Hi Johan,
>> +static int uuid_cmp(const void *a, const void *b)
>> +{
>> + const sdp_record_t *rec = a;
>> + const uuid_t *uuid = b;
>> +
>> + return sdp_uuid_cmp(&rec->svclass, uuid);
>> +}
>> +
>> +static void uuid_to_uuid128(uuid_t *uuid128, const uuid_t *uuid)
>> +{
>> + if (uuid->type == SDP_UUID16)
>> + sdp_uuid16_to_uuid128(uuid128, uuid);
>> + else if (uuid->type == SDP_UUID32)
>> + sdp_uuid32_to_uuid128(uuid128, uuid);
>> + else
>> + memcpy(uuid128, uuid, sizeof(*uuid));
>> +}
>> +
>> +static uint8_t get_uuid_mask(const uuid_t *uuid)
>> +{
>> + if (uuid->type != SDP_UUID16)
>> + return 0;
>> +
>> + switch (uuid->value.uuid16) {
>> + case DIALUP_NET_SVCLASS_ID:
>> + case CIP_SVCLASS_ID:
>> + return 0x42; /* Telephony & Networking */
>> + case IRMC_SYNC_SVCLASS_ID:
>> + case OBEX_OBJPUSH_SVCLASS_ID:
>> + case OBEX_FILETRANS_SVCLASS_ID:
>> + case IRMC_SYNC_CMD_SVCLASS_ID:
>> + case PBAP_PSE_SVCLASS_ID:
>> + return 0x10; /* Object Transfer */
>> + case HEADSET_SVCLASS_ID:
>> + case HANDSFREE_SVCLASS_ID:
>> + return 0x20; /* Audio */
>> + case CORDLESS_TELEPHONY_SVCLASS_ID:
>> + case INTERCOM_SVCLASS_ID:
>> + case FAX_SVCLASS_ID:
>> + case SAP_SVCLASS_ID:
>> + /*
>> + * Setting the telephony bit for the handsfree audio gateway
>> + * role is not required by the HFP specification, but the
>> + * Nokia 616 carkit is just plain broken! It will refuse
>> + * pairing without this bit set.
>> + */
>> + case HANDSFREE_AGW_SVCLASS_ID:
>> + return 0x40; /* Telephony */
>> + case AUDIO_SOURCE_SVCLASS_ID:
>> + case VIDEO_SOURCE_SVCLASS_ID:
>> + return 0x08; /* Capturing */
>> + case AUDIO_SINK_SVCLASS_ID:
>> + case VIDEO_SINK_SVCLASS_ID:
>> + return 0x04; /* Rendering */
>> + case PANU_SVCLASS_ID:
>> + case NAP_SVCLASS_ID:
>> + case GN_SVCLASS_ID:
>> + return 0x02; /* Networking */
>> + default:
>> + return 0;
>> + }
>> +}
>
> I don't think it makes sense to duplicate this code in bluez.git,
> especially not the last function. Instead, I think it should be
> refactored to the appropriate common place (lib/sdp.c maybe? Or maybe
> not if we don't want to pollute the already deprecated libbluetooth - in
> which case I'd expect a proposal where to put this).
I was actually thinking that we just make this static. In Android all SDP records are normally registered ahead of time. They even did this when they used BlueZ. It was essentially a sdptool add for all the records they needed.
This is also means we can just use the service class hint as a static value.
>> +static int add_uuid(uuid_t *uuid, uint8_t svc_hint)
>> +{
>> + struct mgmt_cp_add_uuid cp;
>> + uuid_t uuid128;
>> + uint128_t uint128;
>> +
>> + uuid_to_uuid128(&uuid128, uuid);
>> +
>> + ntoh128((uint128_t *) uuid128.value.uuid128.data, &uint128);
>> + htob128(&uint128, (uint128_t *) cp.uuid);
>> + cp.svc_hint = svc_hint;
>> +
>> + if (mgmt_send(adapter->mgmt, MGMT_OP_ADD_UUID,
>> + 0, sizeof(cp), &cp,
>> + add_uuid_complete, adapter, NULL) > 0)
>> + return 0;
>
> Since you're skipping the "is_supported_uuid" check that we're doing in
> the normal bluetoothd I suppose we at least have a proper check for a
> minimum mgmt version when initializing the android bluetoothd? Otherwise
> your patch is broken.
We should require a minimum mgmt version for Android. One that has all the bugs fixed. There is no point in planning on supporting older kernels anyway.
Regards
Marcel
--
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
