On 2/24/26 10:16 AM, Konrad Dybcio wrote:

On 2/23/26 7:07 PM, Srinivas Kandagatla wrote:

quoted

Buffers are allocated on pcm_new and mapped in the dsp on every prepare call,
which is inefficient and unnecessary.

Add new functions q6apm_[un]map_memory_fixed_region to map it on
to dsp only once after allocation.

Signed-off-by: Srinivas Kandagatla <redacted>
---

[...]

quoted

-int q6apm_map_memory_regions(struct q6apm_graph *graph, unsigned int dir, phys_addr_t phys,
-			     size_t period_sz, unsigned int periods)
+int q6apm_set_memory_map_handle(struct q6apm_graph *graph, unsigned int dir)
+{
+	if (dir == SNDRV_PCM_STREAM_PLAYBACK)
+		graph->rx_data.mem_map_handle = graph->info->mem_map_handle;
+	else
+		graph->tx_data.mem_map_handle = graph->info->mem_map_handle;
+
+	return 0;
+}

This doesn't return a useful value today, will it ever?

return type should be  void.

[...]

quoted

+		case APM_CMD_SHARED_MEM_UNMAP_REGIONS:
+			apm->result.opcode = hdr->opcode;
+			apm->result.status = 0;
+			rsp = data->payload;
+
+			info = idr_find(&apm->graph_info_idr, hdr->token);
+			if (info)
+				info->mem_map_handle = 0;
+			else
+				dev_err(dev, "Error (%d) Processing 0x%08x cmd\n", result->status,
+					result->opcode);
+
+			dev_err(dev, "Memory Map region %x\n", rsp->mem_map_handle);

Looks like a debug leftover (also in the below 'case')

Right, will fix that in next spin.

--srini

Konrad