search

shangdawei / btstack

Automatically exported from code.google.com/p/btstack
0 stars 1 forks source link

SDP query fails after dedicated bonding completes #405

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
What steps will reproduce the problem?
1. Call gap_dedicated_bonding for a device.
2. On a GAP_DEDICATED_BONDING_COMPLETED event, call 
sdp_query_rfcomm_channel_and_name_for_uuid.
3. An SDP_QUERY_COMPLETE is immediately received, with status 0x81 
(SDP_QUERY_INCOMPLETE).

What version of the product are you using? On what operating system?
Present in the latest r2654.

Relevant part of the log:
[2014-06-18 14:43:25.598] LOG -- hci_emit_dedicated_bonding_result 0
[2014-06-18 14:43:25.598] EVT <= E1 07 00 71 1F D4 DB C6 BC
[2014-06-18 14:43:25.601] CMD => 06 04 03 02 00 13
[2014-06-18 14:43:25.602] LOG -- L2CAP_CREATE_CHANNEL_MTU addr 
BC:C6:DB:D4:1F:71 psm 0x1 mtu 668
[2014-06-18 14:43:25.603] LOG -- l2cap_create_channel_internal, hci connection 
already exists
[2014-06-18 14:43:25.603] LOG -- l2cap_handle_connection_complete expected state
[2014-06-18 14:43:25.603] LOG -- l2cap received remote supported features, 
sec_level_0_allowed for psm 1 = 1
[2014-06-18 14:43:25.604] CMD => 34 F4 08 71 1F D4 DB C6 BC 2F 11
[2014-06-18 14:43:25.606] EVT <= 0F 04 00 01 06 04
[2014-06-18 14:43:25.607] ACL => 02 20 0C 00 08 00 01 00 02 01 04 00 01 00 40 00
[2014-06-18 14:43:25.609] LOG -- l2cap_start_rtx for local cid 0x40
[2014-06-18 14:43:25.609] LOG -- l2cap_stop_rtx for local cid 0x40
[2014-06-18 14:43:25.632] EVT <= 13 05 01 02 00 01 00
[2014-06-18 14:43:25.711] EVT <= 05 04 00 02 00 16
[2014-06-18 14:43:25.712] LOG -- Connection closed: handle 2, BC:C6:DB:D4:1F:71

I believe the problem is caused by the sdp query trying to reuse the dedicated 
bonding's hci connection. When the dedicated bonding is done, an hci_disconnect 
is sent, but the connection stays in the connections list. When the sdp client 
tries to create an l2cap channel, l2cap_create_channel_internal finds the 
existing hci connection and tries to use it, but the disconnection will 
complete in a moment.
I think there a few things to consider here:
- Dedicated bonding's hci connection shouldn't be reused, since it will be 
disconnected when the bonding completes.
- I think it will be a good idea for hci_connection_for_bd_addr_and_type to 
check the hci connection's state and make sure its usable. But it means that 
whenever someone sends a hci_disconnect command, the hci connection's state 
must be updated.
- Another option is to a add a reference count to the hci connection.

Original issue reported on code.google.com by kob...@mce-sys.com on 18 Jun 2014 at 12:11

GoogleCodeExporter commented 9 years ago 
Just to make sure this what was causing the problem, I've commented the line 
which sends the disconnection when the bonding completes (in hci_run):
hci_send_cmd(&hci_disconnect, connection->con_handle, 0x13);  // authentication 
done

And the SDP query worked properly.

Original comment by kob...@mce-sys.com on 18 Jun 2014 at 12:28

GoogleCodeExporter commented 9 years ago 
Event "dedicated bonding complete" is now sent after the HCI connection is 
closed, SVN r2745

Original comment by mila@ringwald.ch on 21 Aug 2014 at 1:28