Re: [PATCH] mgmtops: Fix adding UUID when EBUSY error is returned

Hi Szymon,

On Sun, Jun 10, 2012, Szymon Janc wrote:
> +static void mgmt_add_uuid_resend(int sk, uint16_t index)
> +{
> +	struct controller_info *info;
> +	struct pending_uuid *pending;
> +
> +	DBG("index %d", index);
> +
> +	info = &controllers[index];
> +
> +	pending = info->pending_uuids->data;
> +
> +	send_uuid(index, &pending->uuid, pending->svc_hint);
> +}
> +
>  static void mgmt_cmd_status(int sk, uint16_t index, void *buf, size_t len)
>  {
>  	struct mgmt_ev_cmd_status *ev = buf;
>  <at>  <at>  -1460,6 +1479,10  <at>  <at>  static void mgmt_cmd_status(int sk, uint16_t index, void *buf, size_t len)
>  	case MGMT_OP_READ_LOCAL_OOB_DATA:
>  		read_local_oob_data_failed(sk, index);
>  		break;
> +	case MGMT_OP_ADD_UUID:
> +		if (ev->status == MGMT_STATUS_BUSY)
> +			mgmt_add_uuid_resend(sk, index);
> +		break;
>  	}
>  }

It's there a risk that this ends up in an infinite loop if the kernel is
in some state where it always keeps sending STATUS_BUSY? Would it make
sense to add a counter for retries and give up after 5 or 10 attempts?
Would it also make sense to add a small delay before resending the
command?

Johan