Hi Johan
On Tue, Apr 29, 2014 at 2:56 PM, Johan Hedberg <johan.hedberg@xxxxxxxxx> wrote:
> Hi Lukasz,
>
> On Tue, Apr 29, 2014, Lukasz Rymanowski wrote:
>> On Tue, Apr 29, 2014 at 10:13 AM, Johan Hedberg <johan.hedberg@xxxxxxxxx> wrote:
>> > On Tue, Apr 29, 2014, Lukasz Rymanowski wrote:
>> >> +static uint8_t read_request(const uint8_t *cmd, uint16_t cmd_len,
>> >> + struct gatt_device *dev)
>> >> +{
>> >> + uint16_t handle;
>> >> + uint16_t len;
>> >> + uint16_t offset = 0;
>> >> + struct req_data *req_data;
>> >> +
>> >> + DBG("");
>> >> +
>> >> + req_data = new0(struct req_data, 1);
>> >> + if (!req_data)
>> >> + return ATT_ECODE_UNLIKELY;
>> >> +
>> >> + switch (cmd[0]) {
>> >> + case ATT_OP_READ_BLOB_REQ:
>> >> + len = dec_read_blob_req(cmd, cmd_len, &handle, &offset);
>> >> + if (!len) {
>> >> + free(req_data);
>> >> + return ATT_ECODE_INVALID_PDU;
>> >> + }
>> >> + req_data->long_read = true;
>> >> + break;
>> >> + case ATT_OP_READ_REQ:
>> >> + len = dec_read_req(cmd, cmd_len, &handle);
>> >> + if (!len) {
>> >> + free(req_data);
>> >> + return ATT_ECODE_INVALID_PDU;
>> >> + }
>> >> + break;
>> >> + default:
>> >> + error("gatt: Unexpected read type 0x%02x", cmd[0]);
>> >> + free(req_data);
>> >> + return ATT_ECODE_REQ_NOT_SUPP;
>> >> + }
>> >> +
>> >> + req_data->dev = dev;
>> >> + req_data->opcode = cmd[0];
>> >> +
>> >> + if (!gatt_db_read(gatt_db, handle, offset, req_data)) {
>> >> + free(req_data);
>> >> + return ATT_ECODE_UNLIKELY;
>> >> + }
>> >> +
>> >> + return 0;
>> >> +}
>> >
>> > Where is req_data freed in the case that gatt_db_read succeeds?
>>
>> It is freed later in the callback. Will put comment about that.
>
> Which callback is that? E.g. in 13/36 the gap_read_cb function does not
> free req_data.
Your previous comment realized me that we are missing that in read
callback for gatt service, device
information service and service changed. it is now fixed.
>
> In general this kind of API design seems quite strange to me. Probably
> the information contained in struct req_data can be made fixed for the
> gatt-db API instead of something chosen by the user (in this case
> android/gatt.c), and probably doesn't have to be dynamically allocated
> to begin with.
well I agree. I will do something like request_complete_cb which should be
called in the end of request and there we do free.
Also for now this callback will take aswell response pdu and length,
so it will be one place
where msg is send back to remote device.
it will be in v2
>
> Johan
--
To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
