On Mon, Jun 06, 2016 at 01:44:23PM +0200, Hans de Goede wrote: > Hi All, > > While looking at libusb today I ended up looking at the > reap-after-disconnect code. > > What stands out is that libusb expects to be able to > reap all outstanding urbs on a device on receiving > a POLL_ERR status from poll (on supported kernels). > > But the usbfs poll implementation will return POLL_ERR > as soon as ps->dev->state == USB_STATE_NOTATTACHED, > and the kernel usb code sets state = USB_STATE_NOTATTACHED > some time before usbdev_remove() gets called and moves > all pending urbs to the async_completed list. > > So if libusb ends up calling poll between the > setting state = USB_STATE_NOTATTACHED and usbdev_remove() > being called then it ends up not reaping all urbs > since it will stop monitoring the device after > the first POLL_ERR. > > I'll write a libusb fix for this, but I'm wondering if this > is something which we also want to fix on the kernel side? Would fixing this in the kernel mean that libusb fixes are not needed? If so, I would recommend doing this. thanks, greg k-h -- 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
