BLE failures with EIO

[sheindel-llt]

I'm developing a BLE application using several connections, and of course I'm advertising and scanning to establish those connections. I have up to 5 connections at a time and running both peripheral and central roles.

Occasionally I get an error when trying to do a BLE operation and it's not clear to me what the error means for how I should respond. In particular I sometimes get -EIO when calling bt_le_ext_adv_start() or bt_conn_disconnect(). A cursory look through the code in these functions doesn't show an obvious reason why EIO occurs, except that it may be caused by an error when sending an HCI command to the controller in bt_hci_cmd_send_sync(). Looking at the logs, I do sometimes see a warning like "bt_hci_core: opcode 0x200e status 0x0c" which is indeed generated in bt_hci_cmd_send_sync(). Since the EIO error only occurs occasionally I suspect it may be occurring due to a timing window where the host and controller are out of sync, or some other transient condition. In some cases retrying the operation later works eventually, sometimes not (advertising seems to often fail many times in a row and then succeed; disconnecting failures seem to never succeed with retries).

At the end of the day I just want my application to robustly handle error conditions like this. But, with the error code EIO I'm not sure how to respond. If it's dependent on the specific opcode/status code, the application doesn't have access to this information (only EIO). Can you give me any insight into what I should do in situations like this? Is there any way for the application to know more specific error codes so it can respond appropriately?

Details: I am using NCS 2.4.1 on the nRF5340 with the net core running the soft device controller. I think these are the relevant parts of my project configuration.

App core:

CONFIG_BT=y
CONFIG_BT_PERIPHERAL=y
CONFIG_BT_CENTRAL=y

CONFIG_BT_MAX_CONN=5

CONFIG_BT_EXT_ADV=y
CONFIG_BT_EXT_ADV_MAX_ADV_SET=2

CONFIG_BT_GATT_DM=y

CONFIG_BT_BUF_ACL_TX_SIZE=251
CONFIG_BT_BUF_ACL_TX_COUNT=20
CONFIG_BT_BUF_ACL_RX_SIZE=251

CONFIG_BT_FILTER_ACCEPT_LIST=y
CONFIG_BT_USER_PHY_UPDATE=y
CONFIG_BT_USER_DATA_LEN_UPDATE=y
CONFIG_BT_SMP=y
CONFIG_BT_DEVICE_NAME="ABC"

CONFIG_BT_ATT_PREPARE_COUNT=2
CONFIG_BT_GATT_CLIENT=y
CONFIG_BT_GAP_AUTO_UPDATE_CONN_PARAMS=n

CONFIG_BT_L2CAP_TX_MTU=247
CONFIG_BT_L2CAP_DYNAMIC_CHANNEL=y

CONFIG_BT_SCAN=y
CONFIG_BT_SCAN_FILTER_ENABLE=y
CONFIG_BT_SCAN_UUID_CNT=1

Net core:

CONFIG_BT_MAX_CONN=5

CONFIG_BT_BUF_ACL_TX_SIZE=251
CONFIG_BT_BUF_ACL_TX_COUNT=20
CONFIG_BT_BUF_ACL_RX_SIZE=251

CONFIG_BT_CTLR=y
CONFIG_BT_CTLR_ADV_DATA_LEN_MAX=255
CONFIG_BT_CTLR_ADV_EXT=y
CONFIG_BT_CTLR_ADV_SET=2
CONFIG_BT_CTLR_DATA_LENGTH_MAX=251
CONFIG_BT_CTLR_PHY_2M=y
CONFIG_BT_CTLR_PHY_CODED=y
CONFIG_BT_CTLR_RX_BUFFERS=10
CONFIG_BT_CTLR_TX_PWR_DYNAMIC_CONTROL=y

CONFIG_BT_CTLR_SDC_PERIPHERAL_COUNT=1

[MMater]

Was there any resolution to this?

I am running into a similar issue were I get a "BT_HCI_ERR_CMD_DISALLOWED" error, which propagates as a -EIO error when I attempt to start scanning.

Once I get the error the stack appears to freeze, I cannot scan or advertise. We are using a nRF5340.

[Hieu]

Hi MMater,

What NCS version are you using? If you are using a version earlier than 2.5.0, then there is a known issue. You can find more details in this DevZone case:

 nRF5340 - bt_hci_cmd_send_sync hangs with error -11  

Hieu

[sheindel-llt]

I think I resolved this by turning off ACL flow control with `CONFIG_BT_HCI_ACL_FLOW_CONTROL=n`. I believe there is an issue with the flow control that was worked around in 2.5.0, but if you are on 2.4.2 (like me) then the only fix was to disable flow control. I didn't see any negative side effects from turning it off either.

[MMater]

I attempted to disable flow control with no success.

We are on 2.5.2 of the SDK

[sheindel-llt]

Hmm, not sure then. Only other thing I can recall that might be relevant is that EIO can come up if you 'overallocate' the radio. If you ask for more than 100% of the time to be spent scanning, for example asking for 80% duty cycle scan on 1M PHY and 50% duty cycle scan on coded PHY. That might be a different root error code though and it shouldn't show up intermittently.