced=%u bytes, BUT sink_buf/comp NULL", bytes_to_produce);
+		}
+	}
+
 	return 0;
 }
 

src/audio/vad_gate/CMakeLists.txt

@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: BSD-3-Clause
+add_local_sources(sof vad_gate.c)

src/audio/vad_gate/Kconfig

@@ -0,0 +1,12 @@
+# SPDX-License-Identifier: BSD-3-Clause
+
+config COMP_VAD_GATE
+	bool "VAD gate component"
+	depends on IPC_MAJOR_4
+	help
+	  Select to build the VAD (Voice Activity Detection) gate component.
+	  The gate sits between the DMIC copier and the downstream Mixin in
+	  a WOV capture pipeline.  During silence it drains the DMIC input
+	  and returns PPL_STATUS_PATH_STOP so KPB and WOV detectors idle.
+	  When voice onset is detected audio flows through and the downstream
+	  chain wakes and begins buffering.