It looks like I've run into some inconsistency in the USB stack behavior.
The USB stack maintains, among others, two states for the attach
USB device: authorized and owned. Authorization state is accessible to
the user space code through correspondent sysfs files, the ownership
can be set by claiming the hub's port with ioctl call. Both state may
be set before the device is attached, by access the hub settings. When
the new device is attached, both authorization and ownership prevent
the kernel USB stack from setting the newly attached device
configuration, but when the device is authorized, the ownership state
is ignored. It looks like ignoring the ownership state on
authorization make the stack behavior inconsistent; it also prevents
the user space code from completely overriding configuration
selection, important for implementing workarounds for bugs in the
device configuration selection.
The following patch makes the stack behavior more consistent, by
moving ownership test into usb_choose_configuration - the later
function is used both by generic_probe and usb_authorize_device
Signed-off-by: Joseph Hindin <hindin@gmail.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
int insufficient_power = 0;
struct usb_host_config *c, *best;
+ if (usb_device_is_owned(udev))
+ return 0;
+
best = NULL;
c = udev->config;
num_configs = udev->descriptor.bNumConfigurations;
/* Choose and set the configuration. This registers the interfaces
* with the driver core and lets interface drivers bind to them.
*/
- if (usb_device_is_owned(udev))
- ; /* Don't configure if the device is owned */
- else if (udev->authorized == 0)
+ if (udev->authorized == 0)
dev_err(&udev->dev, "Device is not authorized for usage\n");
else {
c = usb_choose_configuration(udev);