Hi Rocky,
On Fri, Dec 27, 2019 at 03:21:30PM +0800, Rocky Liao wrote:
> This patch adds the HCI command timeout handling, it will trigger btsoc
> to report its memory dump via vendor specific events when hit the defined
> max HCI command timeout count. After all the memory dump VSE are sent, the
> btsoc will also send a HCI_HW_ERROR event to host and this will cause a new
> hci down/up process and the btsoc will be re-initialized.
>
> Signed-off-by: Rocky Liao <rjliao@xxxxxxxxxxxxxx>
> ---
>
> Changes in v2:
> - Fix build error
> Changes in v3:
> - Remove the function declaration
> - Move the cmd_timeout() callback register to probe()
> - Remove the redundant empty line
>
> drivers/bluetooth/hci_qca.c | 45 +++++++++++++++++++++++++++++++++++++
> 1 file changed, 45 insertions(+)
>
> diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
> index ca0b38f065e5..026e2e2cdd30 100644
> --- a/drivers/bluetooth/hci_qca.c
> +++ b/drivers/bluetooth/hci_qca.c
> @@ -47,6 +47,8 @@
> #define IBS_HOST_TX_IDLE_TIMEOUT_MS 2000
> #define CMD_TRANS_TIMEOUT_MS 100
>
> +#define QCA_BTSOC_DUMP_CMD 0xFB
> +
> /* susclk rate */
> #define SUSCLK_RATE_32KHZ 32768
>
> @@ -56,6 +58,9 @@
> /* max retry count when init fails */
> #define QCA_MAX_INIT_RETRY_COUNT 3
>
> +/* when hit the max cmd time out count, trigger btsoc dump */
> +#define QCA_MAX_CMD_TIMEOUT_COUNT 3
nit: MAX_CMD_TIMEOUTS?
Similar to QCA_MAX_INIT_RETRY_COUNT on which I commented earlier I don't
think the 'QCA' prefix adds value here. The constant is defined in the driver
itself and isn't related to hardware.
> +
> enum qca_flags {
> QCA_IBS_ENABLED,
> QCA_DROP_VENDOR_EVENT,
> @@ -123,6 +128,8 @@ struct qca_data {
> u64 rx_votes_off;
> u64 votes_on;
> u64 votes_off;
> +
> + u32 cmd_timeout_cnt;
nit: cmd_timeouts?
> };
>
> enum qca_speed_type {
> @@ -1332,6 +1339,11 @@ static int qca_setup(struct hci_uart *hu)
> if (!ret) {
> set_bit(QCA_IBS_ENABLED, &qca->flags);
> qca_debugfs_init(hdev);
> +
> + /* clear the command time out count every time after
> + * initializaiton done
> + */
> + qca->cmd_timeout_cnt = 0;
> } else if (ret == -ENOENT) {
> /* No patch/nvm-config found, run with original fw/config */
> ret = 0;
> @@ -1462,6 +1474,38 @@ static int qca_power_off(struct hci_dev *hdev)
> return 0;
> }
>
> +static int qca_send_btsoc_dump_cmd(struct hci_uart *hu)
> +{
> + int err = 0;
The variable is pointless, just return 0 at the end of the function.
> + struct sk_buff *skb = NULL;
> + struct qca_data *qca = hu->priv;
> +
> + BT_DBG("hu %p sending btsoc dump command", hu);
> +
> + skb = bt_skb_alloc(1, GFP_ATOMIC);
> + if (!skb) {
> + BT_ERR("Failed to allocate memory for qca dump command");
"These generic allocation functions all emit a stack dump on failure when used
without __GFP_NOWARN so there is no use in emitting an additional failure
message when NULL is returned."
Documentation/process/coding-style.rst
hence the logging is redundant, drop it.
> + return -ENOMEM;
> + }
> +
> + skb_put_u8(skb, QCA_BTSOC_DUMP_CMD);
> +
> + skb_queue_tail(&qca->txq, skb);
> +
> + return err;
> +}
> +
> +static void qca_cmd_timeout(struct hci_dev *hdev)
> +{
> + struct hci_uart *hu = hci_get_drvdata(hdev);
> + struct qca_data *qca = hu->priv;
> +
> + BT_ERR("hu %p hci cmd timeout count=0x%x", hu, ++qca->cmd_timeout_cnt);
Is there any particular reason to print the counter in hex instead of
decimal?
Should this use bt_dev_err() since we have a hdev in this context?
> +
> + if (qca->cmd_timeout_cnt >= QCA_MAX_CMD_TIMEOUT_COUNT)
> + qca_send_btsoc_dump_cmd(hu);
> +}
> +
> static int qca_regulator_enable(struct qca_serdev *qcadev)
> {
> struct qca_power *power = qcadev->bt_power;
> @@ -1605,6 +1649,7 @@ static int qca_serdev_probe(struct serdev_device *serdev)
> hdev = qcadev->serdev_hu.hdev;
> set_bit(HCI_QUIRK_NON_PERSISTENT_SETUP, &hdev->quirks);
> hdev->shutdown = qca_power_off;
> + hdev->cmd_timeout = qca_cmd_timeout;
> }
>
> out: return err;
> --
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
