23 Sep, 2008

3 commits

  • The USB transport specification for Bluetooth splits the ACL and SCO
    handling into two separate interfaces. In Linux it possible to probe
    and disconnect these interfaces independently. So make sure that both
    interfaces are tightly bound together.

    This fixes the suspend regression that some people have expierenced.

    Signed-off-by: Oliver Neukum
    Signed-off-by: Marcel Holtmann

    Marcel Holtmann
     
  • The btusb driver contains two typos that result in some buggy behavior,
    but the impact is not immediately visible.

    During initialization the submitting of interrupt URBs might fail and
    then make sure to remove the correct flag and not one of the hci_dev
    flags.

    When closing down the interface make sure to kill the anchor for the
    ISOC URBs and not kill the interrupt URBs twice.

    Also cancel any scheduled work when closing down the interface.

    Signed-off-by: Marcel Holtmann

    Marcel Holtmann
     
  • The newer MacBooks contain a Broadcom based Bluetooth chip and to make
    this work properly, HCI_Reset must be send first. If HCI_Reset is not
    used then a lot of I/O errors show up and its triggers packets from
    non-existent ACL links.

    Signed-off-by: Marcel Holtmann

    Marcel Holtmann
     

18 Aug, 2008

1 commit

  • The new generic driver for Bluetooth USB devices was missing proper
    SCO support. The driver now claims the second interface for these USB
    devices to allow the flow of SCO packets. It also handles switching
    of the alternate setting and re-submission of isochronous URBs.

    The btusb driver is now a full replacement for hci_usb and thus the
    experimental tag has been removed and this driver is promoted as
    preferred one.

    Signed-off-by: Marcel Holtmann

    Marcel Holtmann
     

08 Aug, 2008

1 commit


22 Oct, 2007

1 commit