"H_SDIO_DRV: invalid ret or len_from_slave: 0 0" with coexistence usage (EHM-123)

[nkp123]

Checklist

• Checked the issue tracker for similar issues to ensure this is not a duplicate.
• Provided a clear description of your suggestion.
• Included any relevant context or examples.

Issue or Suggestion Description

Hi,

I am trying to connect ESP32-P4 with ESP32 for simultaneous A2DP streaming and WiFI use.
I am encountering an issue where after a few seconds to a few minutes of streaming the communication between host and coprocessor locks up with a warning printed on host:

H_SDIO_DRV: invalid ret or len_from_slave: 0 0

With default loglevel there is really not much more worth of noting.

It looks like this issue is happening only in a coexistence scenario - with only WiFi or only Bluetooth enabled I cannot reproduce this issue.

My setup:

• ESP-IDF 5.5.1
• Espressif-IDE 3.6.0
• Ubuntu 25.10

Libraries:

• ESP-Hosted: 2.6.4
• ESP-Wifi-Remote: 1.2.0

Host is ESP32P4NRW32, co-processor is ESP32-WROOM-32E-N16R2 module (however I got the same issue on ESP32-WROOM-32E-N4 base model).
Custom PCB, SDIO transport.

I managed to create minimal example:

Host:

• Get A2DP ESP32 example from:

https://github.com/espressif/esp-idf/tree/master/examples/bluetooth/bluedroid/classic_bt/a2dp_sink

• Add esp-hosted and esp-wifi-remote dependencies
• Configure co-processor to ESP32
• Configure SDIO transport and its GPIO pins, enable bluedroid support
• Configure Wifi Remote to reduce memory footprint as defaults doesn't play nice with BT+Wifi scenario:

CONFIG_WIFI_RMT_STATIC_RX_BUFFER_NUM=2
CONFIG_WIFI_RMT_DYNAMIC_RX_BUFFER_NUM=2
# CONFIG_WIFI_RMT_STATIC_TX_BUFFER is not set
CONFIG_WIFI_RMT_DYNAMIC_TX_BUFFER=y
CONFIG_WIFI_RMT_TX_BUFFER_TYPE=1
CONFIG_WIFI_RMT_DYNAMIC_TX_BUFFER_NUM=2
CONFIG_WIFI_RMT_STATIC_RX_MGMT_BUFFER=y
# CONFIG_WIFI_RMT_DYNAMIC_RX_MGMT_BUFFER is not set
CONFIG_WIFI_RMT_DYNAMIC_RX_MGMT_BUF=0
CONFIG_WIFI_RMT_RX_MGMT_BUF_NUM_DEF=5
# CONFIG_WIFI_RMT_CSI_ENABLED is not set
CONFIG_WIFI_RMT_AMPDU_TX_ENABLED=y
CONFIG_WIFI_RMT_TX_BA_WIN=2
CONFIG_WIFI_RMT_AMPDU_RX_ENABLED=y
CONFIG_WIFI_RMT_RX_BA_WIN=2

• In main.cpp replace controller initialization routine with ESP-Hosted init, remote BT Controller init and remote Wifi init routines. I came up with:

	// ------ ESP HOSTED
	esp_hosted_connect_to_slave();
	esp_hosted_init();
	// ------ ESP HOSTED

	// ------ BLUETOOTH
	/* initialize TRANSPORT first */
	if ((err = esp_hosted_bt_controller_init() != ESP_OK)) {
		ESP_LOGE(BT_AV_TAG, "%s esp_hosted_bt_controller_init() failed: %s", __func__, esp_err_to_name(err));
		return;
	}
	// enable bt controller
	if ((err = esp_hosted_bt_controller_enable() != ESP_OK)) {
		ESP_LOGE(BT_AV_TAG, "%s esp_hosted_bt_controller_enable() failed: %s", __func__, esp_err_to_name(err));
		return;
	}
	hosted_hci_bluedroid_open();
	/* get HCI driver operations */
	esp_bluedroid_hci_driver_operations_t operations = {
		.send = hosted_hci_bluedroid_send,
		.check_send_available = hosted_hci_bluedroid_check_send_available,
		.register_host_callback = hosted_hci_bluedroid_register_host_callback,
	};
	if ((err = esp_bluedroid_attach_hci_driver(&operations)) != ESP_OK)
	{
		ESP_LOGE(BT_AV_TAG, "%s esp_bluedroid_attach_hci_driver failed: %s", __func__, esp_err_to_name(err));
		return;
	}
	// ------ BLUETOOTH

