Hi,
it seems to me that we have a subtle bug in usb_reset_device().
If we reset a device we may reprobe an interface. If that causes
another interface to be claimed we may call post_reset() for that
interface without having called pre_reset().
I think it is necessary to do the rebinding in a second pass after
doing the notifications. But this is subtle. Alan, what do you think?
Regards
Oliver
commit e6936977ba72a576a2365502dafb9f92d4260b32
Author: Oliver Neukum <oliver@xxxxxxxxxx>
Date: Wed Nov 11 18:08:26 2009 +0100
usb:fix unbalanced callbacks in usb_reset_device()
if a driver needs to be rebound it may claim secondary interfaces
in probe() for such interfaces post_reset() must not be called.
Therefore rebinding must happen after all notifications.
diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
index 691b0ea..5a3f797 100644
--- a/drivers/usb/core/hub.c
+++ b/drivers/usb/core/hub.c
@@ -3696,13 +3696,15 @@ int usb_reset_device(struct usb_device *udev)
if (!rebind && cintf->dev.driver) {
drv = to_usb_driver(cintf->dev.driver);
- if (drv->post_reset)
- rebind = (drv->post_reset)(cintf);
- else if (cintf->condition ==
- USB_INTERFACE_BOUND)
- rebind = 1;
+ rebind = (drv->post_reset)(cintf);
}
- if (ret == 0 && rebind)
+ cintf->needs_binding = rebind;
+ }
+
+ /* second pass to not call post_reset on new drivers */
+ for (i = config->desc.bNumInterfaces - 1; i >= 0; --i){
+ struct usb_interface *cintf = config->interface[i];
+ if (ret == 0 && cintf->needs_binding)
usb_rebind_intf(cintf);
}
}
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
