Hi,
I created few patches to stop kernel from crashing when SCO channel is requested
to shut down, just after its creation. Some crashed happen because
SCO socket shutdown (sco_sock_shutdown) is called before Synchronous Connection
Complete (sco_connect_cfm) is received or because the thread handling it was
preempted halfway by SCO shutdown.
These are not patches I would like to have accepted immediatelly, but are here
show what had to be done to stop crashes, so we could start discussion how to fix
it properly.
Now, with crashes out of the way I was able to see another defect which is a
Disconnect message sent to quickly, before there is a handle associated with
the connection. Effectively, we fail to disconnect SCO. CS error for Disconnect
is not handled at all.
> HCI Event: Connect Request (0x04) plen 10
< HCI Command: Accept Synchronous Connection Request (0x01|0x0029) plen 21
> HCI Event: Command Status (0x0f) plen 4
< HCI Command: Disconnect (0x01|0x0006) plen 3
Handle: 0
> HCI Event: Command Status (0x0f) plen 4
Status: Unknown Connection Identifier (0x02)
> HCI Event: Synchronous Connect Complete (0x2c) plen 17
Handle: 513
I'm using BT subsystem from 4.1.1 bacpk port on top of 3.8 with PREEMPT.
Kuba Pawlak (5):
Bluetooth: Fix crash on SCO disconnect
Bluetooth: Fix locking issue on SCO disconnection
Bluetooth: Fix locking issue during fast SCO reconnection.
Bluetooth: Fix locking of SCO channel to connection
Bluetooth: Fix crash on fast disconnect of SCO
net/bluetooth/sco.c | 28 +++++++++++++++++++++++-----
1 file changed, 23 insertions(+), 5 deletions(-)
--
1.9.3
Intel Deutschland GmbH
Registered Address: Am Campeon 10-12, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Christin Eisenschmid, Prof. Dr. Hermann Eul
Chairperson of the Supervisory Board: Tiffany Doon Silva
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928
��.n��������+%������w��{.n�����{��� �^n�r������&��z�ޗ�zf���h���~����������_��+v���)ߣ�
