Bluetooth: close HCI device when user channel socket gets closed
With 9380f9eacfbbee701daa416edd6625efcd3e29e1 the order of unsetting
the HCI_USER_CHANNEL flag of the HCI device was reverted to ensure
the device is first closed before making it available again.
Due to hci_dev_close checking for HCI_USER_CHANNEL being set on the
device it was never really closed and was kept opened. We're now
calling hci_dev_do_close directly to make sure the device is correctly
closed and we keep the correct order to unset the flag on our device
object.
Signed-off-by: Simon Fels <email address hidden>
Signed-off-by: Marcel Holtmann <email address hidden>
(cherry picked from bluetooth-next 4c63f83c2c2e16a13ce274ee678e28246bd33645)
BugLink: http://bugs.launchpad.net/bugs/1489327
Signed-off-by: Tim Gardner <email address hidden>
We need to carry and build the HCI SMD driver from now on as part of the backports tree to
keep the dependency on the right bluetooth layer.
The driver itself copied and modified version of the one still shipped as part of the
older bluetooth stack we have in the tree under net/bluetooth. Code was taken from
drivers/bluetooth/hci_smd.c and adjusted to fit the changed HCI driver API.
UBUNTU: backports: respect already existing redefinitions in the flo tree
As the flo tree isn't a clean 3.4 one without further modifications and backported things
from upstream we need to respect already present symbols from newer upstream kernel
versions and drop them from our compatiblity layer.
