On Tue, 24 Dec 2013, Alan Stern wrote: > > This is my analysis, how do you think? > > There is another reason why usb_hub_to_struct_hub() could return NULL: > if usb_get_intfdata(hdev->actconfig->interface[0]) is NULL. This could > happen if recursively_mark_NOTATTACHED() is called _while_ > hub_disconnect() is running. > > There should be locking to prevent this, but there isn't. That's what > we need to fix. It's not an easy problem. Can you figure out a > correct solution? I think this will fix it. Take a close look and do some careful testing. Alan Stern Index: usb-3.13/drivers/usb/core/hub.c =================================================================== --- usb-3.13.orig/drivers/usb/core/hub.c +++ usb-3.13/drivers/usb/core/hub.c @@ -1622,11 +1622,14 @@ static void hub_disconnect(struct usb_in hub->error = 0; hub_quiesce(hub, HUB_DISCONNECT); - usb_set_intfdata (intf, NULL); - for (i = 0; i < hdev->maxchild; i++) usb_hub_remove_port_device(hub, i + 1); + + /* Avoid races with recursively_mark_NOTATTACHED() */ + spin_lock_irq(&device_state_lock); hub->hdev->maxchild = 0; + usb_set_intfdata(intf, NULL); + spin_unlock_irq(&device_state_lock); if (hub->hdev->speed == USB_SPEED_HIGH) highspeed_hubs--; -- 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
