nrf_rpc: remove sending from rx thread

Commit removes any data sending over serial transport from
rx thread and adds sending over allocated pool of threads
for incoming data execution.
The reason of this: if sending requires some acknowledgements
then deadlock of rx thread happens.

Signed-off-by: Aleksandr Khromykh <[email protected]>

Author: alxelax

nrf_rpc/nrf_rpc.c

@@ -85,6 +85,19 @@ struct init_packet_data {
 	const char *strid;   /* Remote group unique identifier. */
 };
 
+enum internal_pool_data_type {
+	NRF_RPC_INITIALIZATION,
+	NRF_RPC_ERROR
+};
+
+struct internal_pool_data {
+	enum internal_pool_data_type type;
+	const struct nrf_rpc_group *group;
+	int err;
+	uint8_t hdr_id;
+	uint8_t hdr_type;
+};
+
 /* Pool of statically allocated command contexts. */
 static struct nrf_rpc_cmd_ctx cmd_ctx_pool[CONFIG_NRF_RPC_CMD_CTX_POOL_SIZE];
 
@@ -102,6 +115,8 @@ static bool is_initialized;
 /* Error handler provided to the init function. */
 static nrf_rpc_err_handler_t global_err_handler;
 
+static struct internal_pool_data internal_data;
+
 /* Array with all defiend groups */
 NRF_RPC_AUTO_ARR(nrf_rpc_groups_array, "grp");
 
@@ -343,6 +358,24 @@ static inline bool packet_validate(const uint8_t *packet)
 	       (addr < (uintptr_t)0 - (uintptr_t)NRF_RPC_HEADER_SIZE);
 }
 
+static void internal_tx_handler(void)
+{
+	struct internal_pool_data copy = internal_data;
+
+	nrf_rpc_os_event_set(&copy.group->data->decode_done_event);
+
+	if (copy.type == NRF_RPC_INITIALIZATION) {
+		if (group_init_send(copy.group)) {
+			NRF_RPC_ERR("Failed to send group init packet for group id: %d strid: %s",
+				    copy.group->data->src_group_id, copy.group->strid);
+		}
+	}
+
+	if (copy.type == NRF_RPC_ERROR) {
+		nrf_rpc_err(copy.err, NRF_RPC_ERR_SRC_RECV, copy.group, copy.hdr_id, copy.hdr_type);
+	}
+}
+
 static int transport_init(nrf_rpc_tr_receive_handler_t receive_cb)
 {
 	int err = 0;
@@ -513,7 +546,11 @@ static uint8_t parse_incoming_packet(struct nrf_rpc_cmd_ctx *cmd_ctx,
 /* Thread pool callback */
 static void execute_packet(const uint8_t *packet, size_t len)
 {
-	parse_incoming_packet(NULL, packet, len);
+	if (packet == (const uint8_t *)&internal_data) {
+		internal_tx_handler();
+	} else {
+		parse_incoming_packet(NULL, packet, len);
+	}
 }
 
 static bool protocol_version_check(const struct init_packet_data *init_data)
@@ -636,14 +673,13 @@ static int init_packet_handle(struct header *hdr, const struct nrf_rpc_group **g
 		 * If remote processor does not know our group id, send an init packet back,
 		 * since it might have missed our original init packet.
 		 */
-		err = group_init_send(*group);
-		if (err) {
-			NRF_RPC_ERR("Failed to send group init packet for group id: %d strid: %s",
-				    group_data->src_group_id, (**group).strid);
-		}
+		internal_data.type = NRF_RPC_INITIALIZATION;
+		internal_data.group = *group;
+		nrf_rpc_os_thread_pool_send((const uint8_t *)&internal_data, sizeof(internal_data));
+		nrf_rpc_os_event_wait(&(*group)->data->decode_done_event, NRF_RPC_OS_WAIT_FOREVER);
 	}
 
-	return err;
+	return 0;
 }
 
 /* Callback from transport layer that handles incoming. */
@@ -765,8 +801,13 @@ static void receive_handler(const struct nrf_rpc_tr *transport, const uint8_t *p
 	}
 
 	if (err < 0) {
-		nrf_rpc_err(err, NRF_RPC_ERR_SRC_RECV, group, hdr.id,
-			    hdr.type);
+		internal_data.type = NRF_RPC_ERROR;
+		internal_data.group = group;
+		internal_data.err = err;
+		internal_data.hdr_id = hdr.id;
+		internal_data.hdr_type = hdr.type;
+		nrf_rpc_os_thread_pool_send((const uint8_t *)&internal_data, sizeof(internal_data));
+		nrf_rpc_os_event_wait(&group->data->decode_done_event, NRF_RPC_OS_WAIT_FOREVER);
 	}
 }