	// ------ SIMPLE WIFI TEST
	esp_netif_init();
	esp_event_loop_create_default();
	esp_netif_create_default_wifi_sta();
	wifi_init_config_t wifi_initiation = WIFI_INIT_CONFIG_DEFAULT();
	esp_wifi_init(&wifi_initiation);
	wifi_config_t wifi_configuration = {
		.sta = {
		    .ssid = "SSID",
		    .password= "PASS"
		  }
	};
	
	esp_wifi_set_config(ESP_IF_WIFI_STA, &wifi_configuration);
	esp_wifi_start();
	esp_wifi_set_mode(WIFI_MODE_STA);
	do
	{
		err = esp_wifi_connect();
		vTaskDelay(100);
	} while (err != ESP_OK);
	// ------ SIMPLE WIFI TEST

Co-processor:

• Clone and compile firmware from:

idf.py create-project-from-example "espressif/esp_hosted=2.6.4:slave"

• The only change is to change SDIO TX/RX queue length to 10 in menuconfig to conserve DRAM.

Test procedure:

• Compile and flash firmwares
• Flood P4 with ping: sudo ping IP -i 0.001
• Play audio via BT. For testing I was using Android phone
• After few seconds to few minutes BT and Wifi should cease to work, and on host there should be a warning message: H_SDIO_DRV: invalid ret or len_from_slave: 0 0
  Flooding with ping is optional, but I noticed that increased activity makes the program lock up faster.

I am attaching:

• modified main.c from a2dp example
• sdkconfig from host and coprocessor
• execution logs from host and coprocessor

coprocessor.log
host.log
main.c
sdkconfig_coprocessor.txt
sdkconfig_host.txt

Any help would be appreciated :)

Best regards,
Norbert

[mantriyogesh]

Can you please confirm if :

1. isolated transport test in tx and rx independently tested is very important.
2. If SDIO is being used, would highly suggest to take camera picture and send us how they are connected.
3. Hardware requirements are stated (here). Are the Hardware requirement fulfilled? It includes but not limited to minimal and equal trace length, correct pull-up registers on expected SDIO lines.
4. As you are using different GPIOs, can you please confirm if all the GPIOs involved for SDIO operation are having 3v3 voltage.

I am also yet to understand the system and features used and what part works independently or fails to perform anything.
I note the wifi buffers are extremely small. you can use esp32-c2 wifi buffer for your case from https://github.com/espressif/esp-hosted-mcu/blob/main/docs/performance_optimization.md#14-esp32-c2-as-co-processor

You can optionally use these same numbers at sdkconfig.defaults.esp32 at slave (however, variable names would change and not contain RMT in their names). If unsure or not understandable this part, you can also ignore. anyway host config overrides slave config.

at host sdconfig.default.esp32p4. c2 buffers are smaller and you can also use even if esp32 is slave. this would reduce peak throughput, but right now I am not much bothered about peak speed and thinking of stability for your use case of Wi-Fi+Classic BT with A2DP.

Please note, ESP has single RF and always shares the RF in between BT and Wi-Fi. So when BT is active, it is expected that Wi-Fi performance would lower. First things first. Is Wi-Fi working fine, if you do not any Bluetooth 'active' operations?

Is Bluetooth A2DP working independently?

If both independently working fine, then we will move to co-ex case.

It would be highly recommended that at slave, the Wi-Fi and Bluetooth are pinned to different cores.
I see in host config, both are pinned to core 0.

Slave side, I can see Wi-Fi is pinned to core 1 in config, her:

CONFIG_ESP_WIFI_TASK_PINNED_TO_CORE_1=y

however, It would be overriden due to host side config,

CONFIG_WIFI_RMT_TASK_PINNED_TO_CORE_0=y

better to ensure we have same understanding at host and slave. i leave to you to decide what goes on which core, but try to be different core.

The default Classic BT + Wi-Fi also involves different things in place. but i will wait for above discussion first to be grasped and confirmed, before moving ahead.