Re: [PATCH] hci-usb bugfix

[Marcel Holtmann <marcel@holtmann.org>]

Hi Sebastian,

> this is a little patch against 2.6.6-rc3 which fixes the Oops when
> unplugging an USB Bluetooth device.
> 
> hci_usb_disconnect() got called recursively which caused
> sysfs_hash_and_remove() finally to dereference a NULL pointer.
> 
> --- SNIP ---
> diff -uNr linux-2.6.6-rc3.old/drivers/bluetooth/hci_usb.c linux-2.6.6-rc3/drivers/bluetooth/hci_usb.c
> --- linux-2.6.6-rc3.old/drivers/bluetooth/hci_usb.c     2004-05-09 20:25:00.000000000 +0200
> +++ linux-2.6.6-rc3/drivers/bluetooth/hci_usb.c 2004-05-09 20:28:30.000000000 +0200
> @@ -986,8 +986,21 @@
> 
>         hci_usb_close(hdev);
> 
> -       if (husb->isoc_iface)
> +       if (husb->isoc_iface) {
> +#if 0
>                 usb_driver_release_interface(&hci_usb_driver, husb->isoc_iface);
> +#else
> +               /* do the same as usb_driver_release_interface would do,
> +                * except calling diconnect().
> +                * usb_driver_release_interface() _does_ check if
> +                *  are in disconnect() or add, but I really dunno 
> +                *  where dev->driver_list or dev->bus_list gets set.
> +                *      -- yath
> +                */
> +               husb->isoc_iface->dev.driver = NULL;
> +               usb_set_intfdata(husb->isoc_iface, NULL);
> +#endif
> +       }
> 
>         if (hci_unregister_dev(hdev) < 0)
>                 BT_ERR("Can't unregister HCI device %s", hdev->name);
> --- SNAP ---
> 
> Please apply it,

thanks for finding the reason for this problem, because the only thing
that I knew, was that enabled SCO support caused this problem. I prefer
to let the USB guys fix it, because this bug should be present for all
drivers that claimed more than one interface.

Regards

Marcel